Documentos de Académico
Documentos de Profesional
Documentos de Cultura
ISBN 92 75 11505 2
0
Publications of the Pan American Health Organization enjoy copyright protection in accordance with the provisions of Protocol 2 of the Universal Copyright Convention. For rights of
reproduction or translation of PAHO publications, in part or in toto, application should be
made to the Editorial Service, Pan American Health Organization, Washington, D.C. The Pan
American Health Organization welcomes such applications.
The designations employed and the presentation of the material in this publication do not
imply the expression of any opinion whatsoever on the part of the Secretariat of the Pan
American Health Organization concerning the legal status of any country, territory, city, or
area of its authorities, or concerning the delimitation of its frontiers or boundaries.
The mention of specific companies or of certain manufacturers products does not imply
that they are endorsed or recommended by the Pan American Health Organization in preference
to others of a similar nature that are not mentioned.
The authors alone are responsible for the views expressed in this publication.
CONTENTS
Page
Preface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
ix
xiii
3
18
20
24
27
31
33
37
42
46
60
67
72
75
80
85
99
103
112
122
125
135
142
149
...
111
167
176
180
184
198
223
235
243
252
27 1
279
290
295
312
319
350
358
368
384
393
40 1
406
C. STUDIES IN EPIDEMICS
415
.................................................................
419
Congenital Cataract Following German Measles in the Mother-N. McAlirter Gregg . . . . . . 426
Rickettsialpox-A Newly Recognized Rickettsia1 Disease-Morris Greenberg, Ottauio J.
Pellitteri, and William L. Jellison . . . , . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 435
Outbreak of Paralysis in Morocco Due to Ortho-Cresyl Phosphate Poisoning-Honor V
Smith and J . M. K. Spalding . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 442
Adenocarcinoma of the Vagina: Association of Maternal Stilbestrol Therapy with Tumor
Appearance in Young Women-Arthur L. Herbst, Howard Ulfelder, and David C.
Poskanzer . . . . . . . . . . . . . . . . . . . . . . . . .
.........................................
446
Salmonellosis Associated with Marijuana: A Multistate Outbreak Traced by Plasmid
Fingerprinting-David N . Taylor, I . Kaye Wachsmuth, Yung-Hui Shangkuan, Emmett V.
Schmidt, Timothy J. Barrett, Janice S. Schrader, Charlene S. Scherach, Harry B. McGee,
Roger A. Feldman, and Don J. Brenner . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
45 1
iv
A Study of the Relation of Family Income and Other Economic Factors to Pellagra
Incidence in Seven Cotton-Mill Villages of South Carolina in 1916-Joseph Goldberger,
G. A. Wheeler, and Edgar Sydenstricker . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Mortality from Lung Cancer in Asbestos Workers-Richard Doll . . . . .
..
An Approach to Longitudinal Studies in a Community: The Framing
Thomas R. Dawber, William B. Kannel, Lorna P. Lye11 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Mortality in Relation to Smoking:
Doll and Austin Bradford Hill .....................
..........................
Final Report of a Prospective Study of Children Whos
ad Rubella in Early
Pregnancy-Mary D. Sheridan .
Radiation-Related Leukemia in Hiroshima and Nagasaki, 1946-1964 I . Distribution,
Incidence, and Appearance Time-0. Joseph Biztozero, Jr., Kenneth G. Johnson, and
Antonio Ciocco, with the collaboration of Takashi Hoshino, Takashi Itoga, Shigeki Toyoda, and
Sho Kawasaki ..............................
Incidence and Prediction of Ischemic Heart Disease in London Busmen-J.N. Morris,
Aubrey Kagan, D.C. Pattison, M.J. Gardner, and P.A.B. Raffle .........................
....
An Overview of the Risk Factors for Cardiovascular Disease-William B. Kannel
Psychiatric Disorders in Foster Home Reared Children of Schizophrenic MothersLeonard L. Heston . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
584
610
619
63 1
668
676
687
699
719
F. EXPERIMENTAL STUDIES
726
73 1
747
753
759
766
780
786
809
829
A. PREVENTIVE SERVICES
..............................................................
838
855
868
874
888
B. MEDICAL CARE
Measuring the Quality of Medical Care Through Vital Statistics Based on Hospital Service
Areas: I. Comparative Study of Appendectomy Rates-Paul A. Lembcke . . . . . . . . . . . . . 894
Case-Fatality in Teaching and Non-Teaching Hospitals, 1956-1959-L. Lipworth, J. A . H.
903
Lee,andJ. N.Morris . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Rehabilitation of Nursing Home Residents-Howard R. Kelman and Jonas N.Muller . . . . . . 909
917
The Iceberg: Completing the Clinical Picture in General Practice-J. M. Last
The Burlington Randomized Trial of the Nurse Practitioner: Health Outcomes of
Patients-David L. Sackett, Walter 0. Spitzer, Michael Gent, and Robin S. Roberts, in
collaboration with W. Ian Hay, Georgie M. Lefroy, G. Patrick Sweeny, Isabel Vandewlist, John
C. Sibley, Larry W. Chambers, Charles H. Goldsmith, Alexander S. MacPherson, and Ronald
. . . . . . . . . . . . . . . . . . 923
G. McAuley . . . . . . .
Health Service Input and Mortality Output in Developed Countries-A. L. Cochrane,
930
A. S . St. Leger, and F. Moore . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
vi
.......................................................................
Appendix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
vii
967
986
PREFACE
Epidemiologists, health planners, and administrators from countries
throughout the Americas Region, among others, met at a seminar in
Buenos Aires, Argentina in November 1983, to discuss and analyze the role
of epidemiology in the developing countries of the Western Hemisphere.
After formulating and analyzing ideas and initiatives on the use and future
prospects of epidemiology in Latin America, the participants made important recommendations for adjusting epidemiology's practice to current
needs.
They agreed that the most important epidemiological issue in the Region
has been the change in the health profile of the population as a result of the
social, economic, environmental and demographic changes. While communicable diseases persist in most countries of the Americas, they have been
increasingly joined by noninfectious diseases that strike mostly adults and
the elderly, by accidents, and by illnesses linked to the workplace and to
environmental pollution.
For the industrialized nations, the evolution of patterns of disease
spanned more than a century and fell into three fairly distinct stages. The
first, marked by infectious diseases associated with poverty, malnutrition,
and poor environmental and personal hygiene, gradually gave way thanks
to better housing and sanitation, greater availability of safe drinking water,
and vaccination services. In the second stage, degenerative diseases such as
heart disease, cerebrovascular accidents, and cancer gradually began to
replace infectious diseases as the leading causes of death. Finally, the third
stage reflects a growing concern with health problems caused by exposure
to environmental pollution and to changing social conditions in families,
communities, and the workplace which foster violence, alcohol abuse, and
drug addiction.
One of the distinguishing features of the health situation in developing
countries is that, whereas developed nations went through all three stages in
more than a century, developing countries must face all three at once.
Consequently, health conditions in the Americas have become a veritable
epidemiological mosaic.
'The report and the working documents of the Buenos Aires seminar were published in
1984, in PAHO publication PNSP 84-47.
ix
Preface
Preface
xi
INTRODUCTION
The Challenge of Epidemiology represents the culmination of a collaborative
effort. Two units of the Pan American Health Organization, Health Situation Analysis and Trend Assessment and Health Manpower Development,
worked on the book. Dr. Clovis H. Tigre, an epidemiologist with the former,
coordinated its production, and a working group of editors comprised of
four eminent epidemiologists-Carol Buck of the University of Western
Ontario, Alvaro Llopis of the Central University of Venezuela, Enrique
Najera of the University of Seville, and Milton Terris of thejournal of Public
Health Policy-selected works for the anthology and provided the background discussions for each section. These discussions, designed to provide
a framework for the anthologized works, were taped and transcribed from
actual sessions held by the editors. They have been kept in their original
style to preserve the editors opinions and the spirit of their discussions.
The editors first met to discuss objectives and overall procedures. At that
time it was decided that selections for the anthology of the book would be
chosen from suggested entries submitted by experts in epidemiology from
throughout the world. Letters were sent to approximately one hundred
such individuals asking them to nominate up to ten works which they
considered landmarks in the development of the discipline, outstanding
contributions to the field, or examples of the advancement of an innovative
concept. From this vast storehouse of articles, chapters of books, and
excerpts of works, the four editors would then carefully select most of the
anthology, seeking a well-balanced, final version. The remaining submissions would be integrated into an extensive epidemiology bibliography.
At a subsequent meeting the editors organized the selected articles in five
sections. The first two, Historical Development and From the Old to the
New Epidemiology, trace the historical evolution of the discipline. Section
3, Etiologic Investigations, and section 4, Health Services and Health
Policies, address the application of epidemiology. Section 5 , Perspectives
and Prospects, offers the editors views on the future of epidemiology. The
first four sections contain both discussions and a collection of reprinted
...
Xlll
xiv
Introduction
PART I
HISTORICAL DEVELOPMENT
DISCUSSIONS
NAJERA:
LLOPIS:
NAJERA:
TERRIS:
Discusions
TERRIS:
NAJERA:
Discussions
TERRIS:
NAJERA:
BUCK:
TERRIS:
NAJERA:
TERRIS:
I am not sure I accept what you are saying. I think the time
when infectious disease became all important was after
Pasteur and Koch, because then they could do something
about it. Then the whole field became infectious disease. It
was the success that created the interest, not the existence
of the problem. The reason that the Latin Americans today
think only in terms of infectious diseases is that they can do
something about them.
Let me put it another way. Noninfectious diseases existed
long before 1940, but there was no noninfectious disease
epidemiology of any great significance until after some
successes were achieved-until, for example, the relation of
cigarette smoking to lung cancer was discovered and the
risk factors for coronary heart disease were demonstrated.
Once there was some success, everyone jumped on the
bandwagon.
I dont accept the thesis that it was the Industrial Revolution which caused an emphasis on infectious diseases. They
were the most important diseases long before the Industrial
Revolution. Look at plague. Look at all the great epidemics
of the Middle Ages. They dominated the field. As a matter
of fact, look at the early books like Epidemiologia Espafiola,
they are all about pestes, plagues. Its all infectious disease; it
never was anything else. The only time they get into occupational disease and toxicology is with the Industrial
Revolution.
NAJERA:
Well, we could talk a lot about this. But what you said about
the focus shifting to noninfectious diseases in the 1940s is
more complicated. And the reason being success? Not necessarily. Take cancer and smoking. Its been 40 years since
the epidemiological discoveries, and we still are about the
same. Its not that we have had actual success, but, rather,
we have the possibility of success.
And as far as infectious diseases not becoming important
until Koch and Pasteur, its not that clear cut. Snows work
was with infectious diseases, and that was 30 years before
Discussions
Koch. So, by the time of Koch and Pasteur the shift had
already occurred.
TERRIS:
NAJERA:
TERRIS:
NAJERA:
TERRIS:
10
NAJERA:
TERRIS:
I could also argue that another big factor, long before the
Industrial Revolution, was the commercial revolution. The
extension of trade on a world scale brought disease from all
over the world. This was a major factor.
NAJERA:
TERRIS:
NAJERA:
TERRIS:
NAJERA:
TERRIS:
NAJERA:
Discussions
11
TERRIS:
NAJERA:
TERRIS:
NAJERA:
In any case, the fact that Snow was able to deal with cholera
through sanitation because he discovered the key role of
the contaminated water pump, is good. But sanitation was
known to be the fundamental thing to avoid epidemics by
the Greeks in the fifth century B.C. and even by the Indus
Valley civilizations of Mohenjo-daro, Harappa, or Taxila
some five o r six centuries before that. Yet sanitation had
only been applied to the very rich. But since overcrowding
was not a problem, they could manage-epidemics came
and went. But the overcrowding of the nineteenth century
was terrible.
Why did Dickens write his stories? Because the conditions of the people were different. They were living worse
than ever in these industrial-town slums. There are descriptions of families of 12 living in one room. This had not
happened before. True, there had always been poor people, but most of them lived in rural areas where there was
more space.
TERRIS:
Well, I must tell you that the plagues of the Middle Ages
were much more devastating than the plagues of the Indus-
12
NAJERA:
This may be, but the descriptions of the Middle Ages and
even of some later periods were narrative accounts without
any data, and most probably were quite exaggerated.
LLOPIS:
I dont think I agree that these were only stories, since these
narratives gave all the historical and economic contexts of
the plague epidemics of the fourteenth century. And, as
Terris has said, whole cities were wiped out. The economic
impact of these epidemics was major. So many people died
that goods and properties, the entire communitys wealth,
were left for far fewer people. The standard of living rose
after each one of these great outbreaks.
NAJERA:
TERRIS:
Discussions
13
LLOPIS:
NAJERA:
TERRIS:
LLOPIS:
Another important investigator was Panum, who addressed the question of incubation periods in his Observations made during the Epidemic of Measles in the Faroe
Islands. He discovered that the age distribution of the
disease in those islands where the virus had not circulated
for a long time was different than in those where it had. In
the former it was the adults who had the measles, which,
under normal conditions they didnt have.
NAJERA:
14
TERRIS:
BUCK:
NAJERA:
Discussions
15
BUCK:
NAJERA:
BUCK:
TERRIS:
NAJERA:
16
was that when Jenner saw that people were dying from the
vaccine, he went back to variolation without telling anyone.
Two or three books have been published recently on this
subject in England. They have copied comments made at
the time about people who died and ones who did not;
what was happening and why; whether the method was
good or not, etc. It was a very, very complicated beginning
for this method. Perhaps Spain backed it more than most
countries, and organized the expedition of Dr. Francisco
Balmis that carried the vaccine around the world. This
expedition should be designated as the first international
health program.
TERRIS:
NAJERA:
BUCK:
NAJERA:
TERRIS:
NAJERA:
TERRIS:
He died?!
NAJERA:
Discussions
17
LLOPIS:
His investigations of childbed fever made in a Vienna maternity clinic in 1846 constitute a solid piece of epidemiological research.
BUCK:
TERRIS:
18
Hippocrates
19
PREFACE
20
though I was on board several other long Channel cruises, one of twelve weeks particularly,
from the 10 August to the 28 October, yet we
had but one scorbutic patient, nor on the other
that I remember had we the least scorbutic appearance. But in those who I have mentioned
the scurvy began to rage after being a month to
six weeks at sea, when the water on board, as I
took particular notice, was uncommonly sweet
and good, and the state of provisions such as
could afford no suspicion of occasioning a general sickness, being the same in quality as in
former cruises. And though the scorbutic people were by the generous liberality of that great
and good commander, the Hon. Captain
George Edgcumbe, daily supplied with fresh
provisions, such as mutton broth and fowls and
even meat from his own table, yet at the expiration of ten weeks we brought into Plymouth 80
men out of a complement of 350 more or less
afflicted with this disease.
Now it was observable that both these cruises
were in the months of April, May and June,
when we had, especially in the beginning of
them, a continuance of cold, rainy and thick
Channel weather, as it is called, whereas in our
other cruises we had generally very fine
weather, except in winter, when, during the
time I was surgeon, the cruises were but short.
Nor could I assign any other reason for the
frequency of this disease in these two cruises
and our exception from it at other times but the
influence of the weather; the circumstances of
the men, ship and provisions being in all other
respects alike. I have more than once remarked
that after great rains or a continuance of close,
foggy weather, especially after storms with rain,
the scorbutic people generally grew worse; but
found a mitigation of their symptoms and complaints upon the weather becoming drier and
warmer for a few days. And I am certain it will
be allowed by all who have had an opportunity
of making observations of this disease at sea, or
will attentively consider the situation of seamen
there, that the principal and main predisposing
Lind
21
* * *
22
...
Lind
23
As I have already said, I have long endeavoured to examine all the symptoms of this dis-
24
Casal
25
26
George Baker
It seems. . . . not to have been without sufficient foundation, that I have for some time suspected, that the cause of this Colic is not to be
sought for in the pure Cyder; but in some,
either fraudulent, or accidental, adulteration.
Upon inquiry, I find that the disease is very
common all over the county of Devon; but that
it particularly infests those parts of the county,
where the greatest quantities of Cyder are
made. I likewise find that it is not only common
among the lower class of inhabitants; but that it
is much more frequent among people of all
ranks, than in other parts of England; and that
it is not intirely confined to the autumnal season. . . .
According to my information, eighty Patients, under the effects of the Devonshire colic,
were admitted into the Bath-hospital in the
course of the last year; forty of whom are said to
have been cured, and thirty-six sent away greatly
relieved. I likewise am informed from the Bathhospital, that the proportion of such Patients
from Devonshire, to that from the counties of
Hereford, Gloucester, and Worcester, is generally as eight to one.
In some letters, which I have lately received
from Dr. Wall, of Worcester, the following facts
are mentioned. The counties of Hereford,
Gloucester, and Worcester, are not, so far as I
know, subject to the colic of Poitou, or any other
endemic illness, unless it may be the rheumatism; which, I think, the inhabitants of Herefordshire are more liable to, than those of some
other counties. There is no Lead, which can
give occasion to that colic, used in any part of
the apparatus for grinding or pressing the apples, or fermenting the liquor. Once indeed, in
a plentiful year of apples, I knew a Farmer,
who, wanting casks, filled a large leaden cistern
with new cyder, and kept it there, till he could
procure hogsheads sufficient to contain the liSource: Excerpted from George Baker, A n E ~ s a yconcernzng
the Cause of the Endemial Colic of Devonshire. London, J. Hughs,
near Lincolns-Inn-Fields, 1777.
27
28
another person, that in many parts of Herefordshire, and the neighbouring counties, the
stones of the mills, are joined together with
putty; (which is whiting, mixed with oil into a
tough paste) and that neither iron nor Lead are
originally used in the construction of them; but
that, if any of the joints, in wearing, happen to
start (which is sometimes the case) they are repaired with iron cramps fastened with lead.
These facts having been well ascertained, I
determined to make use of the first opportunity, which might occur, of informing myself
by experiment, whether or no there are really
marks of solution of Lead in the cyder of Devonshire. Being therefore, in the month of October 1766, at Exeter, I procured some of the
expressed juice of apples, as it flowed from a
cyder-press, lined with Lead, in the parish of
Alfington. On this I made and repeated several
experiments by means of the atramenturn sympatheticurn, or liquor uini probatorius described by
Neuumann; and of the volatile tincture of sulphur. These experiments intirely satisfied me,
that the Must contained a solution of lead. The
same experiments were made on some cyder of
the preceding year. This likewise shewed evident signs of lead contained in it; but in less
proportion than in the Must.
But, being unwilling to make any positive
assertion, solely on the authority of my own
trials, more especially as I had been under the
influence of a preconceived opinion; I brought
with me to London some of the same Must,
which I had examined at Exeter. This Must,
together with some Devonshire cyder of the
preceding year which I purchased of the maker
(who assured me that he used no Lead in any
part of the apparatus for making cyder, except
only what is necessary for composing the
trough, as was mentioned above), were the subject of some experiments, in making which, Dr.
Saunders, an ingenious Gentleman, who
teaches Chemistry, kindly gave me his assistance. . . .
Experiment I.
Baker
29
Observation 111.
There is some nicety required in making this
experiment. The Hepar Sulphurk is not to be
added in any large quantity; for as all the lead is
precipitated upon the first addition, it is easy to
perceive the several successive shades of colour
in the precipitate, until all the lead is separated;
and then the precipitate, upon a farther addition of Hepar Sulphuris, assumes the whiteness
of the precipitate obtained from the Herefordshire cyder, which intitles it to the appellation of Lac Sulphuris. If a large quantity of Hepar
Sulphuris be at once added, the whiteness of the
too copious precipitate is such, as to render the
dark colour of what is first precipitated imperceptible.
Experiment IV.
Some Devonshire cyder was examined by
means of the volatile tincture of sulphur, as in
Experiment 111. A very dark coloured precipitate was obtained. A similar precipitate could
only be obtained from Herefordshire cyder,
after that a weak solution of Saccharum Saturni
was added to it.
Some of the Must (taken from the press in the
parish of Alfington, as was mentioned above)
treated in the same manner with the cyder, produced precipitates of a deeper black colour.
This sufficiently shews, that the solution of
Lead in the Must was stronger than that in the
cyder.
It is a matter of no consequence, whether the
Lead, the existence of which is proved, was applied to the cyder in its state of Must, or in that
of a vinous liquor. However, as the must afforded more considerable signs of impregnation than the cyder, it would seem probable that
the lead was added to the Must; and that, as the
acid, during the fermentation, is in a great
measure converted into alcohol, a proportional
quantity of lead will consequently be precipitated.
The same experiments were afterwards tried
on several other specimens of Devonshire and
the Herefordshire cyder. The result of them
was constantly and uniformly the same as has
been described.
It has been proposed by several Authors, to
detect such adulterations of wines by means of
the vitriolic or of the muriatic acid; which, by
uniting with the lead, will make it precipitate.
30
* * *
May not I presume to hope, that the present
discovery of a poison, which has for many years
exerted its virulent effects on the inhabitants of
Devonshire, incorporated with their daily liquor, unobserved, and unsuspected, may be es-
31
32
Although the Cow Pox shields the constitution from the Small Pox, and the Small Pox
proves a protection against its own future poison, yet it appears that the human body is again
and again susceptible of the infectious matter of
the Cow Pox, as the following history will demonstrate:
William Smith, of Pyrton in this parish, contracted this disease when he lived with a neighbouring Farmer in the year 1780. One of the
horses belonging to the farm had sore heels,
and it fell to his lot to attend him. By these
means the infection was carried to the cows, and
from the cows it was communicated to Smith.
On one of his hands were several ulcerated
sores, and he was affected with such symptoms
as have been before described.
In the year 1791 the Cow Pox broke out at
another farm where he then lived as a servant,
and he became affected with it a second time:
and in the year 1794 he was so unfortunate as to
catch it again. The disease was equally as severe
the second and third time as it was on the first.2
In the spring of the year 1795 he was twice
inoculated, but no affection of the system could
be produced from the variolous matter; and he
has since associated with those who had the
Small Pox in its most contagious state without
feeling any effect from it.
* This is not the case in general--a second attack is commonly very slight, and so, I am informed, it is among the
cows.
In view of the amount and scope of the information presented in this volume, I feel compelled to summarize it here in a few pages. This
summary will present a kind of comparative
description of the main characteristics of workers in wool, cotton, and silk industries as well as
the similarities and differences in workers of
the same trade but who work in different locations and those in workers of the same industry
but who work on different trades. This chapter,
then, will attempt to summarize the information presented in the preceding chapters. . . .
Weavers, who represent a large class, normally work at home in the cellar or in damp,
poorly ventilated ground floors. Most of them
live in the country. They are frequently involved
in agriculture, and, generally speaking, they are
orderly, thrifty, and of good moral character.
However, except for a small number of weavers
who produce fashion fabrics and dry goods,
their earnings are very modest and their eating
habits, health, and housing leave much to be
desired. . . .
Workers at mechanized spinning and cloth
mills work in common areas, where the mixture
of sexes and ages frequently has a negative impact on their inclinations and habits, particularly when they live in a large city or when they
do not return home to their families in the
evening. This work, in addition to being unhealthy for an excessive number of small children, is oftentimes remunerated with very low
wages, which are all the more insufficient considering that these workers have no other
source of income and almost never have standing or savings.
* From Volume I1
33
34
Villerme
be better off than farm workers during ordinary times if it had better habits and methods,
and a spirit of foresightedness.
So as to give a better idea that the working
day for children is, in fact, too long in the mills,
I would like to mention here that usage and
legislation have established for all kinds of
work, including forced labor, a working day of
12 hours, reduced to 10 hours of actual work
with 2 hours of meals. For the workers under
consideration here, the length of the working
day is between 15 and 15% hours, of which 13
to 13% are actually spent working. What a difference! . . .
CHILDREN EMPLOYED IN FACTORIES
35
shown that children who are not under the supervision of their parents in the factories are
more prone than others to accept wayward ideas
and disorderly habits, particularly if they earn
good wages.
Children who work with their parents in the
factories represent between one-tenth and onehalf of the young working force, or an average
of one-third. Even they, however, do not entirely support or help their parents. Furthermore, many parents place their children in
other factories than where they themselves
work so that if the parents' factory stops working, the entire family will not be out of work at
the same time.
All available information indicates that children who begin working at factories at the age
of 6 have no education whatsoever; and normally those children who begin working before
10 or 11 years of age are not able to read or
write. Although some evening and Sunday
schools have been opened, children who work
12 to 14 hours a day, or who have worked the
night before, are too exhausted to attend
classes. Moreover, the indifference of parents in
this respect is normally very high. Religious instruction, we have been assured, is generally
accorded the necessary time. . . .
GENERAL CONSIDERATIONS
It has been said that whenever a great number of persons are kept together in a closed
area, their health suffers. If we were to extend
this statement to factories, the facts stated above
would still not confirm this. N o disease is exclusively characteristic of a given kind of factory, although some are more frequent when
the conditions in which the workers live promote their development.
In cotton-spinning mills, for example, coughing, pulmonary inflammations, a n d t h e
dreaded phthisis attack and debilitate many
workers employed in the threshing or in the
initial carding stage; according to my information, these same diseases also have a strong impact on the workers who fasten, sweep, or unplug wadding, those who inhale cotton dust and
fluff, and among the hand weavers.
Even though there are many victims of pulmonary inflammations and phthisis, their premature deaths do not seem to me any more
deplorable than the development of scrofula
36
all the troubles that preceded and that will follow said change.
The excessive mortality among families of
workers employed in the cotton spinning and
weaving mills in Mulhouse mainly affects the
younger age groups. In fact, one-half of the
children born to the class of manufacturers,
businessmen, and factory managers reach the
age of 29, whereas one-half of the children of
weavers and factory workers in the spinning
mills will die, as hard as it may seem to believe,
before the age of 2. . . .
No one will deny that there are some professions that are essential and that are just as unhealthy as cotton threshing: the skinner of rabbit and hare pelts, the miner who extracts
mercury from the earth, the night watchman,
the sewerman, the preparer of white lead and
other chemical reagents, etc. These professions,
everyone would agree, are neither less dirty, nor
less toilsome, nor do they always give those who
perform them a happier outlook, a broader
horizon, more space, cleaner air, or better
wages than many of the jobs in the manufacturing sector.
Professions most often affect the health of
persons and their families in an indirect, mediate way through the conditions involving food,
clothing, lodging, fatigue, duration of work,
customs, etc. in which the workers exist. This
rule should be regarded as a general one. The
danger presented by dust to some workers that
inhale it in cotton-spinning mills would be an
exception, as would be the rather frequent accidents that occur during the work day. These are
normally injuries to hands and fingers that are
caught in machines or gears. Sometimes bones
are broken, limbs are severed, or even death can
occur. These accidents are always the fault of
either the manufacturer who has neglected to
isolate or surround the dangerous parts of machines with a casing or screen, or of the workers
themselves, especially children, who neglect to
take safety measures. I do not know how frequent they are, but I believe that the very serious ones are not very numerous and generally
result from oversight on the part of the victims.
Most of these could be prevented by use of the
screens that I mentioned above. Some manufacturers have already made this expense, however,
others, and these are the majority, have not
taken this safety measure. Legislation should be
passed to make this compulsory.
37
38
Table 1. Mortality in the respective ages during the measles epidemic of 1846, and a comparison of
these rates with those usual on the Faroes.
Ages
From 1835-1845
inclusive, died
yearly, by average
computation at the
respective ages
Under
1
1-10
10-20
20-30
30-40
40-50
50-60
60-70
70-80
80- 100
18 1/11
7 3/11
5 5/11
6 6111
6 2111
7 4/11
5 5111
8 2111
14 10/11
16 9/11
Total
96 3/11
Percent of persons
of the resp. ages
taken by death
yearly, for
1835-1845,
reckoned from the
census of 1845
50
6
5
8
13
18
28
31
30
26
Percent of persons
of the resp. ages
died in first twothirds of 1846,
counted from
census of 1845 and
my own notes
10 9/11
611 1
511 1
11/22
17\22
1 1/11
10111
2
6 5/10
16 9/11
30
2
2
13
26
Number of times
mortality in first
two-thirds of 1846
was greater than
that usual in an
ordinary whole year
About 2 9/11
0
6/11
411 1
15122
1/11
8/11
419
8/11
1/11
About
About
About
About
About
About
About
14/11
2 112
2 1/2
5
3 314
2
1 112
215
tality rose from the thirtieth year, until it became greatest for the ages between the fiftieth
and sixtieth years, that is, five times as great as
usual; it then descended again after the sixtieth
year, not because the disease was less dangerous
for those still older, which was by no means the
case, but because it was precisely sixty-five years
ago that measles had last prevailed on the
Faroes, and those who had recovered from the
disease at that time were now spared.
Table 2 shows how the measles epidemic on
the Faroes-irrespective of the difference which
might arise from the fact that no age was ex-
Ages
Mortality in Denmark in
the years 1829-1833 incl.
reckoned for 1000 males.
Under
1
1-10
10-20
20-30
30-40
40-50
50-60
60-70
70-80
80- 100
233
281 261
23
37
60
84
130
144
140
121
366
43
56
60
77
105
123
113
57
199
80 279
60
72
68
81
60
90
164
185
Panum
tendency of epidemics as a whole to decimate
the population of a country. Of course, measles
is not wont under ordinary conditions to menace any but children, but on the Faroes it evidently attacked almost the entire population
without respect to age; and the epidemics in the
aggregate, which prevail in other countries but
partially spare the Faroes, also threaten the entire population, without respect to age. I believe
that I have established that the most essential
cause of the favorable rates of mortality on the
Faroes may be looked for in the freedom of
these islands, because of their situation as well
as their isolated condition as regards commerce,
from many diseases which in other places, in
Denmark, for instance, very considerably increase the mortality. . . .
As to the length of the incubation period, accurate and satisfactory observations have hitherto
been lacking, as far as I know, since some authors regard it as eight days, others as from ten
to fourteen days, and others again assume no
definite stadium contagii latentis. This is not
strange, however, inasmuch as observations in
regard to the subject could not well be made
where a very lively intercourse goes on among
the people, and where each individual comes
into contact with a large number of other individuals, each of whom may be carrying the material of infection with him. Here in Copenhagen, for instance, it can very rarely be said of
a measles patient that he was exposed to infection only once, on this or that day; for it can
hardly ever be proved that he was not in anywise exposed earlier or later, without knowing
it, to the influence of the contagion of measles.
To be able to arrive at some definite result in
reference to this question would call for special
circumstances which might render it possible to
make accurate observations, and these circumstances were offered on the Faroe Islands. The
isolated situation of the villages, and their
limited intercourse with each other, made it
possible in many, in fact in most, cases to ascertain where and when the person who first fell ill
had been exposed to the infection, and to prove
that the contagion could not have affected him
either before or after the day stated. . . .
. . . On the 4th of June a boat with ten men
from Tjornevig had taken part in a catch of
grind at Vestmannhavn; and on the 18th of
June, precisely the fourteenth day following,
the measles exanthem had broken out on all ten
39
40
Panum
for I feared that on it measles would show itself
on others in the house; if this did not happen
on that day, they might perhaps have some
hope of being exempt. As it turned out I was
summoned to Fuglefjord again ten days later
and was met with the outcry: What he said was
correct! On the day he pointed out the measles
broke out, with its red spots, on all nine. . . .
. . . Similarly, towards the end of the epidemic, the disease attacked very slowly in Kuna,
Midtvaag, and Sandevaag. At the height of the
epidemic, in Tjornevig, for example, about
fourteen days after one or several persons had
caught measles, the majority of the residents of
the village were attacked, and only a relatively
small number were spared until fourteen days
after the great onset; but the people in the lastnamed villages fell ill gradually, so that only a
few were attacked fourteen days after those who
took the disease first; fourteen days later, others; about fourteen days after these, still others,
and so on; thus the disease lingered longer in
the villages last attacked than in those that were
infected earlier. Nevertheless, measles preserved withal, at least as far as my experience
went, its definite period of development (from
the reception of the infection to the appearance
of the exanthem); and in fact, I know of no case
where, after a pause of more than fourteen
days, measles had appeared afresh in a village
without reinfection from some other place.
Nevertheless, we cannot deny the possibility
that the infective material may be retained for
quite a while after the cessation of measles, in
wool or clothing, for instance, or in other things
41
It would occupy a long time to give an account of the progress of cholera over different
parts of the world, with the devastation it has
caused in some places, whilst it has passed
lightly over others, or left them untouched; and
unless this account could be accompanied with
a description of the physical condition of the
places, and the habits of the people, which I am
unable to give, it would be of little use.
There are certain circumstances, however,
connected with the progress of cholera, which
may be stated in a general way. It travels along
the great tracks of human intercourse, never
going faster than people travel, and generally
much more slowly. In extending to a fresh island or continent, it always appears first at a
sea-port. It never attacks the crews of ships
going from a country free from cholera to one
where the disease is prevailing, till they have
entered a port, or had intercourse with the
shore. Its exact progress from town to town
cannot always be traced; but it has never appeared except where there has been ample opportunity for it to be conveyed by human intercourse.
There are also innumerable instances which
prove the communication of cholera, by individual cases of the disease, in the most convincing manner. Instances such as the following
seem free from every source of fallacy.
I called lately to inquire respecting the death
of Mrs. Gore, the wife of a labourer, from cholera, at New Leigham Road, Streatham. I found
that a son of deceased had been living and
working at Chelsea. He came home ill with a
bowel complaint, of which he died in a day or
two. His death took place on August 18th. His
mother, who attended on him, was taken ill on
the next day, and died the day following (August 20th). There were no other deaths from
cholera registered in any of the metropolitan
districts, down to the 26th August, within two or
three miles of the above place; the nearest being
Source: Excerpted from Snow on Cholera. Cambridge: Harvard University Press, 1949. By permission of the publisher.
42
Snow
43
44
Diseases which are communicated from person to person are caused by some material
which passes from the sick to the healthy, and
which has the property of increasing and multiplying in the systems of the persons it attacks.
In syphilis, small-pox, and vaccinia, we have
physical proof of the increase of the morbid
material, and in other communicable diseases
the evidence of this increase, derived from the
fact of their extension, is equally conclusive. As
cholera commences with an affection of the alimentary canal, and as we have seen that the
blood is not under the influence of any poison
in the early stages of this disease,' it follows that
the morbid material producing cholera must be
introduced into the alimentary canal-must, in
fact, be swallowed accidentally, for persons
would not take it intentionally; and the increase
of the morbid material, or cholera poison, must
take place in the interior of the stomach and
bowels. It would seem that the cholera poison,
when reproduced in sufficient quantity, acts as
an irritant on the surface of the stomach and
intestines, or, what is still more probable, it withdraws fluid from the blood circulating in the
capillaries, by a power analogous to that by
which the epithelial cells of the various organs
abstract the different secretions in the healthy
body. For the morbid matter of cholera having
the property of reproducing its own kind, must
necessarily have some sort of structure, most
likely that of a cell. It is no objection to this view
that the structure of the cholera poison cannot
be recognised by the microscope, for the matter
of smallpox and of chancre can only be recognised by their effects, and not by their physical
properties.
The period which intervenes between the
time when a morbid poison enters the system,
and the commencement of the illness which
follows, is called the period of incubation. It is,
in reality, a period of reproduction, as regards
the morbid matter; and the disease is due to the
crop or progeny resulting from the small quantity of poison first introduced. In cholera, this
period of incubation or reproduction is much
shorter than in most other epidemic or communicable diseases. From the cases previously detailed, it is shown to be in general only from
'
Snow
45
AUTOBIOGRAPHICAL INTRODUCTION
46
Semmelweis
47
Table 1. Annual births, deaths, and mortality rates for all patients at the two
clinics of the Vienna maternity hospital from 1841 to 1846.
First Clinic
Deaths
Rate
3036
3287
3060
3157
3492
4010
237
518
274
260
24 1
459
7.7
15.8
8.9
8.2
6.8
11.4
20 042
1989
Births
1841
1842
1843
1844
1845
1846
Total
Avg.
9.92
Births
Second Clinic
Deaths
2442
2659
2739
2956
324 1
3754
86
202
164
68
66
105
17 791
691
Rate
3.5
7.5
5.9
2.3
2.03
2.7
3.38
48
1841
1842
1843
1844
1845
1846
Births
First Clinic
Deaths
Rate
Births
2813
3037
2828
2917
320 1
3533
177
279
195
25 1
260
235
6.2
9.1
6.8
8.6
8.1
6.5
2252
2414
2570
2739
3017
3398
Second Clinic
Deaths
91
113
130
100
97
86
Rate
4.04
4.06
5.05
3.06
3.02
2.05
Semmelweis
other persons it causes rheumatic fever. In
puerperae, mistakes in diet induce childbed
fever. In others, similar mistakes cause only gastric fever.
Becoming convinced that childbed fever is
not restricted to puerperae and that it can begin
during birth or even in pregnancy, one may
ignore the puerperal state and focus on the
unique composition of the blood during pregnancy. But even if we adopt such an approach,
what predisposes the newborn to puerperal disease? Surely not the puerperal condition of
their genitals. Do both male and female have
the blood composition uniquely characteristic
of pregnancy? The occurrence of childbed
fever among the newborn shows that the very
conception of puerperal fever is erroneous.
Because Vienna is so large, women in labor
often deliver on the street, on the glacis,5 or in
front of the gates of houses before they can
reach the hospital. It is then necessary for the
woman, carrying her infant in her skirts, and
often in very bad weather, to walk to the maternity hospital. Such births are referred to as
street births. Admission to the maternity clinic
and to the foundling home is gratis, on the
condition that those admitted be available for
open instructional purposes, and that those fit
to do so serve as wet nurses for the foundling
home. Infants not born in the maternity clinic
are not admitted gratis to the foundling home
because their mothers have not been available
for instruction. However, in order that those
who had the intention of delivering in the maternity hospital but who delivered on the way
would not innocently lose their privilege, street
births were counted as hospital deliveries. This,
however, led to the following abuse: women in
somewhat better circumstances, seeking to
avoid the unpleasantness of open examination
without losing the benefit of having their infants accepted gratis to the foundling home,
would be delivered by midwives in the city and
then be taken quickly by coach to the clinic
where they claimed that the birth had occurred
unexpectedly while they were on their way to
the clinic. If the child had not been christened
While Semmelweis was in the first clinic, Vienna was still
surrounded by medieval fortifications. The glacis was a broad
earthwork that sloped away from the city and that constituted
part of the fortifications. Between 1857 and 1865 the city
walls were demolished and were replaced by gardens, bode-.
vards, and public buildings.
49
50
Semmelweis
this opinion the number of students was reduced from forty-two to twenty. Foreigners were
almost entirely excluded, and examinations
were reduced to a minimum. The mortality rate
did decline significantly in December 1846, and
in January, February, and March of 1847. But
in spite of these measures, fifty-seven patients
died in April and thirty-six more in May. This
demonstrated to everyone that the view was
groundless. To further the readers understanding, Table 3 shows the mortality figures
for 1846 and for the first five months of 1847.
We will come back later to the fact that from
December 1846 through March 1847 the mortality rate declined, and that it climbed back u p
again in April and May 1847.8
Recommendations based on studies of the
cause of the great mortality in the first clinic all
involved one inexplicable contradiction: given
the concept of an epidemic, and given that the
commissions did not have the power to change
the atmospheric-cosmic-terrestrial conditions of
Vienna, they should have concluded that no
remedies were possible. But they did not draw
this conclusion, even though they considered
the deaths an epidemic. What does one do to
shorten the duration or to prevent the recurrence of a cholera epidemic? They attributed
the disease to one or more of the previously
identified endemic causes. They did not, however, identify it as an endemic disease, which
would have been appropriate, but rather as an
epidemic. In general, the unfortunate confusion in the concepts of epidemic and endemic
disease delayed discovery of the true cause of
childbed fever.
In classifying puerperal disease as epidemic
or endemic, one must disregard entirely the
number of patients who become ill or die. The
cause of the illness o r death determines
whether the disease is epidemic or endemic.
Epidemic puerperal fever is induced by atmospheric-cosmic-terrestrial influences; the concept of an epidemic does not stipulate whether
one or one hundred persons become ill. If
puerperal fever is caused by endemic factorsthat is, by factors whose operation is limited to a
specific location-then puerperal fever is endemic, and it is immaterial whether one or one
hundred individuals become ill. This follows
51
Deaths
Rate
Feb.
Mar.
Apr.
May
Jun.
Jul.
Aug.
Sept.
Oct.
Nov.
Dec.
336
293
31 1
253
305
266
252
216
271
254
297
298
45
53
48
48
41
27
33
39
39
38
32
16
13.39
18.08
15.43
18.97
13.44
10.15
13.10
18.05
14.39
14.98
10.77
5.37
1847
Jan.
Feb.
Mar.
Apr.
May
311
912
305
3 12
294
10
6
11
57
36
3.21
1.92
3.60
18.27
12.24
1846
Jan.
52
The reader can appreciate my perplexity during my first period of service when I, like a
drowning person grasping at a straw, discontinued supine deliveries, which had been customary in the first clinic, in favor of deliveries
from a lateral position. I did this for no other
reason than that the latter were customary in
the second clinic. I did not believe that the
supine position was so detrimental that additional deaths could be attributed to its use. But
in the second clinic deliveries were performed
from a lateral position and the patients were
healthier. Consequently, we also delivered from
the lateral position, so that everything would be
exactly as in the second clinic.
I spent the winter of 1846-47 studying English. I did this because my predecessor, Dr.
Breit, resumed the position of assistant, and I
wanted to spend time in the large Dublin maternity hospital. Then, at the end of February
1847, Dr. Breit was named Professor of
Obstetrics at the medical school in Tubingen. I
changed my travel plans and, in the company of
two friends, departed for Venice on 2 March
1847. I hoped that Venetian art treasures would
revive my mind and spirits, which had been so
seriously affected by my experiences in the maternity hospital.
On 20 March of the same year, a few hours
after returning to Vienna, I resumed, with rejuvenated vigor, the position of assistant in the
first clinic. I was immediately overwhelmed by
the sad news that Professor Uakob] Kolletschka,
whom I greatly admired, had died in the interim.
The case history went as follows: Kolletschka,
Professor of Forensic Medicine, often conducted autopsies for legal purposes in the company of students. During one such exercise, his
finger was pricked by a student with the same
knife that was being used in the autopsy. I do
not recall which finger was cut. Professor
Kolletschka contracted lymphangitis and phlebitis [inflammation of the lymphatic vessels and
of the veins respectively] in the upper extremity.
Then, while I was still in Venice, he died of
bilateral pleurisy, pericarditis, peritonitis, and
meningitis [inflammation of the membranes of
the lungs and thoracic cavity, of the fibroserous
sac surrounding the heart, of the membranes of
the abdomen and pelvic cavity, and of the membranes surrounding the brain, respectively]. A
few days before he died, a metastasis also
Semmelweis
tact with the genitals of these individuals, creating the possibility of resorption. With resorption, the cadaverous particles are introduced
into the vascular system of the patient. In this
way, maternity patients contract the same disease that was found in Kolletschka.
Suppose cadaverous particles adhering to
hands cause the same disease among maternity
patients that cadaverous particles adhering to
the knife caused in Kolletschka. Then if those
particles are destroyed chemically, so that in examinations patients are touched by fingers but
not by cadaverous particles, the disease must be
reduced. This seemed all the more likely, since I
knew that when decomposing organic material
is brought into contact with living organisms it
may bring on decomposition.
To destroy cadaverous matter adhering to
hands I used chtorzna liquida. This practice began in the middle of May 1847; I no longer
remember the specific day. Both the students
and I were required to wash before examinations. After a time I ceased to use chlorina lzquida
because of its high price, and I adopted the less
expensive chlorinated lime. In May 1847, during the second half of which chlorine washings
were first introduced, 36 patients died-this
was 12.24 percent of 294 deliveries. In the remaining seven months of 1847, the mortality
rate was below that of the patients in the second
clinic (see Table 4).
In these seven months, of the 1841 maternity
patients cared for, 56 died (3.04 percent). In
1846, before washing with chlorine was introduced, of 4010 patients cared for in the first
clinic, 459 died (1 1.4 percent). In the second
clinic in 1846, of 3754 patients, 105 died (2.7
Table 4. Monthly births, deaths, and mortality
rates for all patients at the first clinic
of the Vienna maternity hospital
from June to December 1847.
53
Deaths
Rate
May
Jun.
Jul.
Aug.
Sept.
Oct.
Nov.
Dec.
283
29 1
276
305
313
264
269
26 1
312
299
310
373
10
2
0
2
3
3
1
0
3
7
9
5
3.53
0.68
0.00
0.65
0.99
1.13
0.37
0.00
0.96
2.34
2.90
1.34
Total
3556
45
1848
Jan.
Feb.
Mar.
Apr.
~~
1847
Jun.
Jul.
Aug.
Sept.
Oct.
Nov.
Dec.
Total
Births
Deaths
Rate
268
250
264
262
278
246
273
6
3
5
12
11
11
8
2.38
1.20
1.89
5.23
3.95
4.47
2.93
1841
56
3.04
Avg.
1.27
54
Semmelweis
5.5
Table 6. Annual births, deaths, and mortality rate for all patients at the Vienna
maternity hospital from 1784 to 1848.
~~
Year
Births
Deaths
Rate
Year
Births
Deaths
Rate
1784
1785
1786
1787
1788
1789
1790
1791
1792
1793
1794
1795
1796
1797
1798
1799
1800
1801
1802
1803
1804
1805
1806
1807
1808
1809
1810
1812
1811
1813
1814
1815
1816
284
899
1151
1407
1425
1246
1326
1395
1574
1684
1768
1798
1904
2012
2046
2067
2070
2106
2346
2215
2022
2112
1875
925
855
912
744
1419
1050
1945
2062
259 1
2410
6
13
5
5
5
7
10
8
14
44
7
38
22
5
5
20
41
17
9
16
8
9
13
6
7
13
6
9
20
21
66
19
12
2.11
1.44
0.43
0.35
0.35
0.56
0.75
0.57
0.89
2.61
0.39
2.11
1.16
0.24
0.24
0.96
1.98
0.80
0.38
0.72
0.39
0.40
0.73
0.64
0.81
1.42
0.80
0.63
1.90
1.08
3.20
0.73
0.49
1817
1818
1819
1820
182 1
1822
1823
1824
1825
1826
1827
1828
1829
1830
1831
1832
1833
1834
1835
1836
1837
1838
1839
1840
1841
1842
1843
I844
1845
1846
1847
1848
2735
2568
3089
2998
3294
3066
2872
291 1
2594
2359
2367
2833
3012
2797
3353
3331
3907
4218
4040
4144
4363
4560
4992
5166
5454
6024
5914
6244
6756
7027
7039
7095
25
56
154
75
55
26
214
144
229
192
51
101
140
111
222
105
205
355
227
33 1
375
179
248
328
330
730
457
336
313
567
210
91
0.91
2.18
4.98
2.50
1.66
0.84
7.45
4.94
4.82
8.12
2.15
3.56
4.64
3.97
6.62
3.15
5.25
8.41
5.61
7.98
8.59
3.92
4.96
6.44
6.05
12.11
7.72
5.38
4.63
8.06
2.98
1.28
obstetrics in the second clinic when male students were still admitted report that there was,
at that time, no significant difference in mortality between the clinics. The consistently unfavorable health of patients in the first clinic
dates from 1840, when all male students were
assigned to the first clinic and all female students to the second. After what has been reported, it would be superfluous to explain these
facts further.
Table 1 indicates the difference in mortality
rates between the patients of the two clinics
after the first was devoted exclusively to training obstetricians and the second to training
midwives. This would be the place to provide a
similar table for the years during which female
and male students were divided equally between both clinics. It would show that during
this time the mortality rate was not consistently
larger in the first clinic. However, I do not have
56
1839
1840
Births
First Clinic
Deaths
Rate
Births
2781
2889
151
267
5.4
9.5
2010
2073
Second Clinic
Deaths
91
55
Rate
4.5
2.6
Sernmelweic
57
58
Semmelweis
59
charged the foreigners with being more dangerous than the natives because they were rough
in examinations and, consequently, at any one
time only two foreigners were allowed to attend
the course in practical obstetrics. Everyone,
even those who do not share my opinion, will
agree that the commission acted groundlessly in
imputing guilt to the foreigners. In fact, I alone
held that foreigners were more dangerous than
natives, but not because they examined more
roughly. The reasons that foreigners were more
dangerous than natives lies in the following considerations.
Foreigners come to Vienna to perfect medical
training already begun in their own universities.
They visit pathological and forensic autopsies in
the general hospital. They take courses in
pathological anatomy, in surgery, obstetrics, microscopic surgery of cadavers, they visit the
medical and surgical wards of the hospital, etc.
In a word, they utilize their time as efficiently
and educationally as possible. They have, therefore, many opportunities for their hands to become contaminated with foul animal-organic
matter. Thus, it is no wonder that foreigners,
busy in the maternity hospital at the same time,
are more dangerous for patients. Natives take
the course in practical obstetrics after completing two difficult examinations in order to attain
the degree of Doctor of Medicine. The law stipulates that the minimum preparation time for
these examinations is six months. Thus the
natives have already toiled excessively before
they are admitted into the maternity hospital,
and they regard the time there as a rest. While
enrolled in practical obstetrics, natives do not
concern themselves with other activities that
would contaminate their hands. Indeed, while
working at the maternity hospital, they concern
themselves even less with other aspects of medicine because, after completing the course, they
can perfect their knowledge of medicine to the
highest possible degree. Since the foreigners are
generally able to remain in Vienna only a few
months, they are compelled to work simultaneously in more than one aspect of medicine.
Even so, one cannot impute guilt to the foreigners any more than to me or to all the others
who undertook examinations with contaminated hands. None of us knew that we were
causing the numerous deaths.
'
60
Finlav
Such was the hypothesis which led me to undertake the experimental investigation which I
shall here relate. . . .
Although mosquitoes are found in all latitudes, their abundance varies in different localities. Humboldt and Bonpland, in their Travels in Equinoctial America wrote: The annoyance
suffered from mosquitoes and mncudos in the
torrid zone is not so general as most people
think. On the high plateaux more than 400
toises (2500 feet) above the sea-level, and in very
dry plains, far from large rivers, such as
Cumana and Calabozo, gnats are not much
more abundant than in the most populous parts
of Europe. The influence of dryness and of a
long distance from water-courses, pointed out
by those travels, is easily understood, inasmuch
as the larvae and pupae of the mosquitoes are
aquatic, and the winged insect requires water
for the laying and hatching of its eggs. The
impediment to their propagation at high levels
may consist in the exaggeration of the difficulty
which those insects must always experience in
flying upwards after they have filled themselves
with blood; a difficulty which will be much
more marked in a species having such small
wings as those of the C. mosquito. The rarefaction of the atmosphere at those great heights
necessarily increases that difficulty, and, under
those circumstances, the mosquito will instinctively shun those localities. The above mentioned travelers also relate that a missionary
priest, Bernard0 Zea, had built himself a room
over a scaffolding of palm boards, and they
used to go there at night to dry their plants and
to write their Diary, adding: The missionary
had rightly observed that those insects are more
numerous in the lower strata of the atmosphere,
within 12 to 15 feet from the ground. Further
on they write: As one proceeds towards the
plateau of the Andes, those insects disappear
and the air one breathes becomes pure . . . at a
height of 200 toises (1500 feet) mosquitoes and
zancudos are no longer feared.
Historically the mosquito is one of the insects
most anciently observed. Aristotle and Pliny refer to its proboscis which serves both for piercing the skin and for sucking the blood. The
Greek historian Pausanias, according to Taschenberg, mentions the city of Myus, in Asia
Minor, situated on a bay which had formerly
communicated with the sea but was afterwards
cut off from it: when the water in the lake which
61
62
Finlay
also be hindered by the same cause, from going
far from the place where it has accomplished its
last bite, and, in general, from traveling any
considerable distance through the air without
resting. This circumstance will not prevent,
however, its being conveyed, hidden among
clothes, caught under a hat, inside of a traveling
bag, etc., to considerable distance, after a recent
bite, perhaps carrying upon its mouth-parts the
inoculable germ of the disease. . . .
After this long, but necessary account of the
habits of our Cuban mosquitoes, and of the
Culex mosquito in particular, let us consider by
what means that insect might transmit the
yellow fever, if that disease happens to be really
transmissible through the inoculation of blood.
The first and most natural idea would be that
the transmission might be effected through the
virulent blood which the mosquito has sucked,
amounting to 5 and even to 7 or 9 cubic millimeters, and which, if the insect happens to die
before completing its digestion, would be in excellent conditions to retain during a long time
its infecting properties. It might also be supposed that the same blood which the mosquito
discharges, as excrement, after having bitten a
yellow-fever patient, might be dissolved in the
drinking-water, whereby the infection might be
conveyed if the latter were susceptible of penetrating by the mouth. But the experiments of
Firth and other considerations arising from my
personal ideas regarding the pathogenesis of
yellow fever, forbid my taking into account either of those modes of propagation, as I shall
now explain. When the U. S. Yellow-fever Commission took their leave, two years ago, they
presented us with a valuable collection of microphotographs from preparations made by our
corresponding Member, Dr. Sternberg, showing
what, to me, appeared to be a most striking
feature, namely, that the red blood-globules are
discharged unbroken in the hemorrhages of
yellow fever. This fact taken in connection with
the circumstance that those hemorrhages are
often unattended with any perceptible break in
the blood-vessels, while, on the other hand, they
constitute a most essential clinical symptom of
the disease, led me to infer that the principal
lesion of yellow fever should be sought for in
the vascular endothelium. The disease is transmissible, it attacks but once the same person,
and always presents in its phenomena a regular
order comparable with that observed in the
63
eruptive fevers, all of which circumstances suggested to my mind the hypothesis that yellowfever should be considered as a sort of eruptive
fever in which the seat of the eruption is the
vascular endothelium. The first period would
correspond to the initial fever, the remission to
the eruptive period, and the third period would
be that of desquamation. If the latter phase is
accomplished under favorable conditions, the
patient will only show evidence of an exaggerated transudation of some of the liquid elements of the blood through the new endothelium; if the conditions are unfavorable, a
defective endothelium will have been produced,
incapable of checking the figured elements of
the blood: passive hemorrhages will occur and
the patient may find himself in imminent danger. Finally, assimilating the disease to smallpox and to vaccination, it occurred to me that in
order to inoculate yellow fever it would be necessary to pick out the inoculable material from
within the blood vessels of a yellow-fever patient
and to carry it likewise into the interior of a
blood vessel of the person who was to be inoculated. All of which conditions the mosquito satisfies most admirably through its bite, in a manner which it would be almost impossible for us
to imitate with the comparatively coarse instruments which the most skillful makers could produce.
Three conditions will, therefore, be necessary
in order that yellow fever may be propagated:
1. The existence of a yellow-fever patient into
whose capillaries the mosquito is able to drive
its sting and to impregnate it with the virulent
particles at an appropriate stage of the disease.
2. That the life of the mosquito be spared after
its bite upon the patient until it has a chance of
biting the person in whom the disease is to be
reproduced. 3. The coincidence that some of
the persons whom the same mosquito happens
to bite thereafter shall be susceptible of contracting the disease.
The first of these conditions, since Dr. Ambrosio G. del Valle has been publishing his valuable mortuary tables, we may be sure has never
failed to be satisfied in Havana. With regard to
the 2d and 3d, it is evident that the probabilities
of their being satisfied will depend on the abundance of mosquitoes and on the number of
susceptible persons present in the locality. I
firmly believe that the three above mentioned
conditions have, indeed, always coincided in
64
~~
Firzlay
mosquito to the same patient, D. L. R. at Quinta
Garcini, his attack having then reached its fifth
day and proving fatal on the following one. On
the 2d of August I applied this mosquito to
D. G. B., another of my twenty nonimmunes.
Till the present date (12th), this last inoculation
has not given any result; but, as only 12 days
have elapsed, the case is still within the limits of
incubation.
I have to state that the persons mentioned
above are the only ones who were inoculated
with mosquitoes in the manner described; and
that since June IZth, till now (in the course of
seven weeks), barring my first three inoculated
men, no other case of confirmed or abortive
yellow fever has occurred among the twenty
nonimmunes whom I have had under observation.3
These experiments are certainly favorable to
my theory, but I do not wish to exaggerate their
65
66
MORTALITY OF MINERS,
1848-1853, CORNWALL
67
Ages
Metal
miners
Males,
exclusive
of miners
15-25
25-35
35-45
45-55
55-65
65-75
8.90
8.96
14.30
33.51
63.17
111.23
7.12
8.84
9.99
14.76
24.12
58.61
68
Ages
Cornish
metal
miners
Northern
coal
miners
15-25
25-35
35-45
45-55
55-65
65-75
8.90
8.96
14.30
33.51
63.17
111.23
8.50
8.49
10.13
16.81
24.43
65.16
Ages
Metal
miners
Males,
exclusive
of miners
15-25
25-35
35-45
45-55
55-65
65-75
9.44
9.57
15.12
29.74
63.21
110.51
7.50
8.32
10.08
12.50
19.96
53.31
Ages
Metal
miners
Males,
exclusive
of miners
15-25
25-35
35-45
45-55
55-65
65-75
5.77
4.15
7.89
19.75
43.29
45.04
3.30
3.83
4.24
4.34
5.19
10.48
Farr
Table 5. Average annual number of deaths per
1000 miners in Cornwall from all causes and from
pulmonary diseases in 1849-53 and 1860-62.
Ages
15-25
25-35
35-45
45-55
55-65
65-75
AII causes
During
During
the five
the three
years
years
1849-53
1860-62
8.90
8.96
14.30
33.51
63.17
111.23
9.44
9.57
15.12
29.74
63.21
110.51
Pulmonary diseases
During
During
the five
the three
years
years
1849-53
1860-62
3.05
4.42
8.47
24.31
44.46
55.87
3.77
4.15
7.89
19.75
43.29
45.04
69
Ages
Metal
miners
Males,
exclusive
of miners
15-25
25-35
35-45
45-55
55-65
65-75
9.53
12.38
17.64
33.11
78.34
127.52
7.57
9.19
10.13
16.18
29.38
66.10
70
not employed in mining, and also that this excess of mortality among the miners increases
largely and progressively with increasing age,
up to that period of life after which few miners
continue to work underground. Thus if it be
assumed that the rate of mortality among the
non-mining section of the male population at
each successive period of life quoted in the table
is equal to 100, then the rate among miners
between the ages of 15 and 25 years would be
126; between 25 and 35 years, 135; between 35
and 45 years, 174; between 45 and 55 years,
205; and between 65 and 75 years, 193.
From the return prepared by the Registrar
General it also appears that this excess of mortality among miners is mainly due to the greater
prevalence and fatality of pulmonary diseases
among them, as compared with that among the
non-mining section of the male population.
Table 7 shows the average annual rates of mortality per 1000 miners, and per 1000 males
exclusive of miners, from pulmonary diseases,
for the several successive periods of life from
the age of 15 up to that of 75 years.
With regard to Table 7, if it be again assumed
that the rate of mortality from pulmonary diseases among the non-mining section of the
male population at each successive decennial
period of life is equal to 100, then the rate
among miners between the ages of 15 and 25
years would be 88; between 25 and 35 years,
124; between 35 and 45 years, 334; between 45
and 55 years, 445; between 55 and 65 years,
574; and between 65 and 75 years, 308.
Thus it appears not only that the rate of
mortality from pulmonary diseases among the
lead miners in these counties is higher than that
among the male inhabitants of the same districts who do not work in the mines, but also
Table 7. Average annual number of deaths per
1000 lead miners, and per 1000 males exclusive of
miners, from pulmonary diseases, during the three
years 1860-62 inclusive.
North Wales
The returns of mortality relating to North
Wales, prepared for us by the Registrar General, have reference only to the district of
Holywell. Table 8 shows a very considerable
excess in the rates of mortality from all causes
among the lead miners, as compared with the
Table 8. Average annual number of deaths per
1000 lead miners, and per 1000 males exclusive of
lead miners, from all causes, in the District of
Holywell during the three years 1860-62
inclusive.
Males,
Males,
Ages
Metal
miners
of miners
Ages
Lead
miners
15-25
25-35
35-45
45-55
55-65
65-75
3.40
6.40
11.76
23.18
4 1.47
53.69
3.97
5.15
3.52
5.21
7.22
17.44
15-25
25-35
35-45
45-55
55-65
65-75
6.04
15.72
18.05
25.74
55.19
86.96
exclusive
exclusive
of lead
miners
7.46
10.52
12.57
15.19
28.11
75.78
Fuv
71
Ages
Lead
miners
15-25
25-35
35-45
45-55
55-65
65-75
3.02
4.19
10.62
14.71
35.32
48.31
Males,
of lead
miners
3.39
5.79
5.41
7.06
12.21
16.96
invasion proper; 3) eruption; and 4) dessication, regression or atrophy, or mortification, depending on the course taken by the verrucous
tumor.
Synonyms
Verruga de sangre.-Verruga blanda.-Verruga andicola (Dr. Salazar).-Verruga de Castiiki, de zapo 6 de quinua.-Verruga mular.
Definition
Verruga is an irregular, noncontagious, endemic and anemia-causing fever, chiefly characterized by pains and muscular contractions
(cramps), arthralgias with infarct and more or
less severe ostealgias. It produces a polymorphous eruption; it has a generally lengthy
but variable cyclic course, which is not affected
by treatment; and it also can give rise to numerous complications.
Symptoms
The course of this disease may be divided
into four distinct stages: 1) incubation; 2) invasion, subdivided into the prodrome and the
Source: Excerpted from Casimiro Medina, La Verruga Peruana y Daniel A. Carrz6n. Lima, Irnprenta del Estado, 1886.
IIt must be borne in mind that Carrion had not yet completed nor reviewed these notes.
72
Carridn
limbs which, together with the arthralgias
which immobilize the joints, lock victims in unnatural positions.
In the more severe cases, many patients cannot endure the pain without crying out or
moaning; every worsening of the pain provokes
new and vivid suffering. . . .
Now we will concern ourselves with other, no
less important, symptoms which complete the
characteristic picture of the period of the invasion proper.
The verrucous agent undoubtedly attacks the
blood, for nutrition in those suffering from this
disease is profoundly changed to the extent of
producing cachexia. This is manifested by the
anemia that develops at more or less galloping
speed and with greater or lesser energy, depending on the individual. Unfortunately, I
know of no study that has determined the total
number of red cells destroyed by the etiologic
agent.
The skin becomes pale and muddy; the
mucosae, particularly the palpebral and gingivolabial, lose their color and become waxy.
The heart beats weakly, and in most cases a
murmur of greater or lesser intensity can be
heard.
Movements become languid, without energy
or definition, and walking is unsteady.
There is ringing in the ears, confusion, bewilderment, and insomnia.
More or less rapid serous suffusions usually
occur on many occasions.
In most cases the spleen is considerably enlarged, at times descending into the left iliac
fossa (as with the patient in history No. 10); in
addition, the spleen is hard and so it is easy to
locate by palpation if there is no ascites. In
many cases the liver is found to be infarcted.
The anemia becomes more pronounced as
the disease takes its course.
Finally, during this period women experience
menstrual disorders.
Third period: Eruption.
The eruption begins to appear after a period
that varies between twenty days following the
poisoning or invasion and six or even eight
months thereafter.
The eruption begins on the limbs, face, etc.,
spreading immediately to the rest of the body,
and invading some mucosae as well.
73
During this period the general symptoms improve considerably, particularly if the eruption
is somewhat rapid and complete. Only the anemia may persist and increase, particularly when
the hemorrhage that ensues with rupture of the
verrucous tumors is repeated frequently, which
is very common.
Is the eruption constant? Yes, as constant as
that of other eruptive fevers, this being the most
characteristic phenomenon, the most fully
pathognomic symptom of the disease. It is also
remarkable for its constancy, its duration, its
termination and, finally, for many other particularities that I will mention hereafter.
Passing now to an examination of verrucous
tumors in all their phases, I will say that, of
course, their form and development and the
sites at which they appear are most varied. They
emerge now on the surface of the skin, now
beneath it in the subcutaneous form, both varieties constituting what we could call the external
eruption of the disease, which also includes
eruptions on the surface of the mucosae, such
as those of the mouth, nose and eyelid. . . .
Fourth period:
The end of the disease varies with the course
taken by the tumor.
I will, therefore, say a few words about how
the verrucous neoplasm terminates in its various forms.
When the eruption takes place on the surface
of the skin, the tumor, as indicated above, is no
larger than a pea. Having grown to this size, it
remains stationary for some time and then
shrinks very, very slowly, sometimes taking several months to disappear altogether. Meanwhile, its color changes from deep red to a very
markedly blackish red. As the tumor regresses
or reabsorbs, the skin flattens and all that can be
seen are small blackish spots similar to moles,
which first turn yellowish and then lose more
and more color until all that is left are small
whitish spots clearly distinguishable from the
rest of the skin and covered with scales that soon
fall away leaving no trace whatever.
As to the large tumors that break, mortify
and ulcerate the skin, I will say that they are
generally excised by a surgeon, an operation
that is sometimes easy because the neoplasm is
supported only by a thin pedicle.
The end of the disease, therefore, depends
74
Diagnosis
The diagnosis of the disease is as difficult to
establish at the onset of the disease as it is easy
to confirm during the eruptive period. Indeed,
we frequently see experienced practitioners
mistake for an attack of malaria in its various
forms, or for articular, muscular or osseous
rheumatism, what is but the first or second
stage of Verruga Peruana.
Of course, I will note that one of the principal
difficulties in distinguishing this disease from
malaria is that in most of the areas where it
exists, malaria is also dominant, and the two
diseases can attack together or separately.
In the present state of our knowledge, there
is, in my humble view, only one fact on which we
may safely rely to suspect Verruga Peruana before its eruptive phase. I refer to knowledge of
the place or places where the patient has been
or passed through. If we observe fever, whether
Pathogeny
As I see it, Verruga Peruana is a miasmatic,
and probably a parasitic disease. Its essential
nature is still in question owing to the shortage
of serious studies on the matter.
LECTURE I.*
75
76
Ratio of
invalided per
1000 of strength
Invalided
Ratio of deaths
per 1000
of strength
Died
Ratio of cases
of kakke per
1000 of strength
Cases
of kak'ke
Ratio of
invalided per
1000 of strength
3928.45
44 13.70
4604.32
3397.12
2531.77
3063.97
1865.02
992.48
577.46
434.22
400.59
Ratio of cases
per 1000
of strength
17 788
22 426
22 819
15 766
12 074
16 380
10 515
6866
4874
3954
3679
Ratio of deaths
per 1000
of strength
Cases of
dfiease or
injury
4528
5031
4956
4641
4769
5346
5638
6918
8475
9016
9184
Died
Strength
1878
1879
1880
1881
1882
1883
1884
1885
1886
1887
1888
Average ratio of
cases per person
per annum
Year
~~
3.93
4.41
4.60
3.40
2.53
2.90
1.81
0.91
0.52
0.40
0.40
56
119
63
81
103
85
45
49
63
55
65
12.37
23.42
12.71
17.45
2 1.60
15.90
7.98
7.08
7.43
6.04
7.08
44
39
43
29
30
28
44
33
52
56
48
9.72
7.68
8.68
6.25
6.29
5.24
7.80
4.77
6.14
6.15
9.15
1485
1978
1725
1163
1929
1236
718
41
3
-
327.96
389.29
348.06
250.59
404.29
25 1.20
127.35
5.93
0.35
32
57
27
30
51
49
8
7.07
11.20
5.45
6.46
10.69
9.17
1.42
19
8
9
16
17
4
1
1
4.20
1.57
1.82
3.45
3.56
0.75
0.18
0.14
Takaki
results and without discovering the cause of
beri-beri the time quickly passed to the year
1882. In February of the year 1882 I was appointed the Vice-Director of the naval Medical
Bureau.
About that time it was found to be necessary
to provide an extra medical officer besides the
usual number on board the training ships going
for long cruises on account of the abundance of
beri-beri cases among the men during the voyage. In 1882 there was a critical state of affairs
with Korea and three warships were sent to
Ninsen (Chemulpo) and Saibutsu Bay. They
stayed there only 40 days but owing to shortness
of hands caused by the prevalence of beri-beri
among sailors the officers felt quite unfit for
battle and it was a very anxious time for those in
positions of responsibility because those three
ships would have been of no fighting value in
the critical moment. For example, in one of the
ships 195 out of 330 were down with beri-beri.
As a consequence I handed to the chief of the
Naval Medical Bureau on June 24, 1882, a
memorandum describing the facts. Following
this, in August, 1882, H.I.M.S. Fuso, in spite of
anchoring off Shinagawa Bay, had to send half
of its crew ashore in turns for the treatment of
beri-beri. Continuing still further, I examined
the reports of the Tokyo and Yokohama naval
hospitals for 1881 and found out that threefourths of the patients had suffered from beriberi.
In 1883 I received permission from the Minister of the Navy to examine the hygienic conditions of ships, barracks, schools, &c., belonging
to the navy. I found that although working
hours, clothing, dwelling-houses, &c., were similar everywhere, yet in food there was a great
deal of difference. So I now asked the head of
each sectional department to send me in the
reports describing the details of food taken
three times daily for a week. From this I discovered the following facts: 1. Nitrogenous substances contained in the food were not sufficient to make up for the amount of nitrogenous
substances discharged from the body. 2. On the
contrary, the food contained too much carbohydrates. If we now look into the table recognised
by scientists showing the comparative amounts
of nitrogen and carbon discharged daily by a
fair-sized adult we shall see that carbon equals
310 grammes and nitrogen equals 20 grammes-i.e., nitrogen is to carbon as 1 is to 15.5. 3.
77
1906.
May 26,
78
Takaki
report of the investigation committee placed on
board the training-ship Ryujo during its voyage
in 1883. 2. The report of the experimental voyage of the Tsukuba in 1884. 3 . The results obtained through the examination of food supplied in 1883-that is, the very small quantity
of nitrogenous food as against the large amount
of carbohydrates, the proportion being I of nitrogen to 28 of carbon. In addition, the occurrence of numerous beri-beri and general diseases during that year. 4. The good results
obtained by improved diet since 1884 and the
difference brought about in the proportion of
nitrogen and carbon by various changes in the
proportion of diet.
On August 24th, 1885, I made a proposal to
change bread and biscuit for equally proportioned barley and rice which had been supplied
since March of that year, because I recognised
79
Chugas
~
81
~~
PART I1
DISCUSSIONS
BUCK
NAJERA:
In this second section, I see us starting in the early twentieth century, when there already was a fairly well established, scientifically sound, infectious disease epidemiology,
and then moving on to a broader application of epidemiology to all health problems. This would be what I
would call the difference between the old and the new
epidemiology: the shift of interest and concern that occurred in the 1940s and 1950s.
LLOPIS:
TERRIS:
85
86
Discussions
87
work whose content would be very appropriate. I am thinking of Major Greenwood, who preceded Bradford Hill at
the London School of Hygiene. In 1935, he published a
book called Epidemics and Crowd Diseases, and that book not
only contains chapters on tuberculosis and other contagious diseases of great concern, but also chapters on
cancer and psychological causes of illness. Greenwood
made it clear that epidemiological concepts were transferable from one kind of disease to another.
TERRIS:
NAJERA:
88
Disc.rcssions
89
ices by reorganizing its Ministry of Health. So, my contention is that the development of public health services
greatly influenced the transition in epidemiology and that
this, too, is probably what sets England and the United
States apart from the rest of Europe. A countrys political
development serves as background.
Spain is somewhere in between, because at the beginning
of the twentieth century it was influenced by the Rockefeller Foundation and the development of public health services along American or English lines. In 1924, Spain established its National School of Public Health, one of the
first in Europe (the second, I believe, after the London
School of Hygiene), and introduced a public health component into its already comprehensive rural medical care
network. A real school of thought in epidemiology started
there between the late 1920s and the Spanish Civil War in
the late 1930s. That was quite a development. This could
help explain why there is more epidemiology in Spain than
in many other countries in Europe.
TERRIS:
I have often puzzled over why it was England that pioneered in the noninfectious diseases. Why not Sweden,
where the problems of noninfectious disease were felt earlier because of an aging population. Yet Sweden never
developed this field. The big development was in Britain
and the United States. The question is why, and I am not
sure I have all the answers to that. I once discussed this with
Abe Lilienfeld, who thought that it occurred because of the
development of vital statistics in England. My own interpretation is that political factors are important. The reason the
movement began in Britain, and it is difficult to say why it
did not happen elsewhere, was that much of the leadership
of the British movement for social medicine was influenced
by labor and socialist ideology. Major Greenwood was a
founding member of the Socialist Medical Association
(SMA) in 1930; Richard Doll was an active member; and
Ryle himself had close ties to the SMA. Jerry Morris was
certainly pro-labor. J.A.H. Lee once told me at an International Epidemiological Association meeting in Yugoslavia
that those who went into social medicine in Great Britain
fulfilled at least two of three criteria: one, they were prolabor; two, they were Scots; and three, they had done some-
90
BUCK:
Discussions
91
be refined so that it includes specific mechanisms and specific individual responses, perhaps even analogous to the
immune reactions to infectious disease.
NAJERA:
Isnt it also possible that an important factor in the transition rested with the development of trade unions in England, since they developed alongside the socialism of labor?
What I think played an important part in the shift toward
chronic or noninfectious diseases, was that when the union
members started to demand their rights, their health problems were not chickenpox nor any of the infectious diseases
with the exception of, perhaps, cholera. Consequently, people started concentrating on those diseases that affected
adults, or almost only adults, since infectious diseases were
mostly restricted to children. This is what I think caused
noninfectious diseases to gain importance. And this, in
turn, allowed Ryle and others to consider the importance
of sociological factors even for infectious diseases.
BUCK:
How much was the transition to chronic disease epidemiology, apart from the obvious rising importance of
chronic disease, due to the fact that a lot of these physicianepidemiologists of the period had been internists?
NAJERA:
It is possible that this was a factor, but I think that the key
was this radical or socialist ideological thrust, and that its
impact was the extent of health services coverage that began with the insurance system.
Health service coverage, along with the sanitation movement of the nineteenth century, seemed to be enough. The
nineteenth century sanitation theory postulated that
providing safe water was enough sanitation for disease control. At the same time, the workers conditions forced the
expansion of health coverage to include all diseases. At
first, only accidents were covered, but by the forties, most
diseases were covered by the insurance system. By the twenties o r thirties, those countries that had better services, like
England, looked beyond infectious diseases.
TERRIS:
I dont agree. Most countries with highly developed insurance systems didnt do anything in this area. Sweden didnt
experience this change, neither did France or Germany.
Yet the United States, which had no government health
insurance, played a leading role in the development of
noninfectious disease epidemiology.
NAJERA:
Yes, but England did the most. They started the whole
thing.
92
TERRIS:
NAJERA:
Perhaps in the Soviet Union the system evolved too earlyit was at the beginning of the century. And the other
socialist countries followed the Soviet Unions model.
TERRIS:
Discussions
93
TERRIS:
BUCK:
TERRIS:
BUCK:
94
Discussions
95
and Graham, and Levin, Goldstein, and Gerhardt published their work in the Journal of the American Medical
Association. Then, in England, in September of the same
year, Doll and Hills paper appeared in the Lancet. That was
the starter gun for noninfectious disease epidemiology. It
developed almost simultaneously in the United States and
Great Britain. These two countries became the center, and
from there noninfectious disease epidemiology began to
spread everywhere.
My theory about why it happened first in England and
the United States and not in the rest of Europe is that in
other countries they had no public health as an independent discipline. They never worked with statisticians, they
never developed groups of interdisciplinary teams. In England, the London School of Hygiene was the focal point of
the epidemiologic revolution: Major Greenwood had been
both an epidemiologist and a statistician; they had epidemiologists like Richard Doll, Jerry Morris, and Donald
Reid, and statisticians like Bradford Hill and Peter Armitage. Theirs was an interdisciplinary group made up of
more than just physicians. And the United States, too, had
such a powerful group of epidemiologists coming out of
the Public Health Service: Rosenau, Goldberger, McCoy,
Anderson, Frost, and many others. There was a fantastic
growth of epidemiology in the United States; that is why
the United States became a leader in this field.
NAJERA:
TERRIS:
BUCK:
TERRIS:
Well, think of it. There was a tremendous need for noninfectious disease epidemiology because of the aging popula-
96
TERRIS:
NAJERA:
TERRIS:
Why are the Latin American countries now becoming interested in noninfectious disease? Its very simple. Heart disease is the leading cause of death in 28 countries of the
Americas, cerebrovascular disease in 3, and cancer in 1,
compared with diarrhea and enteritis in only 5 , and influenza and pneumonia in only 2. Injuries are second in Costa
Rica and fourth in Mexico. Latin American countries now
have the same problems as the industrial nations. That is
why they are becoming more interested in the noninfectious diseases.
BUCK:
Leaving that aside for the moment, couldnt you still argue
that the new epidemiology developed in countries in which
noninfectious diseases first began to rise in prominence?
TERRIS:
Not really. If that were the case, France and Sweden should
have been the first because they had the most old people.
BUCK:
Discussions
97
I agree. That is true. Its both. They had the problem and
they had the structure, the capability to deal with it.
You know what is crucial in the whole thing? I am now
convinced that it was the collaboration of epidemiologists
and statisticians. At the London School of Hygiene and
Tropical Medicine they had Doll and Hill, and Major
Greenwood himself was both epidemiologist and statistician. In the United States Harold Dorn set u p a statistical
unit at the National Cancer Institute which included half-adozen of the most brilliant young statisticians in the country- including Jerry Cornfield and Nathan Mantel-who
were put to work to develop the methodology.
BUCK:
TERRIS:
LLOPIS:
NAJERA:
98
BUCK:
TERRIS:
BUCK:
Well, its been that way for about 10 years. I notice it every
month. Barrett-Connor pleaded for a stop to this two-class
system, pointing out that infectious disease epidemiology
takes some know-how too.
TERRIS:
BUCK:
CONSIDERATIONS ON PELLAGRA
Joseph Goldberger
The writer desires to invite attention to certain observations recorded in the literature of
pellagra, the significance of which appears entirely to have escaped attention.
At the National Conference on Pellagra held
in Columbia, S. C., November 3, 1909, Siler and
Nichols in their paper on the Aspects of the
pellagra problem in Illinois state that certain
facts would seem to indicate that the exciting
cause of the disease is present within the institution (Peoria State Hospital), and add that at
the same time no nurses, attendants, or employees have shown the disease.
Manning, medical superintendent of the
asylum at Bridgetown, Barbados, on the same
occasion, in arguing against the identity of a
disease that he called psilosis pigmentosa with
pellagra, but which undoubtedly is this disease,
states that he had never seen it develop in an
attendant.
At the same conference, Mobley, from the
Georgia State Sanitarium, in the course of his
discussion of the relation of pellagra to insanity,
presents data showing that at the Georgia State
Sanitarium a considerable proportion of the
cases of pellagra develop in inmates who have
been residents therein for considerable periods,
mentioning one case in an inmate after 10 years
residence. In this connection he remarks, what
must have struck him, as it no doubt must have
appealed to Siler and Nichols at the Illinois
institution, as a curious fact, that so far as can
be ascertained there has never been a case of
pellagra to develop among the nurses, white or
colored, while employed as such in the Georgia
State Sanitarium.
Sambon (1910) in his Progress report states
that in Italy (no precautions are ever taken to
avoid propagation of the malady in any of the
pellagrosari, locande sanitarie, hospitals, insane
asylums, and other institutions in which very
numerous pellagrins are collected every year.
29:1683-1686, 1914.
99
100
29:2354-2357, 1914.
Goldberger
States, and the fact that this disease has so far
baffled all attempts to ascertain its cause and
means of prevention, the following letter from
Surg. Joseph Goldberger, in charge of the Governments pellagra investigations, is of interest.
Evidence seems to be accumulating to show
that pellagra is due to the use of a dietary in
which some essential element is reduced in
amount or from which it is altogether absent, or
to the use of a dietary in which some element is
present in injurious amount.
10 1
102
without himself realizing it, to subsist on a onesided or eccentric diet. Somewhat similar eccentricities of taste are more or less common in the
insane, some of whom, indeed (as the demented), because of apathy and indifference,
will not eat at all. These, for the most part
included in the untidy class, require special
care in feeding. The poorer the institution, the
fewer and of lower grade is likely to be its attendant personnel and therefore the greater the
danger that these very trying and troublesome
types of inmates will receive inadequate attention, and so be improperly (one-sidedly) fed. It
has repeatedly been noted by observers that at
insane asylums the untidy (the group in which
my observations show scurvy and beriberi most
likely to develop) were the most afflicted with
pellagra. By some this supposed excessive susceptibility is explained as dependent on the untidiness which favors filth infection. The true
explanation, however, is that both the untidiness and the supposed excessive susceptibility of these inmates are primarily dependent on the apathy and indifference typical of
most of this group. The deteriorated mental
condition causing apathy and indifference results not only in untidiness of person, but passively or actively in an eccentricity in the diet. I
believe that in this, in conjunction with a diet
admittedly low in the animal protein component we have the explanation of the excessive
prevalence of the disease at the Peoria State
Hospital, a hospital almost all of whose inmates
in 1909 were of the hopeless, untidy, incurable class, drawn from the other Illinois institutions.
While confident of the accuracy of our observations and of the justice of our inferences,
there is nevertheless grave doubt in my mind as
to their general acceptance without some practical test or demonstration of the correctness of
the corollary, namely, that no pellagra develops
in those who consume a mixed, well-balanced,
and varied diet, such, for example, as the Navy
ration, the Army garrison ration, or the ration
prescribed for the Philippine Scouts.
Respectfully,
Jos. GOLDBERGER,
103
104
quacy and extent of resources and their utilization, to say nothing of the medical babel as to
whether or not we know for each of the diseases
anything that can be used effectively for prevention, early diagnosis, cure or even alleviation.
As a result of some years of labor in this field we
have produced perhaps a mouse. But compared
with nothing a mouse is relatively a gigantic
beast. So in this book we offer our findings,
such as they are, in the hope that they may in
some slight way help others striving to find a
solution, and not a hollow panacea, for the present chronic disease dilemma.
The Inevitableness of the Chronic Disease
Problem
We are theoretically touched by what sickens
and kills people thousands of miles away or on
the other side of the globe, and give grudging
pennies to the Red Cross or our missionaries
with the platitude that we are all brothers and
then forget it all as promptly as possible. But
the things that really motivate are those that
sicken and kill our own families and those of
our neighbors, and that is just what chronic
disease is doing. There is hardly a family in
Massachusetts without immediate experience
with cancer, heart disease, or rheumatism (we
prefer this term as more inclusive than arthritis),and therefore immediate realization of the
inadequacies of our resources and their utilization. It is out of the emotional impetus roused
by such a widespread experience with futility
that action in a democracy and revolution in an
autocracy spring, and, unguided, such emotional action may result tragically in even more
costly futility than before.
Social Aspects
105
In chronic disease we are dealing with a problem, on the surface at least, predominantly of
late middle and old age. When you have diseases that are forty times as common over eighty
as under twenty years of age, with frequency
increasing with the age of the group studied,
you are obviously faced with a problem where
the investment of large sums of money is of
questionable value except in terms of the humanities. Overwhelmingly, most of the work of
the world is done by people in the age groups
twenty to forty (unemployed at forty) or shall
we say fifty or sixty. But chronic disease bears
heaviest after fifty-five or sixty. Perhaps the
most economically sound place for a high death
rate is in the cradle since there the community
has made but slight investment, unless we can
talk of hopes and aspirations and love in terms
of investment. But very properly the community is shocked at a high infant mortality with its
106
107
108
109
110
each and 3800 of the former completely disabled; diabetes, 15 000 cases with 800 completely disabled; cancer, 1 1 500 cases, and so
on.
Here we see that rheumatism displaces heart
disease which leads as a cause of death, and
rheumatism exceeds all others in total numbers
completely disabled. Thus, while rheumatism
does not kill, it does attack large numbers, with
heavy disability rates.
Duration
The average untreated case of cancer lives
some twenty-one months, let us say, for round
numbers, two years. The heart case lives some
seven to nine years, while the rheumatic lives
fourteen years or more. The rheumatics in the
survey who had complete disability had had this
same disability for an average of two years. As
these cases were all alive the average complete
disability for rheumatics before death is presumably at least twice this, or four years. If we
add a preceding period of partial disability we
see that the problem of the crippled rheumatic
becomes staggering. Neurasthenia was the only
disease considered which gave longer complete
disability, but these cases number less than 10
percent of the rheumatics. Heart disease crippled for only half as long and in only half as
many cases.
Rheumatism, then, cripples in the largest
number of cases, and kills in the smallest. This
very ability to cripple without killing would
seem to put it in the lead of all other chronic
diseases as of pre-eminent social, economic and
medical importance.
(c) Physicians will respond to educational opportunities and will send their patients to clinics
and hospitals if such institutions can get their
confidence;
(d) Patients will come in large numbers to
private offices, clinics and hospitals when told
of the need of utilizing such services;
(e) There is some evidence that they will utilize these services earlier if repeatedly importuned;
(0All methods of public information must
be employed interminably;
(g) Under governmental direction, clinic and
hospital service can be developed of a quality
not heretofore available;
(h) There are preventive, diagnostic, curative, alleviatory and terminal aspects to the cancer problem as to all others;
(i) There would seem to be no reason until
some better plan is enunciated, why similar
services under such auspices should not be developed for other of the important chronic diseases;
(i) There is reason to suppose that if someone else does not take the initiative, the people
through their legislature will again do so, to the
great shame of the medical and public health
professions.
11 1
Our consideration of the epidemiology of tuberculosis carried us some way from the conventional field of old textbook epidemiology,
but there was at least a thread of connection
inasmuch as tuberculosis is known to be an infectious disease. When we approach the study
of cancer that thread snaps, because, using the
word "infectious" in any ordinary sense, there is
no reason whatever to think that cancer is infectious.
From the statistical point of view there is no
doubt that cancer is one of the most important
crowd-sicknesses, and, if we were able to take
statistical records at their face value we should
have to conclude that it is a sickness which increases with the spread of material civilization.
Source: Major Greenwood, Epidemics and Crowd Diseases, An
Introduction to the Study of Epidemiology. New York: The MacMillan Company, 1935.
President of the Royal Statistical Society; Professor of
Epidemiology and Vital Statistics in the University of London
(London School of Hygiene and Tropical Medicine); formerly Medical Officer in the Ministry of Health.
In civilized States the mortality ascribed to cancer is far greater than among primitive races,
and in any one civilized State the mortality now
ascribed to cancer is far heavier than a generation ago. The Table l of English experience is
typical.
Statisticians are, however, a cautious race, and
very few of them have been prepared to accept
these figures at their face value. Forty years ago
King and Newsholme in a very valuable paper
initiated a discussion which still goes on. The
most striking part of their work was an analysis
of the experience of Frankfurt-am-Main, where
deaths from cancer had been classified by site of
occurrence for many years. It was found that
between 1860 and 1889 mortality from cancer
of those parts of the body in which the disease
was readily detected had not increased. They
urged that the general increase of mortality
from cancer was consistent with the causal fattor being not a greater number Of deaths from
cancer but a more accurate specification of the
Table 1. Annual mortality from cancer (all forms) per 100 000 living in successive
decennia, England and Wales.
Ages
All ages
standardized
0-
15-
25-
35-
45-
55-
65-
75 and
upwards
Males:
1851-1860
1861-1870
1871-1880
1881-1890
1891- 1900
1901-1910
1911-1920
1921-1930
20.5
25.6
33.3
46.7
63.5
78.2
89.6
100.4
1
1
1
2
2
2
2
2
2
2
2
3
4
4
4
5
6
6
7
8
10
11
11
12
18
21
24
30
38
41
42
42
42
54
71
100
130
155
168
163
93
121
159
230
316
390
444
472
150
187
26 1
376
533
668
800
955
174
227
299
393
582
787
973
1276
Females:
1851-1860
1861-1870
1871-1880
1881-1890
1891-1900
1901- 1910
1911-1920
1921-1930
43.8
52.1
61.9
73.9
88.2
94.0
96.0
98.8
1
1
1
1
2
2
2
2
2
2
3
3
2
3
4
14
16
17
17
18
17
16
16
60
67
79
86
89
85
79
76
128
154
176
205
232
232
227
214
186
230
277
338
410
441
438
424
236
281
352
453
583
666
711
774
233
280
352
460
638
790
919
1131
112
Greenwood
causes of death. Twenty-three years later another eminent statistician, Dr. Willcox, of Cornell University, continued the analysis of the
Frankfurt data down to 1915, and surveyed the
whole field. In his judgment this wider survey
confirmed the opinion of King and Newsholme. In his words: The cumulative evidence
that improvements in diagnosis and changes in
age composition explain away more than half,
and perhaps all, of the apparent increase in
cancer mortality rebuts the presumption raised
by the figures, and makes it probable, although
far from certain, that cancer mortality is not
increasing.
A year or two later, however, Dr. T. H. C.
Stevenson recorded in the annual report of the
Registrar-General for 1917 the results of some
analyses which are not easily reconciled with Dr.
1 13
Table 2. Accessible and inaccessible sites of fatal cancer: mortality per 1 million
living 1901-1917, England and Wales?
All Ages
Crude
Males:
Accessible
Standardized
200
367
440
776
54
68
694
1211
211
285
46 1
598
55
56
727
939
1916-19 17
1901-1902
1916-1917
484
563
457
576
45
31
986
1170
463
486
433
493
44
28
940
1007
{ 191-1902
1916-1917
Inaccessible
902
{ 1916-1917
Indefinite
{ 1901-192
1916-191,
All Sites
1901-1902
{ 1916-1917
347
480
449
66 1
48
47
844
1188
345
391
446
541
48
41
839
973
{ 1901-1902
1916-1917
Inaccessible 1901-1902
{ ,916-1917
Indefinite
l-lgo2
{ 1916-1917
A I ~Sites
1901-1902
{ 1916-1917
Females:
Accessible
{ 1901-1902
1916-1917
Inaccessible 1901-1902
{ 1916-1917
Indefinite { 190 1-1902
~ 1 sites
1
0-
35-
65-
75-
85 and
upwards
45-
55-
429
508
92 1
1056
108
115
1458
1679
996
1320
2373
2868
260
227
3629
4415
1819
2380
2509
3517
4178
4052
552 1
6749
369
550
429
324
6240
7108
8354 10 695
2485
4589
2512
3810
682
325
5679
8724
2064
2180
2141
2400
155
107
4360
4687
2600
295 1
3615
4295
264
160
6479
7406
3235
4214
3857
5493
400
215
7492
9922
3962
5038
2364
4228
687
309
7013
9575
1565
1761
2249
2628
204
165
4018
4554
2877
2253
3935
2754
399 1
3809
5996
484 1
463
310
30 1
234
7331
6372
7829 10 232
3416
4879
2419
4080
686
314
652 1
9273
25 35 60 - 78 13 - 19 98 - 132 -
Persons:
Accessible
25-
68
78
63
73
13
14
144
165
92 1
964
929
1001
99
83
- 1949
- 2048
-
114
Table 3. Cancer mortality-rates per 1 million population (standardized) for the more
important sites for each sex, 1901-1910, 1911-1920, 1921-1930, 1926, 1927, 1928, 1929, and 1930.a
Males
Females
Males
Females
All Sites
1901-1910
1911-1920
1921-1930
1926
1927
1928
1929
1930
784
897
1004
1011
1018
1032
1031
1031
1901- 1910
1911-1920
1921- 1930
1926
1927
1928
1929
1930
10.8
12.6
13.1
13.2
12.6
13.8
11.8
1901-1910
1911-1920
1921-1930
1926
1927
1928
1929
1930
942
959
986
995
984
1000
999
987
Tongue
0.8
0.7
0.7
0.6
1.0
0.7
0.6
0.7
43.1
50.8
46.1
43.7
46.6
45.5
41.8
40.6
?
3.0
3.0
3.1
2.8
2.9
2.8
3.2
Esophagus
51.2
14.6
60.6
16.5
64.2
18.1
65.4
17.8
60.7
18.0
64.3
18.7
62.3
18.3
61.8
18.6
167.2
186.4
221.1
222.2
229.0
227.4
237.2
233.7
Mesentery and
Peritoneum
8.2
15.8
6.0
12.0
5.4
8.1
5.6
9.3
4.8
7.3
5.8
7.3
4.4
7.2
4.9
6.6
Intestine
65.3
72.3
96.8 109.2
125.4 129.9
131.5 135.4
132.0 131.8
132.5 138.5
134.3 138.6
136.9 138.4
Pharynx
Males
Females
Lip
12.8
12.6
11.5
10.6
11.9
12.3
10.4
11.3
Males
Females
Males
Rectum
79.8
55.9
93.6
59.3
105.5
59.8
107.2
59.7
105.7
60.3
105.7
58.0
108.0
58.3
110.6
59.9
23.5
28.3
29.6
29.5
30.5
27.6
29.3
Stomach
133.0
139.0
155.5
163.2
157.0
161.5
164.6
162.8
Females
3.0
3.6
4.1
3.4
3.5
3.5
3.8
Liver
87.1
61.0
61.2
55.8
51.8
52.3
47.7
Jaw
22.6
25.1
20.8
21.0
21.1
19.6
19.2
16.7
6.9
7.2
6.4
6.9
6.0
5.5
6.5
5.3
Gall-bladder
?
98.0
60.9
59.8
52.1
52.6
50.6
45.4
6.0
8.8
9.1
8.3
9.5
9.4
9.5
11.6
16.6
17.7
17.6
16.9
17.6
17.1
Uterus
>
174.4
157.9
156.4
155.1
154.9
150.3
143.9
115
Greenwood
Table 3. (Continued.)
Males
1901-1910
1911-1920
1921-1930
1926
1927
1928
1929
1930
Females
Breast
158.4
170.8
189.1
184.3
193.5
196.2
195.7
194.5
1.5
1.6
1.8
1.7
1.6
1.9
1.8
2.3
23.9
31.3
33.5
31.7
31.8
31.4
31.6
1901-19 10
1911-1920
1921-1930
1926
1927
1928
1929
1930
11.8
26.5
47.7
47.9
47.8
53.8
56.4
54.9
6.0
7.1
7.3
6.9
7.6
7.6
8.5
Prostate
Females
Males
Rodent Ulcer
?
6.7
8.4
7.5
6.5
9.0
9.5
9.1
4.3
4.9
4.8
5.2
5.7
5.0
4.6
10.2
12.7
25.2
23.3
26.8
32.0
33.4
40.2
Penis
-
6.6
6.4
6.9
6.4
6.1
5.7
6.3
7.0
7.0
9.6
9.2
9.7
10.4
11.9
13.9
14.5
16.7
26.3
26.0
30.3
28.8
30.3
29.4
Testis
?
4.9
5.8
5.2
7.1
6.3
5.2
6.7
Females
Pancreas
Lung
Larynx
1901-1910
1911-1920
1921-1930
1926
1927
1928
1929
1930
Males
11.8
13.1
19.5
21.2
20.4
21.0
20.0
23.8
15.7
17.6
17.3
18.1
18.6
17.6
17.3
Females
Scrotum
2.4
2.7
2.7
3.0
3.1
2.7
2.3
?
Males
Other Skin
?
17.6
17.6
18.1
18.8
18.2
18.2
16.1
Kidney and
Suprarenals
8.4
9.1
11.7
11.4
12.2
12.5
13.2
13.0
7.6
7.2
8.9
8.8
9.6
9.0
9.6
8.7
Females
10.9
10.2
9.3
10.3
9.9
10.7
9.0
Bladder
?
28.2
30.5
30.0
30.5
32.0
32.3
31.8
9.7
11.4
11.1
11.6
11.9
12.3
11.5
Mediastinum
Bones
Males
12.0
13.5
13.1
11.7
14.6
14.6
12.0
8.1
9.2
12.6
13.3
12.9
13.3
12.1
13.1
4.5
4.6
5.8
6.0
6.0
5.4
5.6
5.3
Taken from Statistical Review o f England and Wales for 1930, Text, Table LII
691.4, an increase of 24.3 percent. The selected ble and of a little less than 30 percent of morcauses account for 92 percent of the total mor- tality from inaccessible sites. Comparing the
tality ascribed to cancer in 1911-1920, and for c o m p l e t e d e c e n n i a of 1 9 1 1 - 1 9 2 0 a n d
1921-1930, the accessiblerates are 269.3 and
94.3 percent of the total mortality in 1929. In
the last thirteen years the mortality from inac- 283.5, an increase of 5.3 percent. T h e inaccessible cancer has increased much faster than cessible rates are 556.1 and 652.5, an increase
that from accessiblecancer. This is a very dif- of 17.3 percent. It will be noticed that the averferent result from that found by Dr. Stevenson age rate for the decennium 1921-1930 is in the
nearly twenty years ago. The figures quoted accessible group slightly higher, and in the
above are not comparable, because, in order to inaccessible group decidedly lower, than for
meet a not very important objection to the usual the year 1929 used in the last comparison. For
method, a different method of standardization
1930 the accessible rate was 282.9, the inacwas used by him. Even the table in the 1917 cessible rate 688.7. It will be felt that these
annual report is not strictly comparable because statistical results are a little confusing. While
the sum of the accessible and inaccessible between 1897 and 1917 accessiblecancer was
site rates in I9 16-1917 has a larger proportion increasing faster than inaccessiblecancer, durof the total than in our series. Still, the differ- ing the last decade the relation has been reence is not important enough to invalidate a versed. Is any reconciliation possible?
comparison. These earlier figures showed an
In connection with the decennial analysis of
increase between 1901-1902 and 1916-1917 of occupational mortality effected on the material
35 percent in mortality from cancer of accessi- of the census of 1921 and the three years death
116
11
111
IV
128.4
102.5
118.1
127.1
123.8
157.8
0.3
3.6
1.3
0.9
0.4
1.6
7.4
17.6
0.5
5.5
1.6
2.3
1.4
Pharynx
Esophagus
Stomach
1.0
7.5
2.2
3.2
1.G
1.8
9.7
29.5
8.8
24.2
0.7
7.1
2.2
3.1
1.5
1.8
10.1
29.4
1.4
7.5
2.2
3.5
1.7
1.8
8.5
31.2
1.7
12.4
3.6
5.2
2.6
2.9
12.6
38.2
2.
Small intestine
Cecum
Hepatic and splenic flexures
Sigmoid flexure
Colon, part not stated
Intestine, part not stated
Large intestine
Total intestine (excluding rectum)
Rectum and anus
0.6
0.9
0.4
2.2
5.7
3.4
9.1
13.3
12.5
0.7
1.1
0.1
3.4
7.5
2.8
12.0
15.4
11.6
0.6
1.1
0.5
2.5
6.2
3.6
10.2
14.2
12.8
0.7
1.0
0.3
2.2
5.6
3.4
0.1
13.2
12.7
0.6
0.7
0.5
1.8
5.0
3.4
7.9
12.0
12.0
0.8
0.8
0.3
2.1
5.7
3.6
8.8
13.2
12.2
3.
Larynx
Skin
Breast
4.6
3.0
0.2
3.3
1.9
-
4.4
2.2
0.3
4.3
3.0
0.2
4.4
3.6
0.2
6.2
4.5
0.4
4.
0.9
3.4
1.6
3.1
2.9
0.9
0.5
2.2
0.9
1.3
3.5
1.1
3.3
3.2
0.8
0.8
1.6
0.9
0.8
3.5
1.7
3.0
3.2
1.5
0.8
2.5
0.8
1.0
3.3
1.6
3.2
3.0
0.8
0.6
2.3
1.0
0.9
3.0
1.5
2.4
2.3
0.8
0.4
2.1
0.8
0.9
3.8
1.4
3.9
2.5
0.7
0.3
1.9
0.9
5.
Lung
Liver
Abdomen
Neck
Lymphatic glands
Mediastinum
Other specified sites
Multiple
Site not stated
3.3
8.8
0.6
0.3
4.1
1.9
2.0
0.2
0.1
3.3
6.2
0.7
0.6
3.6
3.6
2.6
0.3
-
3.6
8.9
0.5
0.1
3.4
2.1
2.1
0.2
0.1
3.2
8.7
0.6
0.3
4.2
1.7
2.0
0.2
0.1
2.6
8.8
0.6
0.4
3.6
1.9
1.7
0.1
0.1
4.1
9.5
0.6
0.6
5.6
1.8
2.5
0.2
0.1
56.8
25.8
8.1
16.4
21.2
33.0
27.1
5.1
16.5
20.8
45.6
27.1
6.7
17.8
20.9
56.0
25.9
7.9
16.7
20.8
57.8
24.0
8.3
14.2
19.8
79.3
25.4
11.4
16.3
24.9
65.0
63.3
37.9
64.3
52.3
65.8
63.9
63.4
66.1
57.9
90.9
66.5
All sites
Lip
Tongue
Mouth
1.
1.
2.
3.
4.
5.
1.4
Annual Report of the R.G. Decennial Supplement, 1921. Occupational Mortality, p. xxiii, Table 4.
Greenwood
1 17
118
1+e-
(5.376+t)
8.074
1900
10
O 15
20
25
30h
never in finite time attaining 1027.6 per million. In each of the years 1928-1930 (inclusive)
the attained rate was higher than this. I think
we may have a good deal of faith that the standardized mortality-rate in our generation will
not exceed 1050, and that it will certainly not
increase fast. But we cannot be more precise
than that.
The course of mortality in females is still
more refractory to any mathematical polishing
process. It is, indeed, a strange picture, suggesting some queer speeding up of the rate of increase just before the war, leading to a falling
away which has been replaced by a fairly steady
or at worst slightly increasing trend. Mainly
owing to the facts that the mortality from cancer
in women is not increasing, or at least is only
very slowly increasing, while the mortality in
males is now definitely greater than in females,
most people seem to be more interested in
doing sums with the data for males. But from
the educational point of view the mortality from
cancer in women is more interesting. At the
present time more than a third of the whole
mortality from cancer in women is accounted
for by recorded primary cancers of the breast or
uterus, and these rates for sites which lend
themselves to relatively easy diagnosis have
changed greatly in the present generation. In
1901, when the all-sites rate was 943 per 1 million (these are age-standardized rates), the rate
for the breast was 148.9 and for the uterus
Greenwood
Table 5. Cancer mortality in females of the genital organs and breast. Rates per 10 000.
Genital organs
Year
1905
1906
1907
1908
1909
1910
1911
1912
1913
1914
1916
England
and Wales
2.43
2.46
2.55
2.50
2.47
Bredst
Italy
Holland
1.50
1.63
1.59
1.57
1.56
1.60
1.48
1.40
1.50
1.51
1.49
1.29
1.24
1.24
1.43
1.28
1.40
1.17
1.27
1.40
1.31
1 19
England
and Wales
Italy
Holland
1.84
1.97
2.02
1.99
2.11
0.57
0.59
0.56
0.57
0.59
0.58
0.55
0.59
0.59
0.59
0.56
0.80
0.89
1.03
0.89
0.97
1.00
0.98
1.19
1.08
1.10
Male
Female
Male
Female
Male
Female
Male
Female
Male
Female
Male
Female
Male
Female
Male
Female
1911
1912
1913
1914
1915
1916
1917
1918
34.75
36.01
34.70
35.02
25.04
23.04
24.03
23.39
24.00
25.05
28.99
33.34
28.66
33.43
29.11
34.05
29.25
33.19
32.86
35.96
36.68
33.70
35.87
35.05
24.54
24.57
28.68
33.36
15.38
19.29
15.77
20.23
15.47
19.58
14.88
20.18
23.63
24.86
22.89
24.39
23.55
25.51
23.70
24.08
23.67
25.69
24.42
26.66
62.44
59.81
63.58
59.97
63.71
58.79
62.20
58.85
63.98
58.84
63.31
60.04
60.63
56.63
97.52
83.23
98.51 31.08
78.34 33.30
97.98 31.65
81.72 32.46
95.20 31.08
77.80 31.82
101.10
85.34
53.12
45.50
52.04
44.32
50.61
45.85
54.24
42.35
Italy
___
97.26 30.67
77.44 33.57
36.89
35.19
23.27
23.05
28.62
33.65
24.36
25.44
52.63
43.10
14.06
19.33
59.52
55.64
37.77
37.98
22.41
23.98
28.35
33.37
25.58
25.98
15.36
19.64
37.48
34.73
23.88
22.27
Italy
27.85
31.85
England
and
Wales Sweden Holland
England
and
Wales Sweden Holland
Prussia
60-70
50-60
50.78
44.74
47.01
43.41
50.95
45.00
53.38
45.41
54.33
47.27
54.77
47.32
53.57
46.60
52.34
46.68
Prussia
_
~
112.59 45.26
94.41 46.02
97.61 43.20
92.59 45.15
93.86
88.17
108.17
111.94
92.86 67.84
86.84 61.69
91.93 65.89
87.39 60.04
Italy
70 and upwards
England
and
Wales Sweden Holland
60.51
56.06
61.59
57.13
65.95
41.75
62.05
54.98
65.85
85.58
65.68
60.77
67.61
60.13
62.74
58.06
Prussia
Greenwood
* * *
Here my sketch of crowd-diseases must end.
A great many important objects have been
brought into the picture, but many have been
omitted.
Among crowd-diseases in the grand manner,
malaria does not yield in importance to plague;
among crowd evils always with us, whoopingcough is not less deadly than measles, while
traffic accidents (which surely come within my
definition of a crowd-disease) are a good deal
more deadly. The trouble is that when one enlarges the definition of epidemiology one is theoretically committed to a treatise on all the bad
12 1
My Review of Recent Statistical Studies of Cancer Problems (Cancer Review, March 1928)
gives a slightly fuller account and numerous
references. I advise the student to read,first, the
discussions of cancer statistics printed in the
annual reports of the Registrar-General, beginning with that contained in the report for 1911 .
122
...
THE FUTURE
Ryle
Some of my friends have rebuked me for
leaving the clinical fold. I reply in effect that I
have merely taken the necessary steps to enlarge my field of vision and to increase my
opportunities of etiological study. My allegiance
to human medicine is in no whit broken. I wish
I could convey to them and to others some of
the sense of stimulation and rejuvenation that
my close association with statisticians and medical social workers and with men and women in
the public health and industrial health services
has brought to me. Thirty years of my life have
been spent as a student and teacher of clinical
medicine. In these thirty years I have watched
disease in the ward being studied more and
m o r e thoroughly-if
n o t always m o r e
thoughtfully-through the high power of the
microscope; disease in man being investigated
by more and more elaborate techniques and, on
the whole, more and more mechanically. Man,
as a person and a member of a family and of
much larger social groups, with his health and
sickness intimately bound up with the conditions of his life and work-in the home, the
mine, the factory, the shop, at sea, or on the
land-and with his economic opportunity, has
been inadequately considered in this period by
the clinical teacher and hospital research
worker. The medicine of the teaching schools
has, as I have suggested, undergone a gradual
conversion to a highly technical exercise in bedside pathology and therapeutic method. The
morbid material of the hospital ward consists
very largely-if we exclude the emergenciesof end-result conditions for which, as a rule,
only a limited amount of relief repays the long
stay, the patient investigation, and the anxious
expectancy of the sick man or woman. With
etiology-the first essential for preventionand with prevention itself, the majority of physicians and surgeons have curiously little concern. Nor have they at present the opportunity,
nor yet the appropriate types of training or
assistance, requisite for the study of etiology or
prevention. Their material is mainly selected by
four factors: the gravity, the difficulty or the
rarity of their cases, or their suitability otherwise for admission to a hospital. Some of the
most common diseases, the less lethal diseases,
and the beginnings of disease are even considered as providing poor teaching material.
Health and sickness in the population and their
possible correlations with significant and mea-
123
PREVENTION OR CURE?
124
istic reforms in teaching, by the better education of the people or direct representations to
government-it is our first duty as physicians to
explore and prepare the way.
I submit that we can only do this effectively by
electing to pursue the study of social man in
sickness and in health as assiduously as we have
hitherto pursued the study of individual man in
the isolation of the consulting room or the hospital bed, when health has finally passed him
by. The quality of our actions and our practice
and of our leadership in social reformation will
depend, as in the past, on many disciplines, but
not least, perhaps, upon the science whose history I have briefly sketched and whose province
I have endeavoured to define.
125
126
In such a study, the effects of these measurable external factors in health and morale were
assessed by the varying incidence of neurosis
and venereal disease among the flying crews.
Men react differently to strain; some find refuge in neurotic illness, while in others the weakening is one of morals rather than of morale.
With the monthly incidence of these two types
of illness as criterion, then, the relative impor-
Read
strength who become casualties during the
month. T h e cumulative effect of hazard is
taken into account by calculating the correlation
coefficient between the neurosis incidence in
one month and the casualty rate both in the
same month and in the month before.
An alternative method of measuring the degree of hazard is to calculate the average number of aircraft lost through enemy action in the
course of every hundred sorties carried out by
the Bomber Group in the month. This percentage has been termed the missing per sortie
rate. It depends in part, of course, both on the
effectiveness of the enemy opposition and on
the skill of the crews defensive tactics. To some
extent it is thus a measure of their operational
efficiency. It is also a measure of the rate at
which casualties are occurring rather than the
total casualties suffered ddring the month.
Inexperience. T h e level of operational experience of the Group is quite easily measured by
calculating the percentage of the average crew
strength for the month who have been posted
into the Group, on completion of training, during the month. In other words, this percentage
expresses the proportion of the squadrons
strength who are novices to operational flying.
Effort. T h e cumulative effect of long hours of
highly skilled activity should be reflected in a n
association between the incidence of neurosis
and features of squadron employment to which
fatigue is often attributed. Among these are the
average number of flying hours per individual
in each month and/or the frequency and average duration of the sorties occurring during
that month. T h e total number of hours, operational and non-operational, flown each month
by all the aircraft in the Group, and the average
number of men on board being known, it is easy
to calculate the mean number of flying hours
done by each man in each month. T h e frequency with which each man operated, i.e. the
average number of sorties per man per month,
is calculated in the same way from the total
number of aircraft sorties carried out by the
Group in the month in question. SimiIarly, the
number of operational hours and the number
of sorties carried out by the Group in any one
month being known, simple division of the former by the latter gives the average duration of
sortie during that month. As when dealing with
the effect of hazard, the cumulative effect of the
operational effort from one month to the next
127
For security reasons it is unfortunately impossible to give in detail the figures upon which
these measures are based, but they derive from
the experience of the whole of Bomber Command at a period of its maximum strength. T h e
correlation coefficients indicating the relationship between these measures and the criteria
are given in Table 1 where they are arranged in
order of their magnitude under the headings as
previously defined.
For venereal disease, similar results were obtained; the only significant correlations were
between the venereal disease incidence and the
casualty and inexperience groups of items
(Table 2).
It would appear from these correlation coefficients that only hazard and inexperience are
significantly related to the selected criteria of
the effects of operation strain. There is no evidence of such a clear relationship between the
criteria in the effort group of items; yet these
items may be fairly termed fatiguing in that
they might be expected to have a cumulative
effect on the mental health of the crews. It is
clear that the monthly variation in effort in
this study has not had an appreciable effect on
the monthly incidence of neurosis.
Some allowance must be made for the association between the influx of new crews in any one
month and the heavy casualty rate sustained by
these operationally inexperienced men in the
succeeding month. At the same time, in the
latter month, these men reach the critical stage
in their tour when breakdown is most likely.
The association between the neurosis rate and
the casualty rate in any month may therefore be
merely an expression of their common relationship to the novice rate in the previous month.
Fortunately this possibility can be taken into
account by calculating the partial o r net correlation between the casualty rate and the neurosis
rate in any month when the novice rate in the
previous month is kept constant. T h e value obtained (r=0.3734) shows that the casualty rate
in any month is significantly more closely associated with the neurosis rate than might be ex-
128
0.4194~
0.331P
0.2455b
0.2404b
0.2906a
0.2869
0.1163
0.1157
0.0933
0.0351
0.0323
0.0267
0.0029
0.0019
A. H a u w d group of items
1. Casualty rate (in previous month)
2. Casualty rate (in same month)
0.4 loga
0.2972
B. Inexperience group of i t e m
1. Novice rate (in same month)
2. Novice rate (in previous month)
0.239gh
0.0350
Reid
Of the measurable environmental factors affecting health and morale, operational hazard is
thus clearly of decisive importance. Conversely,
factors such as the number of flying hours done
by each man and the frequency of operating do
not produce the cumulative effect on mental
health which conventional ideas about fatigue
might have led one to expect. In laying down a
policy of tour limitation, therefore, the duration
of the operational tour to be expected from
those who survive long enough to become due
for relief must take these facts into account. N o
arbitrary number of flying hours can be taken
as the optimum limit in all types of flying. Since
the number of hours flown does not in itself
determine the incidence of neurosis, it would be
unrealistic to translate a tour limit of x hours
from, say, long-range flying boat sea patrols to
the conditions of intensive fighter operations.
Tour limits may be expressed in flying hours, but
they had to be framed to take into account the
hazard involved in the particular type of flying
in question. The basic principle adopted was
the setting of the limit at a level giving the
individual a chance of surviving the particular
hazards entailed which could be faced without
breakdown in health and morale by the resolute
type of man who elected to serve in the air.
Exactly what that limit should have been was
difficult to assess. Within the range of casualty
rates experienced by Bomber Command during the period reviewed, there appeared to be
no point beyond which a sudden serious deterioration took place. The physical measures of
health already mentioned gave no indication of
a decline in health among those who survived
the initial period of adaptation. On the other
hand, the hazard of bomber operations was
such that the incidence of neurosis in Bomber
Command was higher than in any other Command of the Royal Air Force (7). The expectation of life prevailing in that Command could,
therefore, be taken as a minimum rather than
an optimum for general application and the
duration of the operational tour in other forms
of combat flying set to ensure that the individual pilot had at least the same chance of
survival as was general in the bomber offensive
against German targets.
129
The nature of the two critical questions implies that some follow-up of the subsequent operational careers of the sick men is required.
Not only must their own medical and operational histories be obtained and analyzed, but a
suitable standard of comparison must be afforded. This was effected in practice by making
a series of paired comparisons between the
sick case and a corresponding control man
selected from among the others in the squadron. From the Squadron Weekly Aircrew List
for the week in which the sickness occurred was
selected a pair for the man whose illness dur-
130
ing or just before a sortie had caused its abandonment or cancellation. This control man
was selected as far as possible so that each pair
was comparable in respect of these items in the
following order of priority: (1) crew duty; (2)
number of sorties performed before the date of
the sickness; (3) squadron; (4)officer or noncommissioned officer status. Thus it was hoped
to control the influence on morale and efficiency of such factors as the stresses specific to
the task in the air, operational experience,
squadron leadership, and living conditions. In
other words, during their subsequent careers in
the squadron, both the sick man and his control were subject to the same environmental
influences and differed only in the one respect
of this particular type of sickness. In a group of
such pairs, the differences between their subsequent performance should therefore be due
mainly to differences in the psychological
qualities of the individuals themselves, since
both men in each pair are equally likely to be
affected by the chances and hazards of flying in
war.
The subsequent operational career of each of
the men thus paired was followed up through
the records of the Personnel Staff sections in
each Group Headquarters. Two features were
examined in particular:
(1) the cessation of flying duties through illness, whether physical or psychological;
( 2 ) the survival time subsequent to the date
of sickness, measured for each man of the pair
by the number of sorties completed.
These features were recorded, together with
all the relevant medical data about the sick case,
on cards which could then be sorted into
groups. The manner of this division depended
on the factor being studied, for example, nature
of symptom pattern, previous operational experience, etc. Inter-group as well as inter-pair differences could thus be investigated and the relative prognostic importance of both sickness in
general and of some additional factor, such as
experience, assessed. The statistical methods involved are detailed in the appropriate tables in
each section.
In the clinical assessment of the likelihood of
a psychological basis for the complaints which
necessitated the return of the aircraft, full
weight would be given to definite physical signs
of disease; but in so many of these cases no such
definite aid is forthcoming and the medical of-
13 1
Reid
Table 3. Physical and psychological wastages in symptom groups.
Physical wastage
No. of
pairs
Syncope
Nervous
Alimentary
Airsickness
Oxygen lack
Respiratory
Injuries
Totals
Sick
No.
0
2
Sick
Control
70
NO.
No.
No.
0
2
15
0.0
6.5
2.1
14.3
0.0
2.9
0.0
9
6
1
2
3
0
78.9
29.0
12.8
14.3
22.2
8.8
0.0
0
0
0
0
1
0
0
3.2
36
23.2
19
31
47
7
9
34
8
0
0
7
0
0.0
6.5
2.1
0.0
0.0
20.6
0.0
155
10
6.5
Psychological wastage
Control
1
1
70
0.0
0.0
0.0
0.0
11.1
0.0
0.0
0.6
Other features of each individual case, operational experience and time of onset, must therefore be considered as in Tables 5 and 6.
This division of the cases into groups according to the number of sorties completed prior to
pairs
7%
S E of
difference
19
31
47
78.9
29.0
12.8
29.4
28.2
24.9
14.3
11.1
34
8
8.8
0.0
No. of
Syndrome type
Significant
Syncope
Nervous
Alimentary
Not significant
Airsickness
Oxygen lack
Respiratory,
etc.
Injuries, etc.
70
24.9
50.0
0-3
4-9
10-19
20 +
Total
No. of
pairs
Sick
Control
No.
No.
39
41
46
29
10
15
2
25.6
36.6
19.6
6.9
0
0
1
0
0.0
0.0
2.2
0.0
155
36
23.2
0.7
132
0-9
10-29
30-39
40-59
60-90
90 +
Total
No. of
cases
% of total
cases
Psychological wastage
No.
45
20
23
27
27
13
29.0
12.9
14.9
17.4
17.4
8.4
7
3
6
6
9
5
15.6
15.0
26.1
22.2
33.3
38.5
155
100.0
36
23.2
probably related to the likelihood of psychological failure in the subsequent flying career. Despite this association, however, it does not necessarily follow that each of these factors can be
taken as independent items of clinical evidence
in formulating a prognostic judgment. If one
factor overlaps much with another, for example
if all cases of syncope occurred in the later
stages of the sorties, then taking stage of sortie
into account would not enhance the accuracy of
a prognosis based on our experience of syncope
cases alone. On the other hand if the two are
not related, then one can legitimately consider
the stage of sortie as additional information of
prognostic value. The contingency tables giving
the relationships between each of the three factors are not reproduced, but they suggest that
there is no significantly close relationship between them; this is confirmed in each instance
by the x 2 test. It follows, then, that in any individual case the information relevant to each of
these three aspects can be combined in framing
an estimate of the likely outcome of the patients subsequent operational career.
The risk entailed in allowing a man who has
suffered from a sickness which might well be
psychogenic to go back to operational duty can
be measured by the difference, if any, in the
number of sorties subsequently completed by
the sick compared with the control member
of the pair in the same squadron at the same
time. Any gross disparity in favour of the controls would imply that the return of the sick
men to their operational duties entailed an increased hazard for the aircraft in which they
flew. T h e differences actually observed are
given in Table 7.
As Table 7 indicates, there is a slight adverse
difference in the total number of sorties subsequently performed in the group who went on
No. of
cases
S.N.A.a
Others
63
42
* S.N.A.-Syncope,
ders.
Mean
difference
S. error of
difference
1.35
+2.14
21.48
2 1.68
Reid
operating even after a sickness of a type normally associated with a liability to psychological
failure. T h e t test for the significance of this
mean difference between pairs indicates, however, that the difference might easily have occurred by chance. Further subdivision of the
cases was hardly justified by the numbers involved, so that it must be concluded from these
results that if a man carried on without breakdown after an illness in the air which might well
have been psychogenic in origin, his presence in
the crew did not seem to affect appreciably the
aircrafts chance of safe return in subsequent
sorties.
DISCUSSION
133
This paper describes typical field studies, carried out in Bomber Command of the Royal Air
Force during the war 1939-45, on the effects of
operational stresses both on the incidence of
certain diseases and in individual cases of sickness occurring in the course of bomber operational sorties.
A study of the relationship between the incidence of neurosis and venereal disease a n d
various indices of operational hazard and effort
suggests that the acute anxiety caused by a high
casualty rate had a much more decisive effect
on health and morale than had prolonged o r
intensive operational effort.
Operational inexperience, which also plays a
part in determining the incidence of neurosis, is
shown to be relevant in the prognosis of cases of
sickness occurring in conditions of acute stress
011 operational sorties. T h e chance of subsequent psychological failure in such cases is also
seen to have been related to the type of presenting syndrome and the stage in the sortie at
which the sickness appeared. Fainting a n d
symptoms referable to the nervous and alimentary systems, particularly if they arose some
time after take-off, were indicative of impending breakdown. In those instances where this
did not occur, however, no extra risks to the
aircraft in which he flew were incurred by allowing the man to continue on operational duty.
These suggestions of the predominant effects
of anxiety over fatigue (in the sense of the end
results of prolonged effort) can be related to the
findings of laboratory and clinical research
done in the Royal Air Force. T h e value of simple statistical methods in field research as a
complement to laboratory and clinical studies i s
discussed.
I am indebted to Air Marshal Sir Harold
Whittingham, Director General of Medical
Services of the Royal Air Force during the war,
for his encouragement, to the Consultants in
Neuropsychiatry a n d Medical Statistics, Sir
Charles Symonds and Prof. A. Bradford Hill for
their direction and advice, and to the present
Director General of Medical Services, Royal Air
Force, Air Marshal P. C. Livingston, for permis-
134
References
( I ) Symonds, C.P. B r Med J 2:703 and 740, 1943.
(2) Davis, D. Russell. Pilot error. Air Ministry A.P.
3139a. H.M.S.O., 1948.
( 3 ) Birley, J.L. Lancet 1:1147, 1920.
( 4 ) Reid, D.D. Some measures of the effect ofoperational stress on bomber crews. Air Ministry A.P.
3139: 245-258. H.M.S.O., 1947.
(5) Reid, D.D. Fluctuations in navigator performance during operational sorties. Air Ministry A.P.
3139: 321-329. H.M.S.O., 1947.
(6) Hill, A. Bradford, and G.O. Williams. Investigation of landing accidents in relation to fatigue. Flying
Personnel Research Committee Report No. 423 (m), 1943.
(7) Symonds, C.P., and D.J. Williams. Clinical and
statistical study of neurosis precipitated by flying duties. Air Ministry A.P. 3139: 140-172. H.M.S.O.,
1947.
EPIDEMIOLOGY-OLD
AND NEW'
John E. Gordon2
49:194-19Y,
22, 1949.
An Aging Population
Since 1850 the population of the United
States has shown an increasing proportion of
people in the older age groups, and correspondingly fewer children and young adults.
Persons aged more than fifty years included
13.3 percent of the population in 1900; the
proportion is now almost twice as great, and the
135
136
The mass disease problems that affect modern communities have been notably altered
within a century. T h e causes are many, a great
many more than the two just presented. They
relate both to man as an organism and to the
environment in which he lives. Some diseases
now have far less significance; others have come
from a n inconsequential position to r a n k
among the leading causes of death or as major
factors in lost efficiency o r disabling illness.
What has happened, however, is more than a
matter of differences in relative importance of
the numbers of persons involved. In many instances, the nature and character of a mass disease has changed. Less often, new problems
have been introduced by reason of the shifting
ecological state. T h e communicable diseases
provide the best illustration of these several
considerations, principally because custom has
long marked this group as the type of mass
disease problem.
Communicable Disease
Comparing present conditions with those
that existed a hundred years ago, the changes
that have occurred among the communicable
diseases are so great as to constitute almost another world. Intestinal infections are far less
frequent. Diseases transmitted through discharges of the upper respiratory tract find a
much more important place among infections
in general. T h e situation is less definite for dis-
Gordon
main as they are or continue to follow the trend
they now do. If the environment is altered materially or new factors are introduced, such as
another French Revolution, almost anything
could happen, including reversion to epidemic
situations that characterized the world of a century ago. The composite experience of the past
hundred years and especially that of the recent
global war supports the opinion of a favorable
future in respect to major world outbreaks. The
history of a typical American city, Philadelphia,
is examined with profit. The greatest epidemic
of modern times, the pandemic of influenza of
1918, was a small affair in terms of deaths compared with the earlier epidemic visitations on
that city. The experience is not unique. A similar course of events has characterized other
representative American cities, Boston, New
Orleans, Chicago and New York.
The greatest attention and the principal interest in epidemiology continues to be directed
toward the communicable diseases, not because
these diseases are dominantly important, for
actually they are of lesser relative moment than
some others. It is largely because they are better
and longer understood. It is therefore reasonable for the communicable diseases to continue
to be t h e f u n d a m e n t a l concern of epidemiologists. The gains that have been made
must be held. Perhaps the more important consideration is that these diseases offer the most
favorable opportunity to gain familiarity with
the epidemiologic method. The reasonable approach to the broader and less well explored
fields of mass disease is through expansion and
transfer of that method.
137
Functional Disease
Noncommunicable Organic Disease
The important community health problems
of the present are concerned with those organic
diseases which are not communicable from man
to man and not caused by a specific infectious
agent. This is just as true if judgment is based
on another of the major criteria, not on deaths
but on defect or disability.
Heart disease now ranks as the first caise of
death in the United States; and deaths from
cancer, from circulatory disturbances, from
metabolic diseases, nutritional disturbances,
and the degenerative diseases are well up on the
list. A direct relation is seen with the altered
social and economic conditions of the present
138
Gordon
139
140
Gordon
ward other areas ofmass disease, such as cancer
(12),and specifically cancer in industry (13,14).
There is much activity in the field of nutritional
disturbances ( 1 3 , with dental metabolic diseases (16), and currently in diabetes (17), and
thyrotoxicosis (18).Even such unusual fields as
the congenital anomalies (19) now engage the
attention of the experimental epidemiologist.
With full appreciation of the contributions
made to modern medical progress through
skilled observations by the individual practitioner, there is nevertheless an increasing appreciation of the extent to which knowledge of a
total problem can be enlarged through analysis
and study of the facts accumulated by many
observers; that benefit is to be derived through
investigation of disease processes as they affect
groups of people, as a means of supplementing
clinical study of the individual patient and the
knowledge to be derived from experiment. It
likewise becomes increasingly evident that to
understand fully all the variations which disease
may show, it is necessary to draw on the experience of the world and not to reason too broadly
from results obtained in a small section of a
single country. Disease shows many peculiarities
under the multiple influences of varying environment and that applies not alone to the
communicable diseases. Diphtheria and cancer
are almost universally described in terms of the
clinical and epidemiological behavior they evi-
141
For some time I have been confused and irritated by the division of epidemiology and epidemiologists into the subspecialties of infectious
disease and chronic disease. Those who doubt
this division need only to look at the curricula
of schools of public health, chapter headings in
standard textbooks of epidemiology, and job
descriptions as they are advertised in various
media. Although the division has historical rationale and is perpetuated by current federal
funding, as outlined below, I believe the distinction is arbitrary a n d detrimental to epidemiologists and to the epidemiologic study of
disease.
How did this division come about? Over half
a century ago epidemiologic observations had
already been made about chronic diseases of
non-infectious cause, such as pellagra in the
poor and scrota1 cancer in chimney sweeps.
However, most earlier epidemiologic investigations were concerned with the study of infectious diseases. Then, in the 1940s, many prominent epidemiologists stressed the importance of
applying the tools of epidemiology to the study
of chronic diseases and decried the limitation of
epidemiology to the field of infectious disease.
As Murphy (1)has noted, a major act of faith in
the epidemiologic work of the past thirty years
has been that methods which worked well in
infectious disease could be successfully applied
to chronic diseases. T h e successes of chronic
disease epidemiology support the assumption
but have been paradoxically divisive.
DEFINITIONS
Latency
A long interval between exposure to the
putative risk factor(s) and disease onset is believed to characterize most chronic illnesses.
But many infections appear after latent periods
as long as those proposed for chronic diseases.
Thus, infection with a tubercle bacillus acquired
in childhood is often first manifest in late adult
life. Herpes zoster represents reactivation of
childhood chickenpox in many, if not all, cases.
A large proportion of infections in the compromised host undoubtedly reflects activation of
dormant infection. Indeed, the incubation period for the majority of infections afflicting
adults today is either delayed o r poorly defined.
142
Barrett-Connor
Transmissibility
Many infectious diseases are propagated from
person to person. However, this is by no means
true of all infectious agents: food poisoning
caused by preformed toxins, Legionnaires disease and coccidioidomycosis are not -transmitted
from person to person. Some chronic diseases
of as yet unknown etiology may turn out to be
transmissible. Clusters of leukemia and lymphoma suggest a transmissible agent as d o the
recent studies of residents of households in
which victinis of multiple sclerosis reside (3). It
would be premature to divide epidemiologists
into those who deal with transmissible or nontransmissible conditions. If leukemia, cervical
cancer, multiple sclerosis, arthritis and diabetes
prove to be caused by a transmissible agent, as
many now suspect, persons now classified as
chronic disease experts may find themselves to
be infectious disease epidemiologists.
Etiology
At the turn of the century, infectious disease
was the major area of research in medicine. T h e
discoveries of specific agents which produced
specific diseases were straightforward and satisfying, and led to one of the basic tenets of
medicine: a single disease process has a single
causation. Clinical observation, bacteriologic investigation and the development of the anti~riicrobialsin the early 1940s led to the preeminence of infectious disease as a medical
problem whose etiology and management were
established. In contrast, chronic disease epidemiology has attended the study of diseases of
unknown cause, conditions which a r e increasingly recognized as multifactorial in origin.
Thus, the dichotomy became cause-knowniunifactorial vs. cause-unknownimultifactorial.
Although it is true that the necessary cause of
most acute diseases is a known agent, and the
necessary cause of most chronic diseases remains unknown, this is surely more a function
of the state of the art than the nature of disease.
All diseases have multiple causes. T h e necessary
microbial agent is not the sole determinant of
outcome. As Stewart ( 4 ) has written, If two
susceptible subjects are exposed to equal doses
of the same germ, and one develops infection
while the other does not, the factor governing
the development of the infection clearly lies
outside the germ.
143
Behavioral considerations
Evidence accumulated during the last twenty
years indicates that the most important chronic
diseases are caused by a variety of personal and
social habits, such as improper diet, excessive
drinking and smoking, lack of exercise, and
unsafe driving and working practices. Behavioral considerations also determine the distribution of many infectious diseases. For example,
venereal disease, the most important epidemic
infection in the United States today, does not
occur among the chaste, and active tuberculosis
is disproportionately frequent a m o n g those
who abuse alcohol.
In neither acute, infectious, nor chronic disease is a complete understanding of cause required for prevention. Smallpox was prevented
before isolation of the bacterium; lung cancer
can be prevented before identification of the
specific carcinogen in cigarette smoke. When an
infectious disease is transmitted o r maintained
because of attitudes, behavior or surroundings,
a purely germ-oriented approach is unlikely to
provide effective control.
Study design
N o study design is unique to any branch of
epidemiology. T h e epidemiologic study of both
acute and chronic conditions usually requires a
denominator and/or a comparison group, can
be done retrospectively o r prospectively, and
can examine prevalence o r incidence. T h e
search for causality in a food poisoning outbreak, examining the attack rates of those with
and without exposure to the suspect food, applies the same principles as those used in a
comparison of the incidence of uterine cancer
among those with and without the suspect hormone. Cross-sectional o r case-control com-
144
Barrett-Connor
CONCLUSION
145
References
(1) Murphy, E.A. Epidemiological strategies and
genetic factors. Intl J Epidemiol 7:7-14, 1978.
( 2 ) Commission on Chronic Illness. Chronic illness
in a large city-The Baltimore Study. I n Ciironzc 111ness in the United States. Vol. IV. Cambridge, Mass.,
Harvard University Press, 1957.
( 3 ) Schocket, A.L., and H.L. Weiner. Lymphocytotoxic antibodies in family members of' patients
with multiple sclerosis. Lancet 1 :571-573, 1978.
( 4 ) Stewart, G.T. Limitations of the germ theory.
Lancet 1:1977-2081, 1968.
( 5 ) Emery, A.E.H., R. Anand, N. Danford, et al.
Arylhydrocarbonhydroxylase inducibility in patients
with cancer. Lancet 1:470-471, 1978.
( 6 ) Weller, T.H., and C.W. Sorenson. Enterobiasis:
Its incidence and symptomatology in a group of 505
children. N E n g l j Med 224:143-146, 1941.
( 7 ) Kilpatrick, Z.M., P.A. Greenberg, and J.P. Sanford. Splinter hemorrhages-their
clinical significance. Arch Intern Med 115:730-735, 1965.
(8) Price, W.H. Gallbladder dyspepsia. Br Mcd J
2:138-141. 1963.
( 9 ) Dawber, T.R., W.B. Kannel, and L.P. Lyell. An
approach to longitudinal studies in a community: T h e
Framingham Study. A n n N Y Acud Scz 107:539-556,
1963.
(10) Fox, J.P., C.E. Hall, and M.R. Cooney. T h e
Seattle Virus Watch.11. Objectives, study population
and its observation, data processing and summary of
illnesses. A m J Epidemiol 96:270-285, 1972.
(11) Feinstein, A.R. Clinical Bzostutistics. St. Louis,
Mo., CV Mosby, 1977.
(12) Langmuir, A.D. Epidemiology of airborne infection. Bacterzol Rev 25:173-181, 1961.
(13) Morrison, AS., H.O. Jick, and H.W. Ory. Oral
contraceptives and hepatitis. A Report from the
Boston Collaborative Drug Surveillance Program,
Boston University Medical C e n t e r . L a n c e t
1:1142-1143, 1977.
PART I11
ETIOLOGIC INVESTIGATIONS
DISCUSSIONS
NAJERA:
Perhaps we should start by emphasizing the interrelatedness of the factors that cause disease. Today, everybody
talks of multicausation, but if you read the studies, most
researchers still search for a cause, they still think in terms
of a single or a few simple causes of disease. They havent
really begun to understand disease as a result of the interaction of factors working within a real web. It was MacMahon who first talked of a web of causation, but too
often this is still interpreted as a complicated but linear
chain of causation rather than a complicated interrelationship of many factors. A web really means interrelation. I think we have to emphasize this.
BUCK: But, you see, it is so hard to keep the web in mind when you
NAJERA:
TERRIS:
150
TERRIS:
T h e lead content was not enough to produce massive poisoning, just enough to damage the kidneys and raise the
uric acid in the blood.
To test their hypothesis, the Alabama clinicians went to
England and analyzed four bottles of port dating from the
late eighteenth and early nineteenth centuries. They found
large amounts of lead in the old wine, but only traces in
new wine. Who would have thought that the Treaty of
Methuen would have had a hand in the eighteenth-century
gout epidemic? Yet it all fits epidemiologically: only the
Discussions
15 1
rich were affected; they ate a lot of meat, drank a lot of port
wine, and they suffered from gout.
NAJERA:
Using a play as a metaphor is another good way to understand how the concept of the web of causation differs from
other approaches of disease causation. For example, in
some plays one actor, one main character, practically carries
the whole play. This would be analogous to thinking of
disease causation in terms of one agent that is more prevalent, more necessary, more important. In other plays, however, there are many actors with equally important roles;
you need all of them to reach the plays outcome. This is
comparable to approaching disease causation in terms of
how people and other factors interrelate in a complicated
web of causation. Some factors would be more important
than others, of course, just as in most plays you can have
many actors but fewer lead roles. Investigations should aim
at understanding all the factors involved. This should facilitate separating confounding factors in the analysis.
TERRIS:
NAJERA:
TERRIS:
NAJERA:
TERRIS:
NAJERA:
TERRIS:
NAJERA:
152
TERRIS:
NAJERA:
This is what I wanted to come to: the condition of necessary is not so clear, even for the infectious diseases. For
instance, tuberculosis mycobacteria are necessary for a diagnosis of tuberculosis only because we dont call the disease tuberculosis if there are no mycobacteria. But, after
all, what is the real difference between a chronic pulmonary
disease with mycobacteria and a chronic pulmonary disease
without? Not much, except perhaps in the way we study
and describe it, or in how we treat it-if there is a specific
treatment for a specific agent, that is. Aside from this, most
infectious diseases could be considered in a completely
different manner. We could reclassify acute respiratory diseases epidemiologically, rather than accepting pathological
o r therapeutical classifications.
BUCK:
NAJERA:
TERRIS:
NAJERA:
TERRIS:
Discussions
153
Yes. Whats important is how and why-under which environmental conditions including social ones-these factors can reach the host to cause disease. It is not enough to
accept the concept of agent-host-environment, one should
also consider what factors influence these agents to go
through the environment and reach the host. This becomes
much clearer with chronic diseases. Here you find factors-you could call them through factors-that give origin to the agents and also influence the host. In my opinion, these are mostly social factors. Given this, one can
discuss the non-validity of the so-called lifestyles. After all,
lifestyle is a product of the environment, the social environment.
TERRIS:
NAJERA:
BUCK:
I teach it.
Me, too.
TERRIS:
You teach it? Very few teachers give that lecture. My students dont like it. It is too theoretical for them.
NAJERA:
TERRIS:
NAJERA:
154
TERRIS:
NAJERA:
TERRIS:
BUCK: As far as years of life lost goes, I suspect this may have been
established as a methodological approach by Farr when he
used Halleys life table concept.
NAJERA:
Did you know that this Halley was the famous astronomer
for whom the comet that has just visited us again was
named? You are right, Carol, his mortality tables for the
city of Breslau, published in 1693, were one of the first to
relate mortality and age.
TERRIS:
Discussions
155
BUCK:
It would be worthwhile to mention, at this point, that Sydenstrickers early surveys in the United States probably
constitute the landmarks in measurement of morbidity.
In terms of measuring health, some years ago a study
done in the United States compared two groups of children
with different degrees of positive health. It wasnt sick
versus controls in the usual sense, instead the study looked
for determinants of the more positive health of one group.
The idea is good because we seldom ever investigate health,
great or good health versus all other levels. We are still so
disease oriented. There is a problem, of course, in studying
the determinants of good health-the problem of the direction of causality. Hours of sleep, for example, could be
either the result or the cause of your state of health.
TERRIS:
156
TERRIS:
Discussions
157
BUCK:
TERRIS:
BUCK:
TERRIS:
NAJERA:
BUCK:
158
TERRIS:
NAJERA:
BUCK.
TERRIS:
BUCK
TERRIS:
BUCK:
TERRIS:
BUCK:
Discussions
159
TERRIS:
Yet I must admit they may be right. But at this point, its a
little bit like fiber and cancer of the colon; I dont accept
that. I dont accept salt and hypertension, either. If you
really look at the data, its very unclear what the facts are.
BUCK:
But one thing you have to think of (and Im not saying that
this book is the place to put the position forward) is that a
lot of these other factors-diet, fiber, salt, calcium, and all
that sort of thing-may protect target organs from developing a disease that is really a response to stressors. We may
not get the best results if we try to prevent disease by
manipulating these factors one by one. Its conceivable that
if the basic stressors or the stress response can be modified,
the target organs would remain healthy.
TERRIS:
BUCK:
TERRIS:
That is a basic issue. Ive heard this said too many times
now. For instance, Ive heard people at meetings say, Well,
if you cut out one disease, something else takes its place.
Its not true. If you look at the data, nothing takes its place.
There is a decline in the death rate. It was true of infectious
diseases and its true of noninfectious diseases. People die
much later. Fries is right, there is a compression of morbidity.
BUCK:
TERRIS:
BUCK:
TERRIS:
But thats what were doing with heart disease, cerebrovascular disease, accidents. Its all primary prevention.
Its not secondary prevention.
BUCK:
160
TERRIS:
BUCK:
TERRIS:
BUCK:
NAJERA:
TERRIS:
BUCK:
TERRIS:
BUCK:
Hypertension control.
By what method?
By drugs.
I see.
NAJERA:
You will not have later death in a pure sense, though. The
average age at death will remain unchanged. What you will
have is more people reaching that age, but later death, real
later death, you will not have.
TERRIS:
NAJERA:
Up to when?
TERRIS:
NAJERA:
O.K. We will have an age curve that shows that all people
will die at the same time, at the end of the lifespan.
TERRIS:
Right.
DlscussZons
161
NAJERA:
O.K. Then you have prevented all this morbidity, and all
this mortality. People will live healthy lives. . . .
TERRIS:
BUCK:
TERRIS:
NAJERA:
TERRIS:
I dont buy the idea that if we lower the mortality and the
incidence of a disease, something else will take its place.
Nothing else takes its place.
NAJERA:
TERRIS:
Manning Feinlieb showed this very well at one of the International Epidemiological Association meetings. He said
that if you look at the rates, they are going down. Thats all
there is to it. But lets go back to the topic at hand; we were
talking about etiologic investigations and I would like to go
back to Goldberger. I think he was important because his
studies show the similarity between infectious and noninfectious disease methodologies. Goldberger was a master of
observation and experiment.
While he was working in entomology in the United States
Public Service, there was an outbreak of a skin disease
called Schambergs disease. Goldberger was sent to solve
the problem, which he did in a few days. He discovered that
the disease only struck people who slept on straw mattresses. He then experimented on himself and other volunteers by sleeping on contaminated straw mattresses-they
all contracted the disease. Then he sifted particles from the
straw into two clean Petri dishes. The contents from one of
the dishes was applied to the left axilla of a volunteer, and
the skin eruption appeared. The other dish was exposed to
chloroform vapor, and its contents then applied to the right
axilla; there was no eruption. They examined siftings and
found five very small mites which they applied to the axilla
of another volunteer; the characteristic eruption appeared.
They identified the mite and solved the problem.
Goldberger also was part of one of the three groups that
raced to demonstrate that typhus fever is louse-borne:
162
TERRIS:
BUCK:
TERRIS:
BUCK:
TERRIS:
Discussions
BUCK:
163
TERRIS:
BUCK:
164
LLOPIS:
BUCK:
And the fact is, you know, that experimental studies of that
sort can run into serious problems with sample size because
the randomization is of groups of people rather than individuals. In this type of experiment you have to make allowances for clustering, and that leads to much larger sample
sizes.
TERRIS:
BUCK
TERRIS:
BUCK:
TERRIS:
BUCK:
I would agree with you, if we were talking about the etiological implications.
NAJERA:
Discussions
165
logical only in that it tried to find the role that each factor
played.
TERRIS:
BUCK:
TERRIS:
BUCK:
NAJERA:
BUCK:
NAJERA:
BUCK:
166
STATISTICS OF MORBIDITY
Edgar Sydenstricker
Morbidity is one of the terms in the definition of which the dictionary resorts to vague
synonyms. We are told that morbidity is a diseased o r abnormal, not sound, not
healthy, sickly state, and are referred to our
livers in order to illustrate its meaning. Further
reflection might lead us to ask how much morbidity is normal reaction to environment, or
what proportion of illnesses is merely an unavoidable concomitant of the wearing out of
human clocks, to use Pearls metaphor, some of
which are set by heredity to run a shorter time
than others. When is death normal?At threescore years and ten, or at the century mark, or
even at Methuselahs reputed age? How much
of Methuselahs life was occupied in dying?
I am afraid that purely philosophical attempts to define the term will lead to a state of
obfuscation-which might well be regarded as a
form of morbidity in itself. Let us concede at the
outset that morbidity is not as precise a concept
as the statistician would desire; that it is a relative term, since one person may feel ill, stay
away from work longer, be a greater nuisance
than another who has the same objective symptoms; and that morbidity is essentially a subjective phenomenon. But let us take cognizance of
the fact that illness, to use the commoner and
more expressive term, is an undeniable and frequent experience of every person except, of
course, the favored nonagenarian who, after a
career devoted to tobacco, hard liquor, and perhaps other gayer irresponsibilities, is alleged in
newspaper interviews never to have been sick a
day in his life. Unlike birth or death, which can
come but once to an individual, illness may occur often, its frequency depending not only
upon its nature, its causes, and upon the susceptibility of the person concerned, but also upon
its duration in relation to the length of time
Source: Excerpted from Edgar Sydenstricker, The Challenge
ofFacw-, Richard V. Kasius, ed. New York, T h e Milbank Memorial Fund, 1974.
I De Lamar Lecture in Hygiene at the School of Hygiene
and Public Health, the Johns Hopkins University, December
15, 1931.
* * *
The development of morbidity statistics has
been very slow, and they are yet in their infancy.
Their tardy progress may be ascribed to three
principal reasons. One is expressed by the
truism that statistics of a given kind are not
continuously collected on a large scale unless
there is a sufficient demand for their use in
some practical way. A second reason is that the
demand has come for morbidity statistics of
special kinds and for specific population
groups; little, if any, standardization in morbidity statistics has been attained. A third reason follows in some sense from the second-a
confusion as to the concept of morbidity arising
from differences in the uses to which the statistics are put. In addition to this confusion, differences in methods of collecting data, variety in
definitions of a caseof illness, the existence of
peculiar factors that affect the accuracy of the
record, the time element involved, and similar
difficulties, have been deterrents to the accumulation of a large body of homogeneous
morbidity data. It will not be possible upon this
1 ci
168
* * *
Although many kinds of morbidity statistics
exist, their varieties may be classified in five
general groups. I shall refer to each very briefly
in order to present in somewhat greater detail
some results of one study of illness.
1) Reports of Communicable Diseases. In a strict
sense, these are not morbidity data since illness
is not necessarily involved. They exist, or should
exist, for a specific purpose, namely the notification of those diseases for which reasonably
effective methods of administrative control actually have been devised. Only to a limited extent are communicable disease reports useful
for epidemiological studies. As Hedrich and I
have shown ( I ) , not only are the reports of most
diseases extremely incomplete but their incompleteness varies according to age.
2 ) Hospital and Clinic Records. These are of
little use in determining the prevalence or incidence of illness in a population, either in terms
of a gross rate or from any specific disease.
Properly made, as they rarely are, they are valuable for clinical studies and may become more
so as the tendency to hospitalization increases
and as clinicians become trained in analytical
methods.
Sydenstricker
- Company
which pays wages
during disability
____
* * *
169
* * *
I would have liked very much upon this occasion to have been able to bring you fresh reports
upon several field studies of morbidity using or
involving the recording of illness by the method
of continuous observation of population
170
* * *
For an illustration of morbidity studies used
to depict the health of a population we may
turn to the one made in Hagerstown.
The Hagerstown morbidity study ( I I ) included 16 517 years of observation, or an
equivalent of a population of 7079 persons observed continuously for twenty-eight months
beginning December, 192 1. Illnesses were recorded as reported to experienced field investigators visiting each family every six to eight
weeks, the reports being made by the household informant (usually the wife) either as experienced by herself or as she observed them in
her family.
The results of the study indicated that a fairly
accurate record of real illnesses was secured.
Less than 5 percent of the illnesses of exactly
stated durations recorded were one day or less
in duration. Approximately 40 percent were not
only disabling but caused confinement to bed.
It is evident, therefore, that in the main the
illnesses recorded were more than trivial in
their character, in spite of the fact that in some
instances mere symptoms were given as diagnoses. The incidence of acute attacks of specific
and generally recognizable diseases was, we believe, recorded with a satisfactory degree of
completeness. On the other hand, the incidence
of mild attacks, as for example, of coryza, was
quite incompletely recorded as judged by data
on minor respiratory attacks obtained later by
more intensive methods for other population
groups.
For this population 17 847 illnesses were recorded in the twenty-eight month period, an
annual rate of 1081 per 1000 years of life observed, or about one illness per person per year.
This illness rate was over 100 times the annual
death rate in the same population.
Perhaps the most interesting results of this
first morbidity study of a typical population
related to the variations in the incidence of illness according to age. Up to the time the
Svdenstricker
17 1
;:L
2wo
___
- Not sick
g= 1wo
5w
0 0
Ib
20
30
40
AGE
50
60
70
80
I I
AGE
172
Sydenstricker
more accurately described for the higher economic class than for the lower.
173
1000
- Respiratory
900
-.-
----
800
0
0
700
Infections
Organic
Digestive
All others
0
r
5 600
a
2 500
W
5 400
z
z
300
. .
\
200
100
0
10
20
30
40
AGE
50
60
70
80
174
160
140
3
3
120
.\
-.-
----
--
Digestive
Nervous
General
Circulatory
Kidneys and annexa
Bones and locomotion
5 100
1
II
-
2z
a
80
60
40
So, today, we may apply to preventive medicine the test afforded by statistics of illness. It is
true that some of the plagues and pestilences of
Marlowes day have been banished from a part
References
(I) Sydenstricker, E., and A. W. Hedrich. Completeness of reporting of measles, whooping cough,
and chickenpox at different ages. Public Health Rep
64(26):1537-1543, 1929.
(2) Brundage, D. K. T h e incidence of illness
Svdenstricker
among wage-earning adults. Journal of Industrial
Hygiene 12:342, 347, 1930.
(3) Ibid, 340.
(9) Collins, S. D. The place of sickness records in
the school health program. Transactions of the F i f h
Annual Meeting of the American Child Hypene Association, October, 1928.
(5) Downes, J. Sickness records in school hygiene.
A m J Public Health 20:1199-1206, 1930.
(6) Goldberger, J., G. A . Wheeler, and E. Sydenstricker. A study of the relation of diet to pellagra
incidence in several textile communities of South Carolina in 1916. Public Health Rep 35:648-713, 1920, and
later publications.
175
(7) Townsend, J. G., and E. Sydenstricker. Epidemiological study of minor respiratory diseases. Public Health Rep 62(2):112, 1927.
(8) Sydenstricker, E. A study of illness in a general
population group. Public Health Rep 61(39):12, 1926.
( 9 ) Sydenstricker, E. The illness rate among males
and females. Public Health Rep 62(30):1952, 1927.
(10) Sydenstricker, E. Sex difference in the incidence of certain diseases at different ages. Public
Health Rep 63(21):1269-1270, 1928.
(1I ) Sydenstricker E. Hagerstown Morbidity Studies. A Study of Illness in a Typical Population Group.
Public Health Rep, reprints 1 1 13, 1 1 16, 1134, 1163,
1167, 1172, 1225, 1227, 1229, 1294, 1303, and 1312.
each movement upward or downward and assigning to each factor its due share in the
change, then we would be well on the way to
knowing the epidemiology of tuberculosis.
But the record is peculiarly difficult to read
with understanding, because it is immediately
apparent that the most striking changes in mortality rates do not correspond to reasonably
probable changes of like extent in rate of exposure to infection. For instance, nothing that we
know of the habits of mankind and the distribution of the tubercle bacillus would lead us to
suppose that between the first and the second
five years of life there is, in general, a diminution
in exposure to infection which corresponds to
176
177
Frost
the decline in mortality rate. And there is little,
if any, better reason to suppose that the extraordinary rise in mortality from age 10 to age 20,
25, or 30 is paralleled by a corresponding increase in rate of exposure to specific infection.
We are forced, then, to recognize, as at least
highly probable, that the predominant factor in
the up-and-down movement of mortality along
the age scale is change in human resistance.
And this i s a complex of which we have very
little exact knowledge except the plain fact that
age and prior exposure bring no such immunity
against tuberculosis as they establish against
many of the acute infections.
However, my purpose is not to attempt an
interpretation of the age selection of tuberculosis; it is merely to call attention to the apparent change in age selection which has taken
place gradually during the last 30 to 60 years,
and to note that when looked at from a different point of view this change in age selection is
found to be more apparent than real. The agespecific curve of mortality from tuberculosis for
males in the United States Registration Area of
1900 is shown for the years 1900 and 1930 in
Figure 1 and for Massachusetts males for the
years 1880, 1910 and 1930 in Figure 2.
The tuberculosis mortality rates for Massachusetts used throughout this paper are
shown in Table 1. You will note that:
1. At every age mortality is lower in the later
period.
2 . In each period, age selection is generally
similar: mortality is high in infancy; declining
Figare 1. United States Registration Area of 1900, death
rates from tuberculosis, all forms, males by age, 1900
and 1930.
0
0
0
0
400
0
U
W
(I)
I
F
2n
70
A G E IN YEARS
jool\
0
500
W
a
IT
(I)
W
4
F
YEAR 1910
100
10
20
30
40
AGE IN YEARS
50
60
70
178
1880
1890
1900
1910
1920
1930
Males
0-4
5-9
10-19
20-29
30-39
40-49
50-59
60-69
70 +
760
43
126
444
378
364
366
475
672
578
49
115
36 1
368
336
325
346
396
309
31
90
288
296
253
267
304
343
209
21
63
207
253
253
252
246
163
108
24
49
149
164
175
171
172
127
41
11
21
81
115
118
127
95
95
Females
0-4
5-9
10-19
20-29
30-39
40-49
50-59
60-69
70 +
658
71
265
537
422
307
334
434
584
595
82
213
393
372
307
234
295
375
354
49
145
290
260
21 1
173
172
296
162
45
92
207
189
153
130
118
126
101
24
78
167
135
108
83
83
68
27
13
37
92
73
53
47
56
40
They were obtained as follows: For the years 1910, 1920 and 1930-based on U S .
Mortality Statistics-deaths from tuberculosis all forms. For the years 1880, 1890, and 1900
thc rates used are calculated from data compiled by the late Dr. Edgar Sydenstricker from
the state records. Because of differences of classification in deaths, it has been necessary to
base the rates on the deaths recorded as tuberculosis of the lungs to get comparable data
for these years. The rate calculated from the state records for tuberculosis of the lungs has
been multiplied by a factor based on the proportion such deaths bore to those from
tuberculosis, all forms. This factor varied with the year and age considered.
700
0
0
600
Z 500 -
1I
8 500
=
O
ln
w
a
w
a:
400 -
+-I
300 200
4001
300 200
----
100 -
AGE IN YEARS
70
AGE iN YEARS
Frost
Figure 5 shows similarly for males the mortality at successive ages in cohorts of (1870),
1880, 1890, 1900, 1910. Note that terminal
rates for these cohorts make the 1930 curve,
and also that in successive cohorts the age selection has been uniform; with the mortality highest in the first five years and again from 20 to 30
years; thereafter it declines.
This fact was previously noted by K . F.
Andvord ( 1 ) . His interpretation was, in part,
that this regularity of the age curve formed a
basis for extending estimates of future mortality
in the same cohort at higher ages. Such an
interpretation is both tempting and encouraging but perhaps dangerous.
Without attempting to interpret the facts in
detail, certain implications are noted.
1. Constancy of age selection (relative mortality at successive ages) in successive cohorts
suggests rather constant physiological changes
in resistance (with age) as the controlling factor.
2. If, as we may suppose, the frequency and
extent of exposure to infection in early life has
decreased progressively decade by decade,
there is no indication that this has had the effect
of exaggerating the risk of death in adult life
due to lack of opportunity to acquire specific
immunity in childhood.
3. Present day peak of mortality in lute life
does not represent postponement of maximum
179
700
0
0
500
400
I\\
AGE IN YEARS
RECORD LINKAGE
Halbert L. Dunn*
There are many uses for the important records of each person, brought together as a
whole. At times, even now, such a collection is of
sufficient value that it is made at considerable
cost in time and money. Usually, it is the individual who is made to d o the work since he
alone knows where his records are on file. It is
much more difficult for any other person or
organization to assemble the records of his life
since no personal cross-index exists to lead one
to all of a persons records. It is important to
many people and organizations to be able to
assemble this type of information easily and
efficiently.
180
Dunn
IMPORTANCE TO REGISTRARS OF VITAL
RECORDS
The possibility of satisfying his basic responsibilities could be greatly reinforced if the registrar would take on the additional responsibility
of binding the Books of Life into volumes. It is
not necessary for him actually and physically to
assemble the records of a particular individual
and bind them into a volume. For all practical
purposes, the end result will be achieved if he
creates a Life Records Index indicating where all
the most important records of a n individual are
filed.
The accuracy of vital records would be enhanced because of inconsistencies that would
show up. T h e completeness of vital records
would be improved because subsequent documents would show that previous records which
should have been filed had not been placed on
file. Certification would become more secure
from fraud. For instance, birth records of dead
people could not be certified for fraudulent
purposes. Massive certification jobs for government, which of necessity are so often done without charge, could be handled more efficiently
and less expensively through a check-off system. Statistical information would become more
meaningful because it would be linked to other
types of data.
It is not too rash to predict that if the registrars of the country would undertake to prepare and keep up-to-date the Life Records
index on all the people of the country and d o
the job systematically and efficiently, they would
find themselves and their offices to be the focal
point of all records concerning people through-
181
out the country. This would probably be accomplished in a surprisingly few years because the
need for such a record linkage service is very
great. It should be achieved at a relatively small
total cost because the mechanism of the task is a
very simple one, and, in performing this service, registrars would find that they were doing a
particularly fine job in carrying out their primary responsibilities of registration and statistics.
IMPORTANCE TO HEALTH, WELFARE AND
OTHER TYPES OF ORGANIZATIONS
182
edge about their own programs. Most organizations dealing with individuals produce some
type of administrative statistics of the individuals served by them. It would greatly enhance the significance of such statistics if they
could be linked to other facts about the same
individuals, such as, What sort of jobs do they
hold? How many children d o they have?
What sort of illness d o they suffer from?
What kind of social environment do they live
in? In particular health and welfare organizations would find that this type of linked statistical analysis would open new vistas of knowledge to them.
DEATH CLEARANCE
Dunn
183
~~~
We need to know much more about the epidemiology of mental disorders. The
research needed will prove expensive and difficult since it will depend on long
observation and will involve several disciplines, but it is the way we must take to
advance our knowledge of the incidence, duration and prevalence of mental diseases.
This paper gives an opportunity to acquaint
epidemiologists with some of the characteristics
of morbidity data on the mentally ill and some
of the challenging research problems that this
field presents. Because of the complex nature
of the problem, epidemiologic research on the
mental disorders is bringing into the public
health field professional personnel from such
disciplines as psychiatry, psychology, sociology,
anthropology, and psychiatric social work. For
the most part, persons trained in these professions have not been exposed to the philosophy
underlying the epidemiologic approach to the
study of disease and to the statistical methods
used in such studies. This paper, therefore, will
also give an opportunity to provide these professionals with some background material on
two basic morbidity indexes used in studying
the occurrence of disease in population groups.
DEFINITIONS
184
Kramer
resistance of the population and pathogenicity
of the microorganism. This balance may be
called the force of morbidity. The analogy between this concept and concepts that have been
proposed with respect to the incidence of mental disorder in the population is apparent. Incidence of mental disorder would seem to be
dependent upon the balance which exists between resistance of the population and those
forces and stresses-biologic, cultural, psychologic-that produce mental disorder. Doull further states, . . . prevalence is more complex. It
is the resultant of the force of morbidity and
those factors which determine whether the interval between onset and termination shall be
long or short, whether a disease shall be acute
or chronic.
In more formal terms, incidence measures
the rate at which new cases are added to the
population of sick persons and-in conjunction
with the decrement rate, i.e., the rate at which
the disease is arrested, cured, or at which
affected individuals are removed from the population by death-determines the size and composition of the sick population. Thus, the prevalence rate of a disease is a function of the
incidence rate and the duration of the disease.
HYPOTHETICAL EXAMPLES
185
186
Community A. Assumptions: 1000 new cases annually, each of which occurs on January 1 of specified year;
100 patients annually are cured in each cohort of such new cases. The prevalence count will stabilize at 5500
cases on January 1, 1949.
Cohort
of year
1941
1942
1943
1944
1945
1946
1947
1948
1949
1950
1940
1941
1942
1943
1944
1945
1946
1947
1948
1949
1950
1000
900
1000
800
900
1000
700
800
900
1000
600
700
800
900
1000
500
600
700
800
900
1000
400
500
600
700
800
900
1000
300
400
500
600
700
800
900
1000
200
300
400
500
600
700
800
900
1000
100
200
300
400
500
600
700
800
900
1000
0
100
200
300
400
500
600
700
800
900
1000
Total
1000
1900
2700
3400
4000
4500
4900
5200
5400
5500
5500
~
Community B. Assumptions: 1000 new cases annually, each of which occurs on January 1 of specified year; 10
percent of those ill at the beginning of each year are cured during that year. The prevalence count will stabilize
at 10 000 cases on January 1, 2007.
Cohort
of year
1941
1942
1943
1944
1945
1946
1947
1948
1949
1950
1940
1941
1942
1943
1944
1945
1946
1947
1948
1949
1950
1000
900
1000
810
900
1000
729
810
900
1000
656
729
810
900
1000
590
656
729
810
900
1000
531
590
656
729
810
900
1000
478
531
590
656
729
810
900
1000
430
478
531
590
656
729
810
900
1000
387
430
478
531
590
656
729
810
900
1000
348
387
430
478
531
590
656
729
810
900
1000
Total
1000
1900
2710
3439
4095
4685
5216
5694
6124
6511
6859
Community C. Assumptions: 2000 new cases annually, each of which occurs on January 1 of specified year; 40
percent of those ill at the beginning of each year are cured during that year. The prevalence count will stabilize
at 5000 cases on January 1, 1956.
Cohort
of year
1941
1942
1943
1944
1945
1946
1947
1948
1949
1950
1940
1941
1942
1943
1944
1945
1946
1947
1948
1949
1950
2000
1200
2000
720
1200
2000
432
720
1200
2000
259
432
720
1200
2000
155
259
432
720
1200
2000
93
155
259
432
720
1200
2000
56
93
155
259
432
720
1200
2000
34
56
93
155
259
432
720
1200
2000
20
34
56
93
155
259
432
720
1200
2000
12
20
34
56
93
155
259
432
720
1200
2000
Total
2000
3200
3920
4352
4611
4766
4859
4915
4949
4969
4981
Kramer
187
188
Figure 1. Age-specific first admission and resident patient rates per 100 000 civilian population, state hospitals for
mental disease, all mental disorders and selected mental disorders, both sexes, United States, 1952.'
ALL MENTAL
1300 -
1200
1000
1100 900
800
700
600
500
400
SCHIZOPHRENIA A N D MENTAL
DISORDERS OF T H E SENIUM
DISORDERS
SCHIZOPHRENIA
-RESIDENT PATIENTS
.____RESIDENT
PATIENTS
......FIRST ADMISSIONS
RES.DEhT PATIENTS
300 ,,--'
UNDER15
aSOUrCe: Patients in Mental Institutions, 1952, Part V (In preparation), National Institute of Mental Health, Public Health Service
Kramer
189
Fignre 2. Percent distributionby time on books and median time on books of patients residentin New York State Civil
Hospitals for mental disease, both sexem by age and selected diagnosis, March 31, 1955:
NUMBER
OF CASES
SCHIZOPHRENIA
AGE GROUP
75 +
MEDIAN TIME ON
BOOKS (IN YEARS)
2301
20 +
65-74
6346
20 +
45-64
23 608
16.9
25-44
19 001
6.1
15-24
2000
1.6
UNDER 15
204
1.2
75 +
7617
3.1
65-74
5300
2.9
45-64
1572
2.2
PERCENT 0 10 20 30 40 50 60 70 80 90 100
PERCENT 0 10 20 30 40 50 60 70 80 90 100
TIME ON BOOKS
UNDERlYR.
1TO4YRS.
5TO9YRS.
10TO19YRS.
POYRS.&OVER
'Source: Unpublished data requested of model reporting area stales, collected for the 1955 Census of Mental Patients, National
Institute of Mental Health, Public Health Service.
190
W 3. Percent of first admiasions retained in hospital, released alive and dead withh specified pedode following
admission to Warren State Hospital, Warren, Pennsylvania, 1936-1945. Selected diagnoses, aJJ ages, botb sexes.
RETAINED CONTINUOUSLY
100
90
80
60 70
z 60
h,
DIAGNOSTIC GROUP:
-SCHIZOPHRENIA
--
---_
(1271)
MANIC-DEPRESSIVE
PSYCHOSES
(408)
INVOLUTIONAL
PSYCHOSES
(180)
SYPHILITIC
PSYCHOSES
(378)
MENTAL DISEASES
OF SENIUM
(1031)
a:
I-
6 50
I(/]
40
IZ 30
W
0
UZ
w 20
10
RELEASED ALIVE
DYING IN HOSPITAL
c.----
/--
/
/
0@
relatively high release rate, with 45 percent released within one year following admission and
61 percent released within five years, their total
separation rate is lowest because of the relatively
low death rate.
If we compare the percentage distribution of
diagnoses at time of admission among the first
admissions in the five diagnostic categories
shown in Figure 3 with the corresponding distribution for those retained continuously for
five years we find some striking differences.
The schizophrenics constitute 39 percent of this
group at time of admission and 65 percent of
those hospitalized continuously for the five
years, while for the mental diseases of the senium the corresponding percentages are 3 1
and 14 percent, respectively.
Similar differences can be shown by considering the fate of patients in relation to age at first
admission. The youngest patients have the
highest release rates and lowest death rates and
the oldest the lowest release rates and the
highest death rates (Figure 4). Variations in release and death rates by age and diagnosis are
rather interesting but will not be dwelled upon
here (6).
Kramer
191
Figure 4. Percent of first admissions retained in hospital, released alive and dead within specified periods following
admission to Warren State Hospital, Warren, Pennsylvania, 1936-1945. All mental disorders, selected admission
ages, both sexes.
RETAINED CONTINUOUSLY
RELEASED ALIVE
100 3
AGE GROUP:
15-34 Yrs.
(1575)
35-54 Yrs.
(1738)
_---55-74 Yrs.
(1341)
7 5 Yrs. & Over (536)
--.
90
I-
a 80
DYING IN HOSPITAL
0 70
z 60
I-
50
40
I-
5
0
30
20
10
Y
1
l
5
1946-1950 cohort within one year after admission, the percentage of patients released has
increased from 14 percent to 39 percent and
the percentage of patients dying has dropped
from 56 percent to 18 percent. This has resulted in increasing the percentage of patients
retained from 30 percent to 43 percent.
DISCUSSION
192
Table 2. Age-specific first admission rates and resident patient rates for syphilitic psychoses,
United States, 1939 and 1952.
First Admission
Resident Patient
Percent
change
1939
Percent
change
1939
1952
1952
8.0
8.6
2.1
- 73.8
25.9
26.0
2.3
- 73.3
28.0
28.7
1.0
6.7
14.1
13.1
11.0
4.7
0.4
0.8
2.7
4.4
4.1
2.5
- 60.0
-88.1
- 80.9
- 66.4
- 62.7
- 46.8
2.2
14.2
41.2
51.2
45.3
18.7
1.8
5.4
25.4
64.1
57.6
37.1
- 38.3
+ 25.2
27.2
98.4
25276
31 484
+ 24.6
533
3013
7485
7857
4753
1635
346
1259
5573
11 492
7997
4817
15-24
25-34
35-44
45-54
55-64
65 and over
+
+
-
15-24
25-34
35-44
45-54
55-64
65 and over
7781
2532
23 1
1409
2567
2010
1150
414
71
183
594
786
568
330
-67.5
- 69.3
- 87.0
- 76.9
- 60.9
- 50.6
- 20.3
2.5
18.2
- 62.0
+
+
N u m b e r of patients
0.4
-35.1
- 58.2
- 25.5
+46.3
+ 68.3
+ 194.6
Rates per 100 000 population as ofJuly 1939 from series P45, No. 5 and 1952 from series P25, No. 121, Current Population
Reports, Population Estimates, Bureau of the Census, U. S . Department of Commerce, Washington, D. C.
Adjusted to age distribution of population of the United States as of July 1, 1952. Series P25 No. 121.
Figure 5. Percent of first admissions retained in hospital, released alive and dead within specified periods following
admission to Warren State Hospital, Warren, Pennsylvania, 1916-1950. Syphilitic psychoses, all ages, both sexes.
RELEASED ALIVE
100
90
---
:
I-
COHORT OF:
1916-25
1926-35
---- 1936-45
-194640
80
70
DYING I N HOSPITAL
(402)
(320)
(378)
(114)
(3
Z 60
IU
2 50
8 40
0
k-
5
30
0
U
20
10
Kramer
193
194
Krurner
195
Our knowledge of the epidemiology of mental disorders must be extended beyond that
gathered through studies of mental hospital
populations and the studies of either treated or
true prevalence (i.e., treated plus untreated
cases) of mental disorders. To accomplish this,
Figure 6. Pemnt of first admissions retained in hospital, released alive and dead within specified periods following
admission to Warren State Hospital, Warren, Pennsylvania, 1916-1950. All mental disorders, under 65 years,
both sexes.
100
RETAINED CONTINUOUSLY
90
80
0
I
0 70
0
0
I-
a:
50
&
40
I-
----_
-
(2559)
(3273)
(4052)
(2251)
30
5 20
DYING IN HOSPITAL
60
I-
RELEASED ALIVE
COHORT OF:
1916-25
1926-35
1936-45
1946-50
10
5
1
2
3
4
5
YEARS FOLLOWING ADMISSION
196
Figure 7. Percent of first admissions retained in hospital, released alive and dead within specified periods following
admission to Warren State Hospital, Warren, Pennsylvania, 1916-1950. All mental disorders, 65 years and over,
both sexes.
RETAINED CONTINUOUSLY
100
90
F
[r
80
0 70
0
z
Icc
2
(I)
60
50
RELEASED ALIVE
DYING IN HOSPITAL
COHORT OF:
-1916-25
------
(500)
1926-35 (699)
1936-45 (1211)
1946-50 (927)
8 40
+
30
20
10
5
1
2
3
4
5
YEARS FOLLOWING ADMISSION
disorders, pass through various stages of disability, achieve various stages of recovery or die.
It is only through such studies that we will advance our knowledge of the incidence, duration, and prevalence of mental disorders.
References
( 1 ) Dorn, H. Methods of measuring incidence
and prevalence of disease. A m
Pukzc Health
41:271-278, 1951.
(2) Doull, I.A., R.S. Guinto, et al. The incidence
of leprosy in Cordova and Talisay, Cebu, P.I. Int J Lepr
Vol. 10, 1942.
(3) Sartwell, P.E., and M. Merrell. Influence of
the dynamic character of chronic disease on the interpretation of morbidity rates. A m J Public Health
45:579-584, 1952.
(4)Proceedings of the Conferences of Mental
Hospital Administrators and Statisticians: ( 1) First
Conference, February, 1951, PHS Pub. No. 295,
1952; (2) Second Conference, February, 1952, PHS
Pub. No. 266, 1953; (3) Third Conference, April,
1953, PHS Pub. No. 348, 1954. Washington, D.C.,
Government Printing Office.
(5) Kramer, M., H. Goldstein, R.H. Israel, and
N.A. Johnston. Disposition of First Admissions to a State
Kramer
Mental Hospital: Experience of the Warren State Hospital
During the Period 191 6-1 950. Pub. Health Monograph
No. 32., Washington, D.C., Government Printing Office, 1955.
(6) Kramer, M., H. Goldstein, R.H. Israel, and
N.A. Johnston. Application of Lfe Table Methodology to
the Study of Mental Hospital Populations. Psychiatric Research Rep. No. 5, American Psychiatric Association,
1956, pp. 49-76.
(7) Hollingshead, A.B., and F.C. Redlich. Social
197
INTRODUCTION
Cravioto et al.
having experienced a significant degree of malnutrition earlier in life. In this way groups of
children with common ethnic background were
identified and represented the upper and lower
quartiles by height of individuals in the age
groups studied.
Clearly, at least three important variables
must be controlled for when height for age is
being used as an index of prior nutritional risk.
The first relates to parental stature and thus to
familial factors affecting height. Since height at
school age may reflect not only the individuals
nutritional background but also his parental endowment it was necessary in designing the
study to obtain anthropometric information on
parents as well as on children in order to control
for this variable.
A second consideration is that low stature
during the years studied may represent a general maturational lag in the course of which
both height and intersensory functioning may
both be subnormal. To control for this possibility it was necessary to study a second sample
of children of the same ages who exhibited
equivalent differences in height but who had
little or no likelihood of ever having been at
nutritional risk.
Finally, since no integrative capacity is unaffected by environmental influences, comparative information on the social, economic, and
educational status of the families from which
the children derived had to be obtained. Details
of the procedures used for obtaining background data on the children studied will be
presented at a later point.
The indicator of neurointegrative development selected for study was intersensory organization. This was done for two reasons. In the
first place a considerable body of evidence both
in comparative psychology ( I , 2) and evolutionary physiology (3) has accumulated which suggests that the emergence of complex adaptive
capacities is underlain by the growth of increasing liaison and interdependence among the
separate sense systems. Sherrington ( 4 ) in considering this process has gone so far as to argue,
The naive would have expected evolution in its
course to have supplied us with more various
sense organs for ampler perception of the
world. . . . The policy has rather been to bring
by the nervous system the so-called five into
closer touch with one another. . . . A central
clearing house of sense has grown up. . . . Not
199
Thanks are expressed to the administration of the American School in Guatemala for their cooperation in permitting
us to study their pupils.
200
ings obtained through studying the urban lettuce and cabbages, and vegetables including
group could be treated not only with respect to carrots and green peppers grown in small famthe problem of stature but also as a device for ily gardens. Flowers are cultivated for commerteasing o u t the relations between growth and cial purposes in separate plots. There is comfunction in a socially differentiated rural group mercial interchange between the village and the
by comparing it with a socially and economically capital city, as well as with the City of Antigua,
homogeneous urban group in which height dif- the main town in the Department.
ferences existed hut could not be related to conditions of nutrition, health, or social standing. Pattern of Food Consumption
METHOD AND PROCEDURE
The analysis of three dietary surveys conducted during May, June, and November of
1963 and a comparison with the data obtained
for the same village in 1950 by Flores and Reh
(9)revealed a small increase in the consumption
of milk derivatives, greens, bananas, grains,
roots and fats, and a reduction in the intake of
corn and fruits. Nevertheless, the present diet
continues to be protein poor and is not significantly more adequate than that consumed 13
years ago. Table 1 is illustrative both of the poor
nutritional quality of the current diet and of
this lack of improvement (10).
Table 1. Calorie contribution of the protein, fats,
and carbohydrates consumed in Magdalena, M.A.,
1950 and 1963.
Percentages contributed to the
total calorie value of the
diet oeriod studied
~~
Source
Protein
Fats
Carbohydrates
May
1950
bay
1963
June
1963
Nov.
1963
12
12
11
10
12
12
10
80
77
78
78
Migration
Approximately 10 percent of the families
make an annual transient emigration to the
coastal region at the time of the coffee harvest.
During a period of one to four weeks the whole
family leaves the village. Actual emigration, that
is leaving the village permanently, is so rare that
it can he confidently stated that for practical
purposes there is no emigration. The same is
substantially the case for immigration.
Transportation
Buses to and from the capital city are available daily, and twice a week a bus runs to and
from Antigua, the administrative and political
head of the Department of Sacatepequez.
Cruvioto et ul.
Height Measurements
The standing height of all the children aged
6 to 11 was measured by two pediatricians previously trained in standardized procedures and
compared one against the other through a series of exercises until they achieved replicate
measurements varying by no more than 0.4cm
(11).All measurements were made by means of
a firm wall board with a simple counterweight
attached to the head block. T h e child was
helped to stand erect with heels, buttocks, and
shoulders tangentially against the wall board.
The position of the child with the heels together, and feet at an angle of 45 to each other
was assured by drawing the shape of them on
the base of the apparatus and positioning each
child to these drawings. Parents heights were
measured in the same way.
All the measurements were then arranged in
decreasing order of magnitude. Quartiles were
calculated for each age and sex. All the children
who fell in the upper and lower height quartiles were selected to be tested for intersensory development. The number, age, and sex of children included in the rural sample are shown in
Table 2.
Table 2. Age and sex distribution of the rural
children studied.
Age
7
8
9
Boys
21
11
10
11
16
16
9
Total
79
Girls
13
10
9
11
Total
19
31
9
12
20
27
25
21
64
143
201
tification as to ethnic group, the languages spoken in and outside the home with adults and
children.
(b) Factors that may influence health-such
as house sanitation, personal hygiene, presence
of poisonous animals and vectors of disease,
crowding of adults and children, and presence
of domestic animals in the home without separation from the persons.
(c) Educational background-literacy and
schooling of the parents and use of such communications media as books, radio, and newspapers.
(d) Occupations and leisure time practicessources of income for each member of the
household, contribution to the home budget;
the family budget for food and other purposes;
the use of free time; the organizations to which
they belonged; and the attachment to religious
organizations and practices were determined.
(e) Availability of food in the family-the actual food produced by the family, the technique
of production, the disposition of food produced, percent of produce sold, used for animal feed, left for seed, wasted because of inadequate storage; the types and amounts of food
purchased were assessed.
To provide background data three dietary
surveys were conducted in a sample of 57 family
households, with the main objective of assessing
the food consumption pattern and the adequacy of the diets in comparison to figures
available for practices 13 years before.
202
Boys
Girls
10
5
10
11
10
15
9
10
9
10
10
20
20
20
20
20
20
Total
63
57
120
8
9
11
10
11
Total
actively explored the form with his hand outside his field of vision. Kinesthetic information
was provided by placing the subjects preferred
arm behind a screen and, with the arm out of
sight, passively moving it through a path describing the geometric form. This was accomplished by placing a stylus held in normal writing position in t h e subjects hand. T h e
examiner gripped the stylus above the point at
which it was held by the subject and then moved
the stylus and hand through the path of a track
describing the geometric form inscribed in a
linoleum block. The track forms were made
from 4 in. by 5 in. linoleum blocks in which the
patterns were inscribed to a depth of YS in.,
forming a track through which the stylus could
be moved. The outline dimensions of the track
were of the same size and shape as those of the
various blocks used for visual and haptic stimulation.
For all sensory modalities the forms were always presented so that the long axes were paralFigure 1. Geometric forms used for testing
intersensory integration.
Crauioto et al.
lel to the frontal plane of the subject. In a kinesthetic trial the movement was always started at
the topmost point of the figure and continued
in a clockwise direction for the right hand and
in a counter-clockwise direction for the left
hand. In putting the subjects hand through the
motion, a short pause (approximately a second)
was made at each point of the course where
there was a change of direction. For each kinesthetic trial the subjects hand was put through
only one complete circuit from topmost point to
topmost point.
Three cross-modality interactions were explored for intersensory equivalences: visual and
haptic interaction, visual and kinesthetic interaction, and haptic and kinesthetic interaction.
Paired stimuli were presented for comparison,
the first member of the pair to one sense
modality and the second member of the pair to
the other sense modality. In a given series of
presentations the first member of the pair was
held constant as a standard against which varying second members were compared successively. At the end of a complete series of
trials a new standard stimulus was introduced
against which the various forms presented to
the other modality were compared. This procedure was repeated until the subject was examined with each of the eight forms serving as
standard.
The order of presentation is given in Table 4.
In this table the column headings represent either visual or haptic presentations of the standard stimuli. The stimuli listed in the columns
represent successive forms which the subjects
Table 4. Order of presentation of standard and
variable stimuli for testing intersensory
functioning.
Standard stimulia
T R I HEX SQU H-C CRO DIA STA CIR
2
di
Variable
stimuli
st
tr
hc
hx
tr
cr
;:
cr
hx
ci
sq
di
hc
tr
st
tr
sq
di
hx
hc
sq
CI
cr
cr
hc
tr
hx
hc
st
sq
ci
di
sq
hx
di
cr
tr
CI
cr
st
hc
di
tr
di
cr
hc
hc
ci
hx
di
hx
tr
sq
st
sq
st
cr
st
ci
cr
ci
hc
hx
st
ci
sq
tr
di
~~
=
=
203
204
Mean
Visualhaptic
Range
Mean
Range
Haptickinesthetic
Mean
6
7
8
9
10
11
5
3
1.51
1.28
1.64
0.66
0-15
0-10
0-7
0-7
0-4
0-4
1.06
0.28
0.25
0
0
0
Range
~
0-8
0-2
0-4
0
0
0
3.67
2.0
1.5
1.0
1.44
0.76
0-13
0-8
0-6
0-4
0-6
0-3
VisualhaDtic
Haptickinesthetic
Age
Mean
Range
Mean
Range
Mean
Range
6
7
8
9
10
20
6
7.3
1.9
1.36
1.36
1-53
0-30
0-54
0-6
0-9
0-4
13.1
2.5
2.0
1.2
0.96
1.14
1-56
0-6
0-6
0-5
0-5
0-4
15.6
4.5
3.7
0.76
1.24
0.85
0-56
0-31
0-31
0-4
0-4
0-4
11
Cravioto et al.
Figure 2. Mean e m = for the intersensory judgment of
identical geometric forms by rural children at
different ages.
5
_ _ - - - - Visual kinesthetic
,,
----
Haptic kinesthetic
Visual haptic
v)
g 3
U
w
\
\
LL
5 2
,
\
rn
li
dren are presented. As may be seen from Figures 4 to 9, and fromTable 7 the curves of both
groups are markedly similar and differ from
one another primarily in the value of the constants. Further, as may be seen from the illustrations, the calculated theoretical growth curves
and the empirical findings fit one another very
closely.
205
10
11
12
AGE IN YEARS
proved with age and had the form of a logarithmic curve of growth. As was the case for
the New York suburban school children studied
by Birch and Lefford (7), the different pairings
of sensory interrelations did not develop to the
same degree o r at the same rate. In the
Guatemalan rural children, as in the New York
chiidren, visual-haptic integration was significantly more effectively organized at every age
than were either visual-kinesthetic or haptickinesthetic integrative interrelationships. The
error curve for visual-haptic integration
reached an asymptote between the seventh and
eighth year, and performance on this task was
errorless after that age. In contrast, neither visual-kinesthetic nor haptic-kinesthetic integrative performance reached an errorless level
of competence within the range of ages studied.
As may be seen from Figures 2 and 3, both
groups exhibited patterns of improvement with
age which approximated the form of a growth
function whose theoretical value may be expressed in the formula Y = KeCX (where Y equals
errors made, x the age of subjects in years, and
e the base of the natural system of logarithms,
K and c are empirical constants determined by
the data.) In two cases these were not the curves
of best fit and the data were assimilated to a
straight line. In Table 7 the empirical equations
for the development of intersensory integration
with age in the two growth groups of rural chil-
______
----
Visual kinesthetic
Haptic kinesthetic
Visual haptic
1614-
12-
2U 0
10-
:85 6-
42-
9
86
AGE IN YEARS
206
Table 7. Empirical equations of error decrement with age, calculated from the performance of two
groups of rural children with extreme differences in height.
Quartile for
height
Sense modalities
tested
Visual-kinesthetic
Judgment of
nonidentical forms
Y-30.34 + 2 . 5 5 ~ .
Y = 124.3e0 4 5 0 ~
Lower
Upper
Lower
Upper
Lower
Upper
Visual-haptic
Haptic-kinesthetic
Y=20.68+ 1 . 7 9 ~ ~
Y = 12.51eO230~
Y = 241. le 0 5 2 6 ~
y= 14.3-0514~
a Since the better fit for these curves of improvement with age was a straight line, they were assimilated to the equation
Y=a+bx.
Observed figures
figures - - - - - Calculated figures
----
LOWER QUARTILE
FOR HEIGHT
Observed figures
Calculated figures
Differences in the number of errors of equivalence made by the two height groups in judging nonidentical forms also tended to favor the
taller group of children (Figures 5, 7, and 9 and
Table 10). This difference was most notable in
the performances of the youngest groups of
children, the 6-year-olds. Figures 10, 11, and 12
present the data on the cumulative percentage
of 6-year-old children in the two height groups
who made errors of equivalence and clearly
indicate the lag in development of intersensory
competence that was present in the shorter children during the first school year.
1 \
I
UPPER QUARTILE
FOR HEIGHT
Observed figures
Calculated figures
20
cn
p 16U
Observed figures
Calculated figures
g 12a
w
,
6
8
9 1
AGE IN YEARS
8-
4-
->.
1
----
LOWER QUARTILE
FOR HEIGHT
[r
------
8
9 1 0
AGE IN YEARS
-.
1
207
Cravioto et al.
Table 8. Mean and range of height in centimeters of rural school age children at extremes of
difference in stature.
6
98
96-99.5
106
101-108
108
99-110
105-113
113
119
116-122
Age
107
111.5-118
---
LOWER QUARTILE
FOR HEIGHT
119
113-123
127
124-133
124
122-126
134
131-138
u)
----
iB-a
LOWER QUARTILE
FOR HEIGHT
Observed figures
Calculated figures
11
113
114
107-116.5 109-116.5
UPPER QUARTILE
10
fz
Calculated figures
i
L-
-*
7
AGE IN YEARS
8
9 1
AGE IN YEARS
16
5 12
UPPER QUARTILE
FOR HEIGHT
Observed figures
Calculated figures
LOWER QUARTILE
FOR HEIGHT
Observed figures
Calculated figures
Em a
fZ
-----
---~
UPPER QUARTILE
FOR HEIGHT
Observed figures
Calculated figures
LOWER QUARTILE
FOR HEIGHT
Observed figures
Calculated figures
fi
6
8
9 1
AGE IN YEARS
8
9 1
AGE IN YEARS
---- --
----
208
Age
Visual-haptic
quartile for height
UDDer
Mean
Lower
Haptic-kinesthetic
quartile for height
Upper
Mean
Mean
Mean
Lower
N
Upper
Mean
Mean
4.75
(0-13)
10
2.6
(0-8)
5
(0-15)a
10
3.7
(0-9)
1.12
(0-6)
10
1.0
(0-8)
17
3.53
(0-10)
14
2.35
(0-5)
17
0.29
(0-2)
14
0.28
(0-2)
17
2.53
(0-8)
14
1.64
(0-3)
10
2.1
(0-7)
10
1.2
(0-3)
10
0.40
10
0.10
10
1.90
(0-4)
10
1.10
(0-3)
16
1.93
(0-7)
11
0.63
(0-3)
16
11
16
1.46
(0-4)
11
0.54
(0-3)
10
11
1.90
(0-4)
14
0.92
(0-4)
11
14
11
1.63
(0-5)
14
1.28
(0-6)
11
10
1.20
(0-4)
11
0.18
(0-1)
10
11
10
1.20
(0-3)
11
0.36
(0-2)
N = Number of children.
In considering errors it is clear that the examination of group differences can best be explored before performance reaches a common
maximum level. Since visual-haptic integration
tends to reach its asymptote very early, only the
youngest children showed differences in this integrative performance that are associated with
differences in height. At 6 years of age the
shortest of the children made a mean error of
nonequivalence of 1.12 and a mean error of
Table 10. Errors made in the recognition of identical forms by tall and short rural children.
Age
Lower
Mean
Haptic-kinesthetic
quartile for height
Visual-haptic
quartile for height
Visual-kinesthetic
quartile for height
Upper
Mean
Lower
Mean
Upper
Mean
24.50
(3-52)a
10
15.60
(1-53)
19.8
(3-56)
10
6.4
(1-21)
17
4.88
(0-14)
14
7.00
(0-30)
17
2.58
(0-6)
14
2.35
(0-4)
17
10
12.20
(0-54)
10
2.40
(0-7)
10
2.3
(0-6)
10
1.9
(0-5)
10
16
2.00
(0-4)
11
1.72
(0-6)
16
1.66
(0-5)
11
0.81
(0-3)
10
11
2.81
14
1.50
(0-4)
11
1.27
(0-5)
14
11
1.36
(0-4)
10
1.20
(0-4)
11
(0-4)
11
10
1.30
(0-3)
Upper
Mean
19.25
(0-56)
10
12.10
(2-26)
3.0
14
6.07
(1-31)
6.0
(0-37)
10
1.40
(0-4)
16
1.26
(0-4)
11
0.27
(0-1)
0.71
(0-5)
11
1.45
(0-2)
14
1.07
(0-4)
1.09
(0-3)
10
1.30
(0-4)
11
0.45
(0-2)
Figures in parenthesis show the range of errors. Age is given in completed years.
N = Number o f children.
Lower
Mean
(0-11)
Cravioto et ul.
Figure 10. Proportions of tall and short six-year-old
rural children making ermra of equivalence in
visual-haptic judgment.
loo[
70 -
60-
F
2
50 -
Bf
42
I
LOWER QUARTILE
UPPER QUARTILE
209
Figure 12. Proportions of tall and short six- and eightyear-old rural children making errors of equivalence in
visual-kinesthetic judgment.
-----
40-
30-
10
c":
I'
2
NUMBER OF ERRORS
UPPER QUARTILE
NUMBER OF ERRORS
__
2 10
Table 11. Mean and range of height in centimeters of the groups of school age children of an urban
upper social class tested for intersensory organization.
6
10
11
111
106-115.5
115
110-118
121
119-123
130
125.5-133
130
127-132
135
32.5-137.5
126
124.5-129
130
128.5-135
137
132.5-141
141
137-147
147
145-150
156
155-157
Age
manifest in Figures 3 and 4 in which the performance of the total rural and total urban
groups for visual-kinesthetic and haptic-kinesthetic integration are presented. The developmental course is identical for the two groups
and a simp1emodification Of Onstants
clearly result in the superimposition of the agesDecific error curves.
The differences between the two extreme
height groups in the upper social class urban
sample in intersensory integrative ability are
presented in Tables 14 and 15. For neither errors of equivalence or nonequivalence, nor for
.-.-*
Visualkinesthetic
Haptickinesthetic
Age
Mean
Range
Mean
Range
Mean
Range
AGE IN YEARS
2
0.8
1.1
1.3
1.1
0.4
0-6
0-4
0-3
0-3
0-3
0-2
0.7
0
0.1
0.2
0.1
0
0-3
0
0-1
0-1
0.1
0
1.6
1.1
0.5
0.7
1.5
0
0-5
0-3
0-2
0-3
0-4
0-3
7
8
9
10
11
6-
- - X - - -X RURAL GROUP
.-.
X-
2 5-
Ba
URBANGROUP
w 4LL
6
7
8
9
10
11
Mean
1.7
0.9
3.0
1.9
1.3
0.7
Range
0-5
0-3
0-7
0-5
0-3
0-4
Visualhaptic
5m 3-
Haptickinesthetic
Mean
Range
Mean
Range
0.7
1.2
1.2
0.5
0.9
0.4
0-3
0-4
0-5
0-2
0-4
0-4
1.1
0.7
0.5
0.8
1.0
0.3
0-2
0-2
0-3
0-3
0-3
0-2
2 *w
0
2w
>
a
I
4 Ib
AGE IN YEARS
11
I>
2 11
Cravioto et al.
Table 14. Errors made in the recognition of identical forms by two height groups of upper social class,
urban children.
Visual-kinesthetic
quartile for height
Age
6
7
5
5
Lower
Mean
2
(0-6p
1.6
(0-4)
Visual-haptic
quartile for height
Upper
Mean
2
(1-4)
0
Lower
Mean
1.o
Haptic-kinesthetic
quartile for height
Upper
Mean
Lower
Mean
Upper
Mean
0.4
(0-2)
1.8
(0-5)
1.4
(0-3)
(0-3)
5
0
(0)
0
(0)
1.4
(0-3)
0.8
(0-2)
(0)
1.8
03-31
0.4
(0-2)
0.2
(0-1)
0
(0)
1
(0-2)
0
(0)
1.8
(1-3)
0.8
(0-2)
0.4
(0-1)
0
(0)
1
(0-3)
0.4
(0-2)
10
1.6
(0-6)
0.6
(0-1)
0
(0)
0.2
(0-1)
2
(0-4)
1
(0-4)
11
0.4
(0-2)
0.4
(0-1 )
0
(0)
0
(0)
1
(0-3)
0.6
(0-2)
Figures in parenthesis show the range of errors. Age is given in completed years.
N = Number of children.
differences in height as such, when they occurred in children who were not at risk of nutritional deprivation, did not result in differences
in the rate of intersensory development or in
Table 15. Errors made in the recognition of nonidentical forms by two height groups of upper social
class, urban children.
Age
Lower
Mean
1.4
(0-4p
Haptic-kinesthetic
quartile for height
Visual-haptic
quartile for height
Visual-kinesthetic
auartile for heipht
Upper
Mean
2
(0-5)
Lower
Mean
0.6
(0-2)
Upper
Mean
0.8
(0-3)
Lower
Mean
0.8
(0-2)
Upper
Mean
1.4
(0-2)
1
(0-2)
0.8
(0-3)
0.8
(0-3)
1.6
(0-4)
0.8
(0-2)
4.4
(2-7)
1.6
(0-5)
2
(0-5)
0.4
(0-11
0.8
(0-3)
0.2
(0-1)
1.2
(0-2)
2.6
(2-5)
0
(0)
0.4
(0-2)
1.2
(0-3)
1.6
(1-3)
1
(0-3)
1
(0-4)
1
(0-2)
1.2
(0-4)
0.2
(0-1)
0.8
(0-4)
0.4
(0-2)
10
11
5
5
( 1-54
0.8
(0-3)
0
(0)
~~
N = Number of children.
0.6
(0-1)
(0-3)
0.2
(0-1 )
2 12
51
Table 16. The mean height in centimeters of two groups of rural school age children
and of their parents.
Age (years
completed)
Mean height
Children
Fathers
Mothers
a
QLa QUb
98
151
143
107
155
145
QL
106
154
143
QU
113
155
144
QL
108
153
141
QU
119
157
143
QL
113
151
143
10
QU
124
160
143
QL
114
154
144
11
QU
127
156
145
QL
119
151
144
QU
134
157
147
Cravioto et al.
213
Table 17. The mean height in centimeters of two groups of upper social class, urban children
and of their parents.
Age (years
completed)
Mean height
Children
Fathers
Mothers
a
QLa QUb
111 126
163 176
158 158
QU
QL
115
163
130
163
159
1.50
QU
QL
121
162
152
137
175
157
QL
130
168
154
10
QU
141
174
156
QL
130
161
156
11
QU
147
178
153
QL
135
171
153
QU
156
175
160
Table 18. Average annual per capita income in two groups of rural families with school age children
showing extreme differences in height.
6
10
11
79.22
121.0
85.2
100.8
101.7
98.7
132.6
101.5
107.5
92.5
116.9
92.3
Age of children
Table 19. Mean percentage of total expenditure spent on food by two groups of rural families with
school age children showing extreme differences in height.
6
10
11
66.P
61.7
60.2
60.9
64.0
56.9
56.9
60.8
61.2
61.8
65.3
62.0
Age of children
Children in the
lower quartile
for height
40
32
Children in the
upper quartile
for height
29
37
Children in the
upper quartile
for height
34
34
38
31
Childs personal
hygiene
parents or of the child, bore any systematic relation to the childs height during the school
years.
The only strong positive association found
between background circumstances and the
childs height was between height of child and
2 14
Children in the
lower quartile
for height
Children in the
upper quartile
for height
36
32
In
34
34
Children in the
upper 4
quartile
for height
34
31
34
31
Children in the
lower quartile
for height
28
41
41
26
A\
0.5-
8
9 1
AGE IN YEARS
0 1 1
.-
x- - - - - x
3.0-
2.5-
Chi 2 = 5.78
p<0.02.
2.0-
5I 1.5-
:
2
1.0-
1.0-
0L
Children in the
upper quartile
for height
1.5-
fz
g 2.0a
x- - - - - x
,
,
personal hygiene
of fathers
.-.
2.5-
Children in the
lower quartile
for height
Children in the
upper quartile
for height
38
28
31
35
ow
I;
ib
ii
AGE IN YEARS
Cruvioto et al.
height of the child, it was decided to examine
whether the fathers height made any significant contribution to the childs intersensory performance. To carry out this analysis fathers
were grouped in accordance with whether their
heights were above or below the median for the
sample, and the relation of the fathers group
position to the intersensory performance of the
children was plotted (Figures 16 and 17). N o
significant association was found between the
height of the fathers and the level of intersensory competence achieved by the child.
Comment
Two facts clearly emerge from the results presented. For these rural children a difference in
height is accompanied by a difference in intersensory integrative ability. For the upper social
class urban sample, differences in height are
not associated with differences in intersensory
integrative competence. Therefore, height as
such cannot be considered as a determinant of
intersensory integrative organization unless
such difference in height occurred under circunistances in which the height differential developed from causes which affect intersensory
integrative organization.
Differences in growth in the rural children
are most likely to have derived from a failure to
have received appropriate amounts and kinds
of food (primary malnutrition), or to have been
the product of infectious disease or parasitic
infestations which have secondarily interfered
with the individuals nutritional state either directly by increasing tissue catabolism without a
compensatory increase in food intake (12), or
indirectly through anorexia or social custom, in
accordance with which, greatly reduced food
consumption is deemed therapeutic in preschool children during illness and convalescence (13). When group differences in height
do not derive from such a background set of
risks, short stature is most likely to reflect familial differences in stature, and therefore, to be
unassociated with disturbances in growth deriving from malnutrition. Such a view is supported
by the strong relation of paternal height to
height of the offspring in the upper social class
urban group and the presence of a weak association in the rural population.
Whether the inadequacy in intersensory integrative performance in the rural children rep-
2 15
2 16
SCHEME I
SCHEME II
SOCIAL CONDITIONS
SOCIAL CONDITIONS
PRIMARY OR
SECONDARY
MALNUTRITION
POOR
INTERSENSORY
DEVELOPMENT
LOW
STATURE
dR
SECONDARY
PRIMARY
MALNUTRITION
LOW
STATURE
P6OR
INTERSENSORY
DEVELOPMENT
of Comprehension of
Task
217
Crauioto et al.
+
+
+
+
+
+
+
111
IV
V
VI
VII
VIII
+
+
Table 26. Ratio of observed errors to maximum possible number of errors expressed
as a percent (rural children).
Intersensory modalities
Visualhaptic
Haptickinesthetic
Visualkinesthetic
Visual
haptic
Haptic
kinesthetic
Age
Lower
Upper
Lower
Upper
Lower
Upper
Lower
Upper
Lower
Upper
Lower
Upper
6
7
8
9
10
11
31.2
22.1
13.1
12.1
11.9
7.5
23.1
14.7
7.5
3.9
5.8
1.1
7.0
1.8
2.5
0
0
0
6.2
1.8
0.6
0
0
0
29.7
15.8
11.9
9.1
10.2
7.5
19.2
10.2
6.9
3.4
8.0
2.2
41.9
8.7
21.7
3.6
3.2
2.3
27.8
12.5
4.3
3.1
2.7
2.4
35.4
4.6
4.1
3.0
2.3
2.1
11.4
4.2
3.4
1.4
1.3
1.9
34.4
5.3
10.7
2.2
2.6
2.3
21.6
10.8
2.5
0.5
1.9
0.8
2 18
Table 27. Average frequency of occurrence of various patterns of response in two groups of rural
children at extremes of height (judgment of identical forms).
Age in years
Pattern
V-K V-H H-K
+++
+-+
111 + + IV - + +
V +-VI - + VII - - +
I
I1
VIII
a
---
QL
QUb
QL
~~
4.22
0.0
0.55
1.00
0.0
0.77
0.0
0.44
5.5
0.3
0.6
0.9
0.0
0.70
0.0
0.0
6.29
0.05
0.58
0.82
0.05
0.64
0.0
0.0
10
11
QU
QL
QU
QL
QU
QL
QU
QL
QU
6.14
0.07
0.28
0.57
0.0
0.35
0.0
0.0
6.3
0.0
0.5
0.9
0.1
0.3
0.0
7.1
0.0
0.3
0.3
0.0
0.3
0.0
6.31
0.0
0.43
0.50
0.0
0.25
0.0
0.0
7.63
0.0
0.27
0.09
0.0
0.0
0.0
6.81
0.0
0.36
0.54
0.0
0.27
0.0
0.0
7.21
0.0
0.42
0.21
0.0
0.14
0.0
7.3
0.0
0.4
0.0
0.0
0.0
0.3
0.0
7.81
0.0
0.18
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
2 19
Cravioto et al.
Table 28. Average frequency of occurrence of various patterns of response in two groups of rural
children at extremes of height (judgment of nonidentical forms).
Ace in vears
Pattern
V-K V-H H-K
QLa
I +++
I1 + - +
111 + + IV - + +
V +-VI - + VII - VIII - - -
QUb
QL
QU
51.35 43.64
0.82 0.64
1.17 1.64
2.82 2.14
0.58 0.42
1.76 2.42
0.23 0.35
1.00 0.71
22.33 34.6
1.33 2.0
2.4
1.11
7.22
5.6
2.22
6.0
1.88
6.3
2.55
6.0
3.1
10.77
QL
QU
44.1
1.0
0.7
4.6
0.3
4.4
51.8
0.9
0.5
1.3
0.3
0.8
0.2
0.3
0.2
0.5
10
QL
QU
QL
49.62 53.18
0.62 0.54
0.43 0.0
0.75 1.63
0.12 0.09
0.31 0.09
0.37 0.18
0.43 0.0
11
QU
52.27 53.07
0.81 0.35
0.81 0.78
0.72 1.07
0.27 0.14
0.63 0.35
0.18 0.14
0.09 0.07
different ages.
3 6
PATTERN I
0.30
0.10
6 7
AGE IN YEARS
9 10 11 12
Figure 20. Patterns of judgment for all intersensory pairings made when judging
nonidentical forms at different ages.
v)
Go
a
W
t-
IU
U
W
m
I
[
I
///--
PATTERN I
i\
d
W
>
.\
AGE IN YEARS
PATTERN IV
_
~ VI.
PATTERN
QL
QU
55.4 52.90
0.6
0.63
0.7
0.36
1.0
0.81
0.2
0.09
0.1
0.09
0.1
0.18
0.2
0.27
220
ing nerve fibers it has been estimated that protein substance increases more than 2000 times
as the apolar neuroblast matures into the young
anterior horn cell. Perhaps an easier way to
grasp the magnitude of this process may be
simply to recall that at the time of birth the
human brain is gaining weight at the rate of 1 to
2 mg per minute.
Changes in structure of the central nervous
system due to feeding grossly inadequate diets
to animals have been documented by Lowry
(22) and Platt (23). McCance, et al. (24-28) have
shown gross alterations in the content of water
and several electrolytes (28) in the brain substance, and Flexner (29) and associates have
advanced evidence that interference with protein synthesis in the brain produces loss of disorders in mice (29). Ambrosius, et al. (30) have
reported that severely malnourished children
show a distortion of the normal relation between brain weight and total body weight. They
have interpreted their findings as an indication
of arrested growth of the central nervous system. It may well be that so-called critical periodicity in behavior represents the responsiveness of the nervous system when it is at a
given stage of biochemical organization. If this
is the case nutritional inadequacy may interfere
with both staging and timing of development of
brain and behavior.
However, independently of the mechanism
through which intersensory organization is interfered with, there can be little doubt as to the
fact that there is evidence of delayed neurointegrative development in those children at nutritional risk who have grown poorly. It is, therefore, important to consider the potential
significance of this type of developmental lag in
so primary a process as intersensory organization for more complex psychological functioning. In this connection it is of interest to consider two significant features of learningconditioned reflex formation and the acquisition of academic skills.
In most conditioning situations what is being
demanded is the integration between two stimuli, each of which belongs to a different sensory
modality. Thus, in classical salivary conditioning or in the conditioning of a leg withdrawal, a
gustatory or a tactile stimulus is in effect being
linked to an auditory or visual one. The process
of conditioning, when effective, therefore involves the establishment of intimate equiv-
Cravioto et al.
References
( 1 ) Maier, N. R. F. and T. C. Schneirla. Principles of
Animal Behavior. New York, McGraw Hill, 1935.
(2) Birch, H. G. Comparative psychology. In Areas
ofPsychobgy, F. Marcuse, ed. New York, Harper, 1954.
(3) Voronin, L. G. and V. I. Guselnikov. On the
phylogenesis of internal mechanisms of the analytic
and synthetic activity of the brain. Pavlov J Higher
Neru Activity 13:193, 1963.
(4) Sherrington, C. S. M a n o n His Nature. London,
Cambridge University Press, 1951, pp. 287-289.
(5) Birch, H. G . and M. E. Bitterman. Reinforcement and learning: The process of sensory integration. Psychol Rev 56:292 1949.
(6)
. Sensory
221
222
ward. Experimental protein-caloric deficiency, Chapter 2 1. In Mammalian Protein Metabolism. New York,
Academic Press, Inc., 1964.
(24) McCance, R. A. Food, growth, and time. Lancet
2:621, 1962.
(25) Widdowson, E. M. and R. A. McCance. The
effect of finite periods of undernutrition at different
ages on the composition and subsequent development of the rat. Proc R SOCS B 158:329, 1963.
(26) Pratt, C. W. M. and R. A. McCance. Severe
undernutrition in growing and adult animals. 6.
Changes in the long bones during rehabilitation in
the cockerels. B r J Nutr 15:121, 1961.
(27) McCance, R. A. Some effects of undernutrition. J Pediatr 65: 1008, 1964.
(28) Widdowson, E. M., J. W. T. Dickerson, and R.
A. McCance. Severe undernutrition in growing and
adult animals. 4: The impact of severe undernutrition
on the chemical composition of the soft tissue of the
pig. B r J Nutr 14:457, 1960.
Selection of Deaths
In the 12 cities the registration of deaths, at
least of adults, is virtually complete. The international form of medical certificate of cause of
death is always used, except in Guatemala City,
where a proportion of deaths are certified by
police officers, who use a special form for the
purpose. In addition to the medical certificate,
the forms used in all cities give the name, age,
sex, place and date of death, and place of residence; and in all cities but one the civil status of
the deceased is shown. The forms used in several cities include other items as well, such as
occupation, date and place of birth, whether an
autopsy had been performed, duration of residence at last address, and nationality or race.
The cooperation of the registration authorities in each city having been obtained, at
regular intervals throughout the study period
the complete set of certificates of all deaths ocSource: Ruth Rice Puffer and G. Wynne Griffith, Patterns of
Urban MortaliQ-Report of the Inter-American Investigation of
MortulZ@ Washington, D.C., Pan American Health Organization, Scientific Publication No. 151, 1967. Chapters I1 and
xv.
I
223
224
Field Inquiries
During the house visit, the public health
nurse or social worker not only confirmed the
age and residence status of the deceased but
also obtained information concerning the place
of birth, the residence history, the occupations
engaged in by the deceased, and the dates and
places where medical attention had been given,
both recently and in the past. The last items of
information served as points of departure for
the investigating physicians, who assembled the
available clinical information relevant to establishing the cause of death. Hospital records
were traced and surgical and autopsy findings
were recorded, together with the results of ancillary studies such as laboratory investigations,
radiology, and electrocardiography. Physicians
having knowledge of the patient were interviewed. When the home visit proved unproductive for the nurse or social worker, the interviewing physician had an alternative point of
departure, namely, the physician who had
signed the death certificate. If death had occurred in a hospital, the record department was
often the best place to start making inquiries.
Not surprisingly, the degree of success attending these efforts varied from city to city
according to the nature and variety of the obstacles that had to be overcome. The home visitor
had many problems. Repeated visits were often
needed before a relative could be found at
home. The general rule was that, if necessary,
three attempts would be made to interview a
likely informant. In Caracas, although deaths to
be investigated were selected within one week of
registration, the first home visit was frequently
unsuccessful because the family had already
moved to another address in accordance with
an old custom. Sometimes a neighbor knew the
new address, but often no trace of a relative
could be found despite persistent inquiries. In
the poorer quarters of some cities houses are
numbered somewhat haphazardly, and it was
therefore not uncommon to find houses with
225
226
1. Those in which the cause of death appeared to be a condition involving the heart and
circulatory system (Categories 022, 023,
330-334, 400-468, 754, of the International
Classification of Diseases).
2. Those in which more than one cause of
death appeared to be involved in the fatal sequence of events.
3. Those in which the cause of death seemed
to be an ill-defined condition (Categories
780-795).
There was no difficulty in deciding which
questionnaires should be seen by the medical
referees, because in case of doubt the invariable
rule was to refer them.
Questionnaires relating to deaths the cause of
which was not in doubt were not reviewed by the
medical referees (unless death was due to a cardiovascular condition). These questionnaires related to deaths from violent causes, from malignant neoplasms, from maternal causes, and
from specific infections such as tuberculosis, as
well as deaths from other conditions presenting
a well-defined clinical picture. The assignment
of the underlying cause of death was made in
the central office by the reviewing physicians,
and the category numbers were added by expert coders in accordance with the rules of the
International Classification of Diseases. To ensure
uniformity, these coders also assigned the category numbers of the causes of death as stated
on the original death certificates accompanying
all questionnaires.
For deaths from each city there were therefore at least two assignments of cause of death
coded in terms of the full four-digit categories
of the International Classzfication of Diseases. One
was the cause of death according to the death
certificate, coded by the expert coders. The
other was the cause of death on final assignment, selected by the medical referees or by
the reviewing physician without reference to
them. In addition, for all cities except two there
was available the original coding of the death
certificate made in the city concerned. On questionnaires from Cali and Guatemala City this
latter coding was missing, since the inclusion of
the local coding would have caused unacceptable delay. For all cities the coding of the death
certificates in the central office could be compared with the coding of causes given on final
assignment in the light of the information assembled by the field workers, while for 10 cities
another useful comparison was possible, i.e.,
227
228
Standard Population
Where the structure of populations by age
and sex differs widely, as in these 12 cities,
comparisons of mortality are best made in separate age-sex groups. This procedure was followed whenever the numbers of deaths were
229
ro
70.7
42.1
42.4
1.6
-30.8
2.6
1.9
Bristol
14.5
Sio Paulo
0.4
Santiago
-17.8
San Francisco
35.8
37.9
55.2
26.6
46.7
-8.0
4.1
Mexico City
Ribeirso PrCto
18.0
-2.0
47.0
- 19.9
La Plata
Lima
54.3
- 12.2
Guatemala City
Caracas
Cali
37.1
-4.8
12 cities
Bogota
36.3
Total
changesa
City
Excess
or
deficit
Infective and
parasitic
diseases
14.8
9.5
-4.9
-2.5
-0.5
0.1
-5.1
-8.0
-6.4
-2.1
-3.2
15.1
8.5
3.4
0.5
10.9
86.5
-3.7
37.0
15.0
39.5
23.8
53.5
24.4
5.6
- 0.6
4.8
-11.7
29.2
6.2
- 1.7
9.5
15.9
7.9
32.9
17.9
- 3.6
4.7
33.6
33.8
- 8.3
13.7
2.1
13.8
18.0
28.1
39.7
- 7.4
10.7
15.6
26.9
- 0.8
3.6
30.5
Total
changesa
Excess
or
deficit
40.4
73.9
64.0
-3.9
0.0
-10.4
72.0
1.7
-0.4
67.9
75.3
2.6
-7.4
57.5
65.9
-0.9
103.4
63.8
163.7
3.4
-7.4
67.5
88.9
6.3
-2.5
68.0
50.0
- 50.0
44.3
11.9
- 17.0
-3.6
42.0
- 33.9
- 6.5
48.8
19.9
26.0
23.0
20.0
-9.1
-31.6
53.8
32.5
38.7
7.8
-0.6
32.3
- 28.2
29.1
-0.6
33.0
- 33.0
-7.6
-2.2
1.3
12.9
17.0
12.4
9.9
7.4
7.3
1.0
- 32.6
-2.6
29.9
40.1
- 26.9
12.0
37.1
-0.6
14.5
-7.7
5.3
11.7
-2.3
-0.1
Total
changesa
Excess
or
deficit
Accidents and
violence
- 20.4
26.2
159.9
49.8
57.7
32.1
39.5
- 37.3
39.1
33.8
- 25.7
42.9
Total
changesa
Total
changesa
Maternal
causes
Excess
or
deficit
Diseases of
digestive
system
Excess
Total
or
changesa deficit
Respiratory
diseases
27.0
0.7
or
deficit
Excess
Cardiovascular
diseases
5.9
31.4
9.4
-6.7
- 14.4
9.8
11.9
Total
changesa
-3.3
-4.3
deficit
or
Excess
Malignant
neoplasms
89.8
101.3
94.8
- 18.9
-3.3
1.6
97.9
87.1
10.7
96.0
107.7
23.3
7.0
97.4
21.5
72.2
-3.4
123.5
78.9
98.1
96.1
-28.0
10.3
-9.2
7.7
0.7
Total
changess
Other causes
Excess
or
deficit
Table 1. Number of deaths with a change in classification between original and final assignments per 100 final assignments for
eight groups of diseases at ages 15-74 years, in each city, 1962-1964.
23 1
232
dents 20-64 years of age was selected and detailed information was obtained for deaths occurring in Baltimore City. Of 478 death
certificates which gave arteriosclerotic heart disease as the underlying cause of death, 26 were
Table 2. Percentage agreement of reviewers with assignment of cause from information on death
certificates in a study of a sample of deaths from Pennsylvania, 1956, and in the Inter-American
Investigation of Mortality, 1962-1964.
Inter-American Investigation of Mortality
Cause of death
Total
Tuberculosis (001-019)
Malignant neoplasms
(140-205)
Stomach and large intestine (151, 153)
Rectum (154)
Biliary passages and liver
( 155)
Pancreas (157)
Larynx (161)
Trachea, bronchus, and
lung (162, 163)
Breast (170)
Uterus (171-174)
Male genital organs
(177- 179)
Urinary organs (180, 181)
Leukemia and aleukemia
(204)
Lymphomas (200-203,
205)
Diabetes mellitus (260)
Major cardiovascular-renal
diseases
Vascular lesions affecting
central nervous system
(330-334)
Rheumatic fever and
rheumatic heart disease
(400-416)
Arteriosclerotic heart disease (420)
Other heart diseases except hypertension (42 1,
422, 430-434)
Hypertensive diseases
(440-447)
Influenza and pneumonia
(480-493)
a
Pennsylvania
samplea
Most
Total
probable
death diagnosis
certifi- (percentcates
age)
San Francisco
10 Latin American
cities
Bristol
Death
certificates
Percentage
agreement
Death
certificates
Percentage
agreement
Death
certificates
Percentage
agreement
2122
78.7
3865
71.4
4262
78.4
34 521
64.0
27
85.2
39
92.3
21
81.9
1961
89.3
443
85.8
809
89.1
1043
88.4
6572
77.4
120
26
83.3
88.5
125
33
95.8
91.5
174
62
88.4
87.1
1556
132
84.9
88.2
13
16
4
92.3
68.8
100.0
46
7
11
100.0
90.7
97.1
2
36
8
50.0
91.7
75.0
86
214
99
64.3
79.9
87.9
54
32
28
87.0
93.8
96.4
157
72
37
97.2
95.8
98.6
267
123
34
97.3
98.4
91.2
649
429
810
89.0
99.0
95.5
21
16
90.5
100.0
26
37
88.5
93.8
32
46
84.4
91.1
125
211
82.6
89.5
13
100.0
28
96.4
25
88.0
229
88.7
15
61
100.0
86.9
46
52
89.1
62.7
33
36
90.0
71.7
237
999
89.9
74.5
1406
78.2
1748b
71.3
2072b
76.9
10 155b
64.0
254
85.0
279
77.1
507
90.0
2774
86.0
41
82.9
46
87.4
88
85.5
664
80.0
692
78.6
1138
82.7
957
89.6
2740
81.3
167
66.5
72
32.4
193
33.4
1829
43.2
150
76.0
213c
63.5
32P
51.3
2146<
50.5
31
64.5
97
41.0
165
35.7
1139
49.3
233
Table 3. Increase in deaths assigned to selected causes as result of autopsy findings in England and
Wales. and on review of additional information in Bristol and San Francisco, 1962-1964.
~
Bristol
Deaths assigned
Cause of death
San Francisco
Deaths assigned
Deaths assigned
From
Before After Percentage death
Final Percentage From death Final Percentage
autopsy autopsy
change certificate review change
certificate review change
Tuberculosis (001-019)
Syphilis (020-027)
Malignant neoplasms
58
31
95
31
63.8
-
21
8
32
10
52.4
25.0
39
9
50
13
28.2
44.4
2283
253
2378
234
4.2
- 7.5
1043
115
1118
129
7.2
12.2
809
52
853
60
5.4
15.4
288
264
- 8.3
121
127
5.0
106
113
6.6
28
69
146.4
100.0
11
14
27.3
450
534
18.7
267
294
10.1
157
171
8.9
136
153
94
166
147
69
- 3.9
22.1
- 26.6
33
25
36
42
23
46
27.3
- 8.0
27.8
46
28
52
46
30
62
7.1
19.2
1096
886
- 19.2
507
526
3.7
279
312
11.8
179
236
31.8
88
97
10.2
46
59
28.3
(420)
995
1065
7.0
957
986
3.0
1138
1001
- 12.0
(440-443)
103
160
55.3
140
125
- 10.7
87
76
455
504
10.8
187
120
- 35.8
126
144
14.3
504
217
399
324
- 20.8
49.3
165
294
71
369
- 57.0
25.5
97
15
54
30
- 44.3
100.0
210
257
22.4
40
41
2.5
35
40
14.3
(140-205)
Stomach (151)
Large intestine and
12.6
234
235
(5).The estimates in this article are also preliminary. Although they relate to only a single year,
they provide previously unpublished information on whites and other persons and on sex
differences. These estimates are considered
more accurate than earlier computations of
such values.
Results
The two related indices described in this paper are based upon a life table model. They are
(a) the expectation of life free of disability and
( b ) the expectation of disability. Either of these
measures can be calculated using various definitions of disability, and values of each index
based on two alternative definitions of disability
will be presented and compared.
The techniques employ a relatively simple
modification of the conventional life table
model to compute the expected duration of certain defined conditions of interest among the
living population. Somewhat similar methods
have been employed to compute expected values for conditions such as labor force participation and school enrollment (6, 7). In those applications, current mortality rates, summarized
in the life table values, were combined with survey-based rates for events among the living population to produce potentially valuable measures not otherwise obtainable. Calculation of a
summary measure of health status in a somewhat similar fashion was once suggested in a
paper by Sanders, but the more elaborate
health measures which his proposal required
have not yet been developed (8).
The expectations of life and of disability presented in this paper are hypothetical values derived from a period life table. They are the
values which would occur if a birth cohort of
fixed size experienced, age for age throughout
life, the recent age-specific mortality and disability rates used in these life table calculations.
Since the age-specific rates may change considerably over the lifespan of any real birth cohort,
expectations based on a period life table may
236
not reflect accurately the life experience of infants born in any specific period. Hence, these
measures are intended primarily as an index for
comparing the mortality-morbidity experience
of different population groups and should not
be construed as projections or forecasts. Methods of computing these measures are described
in a subsequent section.
The modified life tables provide values of the
expectation of life which omit time lost to disability. In one version of these tables, disability
was defined broadly as institutional confinement for health care, prolonged incapacitation
that does not include institutional care, and
short-term episodes of restriction on a persons
usual activities (Table 1). An alternative version
eliminates only the lifetime duration of periods
of bed disability (Table 2). Bed disability in this
paper includes any periods spent in hospitals or
other institutions for health care and also days
of noninstitutional illness involving confinement to bed for more than half the daylight
hours. Whichever definition is used, eliminaTable 1. Expectations of life and approximate
expectations of life free of disability and of
disability, for whites and other persons by sex, at
birth and at age 65, civilian resident population,
United States, mid-1960s.
Expectation
Life
(1965 US.
abridged
life tables)
Life
free of
disability
Disability
Years at birth
All persons
Male
Female
White
Male
Female
All other persons
Male
Female
70.2
66.8
73.7
71.0
67.6
74.7
64.1
61.1
67.4
64.9
61.6
68.4
65.8
62.5
69.4
58.2
55.1
61.4
5.3
5.2
5.3
5.2
5.1
5.3
5.9
6.0
6.0
Years at age 65
All persons
Male
Female
White
Male
Female
All other persons
Male
Female
14.6
12.9
16.2
14.6
12.9
16.3
14.0
12.6
15.5
11.3
9.4
13.1
11.5
9.5
13.3
9.3
7.5
11.2
3.3
3.5
3.1
3.1
3.4
3.0
4.7
5.1
4.3
Life
(1965 U.S.
Life
abridged free of bed
life tables) disability Disability
Years at birth
All persons
Male
Female
White
Male
Female
All other persons
Male
Female
70.2
66.8
73.7
71.0
67.6
74.7
64.1
61.1
67.4
All persons
Male
Female
White
Male
Female
All other persons
Male
Female
Years at age 65
14.6
13.5
1.1
12.9
12.1
.8
16.2
14.9
1.3
14.6
13.6
1.o
12.9
12.1
.8
16.3
15.0
1.3
1.o
14.0
13.0
12.6
11.7
.9
15.5
14.3
1.2
68.2
65.2
71.4
69.1
66.1
72.4
62.3
59.5
65.2
2.0
1.6
2.3
1.9
1.5
2.3
1.8
1.6
2.2
Sullivan
80
70
79 A
69.4
fifi 1
62.5
60
55.1
50
v)
U
? 40
30
20
10
WHITE
FEMALES
WHITE
MALES
ALL OTHER
FEMALES
237
ALL OTHER
MALES
238
4
U
0
6
Y 3
51
1.5
WHITE
FEMALES
WHITE
MALES
0Expectation of disability
ALL OTHER
FEMALES
ALL OTHER
MALES
239
Sullivan
ences between whites and others in the expectation of bed disability, either at birth or at age 65,
are only negligible. Consequently differences in
the expectation of life free of bed disability are
almost equal to differences in the conventional
expectation of life.
There are noticeable differences in the expectation of disability between whites and other
persons, however. At birth, whites have an expectation of disability of about one-half year
shorter than the remainder of the population.
This differential widens to more than 1Yz years
at age 65. At birth, the difference in expectation
of disability (0.7 years) is a relatively small component of the difference in expectation of life
free of disability (7.6 years). At age 65, however,
the difference in expectation of disability (1.6
years) is the dominant component of the difference in expectation of life free of disability (2.2
years) and results in a difference between whites
and other persons much larger than that shown
by conventional life expectancies.
Thus, not only is the expectation of life
shorter for persons who are not white, but the
expected duration of disability of all types is
greater-both absolutely and also proportionately in relation to length of life. When bed
disability alone is considered, however, no substantial differentials between whites and other
persons in expected duration are observed. Examination of the disability data underlying
these measures confirms that the differences
between whites and other persons in expectation of disability are primarily a result of differences in the prevalence of long-term disability
that is neither bed disability nor institutional
Methods
Data required to compute these indices are a
current abridged life table and a set of current
age-specific rates for disability days applicable
to the population group of interest.
Table 3. Computation of the approximate expectation of life free of disability &t)for white males,
civilian resident population, United States, mid-1960s
Age group
Under 15
15-44
45-64
65-74
75 and over
1%
L,
Disability
weighting
factor
I,b
0
15
45
65
75
100 000
96 767
90639
65 901
39665
1 457 411
2 830 657
1623962
532 960
318095
0.967
,964
,915
,802
,633
Exact
initial
age
-&
Txt
zxt
1 409 316
2 728 753
1485925
427434
201 354
6 252 782
4 843 466
2 114713
628 788
201 354
62.5
50.1
23.3
9.5
5.1
Reference (10).
UI
For each age group, the weighting factor is I,= 1
where w, is the total number of disability days per person
per year in the designated age group.
The dagger symbol, t, is used in this paper to distinguish weighted life table values from the corresponding values denoted
by conventional notation.
a
240
1. Days of health care in long-term institutions. These are obtained as previously described.
2. Days of bed disability among the civilian
noninstitutional population. These include all
reported days of care in general service shortstay hospitals, whether or not the person is actually confined to bed on the day in question.
Days of disability outside of hospitals and institutions are counted as days of bed disability
only when the person is reported as confined to
bed for more than half the daylight hours.
Sullivan
ability measures were derived from data collected by the survey programs of the National
Center for Health Statistics, supplemented by
certain data from the U.S. Census of 1960.
Since it was not possible to derive each component for the same year, the illustrative data
shown are labeled mid- 1960s to indicate they
are synthetic estimates based on data for several
different years. Nevertheless, it is felt that the
results are reasonably accurate approximations
applicable to the United States in mid-decade.
Definitions of terms used in this paper, and a
complete account of procedures used to estimate the total volume of disability are soon to
be published ( 4 ) .
Discussion
The objective of this mortality-morbidity
index is to measure change over time in the
health status of the nation as a whole. Reasons
for using mortality and disability rates as components of a single index which may serve this
purpose have been discussed elsewhere (I). If
such a combination of rates for death and disability is desired, the techniques described have
certain advantages.
Use of the life table model provides one solution to the problem of the relative weights to
assign deaths and episodes of disability when
attempting to measure both phenomena by a
single index. The model is a familiar conceptual
tool, conventionally used in weighting diverse
schedules of mortality rates for comparison
with each other, Its elaboration to permit comparison of disability rates as well may meet with
fewer objections than would any other arbitrary
equation of a death to some specific duration of
disability.
The data in this paper only permit comparisons of whites and other persons and of
sex, but the observed differences in expectation
of disability are sizable enough to indicate the
measure is sensitive to differentials in disability
experience of a magnitude likely to occur in
present-day populations. Since even fractional
differences in this measure represent differences of months in the cumulative average experience of disability in the groups compared,
the measure would seem to be a meaningful
reflection of the impact of disease and injury
among the living.
241
Although observed differences in the expectation of disability are fairly large in absolute
terms, they make a relatively small contribution
to differences cited here in expectation of life
free of disability at birth. This fact may make it
appear that mortality dominates comparisons
based on the index. Where large differences in
conventional expectations of life exist, as they
do between the sexes and between whites and
other persons in this country, they obviously
will be a principal component of differences in
expectations of life free of disability. But the
disability rates also enter into computation of
expectation of life free of disability and may
widen or narrow the gap between populations
for which the index is computed. In the event
two populations approached equality in conventional expectation of life, they might still differ
substantially in expectation of disability and this
difference would be reflected in their expectations of life free of disability.
In this sense, the disability component
emerges as a more prominent component in
comparisons as mortality differences diminish.
This tendency to enhance the role of disability
in comparisons of health status between populations with similar mortality levels seems reasonably analogous to the relative weight frequently assigned to risks of death and disability
when one is assessing the importance of a
health problem (1).
A principal, and probably enduring, disadvantage of these indices is the heavy demands
they make upon available data. Both conventional life tables and data on disability must be
available for a population in order to calculate
these measures. At the national level the required data can be obtained only for the total
population and a few major population categories. Lack of data is likely to preclude application of the indices to states or local areas for the
foreseeable future.
In addition to their data requirements, there
are other problematical aspects of these measures. Problems in interpreting sex differences in
disability resulting from the criteria of disability
used have already been mentioned. Further
studies are also needed to determine the sensitivity of the disability measures to changes over
time. These problems and other limitations on
the indices have been discussed more extensively (4).
242
References
( I ) Sullivan, D.F. Conceptual problems in developing an index of health. PHS Publication No. 1000,
Series 2, No. 17. U.S. Government Printing Office,
Washington, D.C., May 1966.
(2) Chiang, C.L. An index of health: mathematical
models. PHS Publication No. 1000, Series 2, No. 5.
US.Government Printing Office, Washington, D.C.,
May 1965.
(3) Moriyama, I.M. Problems in the measurement
of health status. I n Indicators of Social Change, E.B.
Sheldon and W.E. Moore, editors. Russell Sage Foundation, New York, 1968, ch. 1 1 , pp. 573-600.
( 4 ) Sullivan, D.F. Disability data components for an
index of health. PHS Publication No. 1000, Series 2,
No. 42. U.S. Government Printing Office, Washington, D.C. I n press.
(5) U.S. Department of Health, Education, and
The indicator of Potential Years of Life Lost between ages 1 and 70 (PYLL) is
proposed with the primary objective of ranking major causes of premature mortality. This proposal is based on a review of existing mortality indicators and indices
and of the history of the concept of potential years of life lost. The method of
calculation along with the corresponding rate and the age-adjusted rate are discussed and presented with applications to Canadian data and interpretation. Several
methodological aspects are discussed, particularly the comparison with more sophisticated approaches based on life tables which do not appear to alter the ranking
of major causes of premature death. This indicator fits well into the category of
Social Indicators and can help health planners define priorities for the prevention
of premature deaths. Epidemiological studies could also make use of this indicator
of premature mortality. The simplicity of calculation and ease of comprehension
should facilitate its use.
disability have to be considered if one requires
an overview of major health problems. HowThis paper addresses itself to the basic question ever, in many jurisdictions routine data on these
of How to compare the importance of major other dimensions are lacking.3
causes of death? The indicator of Potential
The objectives of this paper are to review
Years of Life Lost between ages 1 and 70 is a existing or proposed mortality indicators and
concept which has been discussed by a number indices, demonstrate how to calculate Potential
of authors over the past 30 years using various Years of Life Lost between ages 1 and 704 with
methods of calculation. This indicator is an at- associated rates, discuss alternative methtempt to incorporate both theoretical and prac- odological approaches and underlying assumptical aspects already discussed by others in the tions, and finally present some Canadian data
field of mortality analysis, with the primary ob- and interpretation.
jective of helping health planners define priorities, particularly with respect to prevention.
Obviously untimely or premature death conREVIEW OF MORTALITY INDICATORS AND
INDICES
stitutes only one aspect of all health problems,
and other dimensions such as morbidity and
Most mortality indicators or indices have
been proposed in order to compare mortality in
Source: Internatzonal Journal of Epidemiology 6(2):143.151,
different geographic areas, occupational groups
1977. 0 Oxford University Press, 1977.
or for different years. The concept of potential
I This work was first presented at the Canadian Public
Health Association, June 1974; a more detailed version is
years of life lost, however, originated with the
available in French and in English.
* Long Range Health Planning Branch, Department of primary objective of comparing the relative importance of different causes of death for a parNational Health and Welfare, Jeanne Mance Building. Ottawa KIA OK9, Canada.
ticular population.
In Canada, such data will be regularly collected by the
An excellent review of mortality indicators
Canada Health Survey which has been planned since early
was made in 1943 by Woolsey (2) who described
1974 ( I ) and will provide the first annual data for the year
1978.
direct and indirect methods for standardizing
The indicator of Potential Years of Life Lost between
death rates, with the resulting indices being
ages 1 and 70 is referred to subsequently as the PYLL indicacalled Standardized Mortality Ratios. The life
tor.
INTRODUCTION
243
244
The method of calculating PYLL for a particular cause or group of causes consists of a summation of the number of deaths at each age
(between 1 and 70) multiplied by the remaining
years of life up to age 70.
Let di=number of deaths between ages i and
i+ 1
%=remaining years to live until age 70
when death occurs between ages i and
i + l = 70-(i+.5)
assuming uniform distribution of deaths
within age groups, where i represents age
at last birthday
Therefore, PYLL is given by:
69
pyLL
C aidi
i=I
69
(70-i-.5)
di (2.1)
a=I
245
Table 1. Calculation of potential years of life lost between ages 1 and 70 (PYLL), rate and age-adjusted
rate, Ontario, ischemic heart disease, males, 1974.
Standardized PYLL and rate
iaccordinn to Canada age-structure)
Remaining
years
Number of
deaths
Ages
PY LL
Correcting factor
Age-adjusted PYLL
0
0
57.5
52.5
142.5
382.5
975
2892.5
5445
11 002.5
13 510
12 687.5
10 642.5
4075
61 865.0
1.08
1.02
1.03
1.05
1.03
.97
.96
.96
.95
.94
.95
1.oo
1.oo
1.01
0
0
59.2
55.1
146.8
37 1.O
936.0
2776.8
5172.8
10 342.4
12 834.5
12 687.5
10 642.5
41 15.8
60 140.4
d'
0
0
1
1
3
9
26
89
198
489
772
1015
1419
1630
5652
67.0
62.5
57.5
52.5
47.5
42.5
37.5
32.5
27.5
22.5
17.5
12.5
7.5
2.5
1-4
5-9
10-14
15-19
20-24
25-29
30-34
35-39
40-44
45-49
50-54
55-59
60-64
65-69
Total (1-70)
Rate of PY LL
6 1 865
3 791 600
=
60 140.4
3 791 600
x 1000
N = 3 791 600 represents the Ontario male population between ages 1 and 70 in 1974.
N, = 10 531 000 represents the Canadian male population between ages 1 and 70 in 1974.
Note: the corresponding rate of PYLL for Canada 1974 is 15.1 per 1000.
deaths by five-year age groups (with the exception of ages one to four) and the corresponding
values of ai. This shorter method for the calculation of PYLL is preferred as a good approximation of formula 2.1. When both calculations
were made for several causes of death they differed by less than two percent for any one cause.
It should be noted that PYLL is additive for
different causes of death.
If A and B are two causes of death, then:
If one wants to compare PYLL for two populations of different sizes then a rate must be
used. A simple way is to express the rate of
PYLL per 1000 population (Table I), as is usually done for death rates. Hence,
69
rate of PYLL
2 aidi x
i=l
1000
..
(2.3)
When mortality for specific causes is compared between two or more different populations it is usually done by using standardized
death rates to eliminate the effect of different
age structures among the different populations.
The proposed age-adjusted rate of PYLL corresponds to the direct method of standardization (2, 14, 22) and is given by:
246
69
a@)
i=l
(k)
x 1000
where
Pi = number of persons of age i in the actual
population
P, = number of persons of age i in the reference population
N, = number of persons between ages 1 and
70 in the reference population
Formula 2.3 can be written:
The life table method is a very useful demographic technique which allows the calculation
of life expectancies and has been successfully
applied by numerous bioiitatisticians in a variety
of research related to mortality or survival (25,
26). Public health authorities have often used
life expectancy at birth as an overall indicator of
progress. Similar to the PYLL indicator, life expectancy gives high weights to infant or early
deaths, but the weights are not so readily noticeable. Three life table methods will be discussed
and compared to the PYLL method which does
not require life table values.
247
Table 2. PYLL' and corrected PYLL due to mortality before age 70 from other causes (competing risks)
Canada, males, 1967, ten groups of causes.
ICDb
I
I1
111 & IV
VI
VII
VIII
IX
X
XVII
v, XII,
XIII, XIV,
xv, XIV, XI
Infectious diseases
Neoplasms
Allergic, endocrine, & blood
Nervous system
Circulatory system
Respiratory system
Digestive system
Genito-urinary system
Accidents
All other causes
Total
PY LLa
(1-70)
Corrected PYLL
(1-70)
9%
10 200
124 290
14 310
41 517
192 554.5
33 139.5
29 962.5
8784
30 8147
1.29
15.69
1.81
5.24
24.30
4.18
3.78
1.11
38.89
9235.1
11 4916.0
12 980.9
37 928.7
182 229.6
30 259.5
27 108.2
7953.7
283 457.5
1.26
15.67
1.77
5.17
24.85
4.13
3.70
1.08
38.66
29 429.5
3.71
27 125.7
3.70
100
733 194.9
100
Causes
792 334
The calculations are for the year 1967, for sensitivity of the method to the choice of cut-off
which Canadian abridged life tables assuming age. Table 3 indicates the distribution, in pertotal elimination of each of the 10 groups of centage, of Potential Years of Life Lost before
causes of death (27) were available. The cor- ages 65, 70, and 75 for five major causes.
In choosing a cut-off age above 70 one should
rected PYLL was calculated as follows, for all
years of life lost up to age 70. A person having consider that the determination of the underlyjust reached age 40 and dying from cancer was ing cause of death becomes difficult, particunot considered to have lost 30 years of life, but a larly for very old people. On the other hand,
lesser amount corresponding to the average age 65, proposed by several authors, appears
number of years he could live up to age 70, as rather young, since a great proportion of peoderived from the life table assuming elimination ple are still productive at this age.
Recognizing the dependence of potential
of cancer. As can be seen in Table 2, the percentage distribution of PYLL did not change by years of life lost on the choice of cut-off age,
more than 3 percent for any one cause as a
result of this correction, which shows the negli- Table 3. Potential years of life lost between age 1
and age 65, 70, 75: percentages for five major
gible effect of the correction for competing
causes, Canada, 1974, Males.
risks. However, if one is interested in more than
simply the ranking of major causes of preBefore age Before age Before age
mature death (such as cost-benefit analysis or Cause
70
75
65"
expected disability-free life), the calculation of
All causes
100.0
100.0
100.0
life table modifications might be preferable.
Motor vehicle
accidents
Ischemic heart
disease
Other accidents
Suicide
Malignant
neoplasms of
digestive organs
25.2
18.2
17.2
12.2
18.2
8.6
15.1
12.6
6.4
21.0
13.0
6.3
2.5
4.3
4.3
248
one should be flexible in using the corresponding rankings to define priorities but also keep in
mind that all these rankings are quite different
from the ranking based on number of deaths
(Table 4).
Concerning the exclusion of infant deaths
from the calculation of PYLL, two considerations should be kept in mind. Firstly, most cases
of infant mortality are due to causes specific to
this early period of life and often have a different etiology than death later on. Secondly, each
infant death would account for almost 70 years
lost, giving a weight double that of a death
between ages 30 and 40. This appears to be an
overestimation of the value accepted by society
for such a loss in light of the fact that a very
early death is often replaced (28) by another
birth. Therefore, from the point of view of social criteria, infant mortality is less disrupting
than mortality of older children and adults.
APPLICATION TO CANADIAN DATA AND
INTERPRETATION
Table 4. PYLL and deaths between ages I and 70 by major causesb, Canada, 1974.
Causeb
All Causes
PYLL (1-70)
Total
%
1312 675
100
Deaths (1-70)
%
Total
100
77 440
239 283.5
198 327.5
165 264.5
84 195
8.2
5.1
2.6
6.4
8.0
26.2
6.5
3.7
5864
19 205
4795
2716
56 667
50 264
4.3
3.8
7.1
4.7
5186
3425
48 079.5
45 418
34 954
30 919.5
3.7
3.5
2.7
2.4
6.1
5.5
3.0
2.9
4444
4068
2204
2108
29 634
2.3
1.5
1119
249
Figure 1. Distribution of Potential Years of Life Lost between age%1 and 70 (PYLL)by major causes, by a x ,
Canada, 1974.
I00 000
50 000
YEARS .OST
50 000
100 000
150 000
YEARS LOST
I
I
other accidents
I
1
suicide
FEMALES
respiratory diseases
cancer of lung and other respiratory organs
0
1
cerebrovascular disease
cirrhosis of liver
breast cancer
After an extensive literature review, consideration of the basic issues and assumptions and
experimentation with different techniques, it
has been concluded that the indicator of Potential Years of Life Lost between ages 1 and 70
(PYLL) could be used in a variety of ways. The
simplicity of calculation and ease of comprehension should facilitate its use but some of
250
Table 5. Rates of PYLLb and percentage distribution for four major causes, Canada, and Provinces,c 1974.
All Causes
56
Provinces
Canada
British Columbia
Alberta
Saskatchewan
Manitoba
Ontario
Quebec
New Brunswick
Nova Scotia
Newfoundland
62.8
70.8
71.1
63.3
64.6
64.2
63.5
62.9
61.9
57.7
57.2
64.9
65.3
67.7
68.6
67.1
67.1
53.8
57.6
100
100
100
100
100
100
100
100
100
100
Motor Vehicle
Accidents
%
11.5
14.6
15.1
12.8
12.7
13.6
13.8
8.1
8.2
8.5
8.7
13.0
12.7
16.1
15.4
13.9
13.7
11.2
10.4
18.2
20.7
20.2
21.2
12.9
14.8
20.0
23.7
20.7
20.8
Ischemic Heart
Disease
%
Other Accidents
9.5
7.8
7.6
7.0
7.5
6.7
6.1
9.1
8.7
10.2
10.0
10.4
10.6
9.6
10.2
11.0
10.9
7.7
9.6
7.9
12.4
12.7
9.8
9.8
9.9
10.0
10.2
10.3
6.2
6.2
6.5
6.5
9.5
9.5
8.9
9.0
8.5
7.8
15.1
11.1
11.1
10.4
14.4
17.7
16.0
14.2
16.4
14.3
Suicide
12.6
17.6
15.5
15.4
16.2
10.7
10.1
14.0
13.3
4.0
5.2
5.2
5.5
5.6
3.9
4.2
4.5
4.6
4.1
4.1
3.4
3.3
3.1
3.1
3.6
3.8
.3d
15.8
.3
6.4
7.3
8.7
6.0
7.2
7.2
5.2
4.6
5.4
.5
a Under each crude rate per 1000 population between ages 1 and 70 is indicated in italics the age-adjusted rate, according to
the Canadian population of 1974.
Potential Years of Life Lost between ages 1 and 70.
Only provinces with more than 500 000 population (Prince Edward Island excluded).
Significance questionable (less than 50 deaths).
Romeder and M c W h i n n i e
12) Kitagawa, E.M. Theoretical considerations in
the selection of a mortality index, and some empirical
comparisons. Human Biology 38:No. 3, 293, 1966.
(13) Elveback, L.R. Discussion of Indexes of mortality and tests of their statistical significance.Human
Biology 38:No. 3, 322-324, 1966.
(14) Hill, G.B. The use of vital statistics and demographic information in the measurement of health
and health care needs. Chapter 2 o f Methods ofHealth
Care Evaluation, Sackett and Baskin (eds.), McMaster
University, 197 1 .
(15) Dempsey, M. Decline in tuberculosis; the
death rate fails to tell the entire story. A m Rev Tuberculosis 86:157, 1947.
(16) Dickinson, F.G. and E.L. Welker. What is the
leading cause of death? Two new measures. Bureau of
Medical Economic Research, American Medical Association, Bulletin 64, Chicago, 1948.
(17) Greville, T. N. E. Comments on Mary Dempseys article on Decline in Tuberculosis: the death
rate fails to tell the entire story. Tuberculosis 87:417,
1948.
(18) Haenszel, W. A standardized rate for mortality
defined in units of lost years of life. A m J Public Health
40:17, 1950.
(19) Doughty, J.H. Mortality in terms of lost years
of life. Can. J Publzc Health 42:134, 1951.
(20) Martin, W.J. Life table mortality as a measure
25 1
At the beginning of this century it was generally held that typhus fever had disappeared
from the United States except for an occasional
case imported from Europe or from Mexico.2
In 1910 Dr. Nathan E. Brill of New York,
(1-3), called attention to a typhuslike disease
occurring endemically in that city. He hesitated
to identify it as typhus because of its generally
milder course and its occurrence under circumstances different from those usually associated
with that disease. He accordingly believed that
he was dealing with a new clinical entity, an
infectious disease of unknown etiology. Cases
of this type have since been known in the
United States as Brills disease.
In 1912 Anderson and Goldberger (4), who
had previously reported on the experimental
transmission of Mexican typhus (tabardillo)to
monkeys, were similarly successful in the inocu-
252
Maxcy
generally assigned to typhus. He pointed out
that the cases occurred sporadically, without
traceable connection with each other, that they
seldom, if ever, gave rise to new cases among
those in contact with the sick person, that no
localized outbreaks occurred, and finally, that
their seasonal distribution differed from that of
typhus. Later, accepting the identity of the virus
with typhus as indicated by the work of Anderson and Goldberger, Brill (10) was led to raise
the question whether some vector other than
the louse might not be concerned in the transmission. The same question is raised by Allan
(11)as a result of his observations upon a series
of cases occurring in Charlotte, N.C.
In 1922, while detailed as acting State epidemiologist to the State Board of Health of
Alabama, the writer had occasion to observe
with Havens (12) a number of cases which were
identified clinically as the endemic form of typhus described by Brill and which gave a positive Weil-Felix reaction. As the same question
with regard to the mode of transmission arose
in these cases, an epidemiological study was undertaken under instructions from the Surgeon
General in cooperation with local health authorities, and has been continued up to the
present. The opportunity for study has been
especially favorable, since this section of the
United States is little subject to immigration either from Europe or from Mexico; and with
cases occurring in the smaller cities and towns
one could exclude more surely the possibility of
constant reintroduction of the virus from exotic
sources and trace association between cases, if it
existed.
EVIDENCE OF PREVALENCE IN
SOUTHEASTERN UNITED STATES
253
254
residence and place of business, occupation, recent travel, date of onset, clinical course, contact
with preceding cases, secondary cases, presence
of lice or other vermin. A majority of the cases
have been seen personally by Doctor Bassett
during the acute illness. When this was not
done the information desired was obtained either by a personal visit to the patient himself
after convalescence or from the physician in
attendance, or a combination of these. The author has accompanied Doctor Bassett on many
of these visits.
IDENTIFICATION AS TYPHUS
It has been tentatively accepted that the disease with which we are dealing in the southeastern United States is typhus, because of:
(1) Its clinical identification with Brills disease ( I 6).
(2) The Weil-Felix reaction.
(3) The work of Anderson and Goldberger
( 4 ) , identifying the virus of Brills disease with
that of Mexican tabardillo.
(4) The successful transmission of the disease to Rhesus monkeys and to guinea pigs
from cases in Savannah and Montgomery by
the author, and the character of the reaction in
these animals. (Unpublished report.) Further
studies of the activity of this virus in experimental animals and its relation to the European
virus are in progress.
255
Maxcy
Table 1. Distribution of cases of endemic typhus in Alabama during four years of observation.
City or town
Birmingham
Montgomery
Mobile
Atmore
Brewton
Red Level
Andalusia
OPP
Troy
Sampson
Hartford
Dothan
Headland
Kinston
Total
Population
1920
1922
1923
178 806
43 464
60 777
1775
2682
385
4023
1556
5696
1646
1561
10 034
1252
163
1924
1925
1
24
17
1
1
5
1
2
6
2
1
63
Total
7
44
21
1
1
1
6
1
6
4
1
6
4
1
104
Confirmed by
Weil-Felix
4
28
12
1
1
1
6
1
2
2
1
1
1
1
62
256
Savannah, Ga.
1923
1924
1925
Montgomery, Ala.
Other cities
and towns
in Alabama
1922
1923
1924
1925
1922
1923
1924
1925
Grand total
Jan.
Feb.
Mar.
1
1
2
2
2
1
1
2
2
2
5
Apr.
1
3
1
1
May June
3
3
1
1
2
2
2
5
2
3
5
1
1
3
3
9
July
1
0
1
1
7
1
9
17
8
7
15
3
3
3
5
8
5
9
29
2
7
10
33
1
2
2
3
6
Aug. Sept.
Oct.
Nov.
6 1 1
4
2
1
8
11 21
1
5
2
4
11
7
8
2
3
3
2
2
1
8
1
1
4
6
30
2
1
6
9
28
4
6
25
Dec.
Total
1
1
3
5
1
1
2
1
5
38
14
41
93
6
6
8
24
44
2
1
6
9
19
5
8
8
39
60
197
Occupation
Date of
onset
28 Waiter in cafeteria
1922
June 5
2
3
4
W
W
W
F
M
F
35 Housewife
50 Proprietor, bottling works
38 Housewife
WeilFelix day
after
onset
Reaction
result
Oct. 8
Nov. 12
Nov. 20
5th
8th
14th
25th
16th
Neg
Pos.
Pos.
Pos.
Pos.
Nov. 25
15th
Pos. 1-1280
1-800
1-640
1-320
1-160
Remarks
Wife of No.
3
Maxcy
257
Table 3. (Continued.)
Case
No. Race Sex Age
~
Occupation
Datc of
onset
WeilFelix day
after
onset
Reaction
result
Remarks
60 Housewife
Dec. 19
7
8
11
W
W
W
M
M
M
1923
Jan. 7
Feb. 15
Sept. 22
12
45 Housewife
13
14
W
W
M
M
50
51
W
W
52
53
8th
Neg. 1-80
Dec. 1
8th
6th
5th
14th
14th
Nov. 30
Dec. 10
9th
19th
Neg.
Neg. 1-80
Neg. 1-20
POS.1-2560
Pos. 1-1280 Wife of No.
77.
Pos. 1-320
Pos. 1-1280
F
M
34 Housewife
15 Clerk, wholesale grocery
1924
Apr. 28
Aug. 6
W
W
M
M
Aug. 9
Aug. 24
10th
5th
14th
10th
POS. 1-160
Neg.
POS.1-5000
POS.1-160
57
58
W
W
M
F
59
60
W
W
F
F
11 Schoolgirl
36 Housewife
Dec. 6
Dec. 7
6th
5th
17th
13th
8th
POS. 1-160
Neg.
Neg.
POS.1-320
Neg.
61
62
63
64
W
W
W
W
M
M
M
F
52
38
22
24
1925
Mar. 20
Feb. 28
Apr. 19
May 15
11th
7th
11th
10th
Pos. 1-320
Neg. 1-20
Pos. 1-320
Neg.
65
66
67
68
69
70
71
72
W
W
W
W
W
W
W
W
F
M
M
M
M
M
June 13
July 7
July 28
do
Aug. 8
Aug. 12
Aug. 13
Aug. 17
8th
10th
9th
11th
Neg. 1-80
Neg.
POS.1-320
Pos.
M
M
20
25
37
45
46
56
17
43
73
74
75
76
77
W
W
W
W
W
F
M
F
F
M
45
32
17
30
63
Aug.
Sept.
Sept.
Sept.
Sept.
12th
7th
14th
8th
14th
7th
10th
14th
POS. 1-160
Neg.
POS.1-320
Pos.
Pos. 1-640
Pos. 1-160
Pos. 1-640
Pos. 1-640
78
11 Schoolgirl
Sept. 20
79
80
81
82
W
W
W
Col.
M
M
M
M
22
35
24
58
Sept. 30
Sept. 27
Oct. 19
Oct. 30
4th
10th
12th
Neg.
POS.1-160
Pos. 1-1280
83
Nov. 4
POS.1-640
Neg.
POS.1-1280
Pos. 1-640
84
9th
3d
15th
9th
Dec. 2
5th
Pos. 1-100
Bank clerk
Taxi driver
Produce salesman
Employee of restaurant
23
12
17
9
15
Contact with
No. 52?.
Guinea pigs.
Positive.
Typical
clinically
Husband of
No. 12
258
Case
No. Race Sex Age
W
W
W
W
W
6
7
9
10
11
12
13
14
17
8
16
17
18
19
20
21
22
23
24
25
26
27
28
W
W
W
W
W
W
W
W
Col.
W
W
W
W
W
W
W
W
W
W
W
W
W
W
F
M
M
M
28
29
29
30
M
M
M
M
M
F
M
M
M
F
M
F
F
M
F
F
F
F
19
31
60
31
52
40
49
28
21
21
51
37
30
21
35
28
32
25
30
17
35
14
40
Col.
W
W
W
M
M
F
M
38
38
38
10
Painter
Mechanic
Housewife
Child
31
32
32
Nov. 2
33
34
37
40
W
W
W
W
F
F
44 Housewife
26 do
43 do
52 Convict guard
Nov.
Now
Nov.
Dec.
51 Housewife
1924
Mar. 7
W
W
W
W
W
W
F
M
F
M
M
M
40
44
14
35
50
48
May 1
May 12
June 11
July 9
Sept. 1
Sept. 6
F
M
M
hf
do
do
Railroad engineer
Schoolgirl
Shipping agent
Foreman, railroad yards
Turpentine broker
WeilFelix day
after
onset
Reaction
result
Remarks
1923
July 13
July 14
July 16
July 19
July 21
Employee of restaurant
Dealer in hay and grain
Housewife
Employee of restaurant
Watchman, Salvation Army
Industrial Home
Housewife
Salesman, meat packer
Salesman, ship chandler
Salesman, tobacco warehouse
Salesman, wholesale candy
Tailor
Grocer
Butcher, store H
Dairy worker
Unemployed
Housewife
Clerk, wholesale warehouse
Clerk, grocery store
Employee, restaurant
Housewife
Fire department employee
Housewife
do
Clerk, grocery store
Unemployed
Housewife, boarding house
Schoolgirl
Clerk, grocery store
23
45
30
52
38
Occupation
Date of
onset
July 27
July 28
Aug. 3
Aug. 4
Aug. 6
Aug. 14
Aug. 16
Aug. 27
Aug. 29
Aug. 25
Sept. 1
Sept. 2
Sept. 17
Sept. 23
Sept. 24
Sept. 25
Oct. I
Oct. 2
do
Oct. 7
Oct. 9
Oct. 15
Oct. 21
NOV.Oct. 24
Oct. 25
do
7th
40th
1lth
8th
12th
6th
POS. 1-160
Neg.
do
do
POS. 1-160
Neg.
8th
17th
Neg.
14th
10th
do
POS. 1-320
Neg.
Neg. 1-80
do
do
8
11
14
10
Neg.
15th
12th
do
9th
25th
Pos. 1-320
do
do
Neg.
POS. 1-320
14th
Neg.
8th
Neg. 1-80
14th
Neg.
Sonof
No. 29.
Daughter
of No. 29.
Husband
of No. 29.
Guinea
pigs.
Maxcv
Table 4. (Continued.)
Case
No. Race Sex Age
Occupation
Date of
onset
WeilFelix day
after
onset
Sept. 20
do
Oct. 21
do
Nov. 25
Nov. 18
Dec. 18
15th
8
9
11
13
14
15
16
W
W
W
W
W
W
W
F
F
M
M
F
M
F
57
19
62
19
62
28
50
17
18
19
W
W
W
F
M
F
18 Clerk, office
1925
Jan. 30
Mar. 28
Apr. 4
20
21
22
38b
23
24
25
26
27
28
29
W
W
W
W
W
W
W
W
W
W
W
M
M
M
M
F
M
F
M
M
M
M
34
56
36
65
35
47
52
23
29
17
60
Apr. 30
Apr. 18
May 16
Jan. May 7
May 30
July 9
July 15
July 5
June 25
July 22
30
31
32
33
34
35
36
38
W
W
W
W
W
W
W
W
F
M
M
M
F
M
M
F
30
30
25
35
73
28
17
33
39
40
41
42
43
44
45
46
W
W
W
W
W
W
W
W
M
M
F
M
F
M
M
M
54
41
19
16
60
10
27
22
47
48
50
51
52
53
54
55
56
57
58
W
W
W
W
W
F
F
F
M
M
M
F
M
M
M
M
20
56
7
50
30
57
30
52
14
26
40
Unemployed
Clerk, grocery
Employee, X dairy
Employee, restaurant
Housewife
Clerk, feed store S
Employee, X diary
Telephone operator, living over
grocery store A
Clerk, feed store S
Carpenter
Schoolgirl
Schoolboy
Housewife
Schoolboy
Clerk, wholesale grocery
Clerk, wholesale tobacco
warehouse
Clerk, grocery store
Housewife
School child
Molder, living on water front
Manager, ice plant
Engineer
Clerk, physicians office
Merchandise broker
School child
Pipe fitter, railroad shop
Painter
W
W
W
W
W
Housewife
Clerk, department store
Farmer
Barber
Housewife
Employee, filling station
Housewife, living over store
1925
July 28
do
July 9
Aug. 11
Aug. 15
July 27
Aug. 12
Aug. 16
Aug. 7
Aug. 26
do
Sept. 17
Oct. 1
do
Oct. 2
Oct. 5
Oct. 1
Oct. 11
Oct. 19
Oct. 16
Nov. 4
Nov. 9
Nov. 7
Nov. 22
Dec. 13
Dec. 14
Dec. 15
Reaction
result
POS. 1-60
14th
POS.1-160
10th
17th
6th
Neg. 1-80
Pos. 1-1280
Neg.
do
POS.1-320
16th
10th
Pos. 1-640
Pos. 1-100
22d
10th
Pos. 1-320
Neg. 1-40
14th
12th
15th
8th
7th
1Ith
15th
17th
8th
20th
10th
5th
10th
9th
8th
8th
POS. 1-160
do
Neg.
POS. 1-1280
POS. 1-320
Neg.
Pos. 1-10@
doa
doa
POS. 1-320
do
Neg.
POS. 1-160a
POS. 1-320
Pos. 1-160
Pos. 1-320
Pos. l-160a
Pos. 1-640
do
Pos. 1-1280
Pos. 1-160
Remarks
259
260
Table 5. Number of cases and case rate of endemic typhus according to broad occupational groups in
Montgomery, Ala., 1922- 1925.
Total persons
in group
Group
Male
kernale
16 428
13 242
3186
215
24
41 14
2608
3048
402
650
1102
1079
19 498
7020
11 878
26
1
768
131
530
10
57 1
4915
668
Number of
cases in eroua
Male Female
29
29
0
0
0
1
3
18
1
3
3
0
14
3
11
0
0
0
0
3
0
0
0
0
1.77
2.20
0
0
0
,250
1.15
5.90
2.49
4.62
2.73
0
0.72
.39
.93
0
0
0
0
5.67
0
0
0
0
Table 6. Number of cases and case rate of endemic typhus according to broad occupational groups in
Savannah, Ga., 1923-1925.
Group
Total persons
in group
Male
Female
33 676
28 986
4690
273
13
10 816
6573
4810
940
977
1800
2784
35 463
12 880
22 583
24
0
1753
245
878
9
864
7710
1397
Number of
cases in group
Male Female
57
52
5
6
0
10
5
23
2
0
6
0
34
7
27
0
0
0
1
4
0
0
0
2
1.69
1.79
1.07
21.98
0
.92
.76
4.78
2.13
0
3.33
0
0.96
.54
1.19
0
0
0
4.08
4.56
0
0
0
1.43
Maxcy
Population,
United
States
census
1920
Ward
Number of
cases
4
4
5636
9405
4147
7035
5044
4075
8122
1
2
3
4
5
6
7
8
10
4
4
5
1922-1924.
v)
15
Q
0
6a
-1
10
I-
LL
a
W
MAR.
Case rate
per 1000
population
0.71
.43
1.98
1.42
0.74
.98
.62
This division of the city is peculiarly unfavorable for the purposes in mind, inasmuch
as the wards are arranged radially in such manner that all except one (ward 7) include portions
of the central part of the city. Even though this
be true, the tabulation indicates a slight excess
of cases in wards 3 and 4, which include a large
portion of the older residential section bordering upon the business district.
. . . the Savannah cases have in like manner
been shown according to their places of residence. The distribution appears to be rather
general, except perhaps for the newer residential portions and the outlying districts, where
the incidence is apparently light. Population
figures by wards for this city are not available in
the United States census, and it is therefore not
Figure 1. Seasonal distributionof endemic typhus (Brills disease) based on 197 cases
in Alabama and Savannah, Georgia, 1922-1925, compared with typhus in Romania,
261
JUNE
SEPT.
DEC.
262
in restaurants.
Similarly, in Savannah 23 of the 52 males
(44 percent) were in trade, although only 17
percent of the total number of occupied males
are so engaged. The rates in manufacturing
and mechanical industries and in transportation are comparatively low. In agriculture, forestry, and animal husbandry the cases consisted of four employed by dairies and two
retired farmers; in domestic and persona1 service four employed in restaurants, one barber,
and one hotel keeper.
Among employed females the distribution is
much the same in both cities, though the
groups are small. In both instances the highest
incidence is found in trade, the rate being
approximately the same as for males in this
group alone.
Using a different basis of classification, and
the occupations as given in Tables 3 and 4, it is
notable that in Montgomery 10 of the 32 employed persons (31 percent) who had typhus
were engaged in handling foods, groceries,
meat, produce, feed, flour, or were employed in
feed stores and restaurants. In Savannah 20 of
59 employed persons (34 percent) having the
disease were so engaged. The apparent excess
of cases among food handlers is strikingly similar in the two cities, as are the rates for both
males and females in trade.
These analyses of the occupations of persons attacked by endemic typhus suggest very strongly that as
compared with the rest of the population those engaged in trade, and especially those employed in
food depots, groceries, feed stores, and restaurants, are
exposed to a distinctly increased risk of infection.
(i) Social status. The occupational analysis also
brings out the fact that the disease attacked, for
the most part, persons earning a reasonably
good livelihood. There is a notable absence of
cases among unskilled laborers and unemployed males.
From personal observation of the cases and their
surroundings the author and his collaborators are
convinced that the disease did not select the poor and
uncleanly. It occurred among all classes. The
cases so far as they were discovered present a
fair cross section of the social strata of the average American community. This implies that a
great majority of the cases were in persons
cleanly in their homes and in their personal
habits.
There were no cases among the inmates of
MCZXCQ263
jails, prisons, or asylums. There was no particular association of the disease with cheap boarding or rooming houses. The time-honored characteristics of Old World typhus were entirely
lacking in this respect.
(j} Contact between cases. One of the items of
information on the case history form used in
the Alabama series was, History of Contact
with Antecedent Case. In only one instance
among the 44 cases in Montgomery was the
patient or the physician in attendance or the
investigator able to state that there was definite
close association within three weeks prior to
onset with a case of the same disease or a suspected case. The one exception was case No. 4,
who came down eight days after her husband.
Of the 60 cases occurring in other parts of
Alabama for whom a case history form was
filed, in no instance was the patient or his physician aware of contact of the type described with
a preceding case.
The same statement holds true for the Savannah cases with the following exceptions:
Case 32 came down about seven days after his
wife and two children had become ill with the
same disease.
Case 27, onset July 5; case 35, onset July 27;
and case 39, onset August 7, were employed in a
large wholesale grocery store. They were thus in
casual contact at their place of business.
Case 28, onset June 25; case 32, onset July 9;
and case 36 (fatal), onset August 12, worked on
the same dairy farm and were in contact in their
work. It will be noted that these cases occurred
about the same time as those in the wholesale
grocery store noted above. The dairy purchased
feed from this store during the period involved,
but personal contact of the men on the dairy
farm with the men in the store could not be
demonstrated. There were no known cases of
typhus among the 100 or more patrons of the
dairy.
It is thus seen that known contact with a preceding case is a very rare finding. It must be
admitted, of course, that close contacts may
have existed but were undiscovered, particularly
in those cases in which dependence was placed
entirely upon information supplied by the attending physician and his patient. On the other
hand, it seems quite unlikely that any considerable number of actually traceable contacts with
sick persons or convalescents should have been
overlooked.
Moreover, there is evidence from another angle that the disease observed in this study was
not readily communicable from person to person. For each case that occurred there were a
number of persons in intimate contact with the
patient, including other members of the family,
physicians, nurses, and visitors. Notwithstanding the absence of prophylactic measures, infections among these known intimate contacts were
rare.
Among the 197 cases on which this report is
based there were only two instances, noted
above, in which more than one case occurred in
a family in such sequence as to suggest the
possibility that the earlier case might have infected the later one.
Eighteen of the 93 Savannah cases and six of
the 44 Montgomery cases were hospitalized. No
effort was made to delouse the patient upon
admission to the hospital; no precautions whatever were taken with regard to lice. Not a single
case has occurred among nurses, attendants,
physicians, or fellow patients. One physician
had the disease in Montgomery, but he stated
positively that he had not attended a case of
known or suspected typhus for at least one
month before the onset of his illness.
Brill (10) states that in over 500 cases of endemic typhus observed by him in New York
City there have been only two instances in which
more than one member of a family has been
infected with the disease at the same time or
nearly the same time. Many of the New York
cases have been hospitalized, from 15 to 30
being reported in that city each year since 1912,
but no contact cases among patients, nurses, or
doctors have been reported.
Allan (11) was unable to trace any contact
from case to case at Charlotte, N.C. In this
connection mention should also be made of the
numerous other cases reported in the literature
and to the Surgeon General which have been
sporadic and without secondary spread.
By way of contrast attention is called to Boyds
report (6) of a small outbreak of Mexican typhus (tabardillo)in Iowa. During 1915-1918 a
considerable epidemic of tabardillo raged in
Mexico, and as a consequence sporadic outbreaks were originated in American territory by
imported laborers. A Mexican laborer was admitted to the Santa Fe Railroad hospital, Fort
Madison, Iowa. It was later discovered that he
had typhus, and lice were found upon his
264
Maxcy
erable proportion of the cases having occurred
in persons of such habits and living in such an
environment that the harboring of lice is not to
be suspected.
DISCUSSION
265
266
To maintain the disease under these conditions of transmission, therefore, there must be
available a supply of infective lice, renewed at
frequent intervals by the occurrence of cases in
lousy persons, either infected locally or imported. For sustained endemic prevalence, not
tending to decline, the louse infestation of the
population must be sufficient to establish on the
average at least one new human infection for
every one that is terminated by death or recovery. Otherwise the prevalence will decline. To
meet these conditions a certain proportion of
the cases, probably a majority of them, must
occur in persons sufficiently infested with lice to
serve as foci for the infection of others, since the
cases which may occur in uninfested persons
bitten casually by stray lice and living in a clean
environment would not contribute to the further spread of the infection.
As to the communities considered in this
study, it seems doubtful that the louse infestation of their population is sufficient to sustain
an infection subject to these conditions of transmission. Obvious lousiness-heavy infestation
with body lice-is an exceedingly rare condition
268
Maxcy
CONCLUSIONS
1. A disease giving a positive Weil-Felix reaction, and clinically indistinguishable from typhus fever except with regard to its relative
mildness and low fatality rate, is endemic in the
southeastern United States.
2. The epidemiology of this disease appears
to differ significantly from that of Old World
typhus.
3. The epidemiological characteristics afford
no evidence suggesting louse transmission and
are interpreted as being at variance with manto-man transfer by lice, unless it be assumed at
the same time that the disease occurs mostly in
unrecognizable form.
4. It is suggested as an hypothesis which
seems to afford a more probable explanation of
the mode of transmission that a reservoir exists
other than in man, and that this reservoir is in
rodents, probably rats or mice, from which the
disease is occasionally transmitted to man.6
ACKNOWLEDGMENTS
This theory of the source and transmission of the *endemic typhus referred to in this paper does not necessarily
deny the identity of that disease with Old World typhus; for
while it is satisfactorily proven that in its epidemic form typhus is trmsmitted from man to man by the agency of the
louse, there remains the possibility-unsupported by positive
evidence but not yet excluded-that the disease may exist also
in rodents, and that in the intervals between epidemics the
infection may be carried over in this reservoir.
269
270
connaissance du typhus experimental chez les merid& Arch de IInst Pasteur de Tunis 15:267-275,
1926.
(27) Hone, F.S. A series of cases resembling typhus
fever. M e d . l A w t 1:l-13, 1922.
INTRODUCTION
27 1
272
Table 1. Variation in prevalence of dental caries in normal and mottled enamel teeth of three endemic
areas according to McKay.
Locality
Total
number
of per.
Number of manent
children teeth exexamined amined
55a
54a
78a
All teeth
Number Percent
examined carious
879
385
356
797
283
1895
11
9
16
8
22
14
Molar teeth
Number Percent
examined carious
254
101
126
213
99
529
46
42
64
33
81
58
a Age, sex, color, continuity of residence, and constancy of exposure to the mottied enamel-producing waters not recorded in
the report
consecutive monthly water samples were received from each of the cities surveyed, which
permitted the computation of an arithmetic
mean annual fluoride (F) content of the communal water supply. The clinical examinations
in these cities were limited to those children
who were born in the community, had always
resided there, and had continuously used the
common water supply for both drinking and
cooking.
In certain of these cities, in addition to recording the degree of severity of mottled
enamel, each child was examined for other defects of the enamel, such as present caries, past
caries (fillings or extractions), pits and fissures,
hypoplasias, etc. The examinations were made
in a good light with the child seated facing a
window. Mouth mirrors free from blemishes
and new explorers were used. For each child
examined in connection with the caries aspects
of the study, the facts with respect to residence
and continual use of the common water supply
were verified by an interview with the childs
parent or guardian.
The amount of caries recorded may appear
somewhat higher than usual; for, in addition to
definite cavitation, defects in the enamel on caries-susceptible surfaces showing either a discoloration or opacity around the edges and in
which an explorer would cling, were counted as
caries. All examinations were made, however, by
one individual, the writer.
An analysis of these data indicates that a
higher percentage of caries-free children is
found in cities whose water supplies contain
relatively toxic amounts of fluorides than in
Source of data
273
Dean
Table 2. Percentages of caries-free children, 9 years of age, in 6 selected cities classified according to
their continuous use of water of different fluoride (F) concentration.
Domestic water supplya
Fluoride
Total hard(F) content
ness
Number of
children
examined
Locality
Actual community
mottled enamel
index
Pueblo, Colo.
Junction City, Kans.
East Moline, Ill.
Negative
Negative
Borderline
flPm
0.6
0.7
1.5b
PPm
303
277
242
49
30
35
Monmouth, Ill.
Galesburg, Ill.
Colorado Springs, Colo.
Slight
Slight
Slight
1.7
1.8
2.5
288
237
20
29
39
54
Caries-free children
All teeth
Number
Locality
Pueblo, Colo.
Junction City, Kans.
East Moline, Ill.
Monmouth, Ill.
Galesburg, Ill.
Colorado Springs, Colo.
Number
examined
114<
122d
Permanent
Percent
0
6
6
8
13
21
20
24
Deciduous
Number
Percent
18
8
37
26
11
4
1
9
3
33
16
22
22
55
56
41
6
11
21
21
28
40
Number
Percent
27
22
30
60
26
49
13
38
11
31
274
order to study that age group with the maximum number of permanent teeth in the
mouth, the 12 to 14 year group was selected for
study. All children referred to in the tables to
follow are white. The amount of caries was determined by combining the data associated with
the following items: Caries, permanent teeth,
Extraction indicated, permanent teeth, Filled
permanent teeth, and Extracted permanent
teeth. For each of these items, the bulletin gives
the number of carious permanent teeth per 100
children. Adjustment was made for sex, and the
amount of caries for each county was expressed
in terms of the number of carious permanent
teeth per 100 children. In Public Health Bulletin
No. 226, examinations from communities with
a population under 5000 were combined with
examinations from the rural areas and designated balance of county.
The South Dakota counties selected from the
bulletin were classified solely on the basis of the
prevalence of or freedom from endemic mottled enamel as shown by the mottled enamel
study. The cities of Aberdeen, Huron, and
Figure 1. South Dakota, showing distribution of mottled enamel, with shaded prcpo indicating
countim selected from Public Health Bulletin No. 226 for dental carlea analyses.
SIOUX FALLS
susceptible period
ty classification for caries analyses
Counties where endemic mottled enamel is prevalent
Counties where mottled enamel distribution is uneven
Counties where no mottled enamel has ever been reported
Counties with no data or insufficient data on prevalence of caries
Dean
Sioux Falls were classified on the basis of
whether or not the common water supply was
producing mottled enamel. The results of these
computations are contained in Table 3. Similar
computations were also made for four Colorado
cities and eight Wisconsin cities; the results are
shown in Table 4.
275
Observations made on a selected sample of 9year-old children continuously exposed to waThis paper, after reviewing the findings of ters of different fluoride concentrations, with
the earlier workers in the field, submits evi- the history of exposure personally verified in
DISCUSSI 0N
Table 3. Dental caries attack rates in permanent teeth of 12 to 14-year-old white children in selected
South Dakota counties and cities classified according to the prevalence of mottled enamel.
examined
Number of
carious
permanent
teeth per
years)
children
Number of
children
(12-14
County
100
Remarks
332
256
653
203
Faulk
266
149
Marshall
39 1
251
Sanborn
260
103a
1902
201
Total
Mottled enamel general throughout county. Areas ind u d e Hitchcock, Wolsey, Virgil, Yale, Cavour, and
rural districts.
On basis of clinical examinations, old city deep well
water contained fluorides in excess of minimal threshold.
Mottled enamel general throughout county. Areas include Faulkton, Orient, Cresbard, Chelsea, and rural
districts.
Mottled enamel severe in western half of county, including Kidder, Britton, Langford, Newark, Amherst,
and rural districts. No information on eastern half of
county.
Mottled enamel prevalent in county including Artesian
and numerous rural districts.
295
294
Aurora
340
227
Kingsbury
398
330
Day
666
309
Hughes
184
206
McPherson
340
394
Lincoln
536
284
Total
2765
314
276
County
Number of
children
examined
(12-14
years)
Number of
carious
permanent
teeth per
100
children
Remarks
436
398
Campbell
264
368
Denel
Hanson
McCook
Minnehaha:
City, Sioux Falls
212
27 1
344
218
382
407
608
451
Balance of county
Moody
Walworth
Total
584
433
329
348 1
476
498
355
415
a Only caries and extractions indicated of the permanent teeth are listed in Bulletin No. 226 for this county; adjusted for
sex shows 68.6 per 100 children. This figure was raised to 103.1 per 100 children in order to compensate for the unknown
filled and extracted permanent teeth per 100 children. The increase was based on the average ratio that these two missing items
bear to the caries and extraction indicated reported in the 4 counties adjoining Sanhorn County. The percentages that these
items bear to the whole were as follows: Jerauld, 31; Kingsbury, 34; Beadle (less Huron) 34, and Aurora, 35.
Table 4. Dental canes attack rates in permanent teeth of 12 to 14-year-old white children of all Colorado
and Wisconsin cities listed in Public Health Bulletin No. 226.
City
Number of
children
examined
(12 to 14
years)
Number of
carious
permanent
teeth
per 100
children
Fluoride
(F) content of common
water
supply
(PPm)
Reference
Colorado
Colorado Springs
Pueblo
Denver
Fort Collins
203
41 1
637
207
162
1948
342
296
(8)
2.5
0.6
0.5
None
(8)
(12)
(12)
2.3b
0.5
0.35
0.3
0.3
0.3
0.2
0.12
(13)
(13)
(13)
(13)
(13)
113)
(13)
Wisconsin
Green Bay
Sheboygan
Manitowoc
Two Rivers
Milwaukee
West Allis
Baraboo
La Crosse
687
244
66 1
382
2645
160
119
47
215
710
682
646
917
83 1
733
73 1
Extraction indicated for boys Unknown; 4.2 rate for girls used in this adjustment.-Author.
Determination made by Senior Chemist E. Elvove, Division of Chemistry, National Institute of Health. Approximately the
same amount has been reported by DeWitt and Nicho1s.J Am Water Works Assoc 29:980-984, 1937.
a
Note.-For the mineral constituents, other than fluorine, of these Wisconsin waters, see Public Water Supplies of Wisconsin,
Wisconsin State Board of Health, July 1935.
Dean
277
278
The writer desires to express his indebtedness to Senior Statistician Wm. M. Gafafer and
Senior Chemist E. Elvove, National institute of
Health, for many helpful suggestions and criticisms in the preparation of this paper, to the
Wisconsin State Board of Health for supplying
information on the fluoride content of the
water supplies of the seven Wisconsin cities with
high dental caries attack rates, and to Principal
Statistician Selwyn D. Collins, National Institute
of Health, for a review of the paper.
References
( I ) Black, G. V. (in collaboration with McKay, F. S.).
Mottled teeth. An endemic developmental imperfection of the teeth, heretofore unknown in the literature
of dentistry. Dent Cosmos 58:129-156, 1916.
(2) McKay, F. S. (in collaboration with Black, G. V.).
An investigation of mottled teeth. Dent Cosmos
58~477-484,1916.
(3) Masaki, T. Geographic distribution of mottled
teeth in Japan. Shikwa Gakuho 36: October, 1931.
(4) Medical Research Council, Special Report Series No. 97, 11, The Incidence of Dental Disease in Children. His Majestys Stationery Office, London, 1925.
(5) Ainsworth, N. J. Mottled teeth. Brit Dent J
55:233-250 and 274-276, 1933.
( 6 ) Erausquin, R.: Dientes vetados. Rev Odont
(Buenos Aires) 23:296-313, 193.5.
(7) McKay, F. S. The establishment of a definite
relation between enamel that is defective in its structure, as mottled enamel, and the liability to decay, 11.
Dent Cosmos 71:747-755, 1929.
(8) Dean, H. T. and E. Elvove. Studies on the minimal threshold of the dental sign of chronic endemic
fluorosis (mottled enamel). Public Health Rep 50: 17191729, 1935.
(9) Dean, H . T. and E. Elvove. Some epidemiological aspects of chronic endemic dental fluorosis. Am J Public Health 26:567-575, 1936.
(10) Dean, H. T. and E. Elvove. Further studies on
the minimal threshold of chronic endemic dental fluorosis. Public Health Rep 52:1249-1264, 1937.
(11) Dental survey of school children, ages 6 to 14
years, made in 1933-34, in 26 States. Public Health
Bulletin No. 226. U.S. Public Health Service, Washington, D.C. 1936.
(12) Boissevain, C. H. T h e presence of fluorine in
the water supply of Colorado and its relation to the
occurrence of mottled e n a m e l . Colorado Med
30: 142-148, 1933.
(13) Wisconsin State Board of Health, Personal
communication, June 21, 1938.
279
280
A tractate (Niddah) of the Talmud, consisting of 10 chapters, deals with this subject. An English translation of Niddah
is available (2).No translation of Yoreh Deah into a European
language has been made; the relevant portions of it (chapters
183-200) have been summarized by Sorsby (3).
Kennawav
Moslem law
The Zend-Avesta ( 9 )enacts that a woman having any normal or abnormal discharge of blood
must be placed in a separate building as is done
28i
282
Hindu ritual
Dr. Khanolkar (personal communication)
says that the Parsee laws regarding menstruation are almost identical with those practised by
orthodox Hindus and probably both of them
are derived from early Aryan ritual. The Abbe
Duhois ( I l ) , writing about the beginning of the
nineteenth century, quotes from the book
Padrna-purama, reputed to be the work of the
hermit Vasishta, a rule for a three-day period of
isolation for the menstruating woman, followed
by a day of ceremonies and ablutions, including
36 total immersions in a river. During the three
days . . . the mere wish to cohabit with her
husband would be a serious sin. Elsewhere he
says, The mother of the newly-born child lives
entirely apart for a whole month or more, during which time she may touch neither the vessels nor the furniture of the house, nor any
clothes, and still less any person whatsoever.
The time of her seclusion being over, she is
immersed in a bath, or else a great quantity of
water is poured over her head and body.
Women are similarly isolated during the time of
their periodical uncleanness. In all decent
houses there is a sort of small gynaeceum set
apart for them; but amongst the poor, in whose
huts there is no such accommodation, the
women are turned into the street, under a sort
of shed or outhouse, or else they are allowed a
corner of the cowshed. . . . When the time of
uncleanness is passed, all the garments that the
woman has worn are given to the washerwoman. Her clothes are not allowed inside the
house; in fact, no one would dare to look on
them.
Jhaveri (12) says that in all Hindu religious
books minute directions for baths on very many
occasions are given. During the first four days
of the first menstrual period everything which
the girl touches must be washed and . . . persons coming into contact with her have to take a
bath. This author says nothing of subsequent
periods. After childbirth a woman is impure for
10 days, even if given a daily bath by the midwife; . . . no one dares to touch her.; after 40
days she is given a final bath, and is then pure
(cf. the Jewish practice after the birth of a male
child).
Ploss, Bartels and Bartels (I) state, without
giving any original authority, that in Malabar
the first three days are spent in a special room
The reference by Vineberg (13) to the conditions under which the majority of his hospital
patients in New York lived is valuable in this
connection. When one stops to consider that of
the total number of the Jewish women 1995 had
badly lacerated cervices . . . and that they were
living in the worst possible hygienic surround-
Kennaway
283
Number in
series
Married
percent
Married under
age of 20, percent of married
549
848
1509
96.0-97.7
93.3-94.0
81.5-84.4
9.3-10.8
11.3-16.4
16.1-21.7
44.0-49.2
33.0-35.4
11.3-22.9
234
83.3
20.5
21.0
18.6
83.2
83.1
17.2
15.6
284
Genetic Factors
This factor is obviously a possible one, but its
action is difficult to prove or disprove. A similar
problem arises in the case of primary cancer of
the liver in the African negro (19,20). Data are
required from other Semitic peoples most
nearly related to the Jews.6 Weir and Little (21)
consider that the maximum incidence of uterine cancer one quinquennium later in Jewesses
. . . compared with other women, suggests a
genetic basis. Certainly the age at which some
forms of cancer occur is affected by genetic
differences; examples have been given in detail
in another paper (22; xeroderma pigmentosum, familial polyposis intestini, cancer of the
breast, gastro-intestinal tract, and corpus uteri
in certain families). But a higher ratio of corpus
to cervix cancer in Jewesses would have the
same effect (Figure 1).
Circumcision.
Some authors (23,24)consider that no explanation of the lower incidence of cervical cancer
upon Jewesses need be sought beyond the (assumed) difference in the bacterial flora of the
AGE. YEARS
285
Kennaway
Douching
70
60
v)
E
w
4
50-
&
40-
a
w
30-
- - - - Wohlhabende
-Arme
80 -
20
10
,-----
I-----
-----I
20
30
40
50
60
70
80
AGE, YEARS
Table 2. Relation of civil state and age to cancer of cervix and corpus uteri in
various countries (14).
Cancer
Cervix
Of these, unmarried
Mean age, years
Cervix
Corpus
153
2033
25269
Under 30.
5
30667
9
Corpus
389
7986
170 = 2.1%
45.75
52
45
50
104
101
13.4%
53.3
60395
65134
7049
7513
58
25
803
286
r
Rate for social class I = 100
Social
class
35 -
45 -
55 -
65
35-
45-
55-
AGE, YEARS
DISCUSSION
the right-hand side of Figure 3. The social difference becomes less as age advances from
1. Any comparison of the figures for Jews
35-45 to 55-65, which change is compatible with
a decreasing proportion of cervical cancers, and and non-Jews . . . with those from the various
suggests that it is this form which is affected by communities in Bombay . . . is possible only on
the basis of cancers of the uterus, or of the
social factors.
The only authorities quoted in this paper cervix, reckoned as a percentage of all cancers
who give any data for the incidence of cancer of in women. Obviously these data were obtained
the corpus upon Jewish and non-Jewish women under varying conditions, but such as they are,
are Smith (30) and Davidsohn (31).Smith's data they are summarized in Table 3.
2. The Jewish ritual has some unique fea. . . show that cancers of the corpus make u p
practically t h e same p e r c e n t a g e of all tures. More or less complete isolation of mengynecological conditions (7.1 and 7.8) in Jews struating women is practiced by many other
and others, while the percentage of malignant peoples and is laid down in the religious litergynecological tumors made up by cancers of the ature of some of these (Koran, Zend-Avesta,
corpus is much higher (31.1) in Jews than in Dharmasindhu). Some quite primitive peoples
others (10.0), owing to the smaller incidence of ( I ; Aleuts, natives of the Congo, and of Auscancer of the cervix in Jews. Davidsohn's figures tralia) are stated to enforce isolation at the men. . . show much less difference between the per- strual period for as long as 6 to 7 days. But the
centages of specimens examined, made u p by Jewish ritual appears to be the only one which
cancer of the corpus, and of the cervix, in Jew- imposes an exact test for the cessation, and posesses requiring gynecological treatment (0.83 sible recurrence, of the flow, and a post-menand 0.93) than in other women (3.0 and 6.4). strual isolation of 7 days. The 5 + 7 day rule of
This difference is compatible with the idea that the Jews is of peculiar interest in regard to the
cancer of the cervix is the form of uterine can- time of ovulation.
cer which occurs in Jewesses less than in other
The Jewish rule suggests the following calwomen.
culation. Suppose that a Jewish woman is mar-
Kennaway
287
Minimum
Maximum
13.9
3.9
25-35
10.5
Cancer of cervix
Mean
Hindus
Bombay
Moslems
Parsees
Indian Christians
44
20.7
14.8
12.7
288
SUMMARY
Kennawav
289
(11) Dubois, J. A. Hindu Manners, Custom and Ceremonies. (Trans. and ed. H. K. Beauchamp.) Oxford,
1906.
(12) Jhaveri, K. M. J Anthrop Soc Bombay 9:217,
1910.
( I ? ) Vineberg, H. N. Contributions to Medical and
Biolopcal Research dedicated to Dr. William Osler in honour of hw 70th birthday. New York, 1919, Vol. 2, p.
1217.
(14) Lane-Claypon, J. E. Reports on Public Health
and Medical Subjects No. 40. H. M. Stationery Office,
London, 1927.
( 1 5 ) Donaldson, M . Brit Med J 1:291, 1946.
( 1 6 ) Smith, F. R. Am J Obstet Gynec 21:18, 1931.
( I 7) Epstein, I. Kiddzrshin, The Babylonian Talmud.
Seder Nashim. (Ed. I. Epstein, Trans. H. Freedman.)
London, 1936, Vol. 1.
(18) Abramowitz, B. The Law of Israel (Trans. S . D.
Aaronson.) New York, 1900.
(19) Berman, C. S AfrJ med Sci 6:145, 1941.
(20) Kennaway, E. L. Cancer Res 4:571, 1944.
(21) Weir, P. and C. C. Little. J Hered 25:277, 1934.
(22) Kennaway, E. L. and N. M. Kennaway. Yale J
Biol Med 7:139, 1944.
(23) Handley, W. S. Lancet 1:987, 1936.
( 2 4 ) Handley, W. S . Brit Med J 2:841, 1947.
( 2 5 ) Kennaway, E. L. Brit J Cancer 1:355, 1947.
( 2 6 ) Statesmans Yearbook. Fiji. London, 1947.
(27) Brewster, A. B. The Hill Tribes of Fiji. London,
1922.
( 2 8 ) Plaut, A. and A. C. Kohn-Speyer. Sctence
105:391, 1947.
(29) Theilhaber, F. 2. Krebsforsch 8:466, 1910.
(30) Smith, F. R. Amer J Obstet Gynec 41:424, 1941.
(31) Davidsohn, I. Med Leaves 2:19, 1939.
(32) Registrar Generals Decennial Supplement.
England and Wales, 1921. Part 11. H. M. Stationery
Office, London, 1927.
(??) Registrar Generals Decennial Supplement.
England and Wales, 1931. Part Ha. H. M. Stationery
Office, London, 1938.
formed an amputation of the cervix ( 4 ) . Karnaky reports that at Jefferson Davis Hospital,
where 5000 women underwent conization, not
one case of cancer was observed (5).
On the other hand, Schiller and other authors have expressed the opinion that cervicitis,
far from being the cause of the development of
cancer, would, on the contrary, be but secondary to its appearance (6).
Novak declares: There is still considerable
difference of opinion among gynecologists as to
the importance or unimportance of chronic cervical irritative lesions as predisposers to cancer
and Miller has recently published a study in
which this danger is minimized. And he adds,
Cancer can develop in cervices in which no
evidence of previous chronic irritation can be
demonstrated. In my own experience I have, as
a matter of fact, been impressed with the fact
that a considerable proportion of the early cancers which I have seen have been noted in cervices showing no noticeable evidence of previous chronic inflammation or irritation. (7)
In the presence of divergent, even contradictory opinions on the subject, it is not surprising
that too often the treatment of cervicitis as a
preventive measure be advised in listless, almost
academic fashion, when it is perhaps of vital
importance that it be advocated and generalized at all costs.
Two important questions therefore arise. The
first is how to convince doctors, who do not
specialize nor have any particular interest in
gynecology, that they should not tolerate the
presence of cervicitis among their patients. The
second, how to launch a major educational campaign when many theories oppose one another
and so much obscurity on the subject still exists.
Under these circumstances, it is absolutely
necessary that the proof be established that cervicitis is one of the essential factors in the development of this cancer, if so it be. That is the
crucial point, the key to the solution of the
problem.
As previously stated, f o r some time
gynecologists have remarked that they found no
290
Camon
cancer among women who had been treated
and cured of cervicitis. However, reports of observations of cancer developed in women so
treated have or will be published. It would be
difficult to imagine that all these patients were
completely cured or that they never again will
be exposed to the usual causes of cervicitis.
This approach to a vital problem would result
only in endless discussions which would bring
about the great disadvantage of delaying its solution. This is due to the fact that the preventive
effects of such treatment cannot be judged until
every patient has reached the age of at least 65
years, or until death has occurred. The proof
could hardly come from this source. Because
too many years of observation would be necessary before sufficient certainty could be acquired, it was decided to carry out the research
work in another direction.
If, in reality, cervicitis is the basis of the cancer of the cervix, this cancer should not exist
among women whose social state and mode of
life protect them from the usual causes of cervicitis or chronic irritative lesions, namely nuns.
As a gynecologist, I have had, over a long
period of years, the advantage of observing and
studying a numerically important group of
women living in the above-mentioned social
state. This conviction gradually grew that this
variety of cancer did not exist among them, or
that, if it did, it existed only in very exceptional
cases. To make sure that this was not an isolated
fact, extensive research work in a number of
similar groups dispersed over a wide area was
carried out.
Purposely, unmarried women living in the
ordinary outside world were omitted for research purposes, for reasons which it is not necessary to mention.
The investigation was concentrated exclusively on carcinoma of the cervix uteri and of
the corpus uteri, the latter serving as basis of
comparison.
A survey of the medical files of an annual
average of 13 000 women, covering a twentyyear period, was carried out in the archives of
many different convents of nuns. This figure of
13 000 adult women represents, in civilian life,
a city with an unchanging population of approximately 65 000 inhabitants, calculation
based on the last census taken in Canada.
This research, to be sure, covers a still longer
period of time and a much wider range of subjects. Several of the groups concerned, as a mat-
291
In Table 1 are given the results of the research work carried out in the previously mentioned group.
Table 1. Annual average, 13 000 women, period of
time: 20 years.
Carcinoma
of cervix uteri
Carcinoma
of corpus uteri
Histologically confirmed
Diagnosed in the course
of operation
Diagnosed clinically
Total
12
14
292
peat, not a single one was discovered. The inverse ratio of frequency is evident in these statistical data.
The results in the groups set aside are identical to those which have just been given. If they
were omitted, it was because the inquiries could
cover only a period of fifteen years in one case,
or could not be completed in the other.
It seems difficult to ignore the mathematical
strength and power of these figures. To do so, it
would be necessary to assume that the medical
statistics on the subject, widely accepted and
used for at least the past twenty years, were
entirely wrong. This would mean that, if error
there be, such statistics should have maintained
that cancer of the corpus uteri is more frequent
than cancer of the cervix uteri.
This would mean, too, that based on these
results, in a city with an unchanging population
of about 65 000 inhabitants, there would not.
have been a single case of carcinoma of the
cervix uteri during a period of 20 years.
Intrigued by these observations, stupefied,
not to say alarmed, it was decided to study and
attack the problem from another angle. A survey of the archives of pathological laboratories
and radium treatment centers was made, forgetting the archives of the convents which had
heretofore been consulted.
The object in doing this was to verify, over a
vast area in which there existed numerous identical groups but in which sphere no investigation had been made, whether or not the findings would contradict or confirm the results
previously obtained. It was expedient to do this
because the area was much too vast and the
identical groups too numerous to be visited individually.
I therefore made a survey of the pathological
archives of malignant tumors of the uterus in
laboratories of two large hospitals in Montreal
and also in the pathological diagnostic centers
serving the hospitals of approximately the eastern two-thirds of the Province of Quebec, the
Magdalen Islands, the French Islands of St.
Pierre and Miquelon, and a few hospitals in
New Brunswick. The statistical work in this connection covered periods of time varying from
twelve to twenty years.
T h e archives of treatments given at
Montreals Radium Institute for the past twenty
years and those of the Radium Institute of
Quebec, since the latters foundation twelve
years ago, were the object of investigation.
Evidently, a large number of anatomical specimens which do not present any special diagnostic difficulties are not sent to diagnostic centers due to the considerable distance which
exists between the latter and the various regional hospitals concerned. It is also likely that
anatomical specimens are sent to laboratories in
the metropolis not searched in the course of
this investigation.
It is evident, of course, that a few of the
tumors, once discovered in convents archives
by the first method of investigation, are duplicated among those found in the archives of
pathological diagnostic centers where the anatomical specimens from these convents are sent
for diagnosis.
Once again, would this research in hospital
laboratories corroborate or come into conflict
with the facts found in the archives of convents?
The verification of pathological reports resulted in the reclassification of two malignant
tumors of the cervix uteri in the category of
cancer of the corpus uteri. Actually, they were
but the propagation of cancer of the corpus
uteri to the cervix.
RESULTS
After taking these changes into consideration, the results obtained by the second method
of investigation on malignant tumors found in
nuns are shown in Table 2.
Table 2. Results obtained by the second method of
investigation.
Carcinoma
of the corpus uteri
Histologically
confirmed
Total
Carcinoma
of the cervix uteri
19
19
3
3
293
Gagnon
3
1
1
1
5
3
1
1
42
4
2
53
9
2
0
130
0
COMMENT
294
295
296
Tropical Areas
Recent work has indicated that the childhood
development of immunity to a number of common infectious diseases may occur far more
widely in the tropics than may have been suspected. This also applies to poliomyelitis and
Sabins observations in the Far East ( I ) , Gears in
South Africa (7) and Hammons on the island of
Guam (2) are examples of this.
Paul et al.
susceptibility for the Miami children extended
to a later age than was the case with the Havana
children. This would imply that infants in
Havana had more ready access to sources of the
virus than did infants in Miami and, therefore,
had acquired their immunity earlier. A similar
explanation has been brought forward to account for the difference in the average age
which existed between rural and urban cases of
poliomyelitis in this country (9) and elsewhere
( l o ) , with rural cases, as a rule, occurring in
older children than those found in adjacent urban children. This difference in age still exists
in some parts of the United States, notably New
York State (1 1 ) .
297
298
1935
'37
*-.
'39
'41
C a s e Rate
'45
'43
0-
-0
'47
'49
'51
Age of Patients
Figure 2. The record of paralytic poliomyelitis case (attack) rates per 100 O M ) in
Havana Province, Cuba, covering the period 1934-1951. Rather crude estimates of
the median age of the patients are charted for the period 1934-1944. There is little
indication that the median age of patients is Increasing in Havana.
Case Rate
0---
Age of Patients
Paul et al.
299
Table 1. Comparison of the age distribution of Lansing neutralizing antibody with clinical cases of
poliomyelitis within four subtropical communities.
Age
groups
(years)
<1
1-4
5-9
10-14
15-19
20-29
30-39
40-49
50 +
Total
Havana
Serol. tests
Cases
PerCum.
cent
perNo. + b
No.< centd
0
22
16
15
15
14
82
64
75
73
67
64
20
12
111 81
12 88
7 93
4 95
7 99
1 100
1 100
0 100
163
Miami
Serol. tests
Cases
PerCum.
cent
perNO.^ f b
No. cent
11
LO
18
24
35
43
35
31
18
225
36
10
50
71
77
79
83
81
83
11
66
107
58
27
43
11
324
3
24
57
75
83
96
99
100
Cairo
Serol. tests
Cases
PerCum.
cent
perNo. f b
No.e cent
36 15
99 75
37 89
27 93
7 100
15 93
10 100
10 90
7 100
248
141
326
0
0
1
0
0
1
0
469
30.1
99.7
99.7
99.7
100.0
100.0
100.0
100.0
100.0
Texas
Serol. tests
Cases
PerCum.
cent
perNo.
+b
No. cent
10 20
65 56
49 84
31 94
8 100
15 93
10 90
7 100
195
15
41
5
3
0
2
1
0
0
23
84
91
96
96
98
100
100
100
67
a Number of sera collected in 1950 on which Lansing neutralization tests were run. The Texas sera were collected from Latin
Americans living in the Rio Grandc Valley in the spring of 1948.
Percentage of sera with Lansing neutralizing antibodies in each age group.
Number of clinical cases of poliomyelitis reported in 1946-1949, inclusive.
Cumulative percentage of cases.
Number of new cases listed at the Electrotherapy Clinic, Childrens Hospital, Cairo, 1948-1949.
Number of cases among Latin Americans in Rio Grande Valley, Texas, during 1948.
Age of Patients Cumulatively Expressed. In Figure 3 (see also Table I), we have charted the
ages at which clinical poliomyelitis is acquired
and reported in Havana and in Miami in the
form of cumulative percentages of cases in each
age group. There seems to be a correlation between the age of local case acquisition and that
300
Figure 3. The age distribution of Lansing antibody curves compared with the age of the poliomyelitiscases in 8 different populations. The scale at the bottom denoting age has heen arranged to give ample space for recording
findings in those age groups which are usually most important (1-10 years) for the acquisition of this antibody and
for clinical poliomyelitis. An insufficient number of patients in the Alaska series (lower right comer) is responsible
for the omission of the case curve in that chart.
>.
n
.100
v,
80 k
I-
a
Z
z0
z
a
LL
-1
20
F
W
r0
100
MUNICH (ANTIBODIES)
BERLIN (CASES)
oaa
80
a
W
cn
60
LL
2cn
a
LL
5
I-
>
+
z
40 F
a
W
a
20
I
ALASKA
- 100
80
60
40
__----
I
I
*-
20
-.--0
AGE
<I
1-4
5-9
10-
14
1-4
5-9
4
$
Paul et al.
301
Age of Patients Cumulatively Expressed. For purposes of comparison we have again charted the
age at which poliomyelitis is acquired, together
with the age at which the Lansing antibody is
acquired in Cairo by natives. The curve is given
in Figure 3. There is again a correlation here
between the local age acquisition of paralytic
poliomyelitis and the age of local Lansing antibody acquisition.
Comment. Our interpretation of these Lansing
neutralizing antibody findings is that there is
ample Lansing-type poliomyelitis virus present
locally in Egypt to which the high attack rate of
poliomyelitis cases (some of them due to the
Lansing type of virus) in British and United
States troops during World War I1 bore witness
(14). We would further presume that exposure
of all age groups of natives to the Lansing and
other types of poliomyelitis virus is heavy in
Egypt, and that young infants not only share in
this exposure but are the only natives who are
susceptible. As the immunization of these infants proceeds at a much faster rate than is
usually the case in this country, the number of
susceptibles to whom the disease might spread
never becomes large enough to allow a big epidemic to start. This heavy and continual seeding of the population with virus is perhaps re'ponsible for the insignificant reporting Of
as shown in Figure 4, for it is well recognized
302
9
2
40
a
UI
*d
$
ln
30-
'"
5
W
>
20-
Paul et al.
that for the strain or strains causing the epidemic. Other evidence available militates
against the probability of a Lansing type virus
being instrumental in the 1948 epidemic.'*
Arctic Areas
Several strains isolated during the epidemic were inoculated into mice and cotton rats without producing disease in
these rodents. Subsequently, 10 strains from the epidemic
were proven by serological means to be related to the
Brunhilde type.
l2
303
304
Iceland
Serological testsb
Serol. tests
1949-
Age
groups
(years)
No.d
<1
1-4
5-9
10-14
15-19
20-29
30-39
40-49
50 +
0
5
49
45
30
47
31
16
20
Total
243
Percent
Urban and
town
N0.d
0
4
0
10
85
74
87
90
0
8
21
19
16
26
17
12
11
130
Percent
Rural
NO.^
Totalc
Percent
0
0
25
37
60
47
62
69
73
83
100
Clinical cases
193I - I939
incl.
83
34
NO.^
0
14
33
31
21
29
21
17
22
188
Percent
29
49
41
71
69
67
82
86
No.
12
144
110
78
61
74
42
554
Cum.
percent
28
48
62
73
86
94
1945-1947
incl.
No.
8
93
158
141
196
228
139
100
Cum.
percent
10
24
38
56
77
91
100
1058
~
would find in a northern Scandinavian, Scottish, or Irish city. However, it is safe to say that in
1950 (the time of our survey) not more than 20
percent of the individuals questioned had been
out of the country at one time or another and,
epidemiologically speaking, some of the isolated districts from which blood samples were
obtained do represent remote populations.
The island population has ranged in size from
94 000 in 1920 to about 132 750 in 1946, but it
is not, or has not been, of sufficient size to
support for long, measles as an endemic infection, or mumps, or pertussis, so that these childrens diseases have recurred at perhaps longer
intervals than one might have expected in a
larger community with frequent outside contacts. With measles, for instance, the period between epidemics, since 1907, has usually been
from 7 to 9 years, instead of the usual 3 to 6;
with mumps, it has been 6 to 10 years.
Paul et al.
Figure 5. Cape rates for poliomyelitis in Iceland covering the period 1934-1951.
In 1945 the appearance of a disease which can he mistaken for poliomyelitis has
rendered the incidence and median age figures for 1945-1951 sufficiently unreliable
so that they are not recorded.
281
ICELAND
vj
>
..
z
a
I-
10
- 8
g
z
z>
w
U
\d-!*,
- 6
4
I.
- 2
1935
I
37
IIdL
waz
(I)
0
I n
39
\.-.-.Ma
I
41
305
43
I
45
I
47
I
49
51
306
Antibody Patterns. It has been possible to divide our collection of serum samples from Iceland into two groups: (a) urban (including inhabitants of small towns) and (b) rural. The
results of serological determinations have been
listed accordingly in Table 2. Although the data
from rural areas are scanty, no serum with Lansing antibodies was obtained from the 11 rural
inhabitants tested under the age of 10 years. In
charting these results (see Figure 3) rural and
urban groups have been combined. Here the
curve indicating the percentage of those Icelanders with Lansing antibodies rises slowly during
childhood and adolescence, although not at a
much slower rate than is the case with some
other populations included in this survey and
which are much closer to home. The 50 percent positive point on this antibody curve occurs at about 15 years of age, the 70 percent
positive point is not reached until after the age
of 30 years. One might have expected to find a
low level of antibodies during childhood and
adolescence in Iceland in the face of the relatively advanced age of the patients who have
contracted poliomyelitis. As for the correlation
between age of cases and antibody acquisition,
this has been charted in Figure 3 according to
two categories: (a) cases of poliomyelitis occurring d u r i n g the 1930s a n d ( 6 ) those in
1945-1947. During the 1930s, when the diagnosis of poliomyelitis in Iceland was thought to
be no more difficult than in other lands, the
cumulative age curve for cases during that period (see Figure 3) lies above the Lansing antibody curve and shows a correlation comparable
to that seen elsewhere. O n the other hand,
when the cumulative age curve for the cases
from 1945 through 1947 is drawn, it lies below
the antibody curve during the first 20 years of
life. In no other area or population studied by
us have we seen this latter situation. It would be
another argument that during the years of
1945-1947, cases diagnosed as poliomyelitis in
Winston-Salem, N . C.
A previous study (11)has been reported from
this laboratory dealing with the urban population in this North Carolina city of 88 000. In it
Lansing antibody levels from two different socioeconomic groups (upper (B) and lower (A))
have been measured and compared. Details of
the local situation will accordingly not be gone
into here, except to mention that sera were collected from healthy children during the year
(1948) of a poliomyelitis epidemic in that state.
History of Poliomyelitis in North Carolina.16 Poliomyelitis case rates for the past 20 years are
charted in Figure 6, indicating that periodic
epidemics have been coming in shorter intervals since 1943 than before that year. During
the entire 20-year period, the age of the patients has advanced but slightly with a median
age of 3.5 years during the 1930s and nearer 5
years since 1943. This is still far below the median age of the Icelandic and Miami cases and
one would hardly have expected that the Lansing antibody level would be about the same in
these three populations, although this actually
turns out to be the case.
Antibody Patterns. Two hundred and thirty-two
samples were tested for Lansing antibody. Results are listed in Table 3 and charted in Figure
3. In the latter, one sees that the antibody curve
rises slowly and it is not until the age of about 14
years that 50 percent of these children have
acquired Lansing antibodies.
Age of Cases. This, too, is charted in Figure 3.
Parallelism with the Lansing antibody curve is
but approximate in that the antibody levels sag
well below the case curve during the age period
of 5 to 30 years. In other words, Lansing antibody levels do not seem to be a good index of
poliomyelitis here. One might assume from this
l 6 We are indebted to Dr. C. P. Stevick, Director, Division of
Epidemiology, North Carolina State Board of Health, for
supplying us with information relative to poliomyelitis in this
state.
Paul et al.
307
Figure 6. Case rates for poliomyelitis in North Carolina for the period 1931-1950. The median
age of the patients has climbed somewhat during this period.
0-
Case Rate
Munich, Germany
This population was chosen as a representative large mid-European city even though the
history of poliomyelitis in that city has not been
too well documented. Nevertheless, samples of
blood were obtained in 1951 from inmates of
the Kinderklinik of the University of Munich
hospital (service of Prof. Weiskott) and from
other clinics associated with this university hospital.17 The population has been classified as
urban and the living conditions as not primitive
(B). Subjects from whom blood was drawn were,
for the most part, not suffering from infectious
disease.
l 7 We are greatly indebted to Dr. Luitgarde Bungards for
collecting these serum specimens and to Lt. Col. H. Sprinz,
MC, Director of Laboratories at the 98th General Hospital,
European Command, for preparing the serum samples and
holding them prior to shipment to the United States.
0---
Age of Patients
308
Table 3. Age distribution of Lansing antibody levels compared with that of poliomyelitis cases in three
temperate zone communities.
Winston-Salem, N.C.
Age
groups
(years)
<1
Serol. tests
No.
22
Percent
No.
1-4
5-9
10-14
15-19
20-29
30-39
40-49
50
84
63
44
20
36
43
19
90
Total
232
22
25
13
9
6
Munich, Germany
Cases
78
Cum.
percent
4
32
64
80
92
100
No.
19
25
22
15
4
13
11
7
2
118
Percent
20
50
47
60
83
87
Madagascar
Casesb
Serol. tests
No.
1
1
1
1
Serol. tests<
Cum.
percent
693
28
990
297
68
80
42 1
97
74
100
2475
No.
11
9
Percent
t
28
75
20
Madagascar18
A very small number of sera, 20 in all, are
included in the report from Tananarive,
Madagascar. We believe they are important in
that they represent a population living under
what might be called a temperate zone climate
and primitive sanitary conditions but have not
regarded this as a significant survey.
The sera were collected for us under the supervision of Dr. J. E. Smadel. They include
young children and infants only. Antibody levels are listed in Table 3. Two points on a theoretical curve are all that is available from this
small series but they suggest that the antibody
response in this young age group falls in line
with those of Cairo or Havana.
DISCUSSION
The Lansing type neutralizing antibody surveys herein reported give us a picture of the
various rapidities with which infants, children
and young adults acquire these antibodies in
various parts of the world, and the eventual
levels which are attained throughout adult life.
A salient featu-e of the study is that we have not
We are indebted for these sera to Dr. J. E. Smadel of the
Army Medical Graduate Service School, Washington, D. C.;
to Dr. Kenneth Goodner of Jefferson Medical College, Philadelphia, Pa., and to Dr. J. Robic, Director, Pasteur Institute,
Tananarive, Madagascar.
Paul et al.
Figure 7. Lansing antibody levels charted on an age distributional basis for two population
groups (A and B), divided more or less arbitrarily on the basis of: group A, crowding with
primitive sanitary conditions; and group B, rural or urban with moderate crowding but with less
primitive sanitary conditions.
LANSING ANTIBODY LEVELS
309
3 10
100
CASES OF POLIOMYELlTlS
GROUP A - CAIRO, HAVANA, R TEXAS.
GROUP 6 - ICELAND, WINSTON-SALEM, MIAMI, 8 BERLIN
: 80 m
Q
0
LL
LU
60
a
UI
a
w
4
3
r3
40
20
Paul et al.
Kokko, and C. P. Winsor. Age incidence and seasonal
development of neutralizing antibodies to Lansing
poliomyelitis virus. A m ] Hyg 52:323-347, 1950.
(5) Paul, J. R. and J. T. Riordan. Observations on
serological epidemiology. Antibodies to the Lansing
strain of poliomyelitis virus in sera from Alaskan
Eskimos. A m J Hyg 52:202-212, 1950.
(6) Paul, J. R., J. T. Riordan, and L. M. Kraft.
Serological epidemiology: antibody patterns in North
Alaskan Eskimos. ,I Immunol 66:695-713, 1951.
(7) Gear, J. H. S. Polioflyelitis in Southern Africa.
Proc. 4th Internat. Cong. Trop. Med. and Malaria.
Dept. of State, Washington, D. C. 1:555-567, 1948.
See also: Gear, J. H. S., V. Measroch, and J. Bradley.
Poliomyelitis in South Africa. Studies in an urban
native township during a non-epidemic year. S Afr
Med J 25:297-301, 1951.
(8) Paul J. R., F. R. Corria, and D. M. Horstmann.
Analyses from a tropical epidemic of poliomyelitis
which occurred in Florida and Cuba in 1946. A m ]
Trop Med 29:543-554, 1949.
(9) Nicoll, M., Jr. Epidemiologic data in the poliomyelitis epidemic in New York State. A m J Dis Child
14:69-79, 1917.
(10) O h , G. and N. 0. Heinertz. Das Auftreten
der Kinderlahmung in Schweden 1930-39. Ztschr f
Hyg u In~ktionskra.nkhezte7c 125: 153-1 74, 1943. See
also: O h , G. a n d N. 0. Heinertz. T h e epidemiological pattern of poliomyelitis in Sweden,
1925-1944. Acta Pathol SUPPI. 91:139-151, 1951.
(11) Paul, J. R. The peripheral spread of poliomyelitis through rural and urban areas: observations
from a recent epidemic. Yule J Biol Med 19:521-536,
1947.
(12) Melnick, J. L., and N. Ledinko. Social
serology: antibody levels in a normal young population during an epidemic of poliomyelitis. Am J Hyg
541354-582, 1951.
31 1
The problems of diagnosis and of management of coronary heart disease are so numerous
and important that great efforts should be
made toward their solution, but several limitations must be recognized. From all the available
evidence there is no reason to hope that coronary heart disease can ever actually be cured in
the sense that a cancer may be cured by surgical excision, or even that tuberculosis may be
permanently arrested. Moreover, there is some
reason to doubt whether, at present, early diagnosis of coronary heart disease will greatly
change the eventual prognosis.
The basic lesion in almost all cases of coronary heart disease is the deposit of lipid materials in the walls of the coronary arteries. Once
thoroughly established, these atherosclerotic
deposits are nearly, or perhaps totally, irreversible. The real focus of attention for the control
of coronary heart disease must be the problem
of preventing these deposits from developing
futher. An essential part of this problem is the
question of predicting the threat of coronary
disease, which is pretty much the same thing as
recognizing the tendency of atherosclerosis to
progress.
Not so long ago it was widely held that arteriosclerosis, and the particular variety called atherosclerosis, was an essential part of aging. The
result was a defeatist attitude and a regrettable
lack of research interest. Now, however, it is
clear that age alone does not bring with it the
cholesterol deposits in the arteries that currently constitute the basic cause of more deaths
in the United States than any other disease category.
Atherosclerosis is a progressive development
and it seems highly probable that the eventual
appearance of clinical coronary disease usually
Source: American Journal of Public Health 43: 1399-1407,
1953.
I Address presented at the Annual Health Conference of
New York at Lake Placid, N.Y., June 3, 1953.
Director, Laboratory of Physlological Hygiene, and Professor, School of Public Health, University of Minnesota, Minneapolis, Minn.
312
Keys
313
3 14
Spanlaids
Age
Ndples
London
Poor
Rich
20
30
40
50
135
216
23 1
229
187
205
248
255
181
215
223
210
202
217
243
264
Minnesota
Naples
London
Poor
Rich
30
40
50
55
100
112
127
131
100
107
106
109
100
121
124
127
100
103
98
100
100
112
122
127
Table 3. Deaths of men in 1948-1949, rates per 1000 of given age, ascribed to all circulatory diseases
(International Category VII) and to degenerative heart diseases (myocardial diseases and coronary heart
disease, Categories 93 and 94 in the International List of 1938).
~
U. 5. Whites
Canada
, % u. s.
England and Wales
Italy
u. s.
, % u. s.
,%
Japan
. % us.
~~
Degenerative heart
40-44
50-54
40-44
50-54
1.66
1.10
66%
0.70
42%
0.63
38%
0.55
33%
5.65
4.03
71%
2.73
48%
1.86
33%
1.25
22%
1.19
0.89
75%
0.27
22%
0.25
21%
0.14
12%
4.44
3.43
77%
2.04
46%
1.02
23%
0.37
8%
3 16
substantially the same values in the two countries for the ages 40-60 for all neoplasms, for
cirrhosis of the liver, for nephritis and nephrosis, and for intracranial lesions of vascular origin. Among major causes of death, in fact, only
the death rates ascribed to heart disease are
grossly dissimilar. Why? Further analysis of the
death rates from all circulatory disorders shows
that when the degenerative group of coronary
disease, angina pectoris, and myocardial disease
are excluded, the death rates from the total of
all other circulatory diseases are much alike in
the two countries. Why? From every viewpoint
the analysis brings us back to the conclusion
that there is a major difference between these
countries for the incidence and mortality of
coronary heart disease. This problem is greatest
among American men and tends to become
progressively less as we pass to countries where
fats contribute less and less to the total diet.
As far as data have been made available, the
blood cholesterol picture in the clinically
healthy members of the populations are entirely in keeping with the dietary and the mortality data.
It should be recognized that, in addition to
the character of the diet, another factor may
possibly be involved. Differences in habitual
physical activity cannot be ruled out of the picture since, in general, the population groups on
Table 4. Death rates, from all causes, in 17 countries with a total population of about 310 millions; all
values are for the period 1947-1949 and are expressed as percentages of the corresponding rates in the
United States in 1949.
Age
Sex
Australia
Belgium
Canada
Denmark
England and Wales
France
Ireland
Italy
Japan
Netherlands
New Zealand
Norway
Portugal
Scotland
South Africa
Sweden
Switzerland
Mean
40-44
50-54
60-64
75
96
78
59
68
96
80
91
156
52
55
64
139
93
93
61
78
91
89
91
83
78
100
78
100
216
69
72
78
125
97
108
86
97
87
91
76
63
76
91
57
77
111
56
66
53
99
93
102
63
78
96
96
92
88
83
91
86
88
153
76
81
65
96
100
115
85
97
94
97
84
70
93
93
69
75
113
63
85
54
99
97
94
68
88
94
101
96
100
88
91
88
97
207
89
88
68
103
107
104
92
108
84.3
97.5
78.8
93.4
84.5
101.1
Keys
the lower fat diets are also characterized by a
higher level of physical activity. Special studies
on this point are already in progress.
Let us now return to the diet. There are
enough good reasons to suspect that our present high-fat diet in the United States is scarcely
favorable to the health of adult men, even
though it is premature to blame it as the whole
cause of our excessive mortality from heart disease. The changes in mortality in countries
forced to alter their diets during the late World
War cannot be ignored. These changes conform
to the concept that the proportion of fat in the
diet is closely related to the development of
arteriosclerotic heart disease (36-38). Changes
in diabetes mortality among patients over 45
years of age, notoriously related to atherosclerosis, were equally marked during the war and the
parallel with the dietary change is equally striking (39, 40).
From the start of acceptable systematic records of the U.S. Department of Agriculture in
1909, the proportion of fats in the national food
supply has steadily risen from roughly 30 percent of total calories to over 40 percent in
1950-1952. The same data indicate that the
protein contribution to the total calories has
remained substantially constant at around 12
percent, or a trifle less. Because of the uncertainties in regard to kitchen and plate waste, the
actual intake of the several major nutrients cannot be specified with great precision; but there
is no doubt that the per capita consumption of
total fats in the United States is higher than
anywhere else and it has been, and still is, steadily rising. In view of the facts and concepts
expressed above, this constitutes a real problem
for the public health.
Among other immediate questions that arise
is that of the several sources of the fats in the
present national dietary. In the classification of
the U.S. Department of Agriculture, by far the
largest proportion of the total fats, from 45 to
50 percent in recent years, is fats and oils as
such, excluding butter. This means cooking
and salad fats and oils-lard, corn oil, cottonseed oil, hydrogenated vegetable oil shortening,
margarine, mayonnaise, etc. Butter contributes
only 5 percent of the total fat calories. It is
evident that a large reduction in our present
high-fat consumption could be achieved without affecting the fats associated with the more
nutritionally valuable foodstuffs. I n other
words, it should not be difficult to reduce the
317
1950.
(9) Gofman, J. W., H. B. Jones, F. T. Lindgren, T.
3 18
tein-lipid relationships in human plasma. 11. In Atherosclerosis and related conditions. A m J Med 1 1 :480,
1951.
(12) Keys, A. Atherosclerosis: A problem in newer
public health. J M t Sinai Hosp 20:118, 1953.
(I?) Katz, L. N. and J. Stamler. Experimental Atherosclerosis. Springfield, Ill., Thomas, 1953.
(14) Messinger, W. J., Y. Porosowska, and J. M.
Steele. Effect of feeding egg yolk and cholesterol on
serum cholesterol levels. Arch l n t Med 86:189, 1950.
(15) Moses, C. Dietary cholesterol and atherosclerosis. A m J Med Sci 224:212, 1952.
(16) Okey, R. Cholesterol content of foods. J A m
Dietet Assoc 21:341, 1945.
(17) Lange, W. Cholesterol, phytosterol, and tocopherol content of food products and animal tissues. J
A m Oil Chem Soc 27:414, 1950.
(18) Pihl, A. Cholesterol studies. I. The cholesterol
content of foods. Scand J Clin Lab Investigation 4: 115,
1952.
(19) Keys, A. The physiology of the individual as
an approach to a more quantitative biology of man.
Fed Proc 8:523, 1949.
(20) Gertler, M. M., S. M. Garn, and P. D. White.
Diet, serum cholesterol and coronary artery disease.
Circulation 2:696, 1950.
(21) Wilkinson, C . F., Jr., E. Blecha, and A. Reimer.
Is there a relation between diet and blood cholesterol?
Arch Int Med 85:389, 1950.
(22) Heyman, W. and F. Rack. Independence of
serum cholesterol from exogenous cholesterol in infants and in children. A m J Dis Child 65:235, 1943.
(23) Keys, A,, 0. Mickelsen, E. v.0. Miller, and C.
B. Chapman. The relation in man between cholesterol levels in the diet and in the blood. Science 112:79,
1950.
(24) Hildreth, E. A,, S. M. Mellinkoff, G. W. Blair,
and D. M. Hildreth. The effect of vegetable fat ingestion on human serum cholesterol concentration. Circulation 3:641, 1951.
(25) Anderson, J. T. and A. Keys. Dietary fat and
serum cholesterol. Fed Proc 12:169, 1953.
(26) Kinsell, L. W., G. Michaels, L. De Wind, J.
Partridge, and L. Boling. Serum lipids in normal and
abnormal subjects; observations on controlled experiments. California Med 78:5, 1953.
INCIDENCE
'
Moslems. Moslem women have a low frequency of cervical cancer as compared to other
religious groups in their community. A report
from the Institute of Pathology in Indonesia
covering the period 1939 to 1949 shows
1.6 percent of Indonesian female patients and
4.9 percent of the Chinese female patients to
have cancer of the cervix. In Indonesia, 90 percent of the Indonesians, but none of the Chinese, are Moslems (26).We show in Table 2 the
frequency of cervical cancer in two hospital
populations in India among different religious
groups. In both hospitals cervical cancer is
much less frequent among Moslem than among
Hindu women.
Negroes. Recent incidence studies by the U S .
National Cancer Institute show a uniformly
higher incidence of cervical cancer among
Negroes than among whites (1) (Table 3), a
finding supported by several hospital studies
(27-29).
Low-Income Groups. Carcinoma of the cervix
occurs more commonly in the lower income
groups. The Registrar General of Great Britain
(30) shows that the lower the social class, the
higher the mortality of cervical cancer. Similarly, Clemmesen (31) shows that in Copenhagen carcinoma of the cervix occurs more
commonly in the low-priced housing districts.
319
320
Table 1. Relative incidence of carcinoma of the cervix in Jews and non-Jews at three American hospitals.
Hospital
Years
Total cases of
cervix cancer
Incidence of cancer
of the cervix in
non-Jews relative to
that in Jews
1893-1906
1909-1918
1928-1948
1925-1945
1916-1937
18
85
323
1317
3106
17.0:l
12.5:l
5.3:l
5.9: 1
8.5:l
Table 2. Cervical cancer as a percentage of total female cancer admissions at the Tata Memorial
Hospital, Bombay, 1941-1950, and the Premier Radiological Institute and Cancer Hospital, Madras,
1950-1952.
Tata Memorial
Hindus
Christians
Moslems
Parsis
Percent
cervix cancer
3828
575
818
396
45
29
16
13
New Orleans
Chicago
Dallas
Birmingham
Detroit
Philadelphia
Total female
cancer
admissions
White
Nonwhite
59
28
44
52
37
33
72
65
65
73
54
65
~~
Total female
cancer
admissions
280
60
67
-
Percent
cervix cancer
53
29
18
-
There have been numerous attempts to explain these variations among the different population groups by implicating possible etiological factors which vary in intensity in the same
fashion from group to group.
Wynder et al.
subpreputial hygiene and cleanliness, is a menace even more serious to the female sex than to
the sex in which they originate. Khanolkar (39)
also draws attention to the fact that the Moslems
in India, who do circumcise their males between the ages of 6 and 12, have a relatively low
frequency of cervical cancer.
321
the estrogens. On the basis of associated endometrial hyperplasia, Bainborough ( 4 5 ) proposes that excessive estrogen stimulation is a
factor in the production of cervical cancer, a
conclusion denied by Bayly and Greene (46),
who find no significant degree of endometrial
hyperplasia among patients with cancer of the
cervix. Nieburgs (47) advances the contrasting
idea that low estrogen levels might be of etiological significance. Lombard and Potter (35)propose that hormonal factors may account for earlier marriage and high divorce rates among
patients with cervical cancer.
Low Economic Status. Kennaway ( 2 4 ) emphasizes a factor as responsible for a high incidence
of cervical cancer, a factor that is intensified in
those married and single women by descent in
economic scale. Lombard and Potter (35)suggest that low economic status has possible etiological significance, stating, Concomitant with
low economic status are such factors as: poor
obstetrical care, improper housing, and poor
nutrition. Smith ( 1 4 ) believes that poor
obstetrical care and post partum care and neglect of symptoms of the lacerated and ulcerated cervix account for the greater frequency
of cervical cancer among the poorer classes.
Clemmesen ( 4 1 ) also observes a higher incidence with low economic status, but emphasizes
that rural areas in Denmark have a relatively
low incidence of cervical cancer compared to
the low-income city areas. He believes that sexual activity may be the common denominator to
account for this observed incidence difference.
Vinebergs (40) opinion, stated in 1919, is recalled with interest. When one stops to consider that the total number of Jewish women,
1995, had badly lacerated cervices . . . and that
they were living in the worst possible hygienic
surrounding, amidst the greatest squalor and
privation, such as obtained in the lower East
Side of the Metropolis, it is truly remarkable
that so few cases of cancer of the cervix were
detected amongst them.
322
Vaginal Discharge. Hausdorff (56)suggests vaginal discharge as a causative factor in the production of cervical cancer.
THE PRESENT STUDY
323
Wynder et al.
viewed, but are not included in any of the analyses which follow:
White
Negro
Non-Jewish Jewish
14
18
Adenocarcinoma
Carcinoma in situ
15
13
4
3
The Interview. Regeena Goodwyn and Florence Moreno interviewed all patients at the Memorial Clinic and Memorial Wards,G and in
Jersey City and Philadelphia. Dr. Charles Miller
interviewed all the St. Louis patients. The paSome of the Jewish patients with cervical cancer seen on
the Memorial Wards were interviewed by Miss E. Schwab and
by E. L. Wynder.
Table 4. Number of patients with cancer of the cervix and controls by age and hospital of interview,
White non-Jewish,Jewish, and Negro.
White
Non-Jewish
Hospital
New York
Memorial Gym Clinic
Memorial Hospital Wards
James Ewing
Bellevue Radiologic Clinic
Harlem Hospital
Cervix
Jewish
Control
Cervix
Negro
Control
Cervix
Control
264
7
32
13
0
7
8
1
1
0
13
20
95
2
14
12
52
28
86
42
33
31
45
66
11
28
3
13
9
1
0
0
0
3
0
0
6
3
16
32
2
39
15
19
11
Barnes
Barnard
24
28
28
32
1
0
0
0
9
8
15
9
Total
354
594
20
326
215
39 1
14
60
110
110
60
54
107
184
154
95
0
7
4
6
3
52
131
101
29
13
10
58
68
51
28
59
137
354
594
20
326
215
39 1
129
12
16
26
302
4
45
Jersqr City
Margaret Hague Med. Center
Philadelphia
Univ. of Pennsylvania Hospital
Washington
Georgetown
Walter Reed
Gallinger
Warwick Clinic of George
Washington Hospital
St. Louis
Age at interview
Less than 30 years
30-39
40-49
50-59
60 and over
Total
41
56
3
117
51
27
324
Table 5. Number of control patients by diagnosis: White non-Jewish,Jewish, and Negro, Memorial
Clinic and all hospitals combined.
Memorial Clinic
All hosoitak
White
Condition
Uterus, malignant
Uterus, benign
Cervicitis, chronic
Polyp of cervix
Carcinoma of ovary
Carcinoma of vulva
Malignant, misc.
Benign, misc.
No positive gyn. lesions
Total
Non-Jewish
White
Jewish
Negro
Non-Jewish
31
44
71
20
13
9
7
69
38
12
49
60
26
2
7
14
59
35
8
21
26
5
1
5
17
8
71
87
111
31
26
13
25
120
110
302
264
95
594
Jewish
Negro
21
60
63
28
4
9
17
64
60
37
152
43
16
17
3
15
73
35
326
39 1
tients seen in Washington, D.C., were inter- were selected on each clinic day from the clinic
viewed by Drs. F. Ablondi, Tom Higgins, Mary list of those present. The list gave no clue as to
Kiernan, James Leonard, Marion MacLean, the patients condition. The interviewing at the
and Ernest Wynder.
Memorial Clinic was thus completely blind. At
The interview approach was the same for all other hospitals, however, it was not always possipatients, since all had gynecologic complaints. ble to assure such automatic control on the obThe patient was told that this interview was jectivity of the interviewer. For this reason, the
necessary to complete her history and would data obtained at the Memorial Clinic have been
assist in evaluating her particular problem. It analyzed separately.
was stressed that all of the questions had to be
Age-Hospital Standardization. The distributions
answered correctly, although some might seem
of
cervical cancer and control patients by age at
personal, because they might be related to the
interview
are not the same. Thus, only 2 out of
development of her disease. The patient was
assured that her replies would be kept abso- 129 white non-Jewish cervix patients at Memorial Clinic were less than 30 years of age, while a
lutely confidential.
considerably
larger proportion of the conSpecial care was taken to assure accuracy in
the question regarding circumcision. It was rec- trols-22 out of 302-were this young. Simiognized that some women regard a loose and larly, the distributions by hospital of interview
short foreskin as circumcision. The interviewers were not the same. For example, one-third of
were therefore instructed not to accept a simple the non-Jewish cervix cases studied were seen at
yes or no answer, but to probe for evidence the Memorial Clinic and one-half of the conof definite information. When such informa- trols. To eliminate the effects of these differences in age and hospital distribution on the
tion was absent, the woman was asked to quescomparisons between cervix and control pation her husband and then to give us her reply.
If this was not possible, or the information was tients, we have uniformly adjusted the results
for the control patients to the age-hospital disstill inconclusive, the answer was left as Dont
tribution of the cervix cases.
Know. Identical procedures were followed for
The procedure used was designed to give
cervix and control cases.
each age-hospital group in the controls the
After the interview was completed, the charts
same relative weight that it had in the cervix
were reviewed for definitive diagnosis. In many
group. Thus if proportion wi of all cervix cases
of the cases the diagnosis was not available at
fell in the ith age-hospital group and if proporthat particular time and had to be added later.
tion pi of the control cases in this age-hospital
In all cases of cervical cancer, biopsy report was
group have some characteristic, i.e., being maravailable.
ried, the age-hospital standardized estimate of
Blind Interviewing. At the Memorial Gyne- the proportion married is Xwipi,the summation
cological Clinic the patients to be interviewed extending over all age-hospital groups. Rou-
Wvnder et al.
tinely this adjustment was performed by assigning a multiplier to each age-hospital group in
the controls and entering this multiplier on the
punch card for each individual in that group. If
w, denotes the number of cases among the controls in the ith age-hospital group, the multiplier
for this group was w,/ni. The age-hospital standardized estimate of the proportion of the controls having a certain characteristic, i.e., being
married, is then obtained by (a) sorting out the
cards for all control patients who were married,
(b) obtaining on a tabulator the sum of the
multipliers for all married controls. The 594
controls, age-hospital adjusted on this basis,
give results of less precision, however, than
would have been obtained with 594 controls
matched on an age-hospital basis to the cervix
patients. The 594 adjusted controls gave results
of equivalent precision to what would have been
obtained with 330 matched controls. For this
reason, we have taken the control total as 330
and have referred to the number of cases resulting from this procedure in the various tables as
the equivalent number of cases after age-hospital standardization.
The Negro controls were adjusted to the agehospital distribution of the Negro cervix cancer
cases. The 391 controls interviewed and adjusted were equivalent in precision to 287
matched and unadjusted controls. The white
Jewish control cases were not adjusted to the
white Jewish cervix group, which is too small to
justify statistical analysis, but rather to the white
non-Jewish cervix group. For the Indian data,
the results were age-adjusted in the same fashion.
At some points in the following analysis we
found it necessary to standardize out other factors, such as age at first marriage. The procedure employed in these cases is identical with
that used for the age-hospital standardization.
RESULTS (AMERICAN)
325
326
Table 6. Percent distribution of cervical cancer and control patients by marital status, White non-Jewish,
Jewish, and Negro
White
Marital status at time
of interview
Non-Jewish
Cervix
Memorial Clinic
Married
Never married
Divorced and separated
Widowed
Total
All Hospitals Combined
Married
Never married
Divorced
Widowed
Total
Jewish
Control
Negro
Control
Cervix
Control
59
2
12
27
67
9
8
16
71
4
8
17
31
38
29
36
9
28
27
100
00
100
100
100
62
2
14
22
65
9
11
15
40
4
33
23
43
10
27
20
100
00
100
100
Table 7. Percent distribution of cervical cancer and control patients who had ever been married, by
number of marriages, White non-Jewish,Jewish, and Negro.
White
Non-Jewish
Cervix
Number of marriages
Jewish
Control
Negro
Control
Cervix
Control
Memorial Clinic
1
2
3 or more
71
28
1
88
11
1
89
11
0
65
31
4
78
18
4
Total
100
100
100
100
100
70
28
86
12
2
65
30
5
78
20
2
100
100
100
100
Total
Table 8. Percent of cervical cancer and control patients who had ever been married, married two or
more times, by marital status, White, non-Jewish, and Negro, all hospitals combined.
White, non-Jewish
Marital status at
time of interview
Number of
cases
Cervix
Negro
Percent married
two or more
times
Control
Cervix
Control
220
49
76
219
34
49
33
37
19
14
11
16
345
302
30
14
Number of
cases
Cervix
Percent married
two or more
times
Control
Cervix
Control
86
71
49
125
75
59
47
21
35
26
16
19
206
259
35
22
~~
Married
Divorced or separated
Widowed
Total
a
327
Wynder et al.
Age at First Marriage and First Coitus. The cervix patients, both white and Negro, at Memorial
Clinic and all hospitals, show a markedly earlier
age of marriage than do the control patients
(Table 9). Thus, 13 percent of the Memorial
white non-Jewish cervical cancer group were
married by age 16, as compared with 5 percent
of the comparable controls. For Memorial
Negroes, 32 percent of the cervical cancer
group was married by age 16, and 13 percent of
the controls. The differences for all hospitals
Table 9. Percent distribution of cervical cancer and control patients who had ever been married by age
at first marriage, White non-Jewish,Jewish, and Negro.
White
~
Non-Jewish
Age at first marriage
Cervix
Jewish
Control
Control
Negro
Cervix
Control
Memoraal C h i c
16 or less
17-19
20-24
25 or more
13
38
31
18
5
26
38
31
4
15
46
35
32
39
19
10
13
38
25
24
Total
100
100
100
100
100
14
40
29
17
8
25
39
28
32
36
21
11
19
34
29
18
100
100
100
100
Total
Table 10. Percent distribution of cervical cancer and control patients by age at first coitus, White nonJewish, Jewish, and Negro.
White
Non-Jewish
Age at first coitus
Cervix
Jewish
Control
Control
Negro
Cervix
Control
Memorial Clinic
16 or less
17-19
20-24
25 or more
Never
17
38
32
12
1
6
27
34
26
7
4
15
43
34
4
45
41
13
2
0
28
44
16
11
1
Total
100
100
100
100
100
19
41
26
12
1
10
25
36
22
7
55
30
12
2
1
36
39
16
7
2
100
100
100
100
17-19
20-24
25 or more
Never
Total
328
Table 11. Number of cervical cancer and control patients by age at first coitus and number of marriages,
White non-Jewish, and Negro, all hospitals combined.
~~
~~
Cervix
Number of marriages
Age at first
COltUS
Controla
Number of marriages
Total
2+
Total
2+
Whate non-Jewwh
16 or less
17-19
20-24
25 or more
Never
No report
66
143
98
40
3
4
1
1
1
2
3
0
35
93
77
32
0
4
30
49
20
6
0
0
34
83
117
72
22
2
1
1
1
22
0
25
60
107
65
0
2
8
22
9
3
0
0
Total
354
24 1
105
330
29
259
42
Negro
16 or less
17-19
20-24
25 or more
Never
No report
119
64
25
5
0
2
4
3
1
45
19
7
0
0
1
103
112
48
20
5
2
5
9
5
5
5
0
71
82
35
15
0
2
27
21
8
0
0
70
42
17
5
0
1
Total
215
135
72
290
29
205
56
0
0
Wynder et al.
329
Table 12. Percent distribution of cervical cancer and control patients by age at diagnosis, age at first
coitus, White non-Jewish, and Negro at all hospitals combined.
Cervix
Age at first coitus
Age at
diagnosis
16 or
less
17-19
White, rzonyewkh
30 or less
3 1-40
4 1-50
51-60
61
7
30
30
19
13
Total
Median age
Negro
30 or less
3 1-40
4 1-50
51-60
61 +
Total
Median age
a
Control
Age at first coitus
16 or
20-24 25+
20-24
25+
Never
3
15
35
28
18
2
14
29
40
15
3
12
21
38
26
(0)
(0)
(67)
(0)
(33)
2
34
18
38
8
2
21
34
24
19
6
19
37
26
12
3
6
30
41
20
(23)
(20)
(41)
100
100
100
100
100
100
100
100
100
100
44
49
51
54
(47)
48
48
47
54
(56)
6
31
28
25
10
5
18
37
25
15
(0)
(27)
(42)
(19)
(12)
(0)
(20)
(20)
(20)
(40)
5
34
29
24
8
5
22
40
22
11
4
28
24
27
17
(0)
(16)
(26)
(39)
(19)
(3)
(10)
(26)
( 0)
(61)
100
100
100
100
100
100
100
100
100
45
47
(45)
(55)
45
47
48
(53)
(51)
less
17-19
Nevera
(6)
(1 1)
330
Non-Jewish
Cervix
Memorial Clinic
0
1 or more
Jewish
Control
Control
Negro
Cervix
Control
9
91
12
88
11
89
89
11
17
83
100
100
100
100
100
35
32
16
10
7
41
29
19
7
4
42
38
13
4
3
50
18
14
14
4
38
28
19
6
9
100
100
100
100
100
12
88
11
89
18
82
100
100
100
100
37
27
16
10
10
44
29
16
6
5
46
25
14
8
7
46
25
16
5
8
100
100
100
100
Total
Pregnant
1-2 times
3-4
5-6
7-8
9 or more
Total
Total
Pregnant
1-2 times
3-4
5-6
7-8
9 or more
Total
Table 14. Number of cervical cancer control patients who had ever been married by number of
pregnancies and age at first marriage, White non-Jewish, and Negro, all hospitals combined.
Cervix
Age at first marriage
25 or
more
Total
16 or
less
17-19
9
38
27
10
13
4
13
28
9
5
2
1
37
114
78
41
17
14
1
3
6
5
4
4
3
22
23
14
8
6
8
50
35
17
4
3
25
39
14
5
1
1
138
101
58
301
23
76
117
85
7
4
7
22
4
6
1
3
9
11
2
0
0
0
46
97
53
34
12
17
4
16
11
5
27
21
8
6
7
7
5
5
9
30
22
11
5
10
14
32
11
14
2
2
19
19
7
2
0
0
66
74
43
22
259
50
87
75
47
Number of
oremancies
Total
16 or
less
17-19
White, Non-Jewish
0
1-2
3-4
5-6
7-8
9 or more
30
116
87
53
30
30
1
13
12
10
7
6
7
37
39
28
8
19
Total
346
49
23
84
45
25
14
14
2
24
18
205
Negro
0
1-2
3-4
5-6
7-8
9 or more
Total
a
Controla
Age at first marriage
20-24
13
20-24
25 or
more
Wynder et al.
Number of
pregnancies
Negro
Cervix Control
0
1-2
3-4
5-6
7-8
9 or more
9
33
25
15
9
9
9
33
28
16
8
6
11
41
22
12
7
7
15
36
Total
100
100
100
100
Negro
control
~~
Cruded
Standardiiedb
Cervix
Crudel
Standardizedb
First
16 or less
17-19
20-24
25 or more
9
36
37
18
3
23
40
34
6
36
36
22
31
34
24
11
18
32
31
19
26
33
25
16
Total
100
100
100
100
100
100
Last
20 or less
21-25
26-30
31-35
36 or more
12
20
26
25
17
7
18
25
24
26
10
18
25
24
23
30
27
17
14
12
20
24
24
15
17
23
24
21
15
17
Total
100
100
100
100
100
100
13
6
8
White non-Jewish
control
Cervix
22
a Obtained from Table 14 by weighting the percent distribution by number of pregnancies for each age-at-first-marriage class in the control group by the number of cases in that
class in the group with cancer of the cervix.
Table 16. Percent distribution of cervix and control patients who had ever
been pregnant by age at first pregnancy and age at last pregnancy, White
non-Jewish,and Negro, all hospitals combined.
Age at pregnancy
33 1
Standardized for age and hospital of interview, but not age at first marriage.
Standardized for age at first marriage in addition.
332
riage. When the control groups are made comparable to the cervix patients in this respect, by
standardizing with respect to age at first marriage, it will be noted that the differences are
essentially eliminated. Therefore, the differences in age at first and last pregnancies do not
appear to be independent variables, but merely
reflect previous differences in age at marriage.
Miscarriages and Abortions. We have also investigated the distributions of the populations
studied by number of miscarriages and abortions. N o marked or consistent differences are
apparent. This body of data has thus yielded no
evidence that there are any differences between
cervical cancer and control populations, in any
characteristics associated with pregnancy,
whether it be the number of such pregnancies,
the age at which they occurred, or associated
phenomena, such as abortions and miscarriages.
Syphilis. Previous studies have reported twice
as many syphilitic patients among women with
cervical cancer as in the general population
(50-52).We found a similar difference between
our cervical cancer non-Jewish white patients
and the controls. Essentially no difference was
found, however, between the Negro cervical
Table 17. Percent distribution of cervix and control patients with sexual experience by circumcision
status of partners, White non-Jewish, and Negro.
Circumcision status
of partners
Memorial Clinic
Circumcised husbands only:
No other partners
Pre-, extra-, or postmarital partners
Circumcised and uncircumcised husbands
Uncircumcised husbands only
Circumcision status unknown
Total
All Hospitals Combined
Circumcised husbands only:
N o other partners
Pre-, extra-, or postmarital partners
Circumcised and uncircumcised husbands
Uncircumcised husbands only
Circumcision status unknown
Total
a
White
Non-Jewish
Cervix
Controls
Jewish
Controls
Negro
Cervix
Controls
3
3
9
3
3
93
2
1
0
8
2
8
5
5
80
10
75
10
1
3
65
25
47
35
100
00
100
100
00
5
3
14
< 1=
9
9
6
75
11
3
69
10
7
5
66
22
100
100
100
49
29
100
This indeterminateness could have been avoided by obtaining information on pre-, extra-, or postmarital partners
for women with uncircumcised husbands. At the time the
survey was being planned, we were unwilling to ask such
questions of more than the circumcised groups, for whom it
was obviously necessary. We did not, of course, foresee the
present difficulty.
unable to report on circumcision status of partners. For patients who had been married only
once, the proportion unknown was smaller, but
still far from trivial-8 percent for the white and
20 percent for the Negroes. By making sufficiently unfavorable assumptions about the circumcision status of the unknowns, one could
wipe out the difference in circumcision status
between cervical cancer and control groups.
We have attempted to check on the circumcision results shown in Table 17 by using a different method of collecting the information-direct interviews with males in the wards of some
of the hospitals where females were studied.
Out of 489 white males, 80, or 16 percent, were
circumcised; 37 out of 208, or 18 percent, of
the Negro males were circumcised. These percentages agree with those reported by females.
They also confirm the fact that in the ward and
clinic population circumcision is equally common among whites and Negroes. This result is
explained at least in part by the following facts:
(1) at least twice as many native-born white
males are circumcised as foreign-born males;
Negro males are predominantly native-born,
and ( 2 ) many Negroes are circumcised in their
teens.
We show in Table 18 the simultaneous distribution of cervical cancer and control cases by
circumcision status of partner and number of
times married. For those married once, 10 percent of the cervix groups and approximately
20 percent of the controls reported a circumcised husband. This was true for both the white
non-Jewish and Negro groups. For those married twice, the proportion with both husbands
circumcised is, of course, smaller, but the difference between cervix and control groups is in the
same direction and of approximately the same
magnitude for both white non-Jewish and
Negro groups, although only for the former
groups is the difference statistically significant.
In Table 19 we show the simultaneous distribution of cervix and control cases by circumcision status of partner and age at first coitus for
those married only once. This table shows that
differences in frequency of circumcision between cervical cancer and control groups persist
for both white non-Jewish and Negro groups,
even after the effect of age at first sexual intercourse is held constant. There is also a suggestion that circumcised husbands are reported
more frequently by those reporting late first
coitus, although the effect is not large and can-
334
whites and Negroes at Memorial and at all hospitals combined. In all cases the median age was
13 to 14. Comparison of length of flow showed
also no essential differences between the two
groups.
Table 18. Number of cervical cancer and control patients with sexual experience by circumcision status
of partners and number of marriages, White non-Jewish, and Negro, all hospitals combined.
Cervix
No. of marriages
Circumcision status
of partners
White non-Jewish
Circumcised husbands only:
No other partners
Pre-, extra-, or postmarital partners
Circumcised and uncircumcised husbands
Uncircumcised husbands only
Circumcision status unknown
Total
Negro
Circumcised husbands only:
No other partners
Pre-, extra-, or postmarital partners
Circumcised and uncircumcised husbands
Uncircumcised husbands only
Circumcision status unknown
Total
a
Controlsa
No. of marriages
2+
2+
15
9
0
197
20
1
2
21
67
14
0
0
0
0
6
42
10
0
188
21
2
4
8
26
24 1
105
261
42
0
0
0
0
8
1
12
0
96
26
0
3
10
45
14
0
0
0
1
23
23
22
0
113
46
3
4
10
25
13
135
72
24
204
55
0
0
0
0
5
Table 19. Number of cervical cancer and control patients who married once by circumcision status of
partners and age at first coitus, White non-Jewish, and Negro, for all hospitals combined.
Cervix
age at first coitus
Circumcision status
of partners
White non-Jewish
Circumcised husbands only:
No other partners
Pre-, extra-, or postmarital
partners
Uncircumcised husbands only
Circumcision status unknown
Total
Negro
Circumcised husbands only:
N o other partners
Pre-, extra-, or postmarital partners
Uncircumcised husbands only
Circumcision status unknown
Total
a
16 o r
less
17-19
20-24
2
30
3
76
7
35
Controls
age at first coitus
16 or
less
17-19
1
64
6
2
24
3
93
77
48
16
6
31
5
12
4
70
42
17
25+
20-24
25+
17
16
2
20
2
1
48
4
2
80
10
5
39
6
32
26
59
109
66
0
4
1
12
42
14
6
44
22
3
20
4
1
5
5
72
81
33
15
Wynder et al.
Abstinence After Menses. About half the white
non-Jewish and Negro patients reported some
abstention after menses, both in the cervix and
control groups. Of those who said they abstained, about 40 percent reported one or two
days as the usual period of abstinence, but
again this percentage was the same in cervix
and control groups for both white non-Jewish
and Negro patients. Of those abstaining in the
white non-Jewish group, 16 percent of the controls and 11 percent of the cervix group reported abstention for seven or more days.
Among the Negroes the comparable figures
were 5 and 6 percent. Thus, there seems to be
no difference in the practice of abstinence between cervix and control groups.
Only 30 percent of the Jewish controls reported no abstention, while of those abstaining,
40 percent reported abstinence for seven full
days. There was a considerable difference by
age groups. Eighty-three percent of those over
50 reported some abstention, but 62 percent of
those under 50 abstained. Of those over 50 who
abstained, 50 percent reported abstention for
seven full days. For those under 50, the comparable figure was 28 percent. There is a considerable difference, therefore, in the practice
of abstention between the Jewish and non-Jewish population, and an additional difference between younger and older Jewish women. The
Jewish women presumably had the Talmudic
proscription in mind while answering this question, whereas the non-Jewish women did not.
Consequently, the answers may not have the
same meanings for both groups. In particular,
the Jewish women answering yes may have
had a more habitual practice in mind than the
non-Jewish women. In view of the lack of any
reported difference between cervical cancer
and control groups, however, it is difficult to
draw any firm conclusion with respect to the
significance of abstention in the etiology of cervical cancer.
Douching. No difference was found between
cervix and control groups with respect to type
or frequency of douching. One-fifth of the
white non-Jewish patients in both cervical cancer and control groups reported never douching. Forty percent of both groups reported frequent douching. Practices among the Jewish
controls were essentially the same. Negro patients, both cervical cancer and controls, reported somewhat more frequent douching than
335
the whites. Specifically, 52 percent of the controls reported frequent douching, while only
6 percent said they never douched. In the cervical cancer groups the comparable figures
were 50 and 13. Of those that did douche, no
differences were apparent between cervix and
control groups, either in the frequency of
douching, or in the types of douches used. In
particular, the use and frequency of use of a
brand-name coal-tar derivative was reported
with equal frequency by both groups.
336
Age at
diagnosis
53
61
67
59
54
43
52
46
50
46
58
39
32
39
36
45
32
66
36
34
Number
of marriages
Age at
first
marriage
22
1
1
1
1
1
1
1
1
1
1
2
1
1
1
1
1
3
1
2
Age at
first
sexual
intercourse
Abstinence
after
menses,
days
Age at
first
pregnancy
Number
of pregnancies
22
19
20
24
24
21
23
20
21
26
0
0
24
27
24
25
25
23
23
24
27
24
25
15
25
3
3
7
2
0
1
2
4
3
3
1
7
7
0
25
16
27
22
25
16
25
26
33
24
24
18
19
25
16
20
24
19
20
24
24
21
19
18
18
19
18
18
16
18
19
22
21
19
2
3
3
2
2
5
2
7
4
3
5
0
7
7
0
0
0
0
2
3
0
Circumcision
code"
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
3
4
a Circumcision code: 1. Circumcised husband and no other partners; 2. Circumcised husband but other partners; 3. Mixed
circumcision in husbands; 4. Uncircumcised husbands.
Wynder et al.
Table 2 1. Relative risk of developing epidermoid
cervical cancer by number of marriages and age at
first coitus, White non-Jewish and Negro women.
White nonJewish
Negro
Virgins
Other single
0.2
1 .o
0
0.7
Married Once
Age at first coitus:
16 or less
17-19
2.2
1.9
2.0
1.0
Category
20-24
25 or more
1.o
0.7
5.3
3.1
2.8
1.o
0.7
3.4
1.8
-
Negro
Circumcised husband
N o other partners
Other partners
Uncircumcised husband
0.4
0.3
0.9
1.o
0.4
0.1
0.8
1 .o
0.2
0.2
0.2
1.0
1.o
0.2
0.0
0.4
0.6
Married Once
1.o
Risk for women with uncircumcised husbands = 1.0 separately for those married once and twice or more.
337
338
Table 23. Estimated risk of developing epidermoid cervical cancer for Negro women relative to White
non-Jewish women on the basis of differences in age at first coitus and number of marriages.
Percent dismibution of controls
White
nonJewish
Category
Virgins
Other single
Married Once
Age at first coitus:
16 or less
17-19
20-24
25 or more
Married Twice or More
Age at first coitus:
16 or less
17-24
25 or more
Total
Negro
Relative risk
Estimated risk
using white
relative risks
Estirriated risk
using Negro
relative risks
White
nonJewish
(3)
Negro
Whitc
(4)
(1) x (3)
(2)x (3)
(1)x (4)
(2)x (4)
Ncgro
White
Negro
(1)
(2)
6.7
2.1
1.7
8.3
0.2
1 .0
0
0.7
1.34
2.10
0.34
8.30
0
1.47
0
5.81
7.6
18.3
32.7
19.7
24.6
28.5
12.1
5.2
1.9
2.2
1 .0
0.7
2.0
1.0
1.0
0.7
14.44
40.26
32.70
13.79
46.74
60.50
12.10
3.64
15.20
18.30
32.70
13.79
49.20
28.50
12.10
3.64
2.4
9.4
0.9
9.4
10.1
0
5.3
3.1
2.8
3.4
1.8
-
12.72
29.14
2.52
49.82
31.31
0
8.16
16.92
(2.52)
31.96
18.08
0
99.8
99.9
Relative risk
149.01
100
212.75
143
partners, the relative risks due to lack of circumcision are consistent with a fivefold difference between Jewish and non-Jewish women. If
the appropriate group is the entire group with
circumcised husbands, without regard to other
partners, the relative risk due to lack of circurncision is not consistent with a fivefold difference
between the Jewish and non-Jewish population.
Although difference in age at first coitus between the Jewish and non-Jewish population
could account for part of the unexplained difference. it could not account for all of it.
RESULTS (INDIAN)
109.06
100
149.29
137
Hindu
Moslem
Indian Christians
Parsis
Jews
Cervix patients
Complete
Total
255
26
22
146
44
29
22
238
135
80
72
339
Wynder et al.
Table 25. Number of cervix and control patients by age at interview, by religious groups, and by
hospital, Tata Memorial Hospital.
Age at
interview
30 or less
31-40
41-50
51-60
61 or more
Total
Hindu
Cervix
Christian
Moslem
Control
Cervix
Control
Cervix
Parsi
Control
Cervix
Control
7
16
22
17
86
97
39
16
54
66
74
36
8
5
6
9
3
3
22
54
42
7
10
1
3
8
8
2
12
16
24
18
10
255
238
26
135
22
80
30 or less
3 1-40
41-50
51-60
61 or more
2
7
14
22
Total
45
13
62
Table 26. Percent distribution of cervical cancer and control patients by age at marriage and by age at
first coitus by religious groups, Tata Memorial Hospital.
Hindu
Cervix
Control
Age at Mnrringe
13 or less
14
15
16
17-19
20 plus
67
11
12
3
4
3
52
13
12
9
9
5
Total
100
Total
Indian
Christian
Mosleni
Cervix
Control
Cervix
Control
Parsis
Cervix
Control
4
1
5
14
37
38
5
11
32
48
100
100
100
14
15
11
(5)
(5)
(14)
(10)
(33)
(33)
0
0
10
5
55
30
20
80
100
100
100
100
(58)
(8)
(0)
(15)
(15)
(4)
18
17
22
13
20
10
100
100
100
40
17
20
11
8
4
23
26
22
12
12
5
(39)
(15)
(4)
(19)
(15)
(8)
13
25
22
100
100
100
(5)
(5)
(13)
(9)
(32)
(36)
2
2
340
Penile Hygiene. In view of the fact that circumcision showed up as a positive variable in this
study, we have done some investigation on the
extent of penile hygiene among American and
Indian males. If one wanted to study this factor
as a direct influence on the development of
cervical cancer, it would require a study of the
sexual partners of women with cervical cancer.
Because of the obvious difficulty of such a
study, we have merely sampled the general hospital population, breaking it down into private,
clinic, non-Jewish white, Negro, and Jewish patients. In the non-Jewish groups only uncircumcised individuals were considered. The Indian
data were broken down into the various religious groups. The Tata material was, in addition, separated into Deccani, Gujarati, and
other Hindus, because of the fact that the Deccanis are of lower economic standing than the
other two Hindu groups.
Hospital data are to be considered with caution, since many male patients may take a special bath before coming for examination. In
general, the American data show a higher percentage of smegma formation among the
clinic than among private patients. Of some
interest was the observation that a rare Jewish
Table 27. Number of pregnancies of cervix and control patients by religion; the controls have been age
adjusted to their respective cervix groups, Tata Memorial Hospital.
Number ,of
pregnancies
0
1 or more
Hindu
Cervix
Moslem
Control
Christian
Control
Cervix
Control
4
96
6
94
17
7
17
24
31
18
14
13
100
100
97
5
95
4
96
93
0
100
16
17
23
22
22
20
19
26
17
18
16
21
28
19
16
27
18
24
17
14
14
23
41
18
4
100
100
100
100
100
Parsi
Control
Cervix
Total
31
28
Wynder et al.
34 1
None (%)
American
Non-Jewish
White (private)
White (clinic)
Jewish
Negro (clinic)
580
125
980
100
92
70
99.4
54
lndzan (Bombay)
Hindu (Deccani)
Hindu (Gujarati)
Hindu (Other)
Christian
Moslem
Parsi
130
90
70
50
130
50
56
74
70
56
100
94
61
48
31
16
27
58
Indian (Madras)
Hindu (mixed)
Christian
Moslem
Moderate (%)
Marked (%)
1
6
0
8
28
16
29
30
0
4
16
10
1
14
0
2
66
63
42 (slight)
18
10
24
0.6
38
342
Wynder et al.
ported to us by non-Jewish women were completely accurate, the absence of any difference
between cervix and control patients in this regard would be of great significance. The great
difficulties involved in accurately reporting a
practice, the importance of which may change
with advancing age, suggest the necessity of
evaluating this result with some reserve. More
information on the abstinence question may become available when the young Jewish women
who do not practice the law to the same extent
reach the cancer age. Even here, however, earlier onset at first coitus, as well as possible
greater exposure to uncircumcised males, presents an obvious complication.
343
Other Factors. Some of the suggested etiological factors enumerated in the introduction of
this report, such as diet, cervical lacerations,
and chronic cervicitis, cannot be easily investigated by an interview and thus have not been
covered. The frequency of chronic cervicitis,
however, among our white non-Jewish, Jewish,
and Negro controls is about the same, suggesting that, even if it is etiologically significant,
other factors are required to explain the difference in incidence among these groups, as has
been previously stressed by McKelvey (48).
Results Relative to Incidence Data. The effects of
early coitus, remarriage, and circumcision are
consistent with incidence differences between
Jewish and non-Jewish women, white and
Negroes. and the various religious groups in
India. Thus, added significance is given to
these factors. It is scarcely necessary to emphasize that if subsequent investigations confirm
the role of circumcision, the preventive implications for many parts of the world will be far
from trivial.
Multiple Etiological Factors. Carcinogenesis
represents the effect of many factors, some of
which may be endogenous and some exogenous. In the development of epidermoid cancer, exogenous factors seem to be of particular
importance. Yet even this type of neoplasm is
the result of multiple factors, many of which
remain unknown in our current state of knowledge.
In our studies we can only throw light on
some of these factors, hoping that, through
them, we might advance our understanding of
344
vical cancer could be accounted for by the earlier age at first coitus among cervical cancer
patients with syphilis, as compared to patients
with cervical cancer without syphilis.
12. Negative variables include onset and flow
of menses, method of delivery, irritative discharge, frequency of douching, and history of
gonorrhea.
13. Epidermoid cancer of the cervix has
been noted in women exposed only to circumcised males and in virgins. Other etiological
factors than those involving coitus and lack of
circumcision must therefore exist.
14. Examination of penile hygiene among
the various groups studied shows that males of
population groups with a high incidence of cervical cancer have poor penile hygiene.
15. The present results are compatible with
the concept that those population groups having a late age at first coitus and first marriage
and a low remarriage rate, whose men are circumcised, have a lower rate of carcinoma of the
cervix.
16. Possible interpretations of these results
and their preventive implications have been discussed.
We wish to express o u r gratitude to the
Chiefs of Staff of those various services, as
listed in Table 4, who have permitted us to
interview their patients. We also wish to thank
Dr. Willard M. Allen and Dr. Rieva Rosh for
their helpful advice and assistance in the incipiency of this investigation. We are grateful to
the following who have critically reviewed the
final draft: Clinical: Drs. Willard M. Allen, Alexander Brunschwig, Harold Burrows, M. Edward Davis, R. G o r d o n Douglas, Alfred
Gellhorn, John B. Graham, E. H. Homing, and
Ernest L. Vennaway. Statistical: Drs. Irwin
Bross, J. Clemmesen, Richard Doll, Morton L.
Levin, Herbert L. Lombard, and L. D. Sanghvi.
Finally, we feel greatly obliged to our senior
interviewers, Regeena Goodwin and Florence
Moreno, without whose careful interrogation
this study would not have been possible, and to
Marianne Bardeleben Vargish for her help in
compiling the statistical data.
References
( I ) National Cancer Institute. Cancer Illness.
Cancer Morbidity Series, Number 1 : 10, 1950-1952.
Bethesda, Md., National Institutes of Health, United
States Public Health Service.
(2) World Health Organization. Epidemiological
Wynder et al.
and VitalStatistics Report, vol. 5, No. 1-2. Geneva, 1952.
(3) Bleich, A. R. J A M A 143:1054, 1950.
(4) Jefferys, W. H. and J. L. Maxwell. The Diseases
of China, Includang Formosa and Korea, ed. 2, Shanghai,
China Medical Association, p. 479.
(5) Khanolkar, V. R. Personal communication.
( 6 ) Cooray, G. H . Indian J Med Res 32:71, 1944.
( 7 ) Vineberg, H. N. A m J Obst 53:410, 1906.
( 8 ) Theilhaber, A. and S. Greischer. Ztschr
Krebsforsch 9:530, 1910.
( 9 ) Peller, S . Ztschr Krebsforsch 34:128, 1931.
(10) Theilhaber, A Ztschr Krebsforsch 8:466, 1910.
(11) Hoffman, F. L. A m J Cancer 17:142, 1933.
(12) Kennaway, E. L. Personal communication.
(13) Treusch, J. V., A. B. Hunt, and A. A. Rousuck.
A m J Obstet Gynecol 52:162, 1946.
( 1 4 ) Smith, F. R. A m ] Obstet Gynecol 41:424, 1941.
(15) Horwitz, A. Surg Gynecol Obstet 44:355, 1927.
(16) Rothman, A,, L. P. Rapoport, and I. Davidsohn. A m J Obstet Gynecol 62:160, 1951.
(17) Rubin, I. C. Cited by H. N. Vineberg. J Mt
Sinai Hasp 10:33, 1943-44.
(18) Weiner, I., L. Burke, and M. A. Goldberger.
A m I Obstet Gynecol 61:418, 1951.
(is)Kaplan, 1. I . and R. Rosh. A m J Roentgen01
57:659, 1947.
(20) Sorsby, M. Cancer and Race; A Study of the Incidence of Cancer Among Jews, New York, N.Y., William
Wood 1L Company, 1931.
(21) Wolff, G. A m J Hyg Sect. A 29:121, 1939.
(22) Davidsohn, I. Medical Leaves, 1939, p. 19.
(23) Sugar, M. and W. E. Levy. New Orleans M C3 SJ
103:424, 1951.
(24) Kennaway, E. L. B r i t ] Cancer 2:177, 1948.
(25) Handley, W. S. Lancet 1:987, 1936.
(26) Tjokronegoro, S . Personal communication.
(27) Robinson, B. W. A m J Roentgenol66:783, 1951.
(28) Quinland, W. S. and J. R. Cuff. Arch Path
30:393, 1940.
(29) Hynes, J. F. A m J RoentgenoZ60:368, 1948.
(30) Registrar-General, Great Britain. Statistical Review ofEngland and Wales (1936). London, H. M. Stationery Office, 1937-1938.
(31) Clemmesen, J. and A. Nielsen. Brit J Cancer
5:159, 1951.
(32) Stocks, P., cited by W. L. Harnett. Brit J Cancer
3:433, 1949.
(33) Maliphant, R. G. Brit M J 1:978, 1949.
(34) Dorn, H. F. Human Biol 15:73, 1943.
(35) Lombard, H. L. and E. A. Potter. Cancer
3:960, 1950.
( 3 6 ) Gilliam, A. G . J N u t Cancer Inst 12:287,
195 1-1952.
(37) Gagnon, F. A m J Obstet Gynecol 60:516, 1950.
(38) Rojel, J. The Interrelation Between Uterine Cancer
and Syphilis. Copenhagen, Nyt Nordisk, Forlag, 1953.
(39) Khanolkar, V. R. Acta U Internat contra Cancrum 6:881, 1948-1950.
(40) Vineberg, H. N. In Contribution5 to Medical and
Biologxal Research Dedicated to Sir William Osler. New
York, New York, Paul B. Hoeber, vol. 2, p. 1217, 1919.
( 4 1 ) Clemmesen, J. J N u t Cancer I n s t 12:1,
1951-1952.
(42) Gardner, W. U. Surgery 16:8, 1944.
(43) Hofbauer, J. J Obstet Gynaecol Brit Emp 46:232,
1939.
345
DISCUSSION
DR. JOSEPH NATHANSON. In order to appreciate the tremendous work which these gentlemen have presented, especially from the
standpoint of the adherence to ritual laws, I
trust you will allow me to present a resume of
them. In the first place, in no religion in the
history of the human race, have the laws, as they
are known in the Hebrew, Nidah, which means
separation from the husband, been so long existent. They were promulgated in Biblical times,
certainly over 4000 years ago, and in spite of the
very many Diasporas, they had been carried o u t
with a great degree of adherence until about
100 years ago.
Now what are these laws? In the first place,
the laws are that, whether a woman menstruates
normally five days or one day, the minimum
time during which she must call herself absolutely unclean is five days. In other words, if she
346
menstruates only one day a month throughout observed the Mosaic laws of menstruation most
her life, from the standpoint of the Mosaic laws, faithfully. Then in the decade from 1909 to
she is known to be unclean for five days plus a 1918 the rate had dropped to seven and oneperiod of seven days beyond that.
half times as frequent in non-Jews as in the
Again, any woman who has any spotting Jews. Even in Israel, there is now an appreciable
larger in extent than three-quarters of an inch incidence of carcinoma of the cervix. It may be
in diameter at any time in the month immedi- of interest and quite surprising to you to know
ately is looked upon or regarded as an unclean that, in my own 30 years of practice, I have
woman. She is analogous to the menstruating never seen a case of primary carcinoma of the
woman, and is therefore subject to the Mosaic cervix in a private patient. I have seen a few
laws of menstruation. If a woman should men- cases of stump carcinoma at the Womans Hosstruate for three days, she would necessarily pital in Jewish patients, and carcinoma of the
have an unclean period of 12 days. If she spot- cervix in a few Jewish women on another ward
ted after the first coitus, again she would have service.
to go through 12 more days. The implications
I believe that no finality can be placed upon
of such episodes are at once apparent. It re- the Mosaic laws of menstruation as an explanaduces the period of exposure to coitus, as the tion of the low incidence of cancer of the cervix
essayists have said, to a very considerable de- in the Jewish race. At the present time they play
gree. As a matter of fact, this rationing puts to rather an insignificant part I believe, and I feel,
shame the rationing which was prevalent in the therefore, that another aspect deserves serious
halcyon days of the New Deal. Some of these thought. The answer cannot come and will not
women were literally being placed in sexual come until at least two and perhaps three generostracism.
ations have passed, because few Jewish women
Although the question of contraception was are now adhering to the Mosaic laws of mennot touched upon by the essayists, I think that it struation. It is my opinion that about 95 percent
is important to discuss it. If smegma is irritat- of the Jewish women in this city are not observing, and if the male wears a contraceptive de- ing the laws of Nidah. This is not my own estivice then the irritating factor should be re- mate; colleagues who have similar types of pracduced. If the male does not use it, but the tice agree with me. Even rabbis have informed
female employs a contraceptive, is she prevent- me in recent weeks that their impression is siming the smegma from irritating the cervix, or is ilar to ours. One rabbi told me recently that in a
she in turn irritating the cervix by the contra- period of 18 years he had never been consulted
ceptive? That point would be worth investigat- by any Jewish woman in his congregation as to
ing.
how to carry out the Mosaic laws of menstruaIt is important to note that in the Jewish race tion. Thus, at the present time, I do not believe
itself, if a couple adheres to the Mosaic laws of that with the information at hand we are justisexual practices, the male cannot use any device fied in stating that the Mosaic laws per se are the
or contraceptive himself. On the other hand, important factor in the prevention of carcinoma
the female is allowed to use a contraceptive.
of the cervix in the Jewish race.
With regard to how many women really obI believe Dr. Wynder said that one of the
serve the Mosaic laws of menstruation to the more definite observations was that if women
letter of the law, or even partially, at the present married late and had coitus for the first time
time, I should like to give you some figures late in life the incidence of carcinoma is lower. Is
which the late Dr. Hiram Vineberg presented in that correct?
a paper about 30 years ago, on the same subject
DR. WYNDER. That is correct.
which is under discussion, namely, the differDR. NATHANSON. If that is so, it is paraences in the racial incidence of carcinoma of the doxical to note that the Old Testament admoncervix. During these years, 1893-1903, he ished the Jewish race to undertake early marfound cancer of the cervix in New York City was riage. We should therefore have had a higher
20 times as frequent in non-Jews as in the Jews. incidence of carcinoma in the Jews because they
That is significant, because you will remember married earlier and they had coitus much earthat was the period in which the peak of immi- lier. How then can you account for the discrepgration from Eastern Europe occurred. T h e ancy? It is my belief, and of course it is only a
majority of Jewish women during that period belief, this will resolve itself into a question in
Wynder et al.
which the geneticist will finally answer the problem. In other words, I believe, at least to a
degree, that, for some reason or other, whether
you call it biologic, or whether it is because for a
period of over 4000 years Jewish females have
observed the Mosaic laws of menstruation and
have therefore been subjected to less irritation
over generations, say a couple of hundred generations, the cervical epithelium in the Jewish
woman has become endowed with the ability to
resist neoplastic changes of a malignant type
better than can that of her non-Jewish sister,
who has not had the benefit of markedly reduced irritation of the lower genital tract.
DR. FRANK R. SMITH. Although I did not
know that Dr. Wynder was doing this work in
connection with the Gynecological Service at
Memorial Hospital, I was interested to see that
he has followed rather closely the pattern of
study that I undertook at the same hospital in
1927.
In an effort to find out why some women
developed cancer, we took a group of cancer
patients and a group of controls and used a
questionnaire.
The striking point that came from our study
was the racial difference. After I found it appearing in the figures I began looking up the
literature and found that the infrequency of
cancer of the cervix in Jewish women was mentioned in a report from the Mayo Clinic about
five years before. The other factor that came up
was the use of Lysol douches. It was astonishing
to note the number of women who used Lysol
douches in the cancer group and the practical
absence of their use in the control group. We
noticed also in the cancer group that the time
between marriage and the first pregnancy
seemed to be greater than in the controls.
The observation about the Lysol douche illustrates one of the dangers in this type of analysis.
The reason that women with cancer of the cervix used Lysol douches was because they had
leukorrhea. The controls who were taken from
another hospital did not have the necessity for
douches.
I spent four and one-half years before the war
chasing monkeys in pursuit of the answer to the
Lysol problem. Every day one of us applied
Lysol in different strengths to the vagina on
tampons. Every three months we biopsied the
cervices. It was pretty hard to get monkeys that
had had young and were in good health, but we
used about 18 in all. We were just beginning to
347
348
Wynder et al.
nificantly later age at first marriage among
them as compared to our clinic patients.
There is one other point about the histologic
variations that deserves comment. We separated
our results into epidermoid cancer, carcinoma
in situ, and adenocarcinoma. There were not
enough data available on carcinoma in situ to
analyze these statistically,but they appear to go
in the same direction as the epidermoid cancer
data. In adenocarcinoma, however, which we
did not analyze statistically, we found, as has
been pointed out previously, the incidence
among Jewish women to be the same as among
the non-Jewish. I think this reiterates the point
that the etiology of epidermoid and adenocarcinoma is entirely different.
Another point I would like to stress refers to
genetic factors and hormonal drive in respect to
cancer of the cervix. Genetic factors have often
been thought to be of etiological significance in
many types of cancers, such as liver cancer
among the Bantus, until it was shown that certain dietary factors could explain the high incidence of this cancer. Similarly, in cancer of the
lung it was suggested that women have a genetic
resistance until it was shown that the greater
amount of smoking among males could explain
this difference.
I n India, we have similar genetic backgrounds in the Hindus, Moslems, and Chris-
349
Some forms of liver cirrhosis, and more particularly some cirrhosis of sufficient severity to
be a cause of death, have long been widely spoken of as a complication of alcoholism (5).
Indeed, so close has the relation been held to be
that liver cirrhosis death rates have provided the
basis upon which nearly all alcoholism prevalence rates have been estimated (6, 7) even
though it is not known what proportion, within
very wide limits, of such deaths are due to or
associated with alcoholism (8-10).
Given the fact, however, of a strong association between liver cirrhosis deaths and alcoholism prevalence, we might well ask how close
the association is between the death rate from
cirrhosis and the consumption of beverage alcohol.
If, moreover, that association should prove to
be close and positive, then an interest in economics or public health will prompt us to inquire further as to the dependence of alcohol
consumption on the price of alcohol. It is into
these two aspects that this paper is, more narrowly, to inquire.
Data. Readily available (I) for Ontario and
Canada are (a) liver cirrhosis death rates,
(b) dollar sales values of licit alcoholic beverages
purchased, (c) gallonage by type of beverage
(beer, wine, distilled spirits), (d) population 20
years of age and over, (e) disposable personal
350
Seeley
351
Figure 1. Deaths from liver cirrhosis per 1000 deaths from all cause%in Canada, 1921-1956.
6.5
6.0
5.5
1920
1925
1930
1935
1940
1945
1950
1955
1960
ANALYSIS
352
CONSUMPTION (C)
- 150
5
t
LIVER CIRRHOSIS
DEATH RATE (L) ,'
120
- 110
- 100
135
- 90
3
- 80
25
- 70
- 60
'2
- 55
1929
1934
1939
1944
1949
1954
1959
Seeley
353
minimum-or whether the relation is curvilinear, so that death rates tend to rise both
above and below a certain consumption level.
Statistical analysis favors (although it does not
unequivocally establish) the second view. If we
recall that the consumption we are speaking
of is the official consumption (i.e., the consumption of alcohol from licit sources alone),
the conclusion would run contrary neither to
common sense nor to experience at other times
and in other places, since it is credible that below a certain level of licit consumption enough
Table 1. Canada. Average standard relative price of alcohol, average consumption of alcohol, and
average adult death rate by liver cirrhosis.
Year
Price of
alcoholb
Consumption
of alcoholc
1926
1927
1928
1929
1930
1931
1932
1933
1934
1935
1936
1937
1938
1939
1940
1941
1942
1943
1944
1945
1946
1947
1948
1949
1950
1951
1952
1953
1954
1955
1956
,047
,049
.037
.04 1
,042
,046
.049
,065
.065
,052
.049
,046
.044
.044
.040
,037
,034
,035
,039
,037
.035
,033
,028
.030
.028
,025
,025
,028
,028
.026
,024
.62
.62
.77
.85
.87
.76
.65
.46
.46
.56
.63
.67
.74
.70
.74
.82
.91
.98
.82
.94
1.20
1.33
1.46
1.41
1.44
1.47
1.46
1.44
1.46
1.39
1.51
Unstandardized
liver
cirrhosis
death rated
64.34
65.17
64.86
68.38
69.94
67.08
68.55
72.60
71.10
69.13
72.31
76.38
79.09
83.09
87.78
88.59
87.90
91.30
96.33
98.28
100.30
107.99
Standardized
liver
cirrhosis
death rateC
38.62
38.75
38.28
40.05
40.64
38.68
39.35
4 1.44
40.24
38.98
40.72
42.76
44.12
46.33
48.88
49.20
48.61
50.58
53.49
54.33
55.35
a The primary data upon which all rates shown in the table were based are to be found in Statistics of Alcohol Use and Alcoholism
in Canada, 1871-1956. All liver cirrhosis death rates were corrected to allow for the effects of the Sixth Revision of the
354
Table 2. Ontario. Average standard relative price of alcohol, average consumption of alcohol, and
average adult death rate by liver cirrhosis.
Year
Price of
alcoholb
Consumption
of alcoholc
1929
1930
1931
1932
1933
1934
1935
1936
1937
1938
1939
1940
1941
1942
1943
1944
1945
1946
1947
1948
1949
1950
1951
1952
1953
1954
1955
1956
,0400
.0409
,0471
.0589
,0605
,0555
,0434
,0386
,0342
,0342
,0327
,0298
,0273
,0254
,0256
,0278
,0262
.0280
,0265
,0234
,0234
,0225
.0213
,0209
,0203
,0222
,0194
.0183
,765
.752
,662
,515
,446
.518
.605
,764
,840
,922
,892
.908
,997
1.175
1.222
1.107
1.197
1.435
1.601
1.737
1.760
1.766
1.757
1.775
1.850
1.876
1.828
1.917
Unstandardized
liver
cirrhosis
death rated
51.30
5 1.40
55.61
61.34
63.30
61.27
60.75
61.51
58.96
57.18
60.06
66.70
74.79
80.99
89.26
94.63
92.72
94.99
102.42
106.69
108.27
1 15.49
Standardized
liver
cirrhosis
death ratee
27.56
28.30
31.45
34.24
34.78
33.64
33.33
33.46
3 1.78
30.88
32.59
36.07
40.43
43.96
48.42
51.49
50.61
51.99
56.44
58.84
59.64
-
a The primary data upon which all rates shown in this table were based are to be found in Statistics ofAlcohol Use and Alcoholism
in Canada, 1871-1956.All liver cirrhosis death rates were corrected to allow for the effects of the Sixth Revision of the
Seeley
at a minimum (31 deaths per million standard
population per annum) when consumption fell
to an average of .88 gallon of absolute alcohol
per annum. This implies 52 percent of the
deaths, given a 54 percent drop in consumption.
The foregoing relations may be presented
synoptically in Table 3.
Greater consistency than this between the
findings for Ontario and Canada, or, within
each, for the standardized and unstandardized
rates is hardly to be expected, given the heterogeneity across Canada and the different
meanings of the questions answered by the
standardized and unstandardized rates, respectively. What seems clear enough is that death
rates of both kinds in both political units are
closely related to alcohol consumption, and that
reductions by about a third to a half might be
expected if alcohol consumption were cut to a
third or a half or somewhere in between.
The prediction takes it for granted that other
alcohol supply conditions remain the same. It
is for this reason that only the period 1935-56
was chosen to correlate alcohol consumption
and liver cirrhosis death rates; before that date,
alcohol supply conditions were sufficiently different that the stricture all other conditions
being equal did not apply. (7).
From 1929-34, as a matter of fact, liver cirrhosis deaths did not vary intimately with price
and consumption (which did vary inversely for
the whole period). If, however, a longer time
period were chosen-say from the beginning of
prohibition, circa 1915, to the present, close
covariation between alcohol consumption and
liver cirrhosis death rate would be visually ob-
355
Table 3. Canada and Ontario. Alcohol consumption and liver cirrhosis death rates, standardized and
unstandardized.
Liver cirrhosis death rate
Unstandardized
Standardized
Statistic
Canada
Ontario
Canada
Ontario
0.96
92
0.97
94
0.91
83
0.96
92
66
62%
56
48%
39
70%
52%
0.60
40%
0.89
46%
0.65
43%
0.88
46%
31
356
Statistic
Correlation between price and consumption
Percentage variance accounted for
Minimum consumption expected:
(a) in gallons absolute alcohol
(b) as percentage of 1956 consumption
Price at which minimum consumption expected:
(a) as fraction of income.
(b) as percentage of 1956 price
a
Canada
Ontario
- 0.99
- 0.96
98%
92%
0.51
34%
0.49
25%
,057
238%
0.053
293%
Statistic
Unstandardized
Standardized
Canada
p
P2
p
- 0.93
-0.91
(0.83)
- 0.90
(0.81)
Ontario
P2
(0.86)
- 0.93
(0.86)
and quite another to assume a causal relationship, and moreover, a causal relationship that
can be extrapolated to the future.
There is only one way to test such a pair of
assumptions and that is by way of an experiment. If an attempt were made by a suitable
government to change (in either direction) to a
sufficient degree (say 20 percent to 30 percent)
the already, in effect, administered price of alcohol for a sufficient period (say three years), it
could be determined whether indeed licit alcohol consumption and, therewith, the rate of
death by cirrhosis of the liver changed in the
expected direction and to the expected degree.
If the price were changed in the upward direction, we might reasonably expect some saving of life as well as a gain in scientific knowledge, and the opening up of a possibility of
public health gains on the basis of precisely
tailored economic measures. It can be shown
(11)also that, by a curiously happy relationship
among the variables, the increased taxation that
would secure the higher price, and lowered consumption and death rates, would add sufficient
additional revenue to government (even in the
face of the diminished consumption) to pay for
a very substantial increase in public health programs, or indeed, in any other desired governmental program. In Ontario, for instance, a
doubling of price, despite its effect in reducing
consumption, might be expected approximately
to double the governments net revenue from
alcohol. (We might also, incidentally, learn
something as to how illicit alcohol supply and
methods of tension management other than alcohol consumption vary with the price of licit
alcohol supply.) In any case, the anticipated effect on revenue might also be regarded as a
Seeley
357
Over the past four decades North Carolina the nature of some of the social and cultural
has changed from a predominantly agricultural processes of relevance to health ( 1 1 ) . We are
state to one that is rapidly becoming indus- now engaged in a series of studies designed to
trialized. While the population of the state has test the utility of this scheme.
The present investigation, which is the first of
increased from 2 559 123 in 1920 to 4 556 155
in 1960, the proportion of the population en- the series, is concerned with testing a general
gaged in farming has decreased during this pe- proposition derived from this scheme. Accordriod from 58.6 percent to about 27 percent. ing to this proposition, recent sociocultural
North Carolina thus presents an opportunity to change will raise the probability of incongruity
study, in microcosm, some of the sweeping between the culture of the migrant and the
changes that, to a greater or lesser extent, are social situation in which he lives. Such inoccurring throughout the world today. In par- congruities as occur will place excessive adticular, North Carolina offers numerous oppor- justive burdens on the social groups in which
tunities to study the impact of industrialization the migrant interacts and on the personality
system of individual migrants. Insofar as these
on health.
The process of rapid industrialization has af- stresses are not absorbed by the small group
forded many opportunities to study the impact systems and/or the personality system, recent
of a changing physical environment on health. migrants to an industrial milieu are likely to
Fewer attempts have been made, however, to manifest increased rates of psychological,
study the health effects of the dramatic social somatic, and social ill health.
The study was conducted in a small industrial
and cultural changes that accompany industrialization, and it is this aspect of the process city of about 5000 people, situated in the westthat we find most challenging and on which we ern part of North Carolina. A little more than
are focusing a series of epidemiological studies. 50 years ago this city was a mountain village of
Despite the growing body of theory postulat- no more than 100 people. At that time a large
ing the significance of social and cultural factors national corporation located a manufacturing
in disease etiology ( l - l O ) , no unified series of plant in this village, and by deliberate manageconcepts delineating these sociocultural pro- ment policy drew its labor force almost entirely
cesses has as yet been developed. While so- from the surrounding area.
This policy continues to the present. The
ciology and anthropology have provided structural schemes in abundance, there is little population from which the labor force is drawn
agreement as to which processes are relevant to is ethnically homogeneous, being predomihealth, how many crucial processes there are, nantly of British stock and having lived in this
and how these processes are linked to health area for well over a century. Practically no migration into this region has occurred during
states.
As a potential contribution to this field, we this period.
Within the factory it was possible to identify
have presented elsewhere a generalized conceptual scheme outlining what we consider to be two groups according to the recency with which
they had undergone the changes accompanying
industrialization.
Source: Archiues of Environmental Health 3:31-39, 1961.
1. A group of first generation factory employees. This group, presently employed in the
factory, are the children of farmers, and represent the people who have most recently undergone the change from a rural folk culture to
an industrial social situation.
358
The Sample
For this initial investigation the population
studied was restricted to white male, hourly
paid employees of the factory on the active company payroll for the 40-month period, January
1956 through May 1959. There were 968 employees satisfying these criteria, from whom a
stratified random sample of 390 individuals was
drawn for study. The stratification was made on
the basis of age, absence experience, and medical examination status. This last was determined by comprehensive biennial examinations, for which all these individuals had been
eligible.3
This study population of 390 individuals was
then divided into first generation and second
generation industrial workers. The first generation workers had had neither parent employed in the factory. As there were practically
no other sources of employment other than
farming available locally, the vast majority of
the parents of these first generation workers
were, thus, farmers. The second generation
workers had had at least one parent employed
in this factory.
It should be emphasized that the sample as
chosen reflects known selective bias. All individuals either entering or leaving the population during the study period were excluded.
Thus, all deaths, retirements, disability separations, and both voluntary and involuntary separations from employment were excluded from
Three categories of medical examination status were developed: those diagnosed as having any major or potentially
serious medical abnormality, those with minor or no medical
abnormalities, and those not examined.
359
Data Obtained
For each individual member of the sample,
data were obtained on age, generation status
(as defined above), length of service, marital
status, and two general indices of health status.
The first of these was a morbidity rate based on
the number of absences of more than three
consecutive days duration attributed to illness,
and the second the number of positive responses given on the Cornell Medical Index
Health Questionnaire. It was recognized that
these were two relatively crude indices of health
status, but they were considered adequate as a
first test of the hypothesis. Our confidence in
the utility of absences as a general index of
amount of illness was increased by the company
policy requiring that all individuals having such
absences present a certificate from their family
doctor upon return to work and submit to an
examination by the in-plant medical service.
RESULTS
360
Age
Length of service
a
Mean
S.D.b
Mean
S.D.
Mean
S.D.
Mean
S.D.
30.2
5.8
2.9
4.3
42.8
16.3
4.7
7.1
28.5
6.4
3.4
4.3
39.6
16.4
4.1
6.8
Morbidity Rates
As indicated above, morbidity rates in this
study were based upon the number of absences
of more than three consecutive days attributed
to illness. From these data two rates were derived: (1) frequency rate (FR)-the number of
such lost time episodes per 1000 population per
year; and ( 2 ) disability rate (DR)-the mean
number of days lost per person per year.
For the total sample the frequency rate was
570 lost-time episodes per 1000 population per
year, while the disability rate was 13.6 work-loss
days per person per year. These rates are considerably in excess of those found in the nation
as a whole. For men over 17 years, the U.S.
National Health Survey reports disability rates
of only 7.2 work-loss days per person per year
(12). These rates, calculated for persons usually working, include all absences of any duration, whereas in the present study only absences
of longer than three days were included. If absences of three days or fewer had been included
the rates in this population would thus have
been higher. Not only are these rates higher
than the national average, but they are higher
in this Carolina division of the Company than
in any of the other manufacturing divisions of
the same Company. These other divisions, located in other states, are producing products
similar to those of the Carolina division, and the
workers are subjected to identical company policies and practices. Nevertheless, the disability
rate in those divisions is approximately 10 workloss days per person per year in comparison
with 13.6 in the Carolina division.
Possible explanations for these high rates in
terms of the findings of this study will be presented in the comment below.
Generational Dijfirences
Table 2 shows the number of illness episodes
and their duration in the two generation groups
by length of service in five-year intervals for
each of the two age groups. In Table 3 the
frequency and disability rates of these illness
episodes have been calculated for the same
groups. To aid in the visualization of some of
the relationships demonstrated in Table 3, Figure 1 presents the least squares lines of best fit
of all such lost-time episodes for each group. In
computing these lines every absence was considered for each year of length of service.
Frequency Rate
In the second generation, using a multiple linear regression model, age, length of service, and their interaction accounted for 10 percent of the variation in absence experience. The standardized partial regression coefficient indicated that the effect of length of service was approximately six
times greater than the effect of age. Analysis of covariance
demonstrated that the rate of change of absences with increasing length of service differed in the two generations, and
the difference was statistically significant (0.01<p<0.05). The
application of multiple linear regression showed no significant relationship between the variables under study (age and
length of service) and absence experiences in the first generation.
361
Table 2. Number of illness episodes and their duration by generation status and length of service in two
age groups (20-34 years and 35-50 years).
Generation
20-34 years
No. in
sample
No. of illness
episodes
No. of days
of illness
35-50 years
No. in
sample
No. of illness
episodes
No. of days
{:sr
{z
{:
(:
{;
:
of illness
L-ength of service ( y . )
All lengths
of service
1-5
6-15
60
72
113
166
2787
3333
34
36
62
108
1352
2002
26
36
51
58
1435
1331
205
53
382
83
9599
1998
13
5
14
19
256
490
91
24
198
33
5298
782
101
24
170
31
4045
726
16-31
Table 3. Frequency and disability rates of illness episodes by generational status and length of service in
two age groups (20-34 years and 35-50 years).
tieneration
All lengths
of service
1-5
6-15
565
692
13.9
13.9
547
900
11.9
16.7
588
483
16.6
11.1
599
470
14.0
11.3
323
1,140
5.9
29.4
653
413
17.5
9.8
16-31
20-34 years
Frequency rate
Disability rate
35-50 years
Frequency rate
Disability rate
{z
DisabiliQ Rate
This rate, a measurement of the amount of
time lost due to illness, shows a similar pattern
for the two generations as does the frequency
rate. As demonstrated in Table 3, first generation workers in both age groups tend to lose less
time than do second generation workers in the
early years of service, but lose more time with
increasing length of service.
Cornell Medical Index Scores
Three scores have been computed from responses to the Cornell Medical Index, based
upon the total number of symptoms reported,
505
388
12.0
9.1
362
Figure 1. Change in number of absences with length of tional differences have been restricted to the
service by generatlon status and age groups (least scores of individuals who had no absences or, at
squares line of best fit).
l2O0I
9001\,
,,,,.
2nd Generation
3001
34 years or under
's.
35 years or older",,.
10
15
20
25
30
Table 4. Mean total Cornell Medical Index scores by generational status, age, and absence levels.
Age in years
21-30
All ages
All absence
levels
0 or 1
absence
2 or more
absences
31-40
41-50
Mean
Mean
Mean
Mean
390
265
125
13.5
14.4
11.8
80
31
49
11.9
12.8
11.3
160
103
57
13.1
14.0
11.6
150
131
19
14.8
15.0
13.4
2d gen.
233
162
71
10.7
11.4
8.1
37
16
21
6.9
9.3
5.1
104
70
34
9.9
10.6
8.4
92
76
16
12.4
12.6
11.3
Total
1st gen.
2d gen.
157
103
54
18.2
19.0
16.6
43
15
28
16.2
16.5
16.0
56
33
23
19.1
21.1
16.3
58
55
3
18.7
18.4
24.3
Total
lst gen,
2d gen.
Total
lst gen'
363
Table 5. Mean number of organ systems involved in Cornell Medical Index by generational status, age,
and absence levels.
Age in years
All ages
31-40
21-30
41-50
6.3
5.5
6.3
6.8
0 or 1 absence
Total
1st gen.
2d gen.
5.5
5.9
4.7
4.0
4.9
3.4
5.3
5.5
5.0
6.3
6.4
5.9
2 or more absences
Total
1st gen.
2d gen.
7.6
7.8
7.2
6.8
6.7
6.8
8.2
8.6
7.6
7.6
7.6
7.8
Table 6. Mean score for emotional componentPof Cornell Medical Index by generational status, age,
and absence levels.
Age in years
All ages
21-30
31-40
41-50
2.9
2.8
3.2
2.9
0 or 1 absence
Total
1st gen.
2d gen.
2.1
2.4
1.4
1.6
2.1
1.3
2.0
2.2
1.5
2.5
2.7
1.5
2 or more absences
Total
1st gen.
2d gen.
4.0
4.1
3.8
3.8
3.8
3.8
4.7
5.3
3.8
3.5
3.5
3.8
a Positive responses to sections M-R of the Cornell Medical Index (symptoms relating to disorders of mood, feeling, and
behavior).
14
6
4
TOTAL
C.M.I.
SCORES
r:!;t
G
e __*---r y
___---- EMOTIONAL
/---
/.*
2nd Generation
21 - 30
COMPONENTS
OF C.M.I.
1st Generation
31-40
AGE
41-50
364
Figure 3. Percentage distribution of total Cornell M e d i d Index mcorea at quartiie positions for total population.
SECOND GENERATION
0 OR 1 ABSENCE
35 OR OLDER
QA
60
FIRST GENERATION
0 OR 1 ABSENCE
34 OR YOUNGER
SECOND GENERATION
0 OR 1 ABSENCE
34 OR YOUNGER
60
52 %
47%
40
40
20
20
0-6
O
h
60
67 /o'
FIRST GENERATION
0 OR 1 ABSENCE
35 OR OLDER
FIRST GENERATION
2 OR MORE ABSENCES
34 OR YOUNGER
SECOND GENERATION
2 OR MORE ABSENCES
34 OR YOUNGER
61 ?h
QA
60
7-17
10+
FIRST GENERATION
2 OR MORE ABSENCES
35 OR OLDER
0-6
7-17 10+
SECOND GENERATION
2 OR MORE ABSENCES
35 OR OLDER
50Yo
44%
40
40
20
20
0-6
7-17
10+ 0-6
7-17
10+
0-6
7-17
10+
0-6
7-17 10+
365
the first generation workers and the social situation in which they now live. The assumption
would be that their system of values, attitudes,
and framework of knowledge with which they
were equipped during their formative years was
appropriate to the social situation in which they
were reared, the folk society of a mountain
cove. It is inappropriate to the industrial situation in which they now work. Consequently, increasing length of service does not increase adaptation to the factory situation as easily as it
does in their second generation colleagues. For
these the change from home to factory represents far less culture change in that their rearing has equipped them more adequately to anticipate and adjust to the demands made upon
them by factory life.
The Carolina division, being situated in a
rural area, is likely to have a greater preponderance of workers who have recently undergone this culture change than are the divisions
of the Company which are situated in urban
centers. The majority of industries in the country are also likely to have fewer new industrial
workers on their payroll than is the case in this
factory. In terms of the hypothesis developed in
this paper these factors would account for the
disability rates being higher in this population
than in other divisions of the Company or the
nation as a whole.
Despite the consistency of the findings of this
study with our original conceptual scheme we
recognize that they could be accounted for by
various alternate hypotheses. Of these the possibility of selection factors is one of the more
difficult to discount. By selection factors we
mean the possibility that first generation workers differed from those in the second generation, not because of the recency of their culture
change, but because they were sick or more
predisposed to illness before seeking employment in the factory. This selection may have
occurred either because first generation workers were representatives of a sicker population
than were the second generation, or because
those first generation individuals who left the
farm to seek employment were the more maladjusted members of their group. To determine
definitively whether such selection factors were
accounting for our findings, it would be necessary to design a prospective study of the families of potential first and second generation employees and examine such employees before
366
367
The object of this study was to test the hypothesis that young male patients on first ad-
368
Method
The General Register Office receives a card
giving, inter alza, the name, date of birth, occupation and diagnosis of every patient admitted to a mental hospital in England and Wales.
The Registrar General kindly agreed to draw a
sample of these cards and then to search the
birth records for the recorded details of each
patient at birth. These include the occupation
of the father at the time of the patients birth. It
would therefore be possible to establish the social class of the patient at the time of his admission to hospital and the social class of his father
at the time of the patients birth, and to compare the two distributions.
The sample was drawn from first admissions
in 1956 of men aged 20-34 years at the time of
admission. This age group was chosen to correspond with the age group of their fathers when
the patients were born. It was intended to deal
particularly with patients aged 25-34 years who
could be presumed to have reached a reasonably final occupational status, the normal age
of apprenticeship and studentship being well
past.
Since the cards were filed alphabetically the
sample was drawn by selecting all the cards for
surnames beginning with the letters A to H, a
Results
In Table 1 the social class distribution of patients aged 25-34 years at the time of their first
admission is compared with the 1951 Census
distribution of all occupied men of the same
age. The distribution shows the usual excess of
patients in social class V (90 observed against 39
expected). The Table also compares the social
class of the fathers at the time of the patients
birth with that of all occupied men aged 20-44
at the 1931 Census, i.e., it is assumed that the
fathers were in that age group and that the
patients were born around 1930. This comparison shows that the social class of the fathers
is very similar indeed to that of the population
as a whole and thus corroborates the hypothesis.
I
I1
111
Total
Not stated
a
Observed
Expected
Observed
Expected
12
21
178
12
44
203
55
39
14
42
192
66
55
8
42
192
68
59
353
18
353
369
369
52
90
IV
369
370
Discussion
The results of this documentary study show
that the patients aged 25-34 were born into
families with a social class distribution very similar to that of the population as a whole. Thus
the socio-economic environment in which these
schizophrenic patients grew up seems unlikely
to be grossly abnormal and it is unlikely therefore to be a major factor in the development of
the disease. This finding is very different from
that of Hollingshead and Redlich (5)and possible reasons for the difference are discussed
later.
The documentary study leaves many questions unanswered. The clinical study, with fewer
cases, attempts to answer some of these, and to
clothe the skeleton of the documentary study.
Thus it was designed to show whether, when a
more rigorous diagnosis of schizophrenia was
used than was possible in the documentary
study, the families still showed a normal social
class distribution; and to explore in detail the
differences in kinds of occupation followed by
fathers and sons. Moreover, the clinical study
was designed to describe the occupational and
educational history of the patients and to find
Part I demonstrated that the fathers of a national sample of young male schizophrenic patients had at the time of the patients births an
occupational distribution very similar to that of
the general population, whereas the patients
themselves showed an excess in social class V.
This finding suggests that social drift occurs,
since a considerable proportion of the schizophrenic patients in semi- and unskilled work
were not born into social class V homes. However, the documentary study, based on two
points only in a chain of events, cannot throw
much light on the processes which may account
for the drift. For example: are the fathers of
schizophrenic patients somewhat inadequate
and unstable people who tend to deteriorate in
their own work performance later in life and
drift into inferior neighborhoods, thus limiting
the opportunities of their children? Is social
inadequacy noticeable in other male relatives in
the family, in the brothers, the grandfathers
and uncles; or is the drift confined to the
schizophrenic sons?
To learn more about the processes whereby
schizophrenic patients are concentrated in the
lower social classes, we made an intensive study
of a series of consecutive admissions of men
under 30 at two mental hospitals.
Hospital A receives patients from two contrasting boroughs. One is a rapidly expanding
suburban community of increasing affluence in
Greater London, the other a working-class dis-
I
I1
Ill
IV
V
Total
I1
I11
2
3
8
-
7
7
16
4
4
2
10
105
23
43
13
38
183
1v
Total
25
19
19
24
4
24
12
21
178
50
90
65
52
35 1
The Samples
In Hospital A we studied all males under 30
admitted between January, 1958 and December,
1960 whose parents lived in Great Britain. This
sample will be called the A sample. Since we
wanted to investigate inter-generation as well as
personal social mobility, and to compare these
findings with local census data, foreign-born
patients, and also Irish patients whose families
were resident in Eire, were excluded from intensive study. Severely subnormal patients and epileptics were also excluded. On all these excluded cases as much information as possible
was gathered from hospital records and from
the patients themselves, though not from the
parents. (A detailed description of the patients
excluded from the study is given in the Appendix.)
In Hospital B we studied all male patients
under 30 admitted between March 1959 and
April 1960, similarly excluding foreign-born
and Irish and severely subnormal and epileptic
patients. This sample will be called the B sample.
Diagnosis
Psychiatrists at the two hospitals kindly undertook to allocate the patients in the study,
after a suitable period of observation, to three
diagnostic categories: definitely schizophrenic
(S), possibly schizophrenic (PS), and definitely not schizophrenic (NS). In 1961 a follow-
37 1
372
Hospital A
Age
PS
NS
Total
15-19
20-24
25-29
4
14
34
6
5
5
9
16
8
19
35
47
4
16
22
Total
52
16
33
101
Mean age
25
22
22
PS
NS
Total
0
1
4
4
12
9
20
35
42
20
64
25
21
24
Since we are mainly concerned with the problems of social mobility in schizophrenia, we shall
disregard the possibly schizophrenic group as
their diagnosis is uncertain.
Method of Study
The patients were seen in hospital soon after
their admission and as much school and work
history as possible was obtained from them. Relevant details from the patients history were extracted from the hospital records. With the patients and psychiatrists permission the parents
were then visited. The fathers occupational histories were explored in considerable detail, and
we also tried to get information on the main
jobs of the paternal and maternal grandfathers,
the paternal uncles, and the siblings of the patients. The patients own school and work histories were discussed in detail. Follow-up visits
were paid to parents or other key relatives during 1961 and special attention was again given
to the employment situation of the patients. In
addition to the information gathered in interviews, school reports were obtained on most
patients, and the General Register Office kindly
sought their birth records and found them in all
but a few cases.
RESULTS
Table 4. Social class distribution of schizophrenic patients on first admission and of their fathers at
patients birth (clinical sample of schizophrenic male admissions to two mental hospitals).
Fathers at patients birth
Patients at admission
Hospital A
Social class
Hospital B
Hospital A
Observed
Expected
Observed
Expected
Observed
8
23
13
I V and V
2
21
21
16
24
4
21
15
Total
44
44
40
40
I and I1
111
Students
School
N o occupation
Not known
a
1
1
Hospital B
Expectedb
Observed
Expectedb
28
15
8
26
16
1
15
21
4
19
14
50
50
37
37
From 1951 Census distribution of all occupied males in their area of residence.
From 1931 Census distribution of all occupied males in their area of residence.
373
Observed
Expected
I and I1
I11
IV and V
3
22
16
4
21
16
Total
PJ.K.
41
1
41
Table 6. Social class distribution of fathers of schizophrenic patients; (1) at patients birth, (2) main job,
(3) present or last job (clinical sample of schizophrenic male admissions to two mental hospitals),
~
~~~
Hospital B
Hospital A
Job at
patients
birth
Present or
last Job
Main
Job
(3)
(2)
(1)
Job at
patients
birth
Present or
last Job
Main
Job
(2)
(1)
(3)
No
No
No
No
No
No
14
56
30
15
25
12
29
48
23
13
IV and V
7
28
15
22
16
26
43
31
I
15
21
3
40
57
3
22
16
7
54
39
4
17
17
10
45
45
Total
50
100
52
100
51
100
37
100
41
100
38
100
Social class
I and I1
111
Not known
374
Table 7. Social class mobility of fathers of schizophrenic patients in three areas (clinical sample of
schizophrenic male admissions to two mental hospitals).
Hospital A
Suburban borough
Fathersjob
at patients
birth
Working-class borough
Fathers job
at patients
birth
Fathers
main job
No.
No.
I and I1
111
IV and V
7
18
4
24
62
14
Total
29
100
Not known
Social class
Hospital B
No
11
48
52
2
12
9
9
52
39
1
15
21
3
40
57
3
22
16
7
54
39
21
100
23
100
37
100
41
100
45
45
10
10
29
100
I and I1
111
I V and V
9
49
42
9
52
39
100%
100%
Fathers
main job
No.
13
13
3
East
London
Fathers job
at patients
birth
No.
Working-class
Borough
Fathers
main job
No.
Social class
East London
7%
-
375
Table 9. Work histories of fathers in social classes I and I1 (main job) and their housing conditions
(clinical sample of schizophrenic male admissions to Hospital A).
Patient
No.
s2
s3
s9
S15
s20
s2 1
S24
S26
527
S28
s35
S38
s45
S48
S52
Main job
Housing conditions
Council house
Flat tied to shop
Flat tied to off-license
Own house
Own house
Own house
Own house
Own house
Own house
Own house
Own house
Own house
Own house
Own house
Own house
a Although the patients birth certificate could not be found, the occupational history taken at the interview indicates that the
father was the manager of a pub at the time of the patients birth.
Table 10. Social class of fathers, paternal and maternal grandfathers, brothers and schizophrenic
patients (clinical sample of schizophrenic male admissions to Hospital A).
Father
Main job
Paternal
grandfather;
main job
Last job;
1st admission
No.
No.
No.
13
23
10
28
50
22
10
26
15
20
51
29
23
58
11
25
63
12
8
22
21
56
23
2
21
21
4
48
48
46
100
51
100
92
100
39
100
44
100
No.
I and I1
IV and V
15
25
12
29
48
23
Total
52
100
Brothers;b
main job
No.
No.
School
Student
Nil
Not known
Paternal
uncles;a
main job
Social class
111
Maternal
grandfather;
main job
Patient
19
3
3
1
In two families no information was obtainable on the number and jobs of paternal uncles.
46 brothers were related to 30 patients
376
Table 11. Type of school attended by schizophrenic patients and their brothers (clinical sample of
schizophrenic male admissions to two mental hospitals).
Hospital B
Hospital A
Type of school
No.
Grammar
Technical
Secondary m o d e r n or
equivalent
13
5
25
10
20
1
2
2
5
9
3
30
58
34
74
32
76
32
ESN
Maladjusted
2
-
10
2
5
15
8
23
52
100
46
100
42
100
64
100
Other
Not known
Total
a
No.
No.
No.
14a
50
-
6 out of the 9 (9 percent) attenders at Grammar School belong to one family living outside London
dle thirties attaining selected secondary education (7). Furthermore, the schizophrenic patients in the sample compare favorably with
their brothers, of whom only 24 percent gained
places in grammar and technical schools. Yet, as
we saw before (Table lo), the brothers achieved
better occupational status than the patients. In
Hospital B there is no suggestion that the patients did better than the brothers at school.
Up to adolescence, then, there appears to be
no indication that future schizophrenic patients
lag behind in their achievement. However, for
some of the boys the difficulties related to work
and achievement seem to have started in adolescence, as the careers of the grammar school
boys illustrate well (Table 12). Several were reported by their headmasters as having been below average. One boy obtained four A level
passes, but even he failed to get into the University of his choice, slowly deteriorated during his
course of study, and failed in his finals. Another boy embarked on an academic course
with great difficulty and eventually failed. Two
attempted an Arts course, and one trained for
the priesthood but had to abandon it in his
fourth year. Only one patient qualified in a profession and is now practicing it. Thus with one
exception, none of the grammar school boys in
the sample completed any kind of professional
or technical training and most of them ended
up in routine clerical or laboring jobs. There is
some evidence that these patients, despite their
innate ability, seek routine jobs with strictly
limited responsibility. Several were offered promotion; two refused, and two actually tried to
377
Table 12. School and job history of grammar school boys (clinical sample of schizophrenic male
admissions to Hospital A).
Patient
Father
s3
11
Vth A stream; 2 0
levels. Left 16-112.
s4
Nil.
Clerical jobs; little unemployment despite many admissions. Now shop assistant.
55
(mother)
I1
Vth C stream; 3 0
levels. Left 16%.
Nil.
s9
I1
Below average. No
G.C.E. Left 16.
RC Seminary 4 years.
Failed.
s12
I11
S13
111
Apprenticed carpentry.
S15
11
Dentistry.
S24
Crammers college.
Failed matric. twice.
Breakdown after first
part finals.
S26
Assistant in bookshop. 7
years at very low salary. Clerk
after discharge. Now lorry
drivers mate.
S27
S31
111
Nil.
S40
I1
Nil.
S4 1
111
S48
I1
4 A levels.
Laboratory technician.
No.
S.C.
Further training
School report
Job history
378
The picture changes considerably as the patients illness continues; it may become quiescent, erupt from time to time in acute episodes
or be a long, drawn-out, continuous illness.
379
Table 13. Main jobs held by the schizophrenic patients before first admission and at follow-up (clinical
sample of schizophrenic male admissions to Hospital A).
Job at end of survey
period
Students
Art
Dentistry
Law
Theology
Science
Dentist
Apprenticed
trades
Instrument maker
Marine engineer
Mechanical engineer
Shipwright
Welder
Shipwright
Turner
Welder
i}
Other
skilled
trades
Painter
Plumber
Lettering artist
Machine minder (printing)
Butchers cutter
Carpenter
Electrician
q11
Clerical and
selling
Clerk
Sales representative
Shop assistant
Semi-skilled
Capstan turner
Machiner (tailor)
Lathe operator
Laborers
Factory
Outdoor
N o occupation
School
At home
112
q3
114
Radio repairer
Laboratory technicians
Painter
Trainee telephonist
1
1
Clerk
Shop assistant
Glue spreader
Seaman
Loader, British Railways
Presser
Machiner (tailor)
Glass worker
Despatch checker
Platen feeder
Outdoor
At home (over 2 months)
In hospital
52
15
l2 127b
52
a These jobs refer to the patients main activities before 1st admission, i.e., that be held longest, and is not comparable with
earlier social class data which referred to the last job held before 1st admission.
Of the 15 patients in hospital at the time of follow-up, 10 had been in hospital more or less continuously since the sample
admission and 5 had varying periods of employment since their sample discharge. Of the 12 who were unemployed at home, 2
had been working for a short time since discharge.
380
Table 14. Main jobs held by non-schizophrenic patients before first admission and during follow-up
period (clinical sample of not schizophrenic patients admitted to Hospital A).
Follow-up
First admission
Students
Art
Apprentices
Electrical engineering
Engineering
Training course
instrument making
Manager shoe shop
Sales manager
Assistant hotel manager
Managerial
Posts
Skilled
Trades
Welder
Turner
Painter
Bricklayer
French polisher
Engraver
Plumber
Clerical
and
Selling
Service
Steward
Waiter
Sales representative
Shop assistant
Clerk
Semi-skilled
Machine minders
Furniture sprayer
Drivers mate
Laborers
; } 3
1
Welder
Turner
Painter
Bricklayer
P.O. technician
Lorry driver
Waiter
Shop assistant
Clerk
Factory
Outdoor
No occupation
}
I
lo
33
Electricians mate
Guillotine operator
Machine operator
Sprayers
Window cleaner
Mortuary assistant
Factory
Outdoor
Not known
Total
q1
Not known
1
33
38 1
England and
Wales (%)
1 and I1
10.7
23.1
53.8
12.4
15.6
57.8
15.6
11.0
111
IV
V
382
Occupational Factors
The observation in the clinical study that the
potential candidates for classes I and I1 have
done as badly as the patients in classes IV and
V, while the patients in class 111 carrying out
jobs involving some degree of repetitive skill
seem to manage better, raises some speculations.
1. Perhaps a job which gives the patient a
framework within which he has to carry out
certain operations demanding little initiative or
control of others, and little intellectual or emotional effort, is peculiarly fitted to the needs
and capacities of the potential or actual schizophrenic patient.
2. The failure of students and of the few who
tried to runbusinesses suggests that the disciplined intellectual effort and application required in most academic studies, advanced apprenticeship courses or managerial jobs require
more ego-strength, control and initiative than
even the most intelligent and gifted young
schizophrenic patients can command.
3. It seems that the unskilled worker,
whether he is in this position from the beginning of his working life or as a result of occupational drift, has little margin for maneuver and
is thus at much greater risk than a skilled
worker of dropping out of the labor market
altogether. It is also possible that the unstructured nature of most unskilled work gives minimal support to a patient who appears to need
something in the nature of a repetitive stimulus.
Many forms of unskilled work may thus be distressful and unhelpful to schizophrenic patients.
These considerations are of course additional
to the many other factors which must influence
the patients capacity to work: for example, the
level of his intellectual and personality development before he becomes ill, the severity of the
illness itself, as well as the degree of residual
disability and the nature of his immediate environment.
Further Studies
Certain pointers towards studies in secondary
prevention of schizophrenia, i.e., by early detection and treatment, emerge.
1. Special attention might be devoted to adolescent schoolchildren and to students who deteriorate in their performance and/or who fail
in examinations without any apparent reason.
2. Further retrospective and prospective
studies of schizophrenic students and young
men already established in professional or managerial jobs may throw light on the factors favoring survival or leading to downward drift. Such
studies may also help to clarify the rather different findings in the Bristol and the present study
regarding the fate of class I and I1 patients.
3. Studies of schizophrenicpatients at work in all
classes might be as useful as studies carried out
among patients in their family settings. While
these studies will not help with the problems of
etiology, they might specify those conditions of
work which are suited to the needs of schizophrenic patients a n d those which seem
harmful.
383
Hospital A
PS
NS
Students
1
8
6
2
10
15
-
N o occupation
14
25
1
24
20
25
Total
S
-
Hospital B
PS
NS
NK
1
4
1
1
4
1
28
24
6
24
14
22
2
26
5
9
15
5
1
35
3
3
29
10
5
50
stantiated, it would clearly be a fact of the greatest importance, particularly although not exclusively with respect to theories of virus
etiology. I have tried elsewhere (4) to analyze
the methodological and conceptual problems of
the space-time cluster, particularly when we are
dealing with a low intensity of events, as we are
in leukemia. Briefly, it is proposed that the examination of such events requires a separate
examination for the three components of epidemicity: (a) concentrations in space, over the
whole of the time of the study: (b) concentrations in time over the whole of the area of the
study; (c) interactions between space and time
concentrations. Examination for the last component amounts to a search for movements of
high concentration areas and the method proposed is the examination of all possible pairs, or
a selection of them, to see whether short geographical distances are positively correlated
with short time intervals.
The present paper is an analysis, both in
these terms and by more orthodox methods, of
the space and time distribution of childhood
leukemia in the North of England over a period
of 10 years.
MATERIAL AND METHODS
Knox
the Appendix; (b} scrutiny of the records of the
regional Cancer Registration Bureau with secondary reference to the hospitals which registered cases not already ascertained; (c) examination of certificates of death due to leukemia in
the years 1951 to 1960 up to the 17th birthday,
with secondary reference to the hospital notes.
When death had occurred in hospital this last
examination was done directly, and when death
had occurred at home by telephoning the doctor who signed the certificate and finding out
which hospital had undertaken investigation
and treatment.
In this way we traced 185 cases, probably
every case in the region. However, the details
were not uniform. There were four coroners
cases for which there were no hospital notes and
for these the day of death was accepted as the
date of onset. There were also two cases whose
notes were lost or destroyed but for whom we
obtained from the admission index the date,
age, and address at the time of the first admission for leukemia. In these cases we used the
address given and accepted as the day of onset a
date one month before admission. Finally, there
was a small group of eight patients, who were
treated in hospitals which did not reply to our
requests, or who died at home and for whom we
could not trace or could not read the name of
the doctor signing the certificate and could find
no record of a hospital admission despite a
search of the alphabetical indexes at the likely
places. For these few we accepted the diagnosis
on the certificate, the home address as stated,
and a date of onset six months before the date
of death.
The date of the first symptom mentioned in
the history taken at the first admission was accepted as the date of onset. This was usually
reasonably precise, being related for example to
the onset of pallor or bruising, or to a fairly
abrupt onset of malaise and anorexia followed
by other symptoms or failure to recover from
infections such as otitis media. In a few cases it
was more difficult as in one case when it was
superimposed upon a pre-existing acholuric
jaundice. In five children in whom the leukemia
was a terminal phase of a lymphosarcoma, the
onset of the primary illness was used as the date
of onset.
In each case an exact if sometimes arbitrary
date was chosen. If the first symptom was recorded one month ago, exactly one month
before the admission date was accepted. Ob-
385
RESULTS
386
Ly mphoblastic
Cytology
M yeloblastic
{ E:;
10 11 12 13 14
1
3
6 8 7 1 6 8 6 9 5 6
5 5 1 1 7 4 6 2 2 1
Total
4
0
3
0
0
1
5
0
88
51
4
1
2
0
2
0
3
2
33
13
8 8111710 9 5 7 7 7
6 8 8 1 2 5 5 7 3 2 4
7
6
3
3
2
2
3
0
5
5
109
76
14 16 19 29 15 14 12 10 9 11 13
3 10
185
3 3 1 4 1 1 1 3 1
0 0 0 2 2 1 0 0 1
Cytology
Place
Lymphoblastic
Urban
Rural
37.5
36.6
18.1
8.3
M yeloblastic
Urban
Rural
9.6
4.8
9.8
5.1
Knox
387
Ly mphoblastic
10
11
12
Total
12
10
13
23
10
10
14
14
10
139
46
16
13
11
11
16
27
13
15
19
21
14
185
M yeloblastic
Total
Month of onset
6
7
Cytology
Season
of
onset
{ May-October
November-April
Summer:Winter ratio
3 4
10
I1
12
9 8 1 2 1 6 6 8 5 3 4 3
2 5
6 7 6 4 6 4 3 1
3
5
2
2
1
2
13 14
1
0
3
2
Total
84
55
388
1951 '52
Year of onset
onset
May-October
November-April
8
3
5
1
'53
'54
'55
'56
'57
'58
'59
1960
Total
5
3
2
3
3
3
5
1
3
0
6
3
8
1
3
2
48
20
Time apart
(days)
Total
0-59
60-3651
0-1
Over 1
Total
5
20
147
4.388
152
4.408
25
4.535
4.560
Knox
389
OTHER FACTORS
DISCUSSION
390
~~
Adjacent cases
1 to 4 Intervening cases
5 to 8
All possible pairs
Expected numbers in parentheses
<I
8( 4.6)
20( 18.1)
14(17.3)
221
-2
Total
-4
-8
5
35
34
28
118
101
47
170
182
6
23
19
94
366
350
445
1423
2094
282
4465
>S
Knox
391
392
APPENDIX
Royal Victoria Infirmary, Shotley Bridge Genera1 Hospital, South Shields General Hospital,
Stockton Childrens Hospital, Sunderland Childrens Hospital, Walkergate Hospital, West
Hartlepool General Hospital.
393
394
SINGAPORE
LAGOS, NIGERIA
TANGANY I KA
80
HONG KONG
ALEXANDRIA, EGYPT
70
CARACAS. VENEZUELA
ROME, ITALY
MELBOURNE, AUSTRALIA
CHRIST CHURCH, N. ZEALAND
NEW YORK, U.S.A.
AMSTERDAM, HOLLAND
LONDON, ENGLAND
GLASGOW, SCOTLAND
MONTREAL. CANADA
- \
60
50.
40-
1
2
3
4
MULTIPLE SCLEROSIS DEATHS PER 100 000
POPULATION
aModified from Limburg. C. Multiple Sclerosis and the Demyelinating Diseases. Williams and Wilkins, 1950, Chapter II.
cover cases of multiple sclerosis and allied disorders. Duplicates were consolidated, the latest
diagnosis and living or dead status was determined, and the prevalence rates were computed
for a date about one year prior to the study to
partially compensate for the long delay from
onset to diagnosis. Table 1 shows that multiple
sclerosis prevalence was much greater in the
northern cities than in New Orleans. Although
there appears to be a north to south gradient,
there was no obvious focalization of cases within
these cities. There were so few conjugal or familial cases that ordinary transmissibility or a
common source of exposure seemed unlikely.
In spite of the differences in prevalence, the
clinical features of the disease were similar in all
cities; however, the mean age of onset appeared
to progressively decrease as the prevalence rate
395
Figure 2. Average mortality rates/lOO,OOO population, Canada and the United Stater, for white population,
1949-1951.
Latitude
"N
Temp. mean
Jan.
"F
-3.5
28
31
4Za (40)b
Denver
50
42
40
Francisco
New Orleans
37
30
50
55
30
13a (6)b
Community
Winnipeg
Boston
Sail
41
38
Preliminary result.
Final result after detailed neurological examination of
patients.
a
396
Winnipeg,
m
s- 110
9-I
100
90,
a
80Z
70-
0
0
::I
6050-
New Orleans
White
40U
a
v)
v)
10
AGE IN YEARS
%ased on probable cases known to be living in New Orleans or Winnipeg on January 1,1949.
bNew Orleans population estimated, 1947: Winnipeg population from 1946 census.
to animals, vaccinations and inoculations received prior to onset, previous illness, back injuries, and head injuries.
Ethnological differences appeared to be of
little significance. In Winnipeg there was no
difference in prevalence for persons of English,
French, or Ukrainian descent. In the earlier
study in Boston, the rate among Caucasians and
Negroes was similar and was considerably
higher than the corresponding rates among
Caucasians and Negroes in New Orleans.
In recent years, other population surveys
have been completed in Ireland, Scotland, Denmark, Canada, and the United States (1, 6-8).
These surveys also attempted to enumerate all
the living, medically attended multiple sclerosis
patients in a specified population. In most instances, patients were examined by the investigator or a neurologic consultant and the diagnostic criteria of multiple sclerosis which were
expressed or inferred are believed to be sufficiently alike to justify a comparison.
The results of these and many other investigations could be summarized as follows: multiple sclerosis is found in many parts of the world
and among all major racial groups. In the
Northern Hemisphere it is more prevalent
among the people of northern Europe, Great
Britain, Canada, and the northern United
States where the rates range from about 35 to
65 per 100 000 population. In these areas the
rates are in the high to moderately high range,
and it is difficult to see any pattern which would
suggest a difference in risk for the northern
Europeans as compared to those in Canada and
the northern United States. The rates are generally higher (within the same geographic latitude) in the smaller communities, but this is
probably due to the greater intensity of the
study which is possible in such homogeneous
populations. Among the people of the tropics
and subtropics, the rates are considerably lower
than in the temperate zones.
The one area in the Northern Hemisphere
which does not show the north to south gradient is Japan (9). An effort was made to conduct a survey, using the same techniques as
used in the United States, in two Japanese communities of over 300 000 population separated
by 10 latitude. The results failed to show a
variation in multiple sclerosis prevalence rates
which were approximately two per 100 000 in
each city. This may be an accurate picture, but
the low rates may also be due to a number of
limitations and difficulties inherent in the
study. Neurology has only recently developed as
a separate medical discipline in Japan and case
finding is always influenced by the mode of
medical diagnosis and the differences in neurologic training of physicians. We prefer to consider these rates as minimal values until further
surveys now in progress are completed.
Data on the Southern Hemisphere are
limited but Acheson (10)has studied mortality
rates of multiple sclerosis in Australia, New Zealand, and South Africa and found a relationship
to latitude similar to that in the Northern
Hemisphere. Dean (11) had first observed that
multiple sclerosis is far more common among
European immigrants to South Africa than
among persons of European stock born and
raised in South Africa.
Deans more recent studies (12) show that
multiple sclerosis does occur in South African
born persons who have not been overseas, but
the rate is very low-about two per 100 000.
This area is about the same distance from the
equator as New Orleans and the rates compare
quite well with those of similar latitudes in the
Northern Hemisphere. Among the immigrants
from European countries the prevalence of
multiple sclerosis is nearly as high as in the
country of their origin. Deans most interesting
finding is that the prevalence of multiple sclerosis in South African born persons who visit
Europe is about seven and a half times greater
than in South African born persons who do not
leave Africa.
The phenomenon of mass migration offers
unique opportunities for investigating the question of geographic influence on multiple sclerosis. First Rozanski (13) and later Alter ( 1 4 )
took advantage of the unusual population composition of Israel to study this question.
A large segment of the Jewish population of
Israel has been exposed to the influences of
diverse climatological and geographic condi-
397
398
City
Seattle
San
Francisco
Los Angeles
Multiple
sclerosis cases
Degrees north per million
latitude
population
Average
annual
sunshine
hours
48
31
2049
38
34
12
4
2935
3217
399
400
tality rates in the temperate zones of both hemispheres and with decreasing rates in the subtropics and tropics. This difference is associated
with geography rather than with race or national origin. No specific exogenous or genetic
basis for the geographic pattern has been identified, but it is speculated that some climatologic
condition influences the frequency of the disease. It is unknown whether this effect is a direct one on the patient or an indirect effect on
the animal or plant life in his environment.
Migratory populations have been especially
useful in this research and indicate that the rate
among those migrating from a high to a low risk
area exceeds that of the population into which
they have immigrated. In the studies of migrating populations, the average minimal latent period from presumed exposure in their prior
home t o the onset of symptoms has been estimated to be about 9 to 12 years.
It will be essential to clarify many of the features referred to above. Communities on the
same latitude, but having different climates with
respect to temperature and sunshine, should be
compared. Similarly, investigation of communities having different altitudes could be rewarding.
It may well be that the causes of multiple
sclerosis are determined by several factors, including a genetic predisposition, and the importance of each factor may vary in different geographic areas. It is still reasonable to expect that
future etiologic hypotheses should be consistent
with the known geographic distribution.
References
(1) McAlpine, D., N. Compston, and C. Lumsden.
Multiple Sclerosis and the Demyelinating Diseases. Baltimore, Williams and Wilkins, 1955.
(2) Limsburg, C. The Geographic Distribution of
Multiple Sclerosis and Its Estimated Prevalence in the
United States. Chap. I1 in Multiple Sclerosis and the
Demyelinating Diseases. Baltimore, Williams and Wilkins, 1950.
(3) Kurland, L. T. and K. Westlund. Epidemiological factors in the etiology and prognosis of multiple sclerosis. Ann N Y Acad Sci 58:682-701, 1954.
( 4 ) Westlund, K. and L. T. Kurland. Studies on
Zoonoses, Anthroponoses,
Zooanthroponoses, Transmissible Diseases
There are a great many diseases of man, animals, and plants on Earth. The diseases of animals are grouped under the term zoonoses
(derived from the Greek zoion-animal and
nozos-disease). Many zoonoses are common
only to animals and do not occur in man (for
example, plague in cattle, pigs, etc.). Some zoonoses, however, can under certain circumstances be transmitted from sick animals to man
directly or by a vector. The vectors are various
invertebrate animals, mainly bloodsucking ticks
and insects which, on biting a sick man or animal (or a healthy parasite host) ingest a diseaseproducing agent. When they subsequently bite
a healthy man or animal, this agent is transmitted to the latter. Such diseases are known as
transmissible diseases.
By this mode of transmission of a diseaseproducing agent by vectors, certain zoonoses,
for example, rabies, etc., are communicated to
man. Also listed among such zoonoses are cases
of asymptomatic invasion by microbes which,
when transmitted to humans, exhibit obvious
virulence.
The diseases in this group are called zooanthroponoses or anthropozoonoses, i.e., diseases
common to animals and man (from the Greek
anthropos-man, zoion-animal, nozos-disease)
irrespective of their route of transmission to
man.
But there are also certain diseases which at
the present stage of evolution of the animal
kingdom occur exclusively in man. They are
Source: Excerpted fmm E . Pavlovsky, Natural Nzdalzly of
Trammwszble Dzseases zn Relatzon to Landscape Epzdemzology of
Zooanthropmoses. Urbana, University of Illinois Press, 1965.
0 1965 by the Board of Trustees of the University of Illinois.
40 1
402
before Leishman this parasite had been identified in Tashkent by Prof. P. F. Borovsky, who
published a paper on his discovery in the Journal of Military Medicine (1898). Therefore, cutaneous leishmaniasis began to be called Borovskys disease.
It owes its name, Penjdeh sore, to the Turkmenian oasis of Penjdeh, where this disease was
frequent. It is also widespread in the tropical
countries of the East (India, Iraq, etc.).
After the Great October Socialist Revolution,
reports began to arrive about severe and apparently infectious diseases involving the brain in
people who had stayed in wild taiga regions.
These diseases affected woodcutters, topographers, road-builders, and residents of
new villages which sprang up close to the taiga
in cleared spaces. Some cases proved fatal.
Many survivors were found to have paralyses of
various groups of muscles-of the arms, neck,
etc.-these lesions disabled them for the rest of
their lives.
These examples of human diseases in natural
surroundings have the following characteristics:
connection of a disease with a definite geographical
landscape and seasonal outbreaks of a disease (in the
warm season of the year, in spring, and early
summer, etc., Figure 1).
The circumstances contributing to the occurrence of diseases in natural surroundings
needed an explanation, even if preliminary,
which was based on certain assumptions or
working hypotheses.
In uninhabited localities, infection of man by
man is out of the question; therefore, the source
of a disease should be sought among the local
fauna. Various geographical landscapes must
apparently have the following inhabitants:
(1) causative agents of transmissible diseases,
i.e., diseases spread by the agency of vectors,
mainly bloodsucking insects and ticks. By biting
an animal or man with a disease-producing
agent in their peripheral blood, a vector also
becomes infective, receiving the pathogen in the
ingested blood; within the vector the diseaseproducing agent may either multiply directly or
pass through a part of its complex life-cycle, at
the end of which the vector becomes infective
and capable of transmitting the pathogen (i.e.,
of infecting healthy susceptible humans or animals) through a bite;
(2) animals which may be the donors of the
causative agents of these diseases to vectors;
Figure 1. Changes in disease incidence according to decades In a transmissible outbreak of tularemia. From the
monograph Tularemia by N. G . Olsufyev and G. P.
Rudnev (1960).
28
22
20
18
DECADES
11 111
JUNE
II
JULY
Ill
I
I1 111
AUGUST
L
I
IX
(3) vectors themselves, specifically those species of bloodsucking ticks or insects, which
transmit the pathogen they receive from a donor to susceptible animals called reczpzents. Under certain conditions, man can also be a recipient of a causative agent.
The study of the mysterious diseases people
are likely to be infected with in natural surroundings made it necessary to send expeditions to corresponding localities to carry out
surveys on the spot.
The basic centers of such research carried out
by the author and his associates were successively the S. M. Kirov Academy of Military
Medicine, the former A. M. Corky All-Union
Institute of Experimental Medicine, the Department of Parasitology and Medical Zoology
at the Academician N. F. Gamaleya Institute of
Epidemiology and Microbiology under the
U.S.S.R. Academy of Medical Sciences, the Institute of Zoology under the U.S.S.R. Academy
of Sciences, the Tadjik branch of the U.S.S.R.
Academy of Sciences, now the Academy of Sciences of the Tadjik Republic, and other institutions within the system of the U.S.S.R. Ministry
of Health and the Academies of Union Re-
Paulouskv
publics, particularly of the Kazakh Republic,
etc.
Guided by the above prerequisites, personal
investigations, field and experimental studies
were carried out by the author and many other
research workers during numerous scientific
expeditions. The results of this research, together with an analysis of the available scientific
literature on medicine, veterinary medicine,
and biology, led the author to summarize his
views of the natural nidality of transmissible diseases
in the form of a theory, which is basically applicable to parasitic diseases also. Moreover, the
problems of studying the focal character of certain nori-transmissible infections have also been
raised of late.
Non-transmissible diseases is the collective name
given to those infections whose agents are not
vector-borne, for instance, smallpox, diphtheria, scarlet fever, etc. Their agents may be communicable directly from a sick to a healthy person by contact, by inhaling air contaminated
with a disease-producing agent when talking,
coughing, or sneezing, (respiratory route of i n f c tion), etc.
The agents of certain diseases are spread with
and without the aid of vectors. For example, the
plague-like disease tularemia is transmitted not
only by at least 40 species of vectors but also in
the process of skinning tularemia-affected water
rats, and from contact with water containing
dead water rats; anthrax may be communicated
by wearing felt boots and coats made from the
wool and skins of animals killed by anthrax, etc.
It is characteristic that the agents of most
transmissible diseases are spread only by vectors
(yellow fever by mosquitoes, malaria by
Anopheles mosquitoes, tick-borne encephalitis by
the Ixodes persulcatus and other Ixodes ticks, etc.).
Such diseases are known as specific oblzgato-transmissible diseases. The pathogens of other diseases are transmitted by various routes, including vectors. Such diseases, the so-called facultative or potentially transmissible diseases are, for
example, enteric fever, cholera, tularemia,
plague, etc.
A number of transmissible diseases are characterized by their foci being confined to an environment untouched by man or modified by
man, intentionally or unintentionally, in the
past or at present. The foci of such diseases are
known as natural foci and the corresponding
diseases as diseases with natural foci.
403
What we call the natural nidality of transmissible diseases is characterized by the following features:
(a) The existence of any transmissible disease
depends on the successive transit of its caucatiue
agent from the body of the animal donor (sick animal, asymptomatic virus carrier, or parasite
host) to the vector body. This transmission usually
takes place when the vector draws blood from the
donor and subsequently transmits the causative agent
to the animal recipient, commonly when drawing its
blood also; the infected recipient may in its turn
become a donor,for another group of vectors, etc. In
this manner, there occurs, as we say, circulation
(spiral-type circulation, Figure 2 ) of the
causative agent from organism to organism in
one and the same or another population of the
natural focus of a disease. Ihe circulation of a
pathogen, beginning from its reception by a
vector from an animal donor and including the
time required for the vector to become infective
and infect the animal recipient which in its turn
becomes a donor of the pathogen for a fresh
vector, is called the complete cycle or tour of the
pathogen during its circulation. But such a cycle is not a complete round, since the next time
the pathogen is transmitted by a vector it goes
not to the animal from which it came originally
but to another animal of the same population
or an animal of another systematic position
(from another specific population). This consecutive transmission of the pathogen occLirs as
though in a spiral form. The vector which has
received the pathogen from a wild donor may
transmit it also to man who then becomes infected and develops the corresponding disease.
(b) This circulation takes place only when the
environmental conditions are favorable (for instance, at a definite temperature) or at any rate do
not obstruct any of the stages.
A natural focus of disease exists when there
are a specific climate, vegetation, soil, and favorable microclimate in the places where vectors,
donors, and recipients of infection take shelter.
In other words, a natural focus of disease is rehted
to a specific geographical landscape, such as the
taiga of a certain botanical composition, a hot
sand desert, the steppe, etc., i.e., a biogeocoenosis.
M a n falls victim to a n animal disease with natural
foci only when he stays i n the territory of its natural
focus at a definite season of the year, and is attacked as
prey by hungry bloodsucking vectors which have
already acquired infection from biting wild animals-the carriers and donors of this disease.
This accounts for the seasonal occurrence of
404
Figure 2. Diagram showing the spiral transition (circulation) of tick-home encephalitis virus from organism to
organism (D,RD) in a natural focus of disease by the agency of tick vectors (TV). Drawing by E. N. Pavlovsky.
(M-infection of man kom ticks (T) having acquired the virus from a wild animal, the virus donor (D); (RD)-one
and the same animal: (R) as a virus recipient, (D) as a v i m donor in the period of virusemia; (M) cul-de-sac in
subsequent vlrus circulation (man); (TV)one and the same tick individual: (T) as a virus recipient, (V) as a virus
vector; I, 11, III, tours of circulation.)
II
such diseases; indeed, vectors in natural surroundings are not, as a rule, active round the
year, but only in a fairly warm season.
Not every individual of a given vector species
transmits infection but only those which have
acquired it from an animal donor and, under
the influence of environmental factors, have
reached an infective stage, i.e., have become
capable 0: infecting other animals and man. In
such cases, a vector acquires natural or spontaneous infectivity and an infective power.
A natural focus of disease harboring hungry,
naturally infected vectors is said to be in a latent
state, i.e., capable of infecting both animals and
man.
The natural foci of certain diseases may occur
in geographical landscapes of varied character,
in which they are connected with specific biotopes, otherwise called habitats. I n mountainous areas, important features are the elevation above sea level, and whether the mountain
slopes face north or south; the disease foci are
Ill
Paulousky
405
Figure 3. Diagram showing summary epidemiological significance of the rodent burrow hiwoenosis as a
natural focus of various diseases of man and animals (anthropozoonoses).Drawing by E. N. Pavlovsky Arrows
indicate members of biocanosis, venomous ( S C O ~ ~ ~ Oand
M spiders) or harmless to man; oval lines show the
route of transmission of pathogenic agents from rodents to man.
the causative agent of pseudotuberculosis of rodents along with plague cultures in the TsaganNur district of the Mongolian Peoples Republic
from the narrow-skulled vole gives one reason
for attributing pseudotuberculosis to diseases
with natural foci also.
The number of known diseases with natural foci is steadily growing and will continue to grow as research advances, particularly
in the tropical countries. There are frequent
cases of the discovery of new viruses transmitted by Arthropoda-ticks o r mosquitoes (arboviruses).
Animals which are donors and recipients of
the causative agent of a transmissible or parasitic disease, the agent itself (of whatever
nature), and infection vectors are all members
of a pathobiocoenosis (E. N. Pavlovsky) associated with a specific biotope (or several biotopes)
of a particular geographical landscape, or of a
biogeocoenosis (as defined by Academician V.
N. Sukachev).
1507 in 1957, dropped with the new classification to 1214 in 1959, and rose subsequently to
2040 in 1966. Between 1959 and 1966 the
death rate increased by 56 percent, from 2.7 to
4.2 per 100 000 persons.
The trends in 19 countries between 1951 and
1964 have been summarized by the World
Health Organization (7). These show that:
(1) In virtually every country there was a
sharp decline in the crude mortality rate between 1957 and 1959 which presumably reflected the change from use of the sixth revision
of the International Classfication to the seventh
revision.
(2) With the exception of Venezuela, every
country with data available before 1958 showed
an excess mortality in males, and since 1958 this
excess has been reduced. Before 1958 the excess is likely to have been due to the inclusion of
a high proportion of deaths from bronchitis, a
condition which is appreciably commoner in
males.
(3) In 1964 the crude mortality varied between the countries from 1.1 to 9.7 per 100 000
persons, a variation which is likely to be due, in
part, to variation in diagnostic criteria.
(4) Between 1959 and 1964 several countries
showed sporadic increases in mortality, but the
general picture is of a constant rate with a slight
tendency to decrease.
(5) England and Wales alone showed a steady
increase.
406
Sbeizer et al.
407
112.1
112.2
112.3
112.4
112.5
112.6
Asthma
Asthmatic bronchitis
Hay asthma
Bronchial asthma
Hay fever
Bronchitic asthma
Spasmodic asthma
Catarrhal asthma
With influenza as a contributory or secondary cause
With chronic endocarditis as a contributory or secondary cause
With myocardial disease as a contributory or secondary cause
With arteriosclerosis as a contributory or secondary cause
With chronic nephritis as a contributory or secondary cause
Without any of the complications here specified (1-5)
1948-1957-6th Revision of I.C.D.
24 1
Asthma (bronchial)
Allergic (any cause)
Bronchitis, allergic
Sporadic
Hay asthma
Hay fever with asthma
Asthmatic bronchitis
This title excludes cardiac asthma (434.2) and pneumoconiotic asthma (523-524)
24 1
Asthma (bronchial)
Allergic (any cause)
Sporadic
1958-
-7th
Revision of I.C.D.
Bronchitis, allergic
Hay asthma
Hay fever with asthma
This title excludes cardiac asthma (434.2)and pneumoconiotic asthma (523-524.)It also excludes asthma not
indicated as allergic with mention of bronchitis (acute) (chronic) (500-502).
408
'61
'67 ' 6 3
64
65
66
Table 2. Number of deaths and death rates from asthma per 100 000 persons, by age: England and
Wales, 1959 to 1969..
Age in
years
Deaths
1959
1960
1961
1962
1963
1964
1965
1966
5-9
No.
Rate
6
0.18
12
0.37
6
0.18
6
0.18
9
0.27
23
0.67
18
0.51
17
0.47
10-14
No.
Rate
12
0.33
13
0.35
24
0.65
29
0.84
41
1.21
53
1.60
66
2.02
80
2.46
15-19
No.
Rate
19
0.64
17
0.55
24
0.77
39
1.11
36
0.99
62
1.67
73
1.96
91
2.45
20-24
No.
Rate
24
0.85
27
0.94
28
0.97
32
1.09
56
1.87
51
1.66
79
2.49
83
2.52
25-29
No.
Rate
32
1.12
28
0.98
39
1.37
38
1.31
45
1.53
69
2.33
90
3.00
79
2.64
30-34
No.
Rate
30
0.99
31
1.04
48
1.61
47
1.57
83
2.79
84
2.85
77
2.63
81
2.78
5-34
No.
Rate
123
0.66
128
0.68
169
0.89
191
1.00
270
1.40
342
1.76
403
2.05
431
2.18
35-64
No.
Rate
597
3.32
594
3.29
568
3.13
655
3.60
845
4.64
903
4.96
1072
5.89
970
5.34
5-64
No.
Rate
720
1.96
722
1.96
737
1.99
846
2.26
1115
2.97
1245
3.30
1475
3.90
1401
369
All
ages
No.
Rate
1214
2.67
1188
2.60
1269
2.75
1352
2.89
1655
3.52
1800
3.80
2080
4.35
2040
4.24
~
From theRegrstrar GeneraE'sStatwticalReuzewsof Englandand Walesfw 1959 to 1965, the Quarterly ReturnforEngland and Wales,
3rd Quarter 1966, and personal communication.
a
Sbezzer et ol.
409
Table 3. Ten major causes of death in England and Wales in children aged 10-14 years for the year
1965.~
No.
Percentage of
total deaths
in age group
Male
127
66
193
16.6
85
39
124
10.7
50
51
101
8.7
53
37
90
7.8
39
43
27
44
35
23
36
9
74
66
63
53
6.4
5.7
5.4
4.6
31
19
50
4.3
14
33
47
4.1
190
109
299
25.8
703
457
1160
100.0
Female
Total
4 10
there was no further increase in 1965. In contrast to England and Wales and Scotland, comparable increases also occurred at ages 10 to 14
years and at ages 35 to 64 years.
InJapan the rates have shown a somewhat
different trend. At ages 5 to 34 years there has
been a steady but rather slow increase in mortality, which in 1964 had risen by about 40 percent, in contrast to the rise of over 250 percent
in England and Wales. This, however, was accompanied by a decrease in mortality at ages 35
to 64 years.
New Zealand alone had an asthma death rate
at ages 5 to 34 years of over 0.5 per 100 000
persons in 1959 and showed no increase in mortality in the next six years.
Different results are obtained when the comparison is limited to ages 10 to 19 years-the
decade in which the largest increase was recorded in England and Wales. Five-year agespecific rates are available for the whole period
1959 to 1964 for only 11 countries, and these
are shown in Table 4. The individual results are
irregular, but when the countries are grouped
regionally, to reduce the effect of random fluctuation of small numbers, it is found that there
Table 4. International death rates from asthma for ages 10-14 and 15-19 years: 1959-61 and 1962-1964.'
1959-1961
Country
Britain:
England and
Wales
Scotland
Australasia:
Australiab
Europe:
Germany
Sweden
Denmark
Netherlands
Belgium
Japan
U.S.A.
1962- 1964
Percent
increase
Age in
years
No. of
deaths
Average
yearly
death rate
per 100 000
No. of
deaths
Average
yearly
death rate
per 100 000
1959.196 1
to
1962.1964
10-14
15-19
59
69
0.48
0.67
143
151
1.26
1.23
162.5
83.6
10-14
21
30
0.56
0.94
44
40
1.14
1.16
103.6
23.4
10-14
15-19
29
31
0.21
0.24
54
47
0.39
0.35
85.7
45.8
10-14
15-19
87
107
0.26
0.38
147
172
0.50
0.55
92.3
44.7
10-14
15-19
104
124
0.21
0.30
140
153
0.26
0.33
23.8
10.0
Data compiled from the Official National Vital Statistics Record for each country for the years 1959-1964, with the
exception of Australia.
Data from Dr. B. Gandevia (personal communication).
Speirer et al.
41 1
by coroners after riecropsy examination increased even more rapidly than the total (from
43 at ages 5 to 34 years in 1959 to 237 in 1966).
If the increase were an artifact we should have
to postulate that there had been an even greater
change in the diagnostic criteria used by pathologists than in those used by clinicians, or that
there had been a change in the type of case
referred to coroners.
We conclude, therefore, that the increase in
mortality attributed to asthma at ages 5 to 34
years is, in large part, real and represents a true
increase in the annual number of deaths frQm
the disease.
TWO EXPLANATIONS
One explanation of a true increase in mortality could be that the number of patients suffering from asthma had risen. This cannot be
tested directly, but an indication of its validity
can be obtained by comparing the frequency
with which patients have consulted their general practitioners. In 1955 to 1956 the Royal
College of General Practitioners in cooperation
with the Registrar General collected figures on
morbidity from a variety of diseases, including
asthma (8), and similar data have been made
available to us for the years 1961 to 1966 for a
selected group of practices covering approximately 20 000 patients (Table 6). Over this 10year period there has, in fact, been a tendency
Table 5. Death rate per 100 000 persons aged 5-34 years from selected respiratory diseases: England and
Wales 1959-1965.1
Deaths
1959
1960
1961
1962
1963
1964
1965
No.
Rate
121
0.65
107
0.57
124
0.66
128
0.67
126
0.65
121
0.62
131
0.66
No.
Rate
a4
0.45
92
0.49
73
0.39
76
0.40
73
0.38
76
0.39
78
0.40
Pneumonia
(I.C.D. 490-493)
No.
Rate
486
2.61
403
2.15
423
2.24
474
2.47
416
2.15
398
2.04
379
1.93
No.
Rate
123
0.66
128
0.68
169
0.89
191
1.00
270
1.40
342
1.76
403
2.05
Diagnostic catezory
From the Registrar General's Statistical Reuiew of England and Wales for each year 1959-1965.
See text for definition.
4 12
'
1955-1956a
1961b
1962
1963
1964
1965
1966
5-14
15-44
45-65
10
10
9.2
6.6
5.2
4.9
3.6
9.3
6.8
3.0
5.1
3.9
5.1
5.2
4.9
6.4
4.5
4.3
6.6
9.6
All ages
5.8
3.8
5.6
5.2
5.3
4.6
Fry (8).
1961 to 1966, Dr. D. L. Crombie, Royal College of'Genera1 Practitioners (personal communication).
produce temporary symptomatic relief, but increase the hazards of dying from the disease
later.
Morbidity studies in New Orleans, U.S.A.,
have suggested that asthma can reach epidemic
levels under particular conditions of atmospheric pollution (9),but it is difficult to believe
that this could be a factor in England. Certainly
the increase could not be due to smoke pollution, which has decreased in English towns over
the last decade, nor could it be attributed to
pollution with sulphur gases, which has remained approximately constant ( I 0). Motor
traffic has increased considerably, and one of
the contituents of motor fumes could perhaps
have had a harmful effect. If this were the case,
however, a substantial difference in mortality
would be expected between urban and rural
areas, and we have failed to find any evidence of
this in the national mortality data for 1966. The
death rate was 2.0 per 100 000 persons aged 5
to 34 years in conurbations, 3.0 in urban areas
of more than 100 000 population, 2.2 in urban
areas of under 50 000 population, and 1.9 in
rural districts.
Other environmental hazards are associated
with smoking and occupation, but these cannot
be responsible for changes that have been observed characteristically at ages 10 to 14 years in
both sexes.
ENVIRONMENTAL HAZARDS
Speizer et al.
SUMMARY
The mortality attributed to asthma has increased annually in England and Wales from
1960 to 1965. The increase is more pronounced
at ages 5 to 34 years than at older ages and is
most pronounced at ages 10 to 14 years. In this
last age group the mortality increased nearly
eight times in seven years, and in 1966 asthma
accounted for 7 percent of all deaths.
4 13
4 14
Dr. F. E. Speizer was supported by the National Center for Air Pollution, Bureau of Disease Prevention and Environmental Control,
U.S. Public Health Service.
References
(I) Gandevia, B. Br Med J 2:441, 1967.
(2) Richards, W. and J.R. Patrick. Am J Dis Child
110:4, 1965.
(3) Smith, J.M. Lancet 1:1042, 1966.
(4) Ford, R.M. M e d J A w t 2:196, 1966.
(5) Kessler, A. and C. Geller-Bernstein JAMA
196:458, 1966.
C . STUDIES OF EPIDEMICS
John Snow
Source: Excerpted from Snow on Cholera. Cambridge, Harvard University Press, 1949. By permission of the publisher.
415
4 16
sionally. In six instances I could get no information, owing to the death or departure of every
one connected with the deceased individuals;
and in six cases I was informed that the deceased persons did not drink the pump-water
before their illness.
The result of the inquiry, then, was that there
had been no particular outbreak or increase of
cholera, in this part of London, except among
the persons who were in the habit of drinking
the water of the above-mentioned pump-well.
I had an interview with the Board of Guardians of St. Jamess parish, on the evening of
Thursday, 7th September, and represented the
above circumstances to them. In consequence
of what I said, the handle of the pump was
removed on the following day. . . .
The additional facts that I have been able to
ascertain are in accordance with those above
related; and as regards the small number of
those attacked, who were believed not to have
drunk the water from Broad Street pump, it
must be obvious that there are various ways in
which the deceased persons may have taken it
without the knowledge of their friends. The
water was used for mixing with spirits in all the
public houses around. It was used likewise at
dining-rooms and coffee-shops. The keeper of
a coffee-shop in the neighborhood, which was
frequented by mechanics, and where the pumpwater was supplied at dinner time, informed me
(on 6th September) that she was already aware
of nine of her customers who were dead. The
pump-water was also sold in various little shops,
with a teaspoonful of effervescing powder in it,
under the name of sherbet; and it may have
been distributed in various other ways with
which I am unacquainted. The pump was frequented much more than is usual, even for a
London pump in a populous neighborhood.
There are certain circumstances bearing on
the sdbject of this outbreak of cholera which
require to be mentioned. The Workhouse in
Poland Street is more than three-fourths surrounded by houses in which deaths from cholera occurred, yet out of five hundred and
thirty-five inmates only five died of cholera, the
other deaths which took place being those of
persons admitted after they were attacked. The
workhouse has a pump-well on the premises, in
addition to the supply from the Grand Junction
Water Works, and the inmates never sent to
Broad Street for water. If the mortality in the
workhouse had been equal to that in the streets
immediately surrounding it on three sides, upwards of one hundred persons would have died.
There is a brewery in Broad Street, near to
the pump, and on perceiving that no brewers
men were registered as having died of cholera, I
called on Mr. Huggins, the proprietor. He informed me that there were above seventy workmen employed in the brewery, and that none of
them had suffered from cholera-at least in a
severe form-only two having been indisposed,
and that not seriously, at the time the disease
prevailed. The men are allowed a certain quantity of malt liquor, and Mr. Huggins. believes
they do not drink water at all; and he is quite
certain that the workmen never obtained water
from the pump in the street. There is a deep
well in the brewery, in addition to the New River
water.
At the percussion-cap manufactory, 37 Broad
Street, where, I understand, about two hundred
workpeople were employed, two tubs were kept
on the premises always supplied with water
from the pump in the street, for those to drink
who wished; and eighteen of these workpeople
died of cholera at their own homes, sixteen men
and two women. . . .
Dr. Fraser also first called my attention to the
following circumstances, which are perhaps the
most conclusive of all in proving the connection
between the Broad Street pump and the outbreak of cholera. In the Weekly Return of
Births and Deaths of 9th September, the following death is recorded as occurring in the
Hampstead district: At West End, on 2nd September, the widow of a percussion-cap maker,
aged 59 years, [suffered] diarrhea two hours,
cholera epidemica sixteen hours.
I was informed by this ladys son that she had
not been in the neighborhood of Broad Street
for many months. A cart went from Broad
Street to West End every day, and it was the
custom to take out a large bottle of the water
from the pump in Broad Street, as she preferred it. The water was taken on Thursday,
31st August, and she drank of it in the evening,
and also on Friday. She was seized with cholera
on the evening of the latter day, and died on
Saturday, as the above quotation from the register shows. A niece, who was on a visit to this
lady, also drank of the water; she returned to
her residence, in a high and healthy part of
Islingt6n, was attacked with cholera, and died
also. There was no cholera at the time, either at
West End or in the neighborhood where the
417
Snow
niece died. Besides these two persons, only one
servant partook of the water at Harnpstead
West End, and she did not suffer, or, at least,
not severely. There were many persons who
drank the water from Broad Street pump about
the time of the outbreak, without being attacked with cholera; but this does not diminish
the evidence respecting the influence of the
water, for reasons that will be fully stated in
another part of this work. . . .
The limited district in which this outbreak of'
cholera occurred contains a great variety in the
quality of the streets and houses; Poland Street
and Great Pulteney Street consisting in a great
measure of private houses occupied by one family, whilst Husband Street and Peter Street are
occupied chiefly by the poor Irish. The remaining streets are intermediate in point of respectability. T h e mortality appears to have fallen
pretty equally amongst all classes, in proportion
to their numbers. . . .
Table 1 exhibits the chronological features of
this terrible outbreak of cholera.
It is pretty certain that very few of the fifty-six
attacks placed in the table to the 31st August
occurred till late in the evening of that day. The
irruption was extremely sudden, as I learn from
the medical men living in the midst of the district, and commenced in the night between the
31st August and 1st September. There was
hardly any premonitory diarrhea in the cases
which occurred during the first three days of
the outbreak; and I have been informed by
several medical men that very few of the cases
which they attended on those days ended in
recovery.
The greatest number of attacks in any one
day occurred on the 1st of September, immediately after the outbreak commenced. The following day the attacks fell from one hundred
and forty-three to one hundred and sixteen,
and the day afterwards to fifty-four. A glance at
Table I will show that the fresh attacks continued to become less numerous every day. On
September the 8th-the day when the handle
of the pump was removed-there were twelve
attacks; on the 9th, eleven; on the loth, five; on
the Ilth, five; on the IZth, only one; and after
this time, there were never more than four attacks on one day. During the decline of the
epidemic the deaths were more numerous than
the attacks, owing to the decease of many persons who had lingered for several days in consecutive fever.
August 19
August 20
August 21
August 22
August 23
August 24
August 25
August 26
August 27
August 28
August 29
August 30
August 31
September 1
September 2
September 3
September 4
September 5
September 6
September 7
September 8
September 9
September 10
September 1 1
September 12
September 13
September 14
September 15
September 16
September 17
September 18
September 19
September 20
September 21
September 22
September 23
September 24
September 25
September 26
September 27
September 28
September 29
September 30
Date unknown
Total
No. of
fatal attacks
1
1
0
1
1
0
1
1
1
1
8
56
143
116
54
46
36
20
28
12
11
5
5
1
3
0
1
4
Deaths
1
0
0
0
2
0
0
1
0
1
2
3
70
127
76
71
45
37
32
30
24
18
15
6
13
6
8
6
5
2
3
0
0
2
3
0
0
3
0
0
2
1
1
1
1
1
1
0
0
0
45
-
616
616
0
2
1
0
0
418
been opened, and I was informed by Mr. Farrell, the superintendent of the works, that there
was no hole or crevice in the brickwork of the
well, by which any impurity might enter; consequently in this respect the contamination of the
water is not made out by the kind of physical
evidence detailed in some of the instances previously related. I understand that the well is
from twenty-eight to thirty feet in depth, and
goes through the gravel to the surface of the
clay beneath. The sewer, which passes within a
few yards of the well, is twenty-two feet below
the surface. The water at the time of the cholera
contained impurities of an organic nature, in
the form of minute whitish flocculi visible on
close inspection to the naked eye, as I before
stated. Dr. Hassall, who was good enough to
examine some of this water with the microscope, informed me that these particles had no
organised structure, and that he thought they
probably resulted from decomposition of other
matter. He found a great number of very minute oval animalcules in the water, which are of
no importance, except as an additional proof
that the water contained organic matter on
which they lived. The water also contained a
large quantity of chlorides, indicating, no
doubt, the impure sources from which the
spring is supplied. Mr. Eley, the percussion-cap
manufacturer of 37 Broad Street, informed me
that he had long noticed that the water became
offensive, both to the smell and taste, after it
had been kept about two days. This, as I noticed
before, is a character of water contaminated
with sewage. Another person had noticed for
months that a film formed on the surface of the
water when it had been kept a few hours.
I inquired of many persons whether they had
observed any change in the character of the
water, about the time of the outbreak of cholera, and was answered in the negative. I afterwards, however, met with the following important information on this point. Mr. Gould, the
eminent ornithologist, lives near the pump in
Broad Street, and was in the habit of drinking
the water. He was out of town at the commencement of the outbreak of cholera, but came
home on Saturday morning, 2nd September,,
and sent for some of the water almost immedi-
Detailed clinical and epidemiological observations were made by members of the staff of
the Division of Communicable Diseases of the
New York State Department of Health in 7
milk-borne epidemics of streptococcus infection
comprising 1529 cases and 24 deaths occurring
in the 3 year period 1934-1936. Three of the
epidemics studied, consisting of 806 cases and
16 deaths, were classified clinically as scarlet
fever, and 4 epidemics, consisting of 723 cases
and 8 deaths, were classified as septic sore
throat. An analysis was made of data obtained
during the investigation of these seven epidemics, with special reference to the clinical and
immunological characteristics of the streptococcus infections observed.
All the epidemics occurred in villages of less
than 6000 population, and in each instance the
incriminated milk supply was one of raw milk or
cream. The generally accepted identifying characteristics of milk-borne epidemics were observed in each outbreak. All the epidemics were
explosive in character; in each outbreak elimination or pasteurization of the incriminated
milk supply was followed by a marked decline in
case incidence. As shown in Table 1, each epidemic was characterized by an age distribution
typical of milk-borne outbreaks, a higher proportion of cases occurring among adults than
among children. An overwhelming majority of
the cases in each epidemic occurred among patrons of a single dairy as shown in Table l by
case rates per 100 quarts of milk among customers of the implicated dairy and among customers of all other dairies.
Efforts to determine the source of contami-.
nation of the milk supply resulted in the discovery of acute mastitis in members of the producing herds in six of the seven epidemics. The
Source: American Journal of Publzc Health, 21:1259-1266,
1937.
Read before the Epidemiology Section of the American
Public Health Association at the Sixty-sixth Annual Meeting
in New York, N.Y., October 6, 1937.
* Division of Communicable Diseases, State Department of
Health, Albany, N.Y.
419
420
Table 1. Summaries of three milk-borne scarlet fever epidemics and four milk-borne septic sore throat
epidemics, New York State, 1934-1936.
~~
~~
Percent of
cases
15 years of
age or over
Number of
cases
Number of
deaths
Scarlet fever
Owego
Wellsville
Red Creek
532
20 1
73
8
6
2
71.4
68.8
69.8
145.0
27.2
45.0
5.9
1.1
2.5
500
112
56
55
7
0
1
0
75.3
75.0
87.5
70.7
107.1
88.7
62.4
51.0
4.3
8.2
1.9
0.3
CLINICAL CHARACTERISTICS
The clinical manifestations of the illnesses observed in the various outbreaks were strikingly
similar with the exception of the presence or
absence of a characteristic scarlet fever rash and
desquamation. Almost without exception the
patients suffered from sore throat with fever of
from 100F. to 104"F., general malaise, and
varying degrees of prostration. The throat was
almost invariably red and frequently edematous. A punctate rash was frequently seen on
Table 2. Secondary attack rates by age among household contacts (1) in a contact epidemic of scarlet
fever and (2) in a milk-borne epidemic of scarlet fever where the primary case occurred in a consumer
of the incriminated milk supply but where the members of the household had not consumed the
incriminated milk.
Contact epidemic
Binghamton
Age group
0-14
15 and over
Total
Persons
exposed
Number of
cases
Milk-borne epidemic
Owego
Persons
exposed
Number of
cases
Contact
epidemic
Milk-borne
epidemic
465
773
93
41
73
192
10
14
20.0
5.3
13.7
7.3
1238
134
265
24
10.8
9.0
Stebbins et al.
throat. Scarlet fever cases without rash and
cases of septic sore throat were also clinically
indistinguishable from the type of case frequently seen in the absence of any epidemic
and usually diagnosed as severe tonsillitis.
COMPLICATIONS
421
SUSCEPTIBILITY
Table 3. Complications observed in milk-borne epidemics of septic sore throat in Baldwinsville and
Dryden and of scarlet fever in Owego and Wellsville.
Scarlet fever
Complications
Number
of cases
Percent
59
48
26
2
11
6
24
11
8.2
6.6
3.6
0.3
1.5
0.8
3.3
1.5
723
180
24.9
Number
of cases
69
50
48
Percent
4
6
13.1
9.5
9.1
2.8
1.1
1.o
0.8
1.1
526
130
24.7
15
6
5
422
Table 4. Attack rates according to age and quantity of milk consumed per day among regular consumers
of the incriminated milk-Owego scarlet fever epidemic.
0-14
Average daily
milk consumption
None
1-7 OZ.
8 oz. and over
Total
15 and over
Persons
exposed
Cases
17
29
203
8
84
249
94
Persons
exposed
tion in a considerable proportion of those attacked, but it has not been thought likely that
scarlet fever produces immunity to septic sore
throat. The effect of a previous attack of scarlet
fever upon susceptibility to scarlet fever or septic sore throat was studied by means of household attack rates. Attack rates among persons
living in households in which one or more cases
occurred are given in Table 5 by age and according to history of previous scarlet fever for
three epidemics of scarlet fever and two of septic sore throat. A previous attack of a streptococcus infection, diagnosed scarlet fever, apparently produced little o r no immunity to
subsequent infection with the hemolytic streptococcus associated with either milk-borne scarlet fever or septic sore throat epidemics. Scarlet
fever patients who gave a history of a previous
attack were apparently as severely ill as were
those who denied having previously suffered
from scarlet fever, case fatality rates and the
incidence of complications being the same in
the two groups. The only difference between
these two groups appeared to be the proportion
Cases
0-14
15 and over
90
227
230
15
57
95
11.8
27.6
41.4
16.7
25.1
41.3
547
167
37.8
30.5
Table 5. Attack rates among persons living in households where one or more cases occurred according
to age and history of previous scarlet fever in milk-borne epidemics of scarlet fever in Owego,
Wellsville, and Red Creek and of septic sore throat in Dryden and Waterloo.
Scarlet fever epidemics
No previous
scarlet fever
c
Attack rate
percent
Cases
Number of
Cases
Previous
scarlet fever
No previous
scarlet fever
36
263
9
101
459
1000
223
434
25.0
38.4
48.6
43.4
9
45
20
43
129
16
59
55.5
44.4
37.2
45.7
299
110
1459
657
36.8
45.0
54
25
172
75
46.3
43.6
persons
Number of
persons
No previous
scarlet fever
Previous
scarlet fever
Cases
CI
Previous
scarlet fever
Number of
persons
Total
Cases
0-14
15 and over
No previous
scarlet fever
Number of
persons
Age group
Previous
scarlet fever
Stebbins el al.
423
Table 6. Percent negative skin sensitivity tests' according to age and history of attack during the milkborne epidemics of scarlet fever in Owego and Wellsville and of septic sore throat in Baldwinsville and
corfu.
Septic sore throat epidemics
Not ill
Number
tested
Number
negative
Number
tested
Number
negative
I11
Not ill
Percent
negative
I11
Not ill
Number
negative
I11
Number
tested
Total
Percent
negative
Number
negative
0-14
15 and over
Not ill
Number
tested
I11
112
164
82
142
1052
410
361
193
73.2
86.6
34.3
47.1
121
116
50
61
412
190
145
100
41.3
52.6
35.2
52.6
276
224
1462
554
81.2
37.9
237
111
602
245
46.8
40.7
Table 7. Percent negative skin sensitivity tests. according to age and history of scarlet fever in various
New York State communities, 1934-1937.
History of
scarlet feverb
Age group
0- 4
5- 9
10-14
15-19
20-29
30-39
40-49
50 and over
Total
a
No history of
scarlet fever
Percent negative
Number
tested
Number
negative
Number
tested
Number
negative
History of
scarlet fever
No
history of
scarlet fever
232
284
83
23
16
13
17
162
216
66
20
14
13
15
132
2816
2713
751
139
62
32
31
30
882
1194
35 1
62
30
21
24
69.8
76.0
79.5
87.0
87.5
100.0
88.2
22.7
31.3
44.0
46.7
44.6
48.4
65.6
77.4
668
506
6676
2594
75.7
38.8
424
Table 8. Percent positive skin sensitivity testsPby age among persons tested in New York State,
1934-1937, and percent of cases with rash among scarlet fever patients in milk-borne epidemics in
Owego, Wellsville, and Red Creek.
Scarlet fever cases in milk-borne
epidemics according to rash
Number
tested
Number
positive
Percent
positive
Number
of cases
Number
with rash
Percent
with rash
135
3080
3046
842
167
82
48
51
104
2030
1611
422
82
36
11
9
77.0
65.9
52.9
50.1
49.1
43.9
22.9
17.6
61
97
75
93
199
111
79
72
53
76
53
55
105
63
35
16
86.9
78.4
70.7
59.1
52.8
56.8
44.3
22.2
745 1
4305
57.8
787
456
57.9
0- 4
5- 9
10-14
15-19
20-29
30-39
40-49
50 and over
Total
a
skin tests by age observed in the general population, indicating a positive correlation between
sensitivity as measured by the skin test and the
development of a rash.
SUMMARY AND CONCLUSIONS
An analysis was made of 1529 cases of streptococcus infection occurring in seven epidemics
in New York State d u r i n g t h e p e r i o d
Figure 1. Percent positive skin sensitivity tests by age among 7451 persons tested in
New York State, 1934-1937, and percent of cases with rash by age among 787 scarlet
fever patients in milk-borneepidemics In Owego, Wellsville, and Red Creek.
90 '80
. _ _ _ Percent
_
of positive skin sensitivity
70
L
60
50
40
30
20
-----___._
l0 0
10
20
30
AGE IN YEARS
40
50 and over
Stebbins et al.
There was evidence in epidemics both of scarlet fever and septic sore throat of transmission
by personal contact comparable to that observed in contact epidemics of scarlet fever.
The appearance of a rash was the only distinguishing characteristic between cases in epidemics classified as scarlet fever or as septic sore
throat. All other clinical characteristics, including frequency and type of complicating condition, were practically the same.
The effect of various factors which might be
expected to influence susceptibility to infection
were studied. No evidence of age o r sex selection was found. Dosage of the infectious material seemed to be an important factor in determining whether or not a given individual
became ill.
A previous attack of scarlet fever seemed to
produce little or no immunity to an attack of
milk-borne streptococcus infection but did materially lessen the probability of the development of a scarlet fever rash.
The effect of an attack of milk-borne scarlet
fever or septic sore throat upon skin sensitivity
to standard streptococcus toxin was tested in
communities in which epidemics occurred, and
a reduced skin sensitivity was observed following milk-borne scarlet fever comparable to that
observed after scarlet fever transmitted by contact but no reduction of skin sensitivity was observed following septic sore throat.
425
ACKNOWLEDGMENT
of
Aw-
426
Gregg
these types were present, The first type in the
right eye and the second type in the left eye. In
my opinion these variations and those described by other observers are not essentially
different from each other, and the apparent
differences are due merely to a variation in intensity and duration of action of the same noxious factor.
The appearance of the cataract does not, in
my opinion, exactly correspond to any of the
large number of morphological types of congenital and developmental lenticular opacities
that have been described. I do not wish to add
to what Duke Elder ( I ) has described as the
confusion which has arisen from the enthusiasm of various observers in the multiplication
of types which differ but little in their essential
pathology and vary only in their shape and
position. I shall, therefore, merely describe the
cataract as subtotal. Other descriptions by my
colleagues in notes on their cases have been:
central nuclear, complete, discoid, nuclear plus,
anterior polar, dense central with riders, complete pearly, mature, and total lamellar. In sixteen cases of the whole series reviewed, the cataract was unilateral.
Vision. In all cases the response to light was
good; the babies appeared to follow readily any
movement of the light stimulus.
Nystagmw. In the very young patients nystagmus was not noted, but in older babies or in
cases in which treatment had to be delayed it
was present. The movements were of a coarse,
jerky, purposeless nature rather than a true
nystagmus. It was a searching movement of the
eyeballs and indicated the absence of any development of fixation. In my own cases it was
always present if treatment had been delayed
beyond the age of three months. In one case, in
which the parents deferred operation in order
to try some other form of treatment of which
they had been informed, it developed before
they consented to operation. In another case it
developed after operation during the process of
absorption. This development during the waiting period before operation has been noticed by
other observers.
Variations. One case in my series was particularly interesting. The baby was referred to me at
the age of three weeks with a diagnosis of bilateral keratitis. The corneas were quite white at
birth and both parents had been subjected to a
427
Wassermann test with negative results. At examination I noted a peculiar corneal haze,
denser in the center than in the periphery. The
iris was just visible through this haze in the
peripheral zone. The tension was normal and
there was no inflammation. I advised reexamination under anesthesia. This was done two
weeks later. By this time the corneas had cleared
and the typical white cataracts were seen in the
pupillary areas. This baby subsequently became
very ill and it was only a few weeks ago that I was
able to operate. At operation mydriasis was
fuller than usual in these cases and the cataracts
were the largest observed in this series.
Two other cases with similar corneal involvement have been noted-namely, by A. Odillo
Maher and H.E. Robinson. Involvement was
unilateral in Mahers and bilateral in Robinsons
case. In these cases there had apparently been
some temporary interference with the nutrition
of the cornea. Mahers case is also interesting in
that the mother developed cataract during
pregnancy at the age of twenty-seven. This is
the only instance throughout the series of any
familial history of cataract.
In another case, reported by S.R. Gerstman,
there was bilateral subluxation of the lenses,
mature cataracts, accompanied by arachnodactyly and large fontanelle. Hip regions appeared
normal.
Other complications reported have been cleft
palate, one; congenital stenosis of naso-lachrymal duct, three; calcaneus varus, one; although it is not certain whether these are above
the average incidence in any group of infants of
similar numbers.
Monocular Cases. The monocular cases merit
special consideration. Sixteen of these have
been reported, and in ten of them definite microphthalmia has been described.
In one of my cases-there were three in allthe cataract was noted by the mother only when
the child was seven weeks old, though she stated
that it may have been present before that date.
The affected eye was definitely microphthalmic,
and examination of the other eye under mydriasis revealed a large pale area with some
scattered pigmentation in the lower half of the
fundus suggestive of a coloboma.
In another case the mother gave a history
that both eyes were said to have had conjunctivitis at birth. This inflammation, she stated,
cleared up under treatment in three weeks, and
428
this respect the following measurements obtained at autopsies in other cases at Royal Alexandra Hospital for Children are interesting:
B.S., uetutis five months. Right eye: anteroposterior diameter, 1.5 centimeters; transverse,
1.7 centimeters. Left eye: antero-posterior diameter, 1.4 centimeters; transverse, 1.7 centimeters.
M.M., aetatis three months. Right eye: anteroposterior diamter, 1.5 centimeters; transverse,
I .5 centimeters. Left eye: antero-posterior diameter, 1.6 centimeters; transverse, 1.6 centimeters.
M.O.S., aetutis five and a half months. Both
eyes: antero-posterior diameter, 1.6 centimeters; transverse, 1.8 centimeters.
J. Maude described one case as bilateral microphthalmos, right eye smaller. According to
Scammon and Armstrong ( Z ) , the average
measurements of the eyeball at birth are: sagittal diameter, 17.6 millimeters; transverse diameter, 17.1 millimeters; vertical diameter, 16.5
millimeters. Postnatal growth is very small in the
first six months, but they stated that it is most
probable that the figures for this period are too
low because of the inclusion of premature cases.
By comparison with these average measurements of the normal eye at birth the figures
quoted above show a definite diminution in the
antero-posterior diameter and a reversal of the
normal relationship between the respective
lengths of the antero-posterior and transverse
diameters.
In the cases under consideration here it [nust
be remembered that many of the babies are
generally undersized, so that any estimation of
the size of the eyes must be considered in relation to the general size and body weight of the
baby.
Gregg
over the base of the heart and down the sternum in all. Some had a thrill. All had signs
suggesting the continuance of a fetal condition
or of a malformation of the heart.
In my own series this condition was present in
all but one case. In the whole series it has been
present in forty-four cases; in eleven cases there
is no record of the cardiac condition; in ten
cases it has been recorded as normal or apparently normal; in four cases in which the condition was not reported upon, the babies died and
death was sudden; in another the baby was illnourished; and in three cases the report was
no defect noted.
Autopsy in three cases at the Royal Alexandra
Hospital for Children revealed a wide patency
of the ductus arteriosw, and I understand that in
autopsies performed elsewhere a similar condition has been found. The reports on the cardiac
condition from autopsies in three cases at the
Royal Alexandra Hospital for Children are as
follows:
M.O.S.: There was hypertrophy of the ventricular muscle; the left measured 0.9 centimeters and the right 0.5 centimeters. A few
petechial hemorrhages were detected on the
surface of the myocardium. The endocardium
and valves were normal and all the septa intact.
The ductus arteriosw, however, was widely patent.
B.S.: There was no free fluid in the pericardial sac. T h e heart was enlarged, with particular
hypertrophy of the right ventricle. Right ventricle measured 0.7 centimeters and the left 0.8
centimeters. There were a few petechial hemorrhages visible on the surface of the myocardium
and one fairly large milk spot. The membranous portion of the interventricular septum
was patent. T h e foramen ovule was not completely occluded, although it appeared to have
been functionally closed. The heart valves and
great vesseis were normal, but the ductw arteriosw was widely patent.
P.F.: The right heart was somewhat dilated.
The right ventricle wall was 0.35 centimeters in
its thickest part. The left ventricle wall was 0.5
centimeters in its thickest part. All valves were
normal. No septa1 defect was present. Vessels
were normal except for a wide patency of the
ductus arteriosus.
429
Both lungs had a considerable degree of hypostatic congestion at the bases. Throughout
the remainder of the lungs there were a very
large number of hemorrhagic spots, some of
which were confluent and covered considerable
areas. Hemorrhagic spots were detected on the
inner surface of the pericardium and on the
surface of the myocardium. In addition, the
visceral pericardium over the upper anterior
aspect of the left ventricle bore a milk spot.
The right kidney was situated in such a position
that the ureter entered the pelvis on the lateral
side of the kidney after coursing across its anterior surface. The right kidney consisted of two
distinct lobes, the upper one about twice as
large as the lower. Each lobe had its own separate peivis, and the ureter divided outside the
kidney into two branches, one to each lobe.
Both ovaries were cystic. The uterus was bicornuate in type.
Another complication noted in a few cases
was the development of a dry, scaly eczematous
condition, involving the face, scalp and limbs,
which was very resistant to treatment.
Sex. Thirty-three of the patients were males,
thirty-five were females. In the remaining ten
cases the reports did not specify the sex of the
child.
430
Gregg
431
432
When operation has to be deferred it is essential to maintain the fullest possible degree of
mydriasis, by atropine if tolerated. If atropine
cannot be employed, then repeated instillation
of homatropine must be substituted for it.
?he value of early operation is well illustrated
by one case reported by E. Temple Smith in
which he performed discission on a baby aged
three weeks. Clear pupils resulted and there has
been no sign of nystagmus developing.
Operation
Discission has frequently proved more difficult than usual. The anterior chamber is particularly shallow, and in many cases the very dense
central portion of the lens has proved very resistant to the needle. Sometimes it has separated off as a firm disk, in others the whole lens
has tended to move away from the point of the
needle and one has obtained the impression
that it would have been possible to perform an
ordinary extraction. In other cases, on the
other hand, discission has been straightforward
and easy.
Gregg
I look forward to further improvements in
contact glass development, for herein lies the
greatest possibility for help in the future.
If we agree that these cases are the result of
infection of the mother by German measles,
what can we do to prevent a repetition of the
tragedy in any future epidemic? Is the mass of
modern research into the causation of senile
cataract going to be helpful by the discovery of
some remedy whch could be given to the
mother to inhibit the formation of opacity in
the developing lens of the embryo?
In the present state of our knowledge the
only sure treatment available is that of prophylaxis. We must recognize and teach the potential dangers of such an epidemic or, I think,
any other exanthem, and do all in our power to
prevent its spread and particularly to guard the
young married woman from the risk of infection.
As to confirmation of the theory of causation
put forward in this paper, I suggest that the
following line of investigation may be helpful.
In all prenatal clinics and maternity hospitals
very careful histories should be taken and recorded of exposure of the mother to infection
of any kind during the entire period of pregnancy.
ACKNOWLEDGMENTS
I wish to thank all those colleagues, too numerous to mention, for the reports they have
furnished me of their cases and for their permission to include them in this review.
I am also indebted to Dr. J. Ringland Anderson for his help with the literature on the subject; to Dr. Margaret Harper for her report on
the cardiac Condition; to Dr. B. Van Someren,
of the New South Wales Government Health
Department, for placing the records of his department at my disposal, and particularly to the
sisters in charge of several of the baby health
centers for the excellent reports they so kindly
furnished; to Dr. Douglas Reye for his reports
on the autopsies; and to Professor Harold Dew
for his timely and helpful criticism on the presentation of this paper.
References
433
***
D.R. GAWLER
(Perth) referred to a child with
this disease. It was seen when four months old
and was ill-nourished and suffering from impetigo. It showed intolerance to atropine and
mydriasis was poor. The cataracts were nuclear
and bilateral. The irides were blue and atrophic
around the pupil. The Wassermann test applied
to blood and cerebro-spinal fluid produced no
reaction. No inquiry was made regarding maternal German measles. D.R. Gawler needled one
eye and found the cortex and nucleus resistant.
The cortex flaked off the anterior surface.
There was little reaction to the needling. There
was no epidemic of German measles in the district during the early months of pregnancy, but
the mother said that there was another child in
the town similarly affected. D.R. Gawler had no
particular theories, but there might have been
an endocrine deficiency, possibly involving the
parathyroids.
ARCHIES. ANDERSON
(Melbourne) had seen
a few cases of this type and in every instance the
mother had had German measles during the
second month of pregnancy. He congratulated
N. McA. Gregg on his striking and original
inquiry.
G.H. BARHAM
BLACK(Adelaide) said that he
had seen one case in which monocular cataract
and nystagmus were present. The mother had
German measles six weeks after the last menstrual period. N o inquiry had been made into
the childs heart condition. The epidemic of
German measles had occurred about the same
time as in other states. There had been a number of severe cases. A soldier had died of encephalitis at Renmark. In South Australia an
investigation had been made of streptococcal
infections of the throat, but no streptococci had
been found. Volunteers had submitted to inoculations from camp throat infections, but the
results were inconclusive.
A.W. DOMBRAIN
(Newcastle) had seen four
patients, two of whom had heart disease. He
asked why the infection was described as socalled German measles. One mother had Ger-
434
RICKETTSIALPOX-A
The clinical features have already been described elsewhere (3). In brief, there occurred
an initial lesion at the presumed site of bite by a
mite, a papule, which when fully developed
measured ?42 to 1?42 cm in diameter. The papule
became vesicular in the center and dried. leav-
435
436
25
20 -
15-
WEEK
0-1
1-4
5-9
10-14
0-14
15-19
20-24
25-29
30-39
40-49
50 and over
15 and over
Total
Number
of cases
Population
Incidence
per hundred
600
5.3
2
16
11
3
32
3
9
42
23
15
92
1400
6.5
124
2000
6.2
Greenberg et al.
437
438
439
Greenberg el al.
~
~~
~~
ings were not cleaned as well as formerly; basements were dirty and cluttered up, and the incinerators were not regularly fired. Combustible material was not sufficient to complete
incineration of the garbage. The accumulation
of garbage in the upper compartments of the
incinerators and even the debris in the lower
compartment furnished an excellent food supply for mice. Both living and dead mice were
found by us on our inspections. When disturbed by the opening of the door to the compartment, the mice would scamper about and
disappear in the crevices between the bricks.
Many of the tenants interviewed complained
of the presence of mice. Some had seen them
only in the basement, others complained in addition of hearing them at night in the walls.
Some found mice in their apartments. In two
instances, mice were actually found by the tenants in the beds used by patients. In one apartment, mouse droppings were observed in a patients bed between the sheet and mattress, and
mice were caught in traps laid in this apartment.
Not all houses were equally afflicted nor were
the three blocks of houses equally infested.
Block I was apparently the greatest sufferer.
Whether this was due to less adequate and conscientious janitorial service in this block compared to the others, we are not prepared to
state. It should be mentioned that there was no
regular exterminating service in the houses at
the time of our investigation.
Complement-fixing antibodies of rickettsialpox were demonstrated in sera from trapped mice. Such antibodies could not be demonstrated in laboratory mice nor in mice trapped
in the Washington area. An organism identical
with the MK strain of rickettsialpox was recovered from one of the mice trapped in the
infested apartments (6).
A routine search for mites was made in the
houses of the development, particular attention
Table 2. Rickettsialpox cases distributed according to residence and presence of mites in building.
Building
Mites found
No mites found
N o search made
Total
Block I
Number of
Buildings
Cases
Block I1
Number of
Buildings
Cases
Block 111
Number of
Buildings
Cases
Total
Number of
Buildings
Cases
12
6
5
47
13
15
3
19
1
20
0
4
16
3
16
9
0
19
41
9
67
42
15
23
75
23
24
23
25
69
124
440
The results of the investigation of the epidemic in Queens indicated that we were dealing
with a hitherto undescribed disease, rickettsialpox, caused by an organism, R. akari, which
was recovered from the blood of two patients
early in the disease. The concomitant finding of
bloodsucking mites, A. sanguineus, in the group
of buildings where the patients lived, and the
recovery of R. akari from two pools of these
mites justified the belief that the mites were the
vectors of this disease. The only other parasites
Greenberg et al.
or hematophagus insects found in the buildings
were adult mosquitoes. Several hundred were
collected from dark basement corners. Some
were tested for infection, with negative results.
Specimens sent to the National Museum were
identified as Culex pipiens L. by Dr. Alan Stone.
The only rodents found in the housing development were house mice, which were present in
large numbers. The finding of mites as ectoparasites of mice, the presence of mammalian
erythrocytes in smears made from engorged
mites, the laboratory determination of the presence of complement-fixing antibodies of
rickettsialpox in the blood of trapped mice, and
the isolation of R . akari from one of them, indicated that the mice acted as the animal reservoirs. The occurrence of a significantly greater
number of cases in buildings where mites were
readily found and the finding of multiple cases
in families strengthened the hypothesis that the
mites were the vectors of the disease.
The initial lesion probably represents a reaction to the bite of an infected mite. That no
patient remembered being bitten is not surprising in view of the small size of the mite and the
fact that no itching or pain was caused. It is not
unusual for persons to be bitten by ticks and to
be unaware of the fact until the acarid is discovered adhering to the skin or scalp by another
person.
It is interesting to observe that rickettsialpox,
as here observed, is a domiciliary disease. In the
epidemic described, incinerators played an
important role in maintaining the infection.
This was not due to anything inherent in incinerators but to the fact that if not fired frequently and regularly they become excellent
harborages for mice due to the accumulation of
garbage. Also, they are warm and the particular
mites identified as vectors of rickettsialpox
thrive well in warm places. However, cases of
rickettsjalpox may occur in homes where there
is no incinerator service. We have observed several such in other parts of the city. In all instances, however, mice harborages existed.
44 1
SUMMARY
Note: We have omitted the map of the affected area which appeared
in the original. Ed.
In the typical case, the illness began with aching pain and tenderness in the calf, followed by
paresthesia and loss of superficial sensation of
stocking-and-glove distribution. After a day or
two the disturbance of sensation decreased and
might disappear, and at about the same time
motor weakness appeared, involving first the
muscles of dorsiflexion and eversion of the foot
and a little later the calf muscles. Later still the
muscles of the hand were commonly affected
also. Only the more severe cases were admitted
to hospital, and victims of the disease who were
442
443
FINDINGS
444
445
References
( I ) Burley, B.D.JAMA 98:298, 1932.
( 2 ) Hunter, D. The Diseases of Occupations. London,
1955.
( 3 ) Jordi, A . U . J Auiatzolz Med 23, 623, 1952.
Adenocarcinoma of the vagina in young women had been recorded rarely before
the report of several cases treated at the Vincent Memorial Hospital between 1966
and 1969. The unusual occurrence of this tumor in eight patients born in New
England hospitals between 1946 and 1951 led us to conduct a retrospective investigation in search of factors that might be associated with tumor appearance. Four
matched controls were established for each patient; data were obtained by personal
interview. Results show maternal bleeding during the current pregnancy and previous pregnancy loss were more common in the study group. Most significantly,
seven of the eight mothers of patients with carcinoma had been treated with
diethylstilbestrol started during the first trimester. None in the control group were
so treated (p less than 0.00001). Maternal ingestion of stilbestrol during early
pregnancy appears to have enhanced the risk of vaginal adenocarcinoma developing years later in the offspring exposed.
Cancer of the vagina is rare, occurring usually as epidermoid carcinoma in women over
the age of 50 years ( I ) . Between 1966 and 1969,
however, seven girls 15 to 22 years of age with
adenocarcinoma of the vagina (clear-cell or endometrial type) were seen at the Vincent Memorial Hospital (2). Although isolated case reports
of histologically similar adenocarcinomas of the
vagina had previously been published (3-8),
these carcinomas, too, were usually in older patients. No such case in the younger age group
had been seen at this institution before 1966.
The tumor typically caused prolonged vaginal bleeding that, occurring in young women,
was mistaken for anovulatory bleeding and delayed the correct diagnosis. Routine vaginal
cytology was often negative, and the tumor was
not palpated on rectal examination. The correct
diagnosis was arrived at only after vaginal examination had been performed.
Histologically, one of the tumors resembled
endometrial carcinoma, but the remainder were
characterized by tubules and glands lined by
clear cells containing glycogen or hobnail
cells. The clear cells also appeared in solid
Source: New-England Journal of Medicine 284(16): 878-881,
1971.
I From the Vincent Memorial Hospital (Gynecological
Service of the Massachusetts General Hospital. Supported by
a grant (1393-C-1) from the American Cancer Society (Massachusetts Division), Inc.
446
Herbst et al.
447
Age at first
symptoms (year)
Year of
birth
Year of
treatment
20
15
1949
1951
1969
1967
14
15
19
1950
1950
1949
1968
1966
1969
16
1951
1967
18
1949
1968
22
1946
1968
Therapy
Status
1971
448
Maternal
age (year)
Mean
of 4
Case Controls
Case
No.
25
30
22
33
22
21
30
26
1
2
3
4
5
6
8
Total
32
30
31
30
27
29
27
28
Maternal
smoking
Bleeding in
this
pmgnancy
Any prior
pregnancy
loss
Estrogen
given in
this
pregnancy
Breast
feeding
Intra
uterine
X-ray
exposure
Case Control Case Control Case Control Case Control C a s e Control Case Control
No
5/32
718
0132
1/32
21/32
0.53
0.50
(N.S.)
618
318
718
Yes
014
014
014
014
014
014
014
No
No
Yes
Yes
Yes
No
Yes
Yes
Yes
Yes
Yes
Yes
Yes
No
Yes
Yes
Yes
Yes
Yes
Yes
Yes
No
Yes
Yes
114
114
114
014
114
014
114
014
014
014
014
014
114
014
014
014
214
314
114
314
314
314
314
314
Yes
Yes
Yes
No
No
4.52
<0.05
No
7.16
<0.01
014
23.22
<0.00001
No
014
114
014
214
014
014
014
014
No
No
No
No
No
No
No
Yes
114
014
014
014
014
114
114
114
318
3132
118
4132
No
No
Yes
Yes
No
No
Yes
2.35
0.20
(N.S.)
(N.S.)
Matched control chi-square test used as described by Pike & Morrow (9).
Standard error of difference 1.7 yr (paired t-test); N.S. = Not statistically significant.
Birth weight
Age at onset of menses
Complications & outcome of study pregnancy
Ingestion of other medications during
pregnancy
Childhood diseases of mothers & patients
History of tonsillectomy
Childhood ingestions
Household pets
Noteworthy illnesses of patients & parents
Cosmetic use in patients & mothers
Cigarette smoking in patients
Alcohol consumption in parents
Occupation & year of education of parents
a Events compared before date of onset of present illness
for each study patient & her matched controls.
DISCUSSION
Herbst et al.
or bleeding during the study pregnancy. Furthermore, these tumors were known to occur,
though rarely, in women born before the availability of oral estrogens. Thus, factors other
than maternal stilbestrol ingestion appear to be
operative in their development. Moreover, the
stilbestrol pills prescribed for these mothers
were those available between 1946 and 1951.
The ingredients of these tablets, the estrogenic
potency of stilbestrol and its other chemical
properties must all be recognized as possible
elements in the association observed. Finally,
among four of the eight families there are five
female siblings, ranging in age from 18 to 22
years, who are also products of pregnancies
d u r i n g which t h e i r m o t h e r s took d i ethylstilbestrol. Up to the present, a vaginal tumor has not developed in any of these girls.
To try to estimate the frequency of stilbestrol
administration and the risk of development of
these tumors in female offspring whose mothers took stilbestrol during pregnancy, we have
examined the files of one of the hospitals in this
study for the years 1946 through 1951. During
this interval there was a special high-risk pregnancy clinic at the Boston Lying-in Hospital in
which stilbestrol was prescribed to 675 ward
patients. There were approximately 14 500
ward deliveries, indicating that at that time
roughly one in 21 ward patients at the Boston
Lying-in Hospital were treated during pregnancy with stilbestrol. Thus, it appears to be
well within the range of statistical expectation to
have a control group in which the frequency of
stilbestrol use was 0 in 32. In the interval 1946
to 195 1 the private service at the Boston Lyingin Hospital had more deliveries than the ward
service. We have knowledge of only one case of
clear-cell adenocarcinoma developing in a patient born at the Boston Lying-in Hospital, and
she was delivered on the private service. Whatever the risk of tumor development in the exposed offspring, it appears to be small.
The high concurrence of benign vaginal adenosis with these adenocarcinomas suggests that
an anomaly of vaginal epithelial development
may be a predisposing condition. Previous reports have described an association between adenosis and this tumor in older women (3,8),
and
their concurrence in younger patients was initially noted in the present cases (2). It may be
that an increase in adenosis occurs at menarche
in these patients and results in greater quantities of benign tissue at risk for malignant
449
ACKNOWLEDGMENTS
450
ABSTRACT
45 I
452
Epidemiology
We identified patients with S . muenchen infection and obtained case histories by reviewing
hospital and health-department records and by
interviewing patients in Ohio and Michigan. We
defined a case as the isolation of S. muenchen
from blood or stool during December 1980
through February 1981, from the first person in
a household to become ill. These initial studies
led to a case-control study in which a questionnaire on foods, restaurants, recreation, and contact with animals was administered by telephone
to 32 case and control households in Ohio. Control households were randomly selected from
case neighborhoods by means of a street directory. Control households were excluded if any
family member had been ill with a diarrheal
illness lasting for more than two days in the past
two months. The initial case-control study led to
a second survey of case households, in which
specific information on interactions with other
case patients and the use of illicit drugs was
obtained.
To determine whether marijuana was more
frequently used in case households than in control households, we conducted a case-control
study of 17 cases in Michigan and 34 control
subjects matched for age and neighborhood.
Statistical testing of the responses from case and
control patients was performed according to the
method of Miettinen (cases matched with a variable number of controls) (8).Samples of marijuana (less than 5 g) were requested from case
households in Ohio, Michigan, Georgia, and
Alabama and were submitted for microbiologic
testing.
Laboratory Investigations
Salmonellae were isolated from fecal and
blood specimens by hospital laboratories using
standard isolation techniques a n d were
serotyped by state health departments (9).Biochemical reactions (9) and antibiotic susceptibilities (20)were determined for all isolates,
including control strains of S . muenchen not related to the epidemics or to marijuana use. A
quantitative estimate of salmonella contamination in the marijuana samples was made by
plate counts of S . muenchen isolates. To confirm
that the samples were marijuana and to deter-
Tavlor et al.
453
Figure 1. Isolates of Salmonella muenchen from cases of infection, in the United States,
January 1978 through August 1981.
~
10090 -
AND MICHIGAN
---OHIO
80 -
STATES
-UNITED
1\
70 70-1
z 602 50-
30 20 -
-____- --__
10-
o-<
J
--.
_---/
1978
1979
-___-
---_/
,
1980
- - - / J
,D , J
I
1981
-TTT
JO N
~ D~
cent) from whom S. muenchen was isolated in posed to marijuana: 13 patients (21 percent)
1980. This increase was significant (P = 0.04, admitted to personal use of marijuana, 22 (35
Fishers exact test, two-tailed)and was associated percent) had been exposed to marijuana in the
with a decrease in the proportion of persons 50 household but had not smoked it, and 14 (22
years old or older, from 20 percent to 8 percent. percent) had been exposed to marijuana outThe predominant symptoms were diarrhea (90 side the household but denied smoking it. Fourpercent), fever (81 percent), abdominal pain (73 teen (22 percent) had no known exposure to
percent), bloody diarrhea (54 percent), and marijuana.
nausea and vomiting (44 percent). Thirty-nine
We conducted the second case-control study
of the cases (62 percent) were hospitalized. The in Michigan soon after the association in Ohio
median duration of illness was eight days.
was discovered. Seventy-six percent of the paThe case-control study in Ohio failed to cor- tients, as compared with 21 percent of the conrelate a place (such as a supermarket, restau- trol subjects, reported personal or household
rant, bar, or recreational center) or a food (such exposure to marijuana (Table 1). Patients were
as a meat product, cheese, dressing, spice, or 20 times as likely as their matched control subholiday food-chocolate or fruitcake) with the jects to have been exposed to marijuana in the
outbreak. However, we found that 44 percent of two weeks before the patients illness (P<O.OO 1).
the case households, but none of the control The patients firmly denied ingestion of marihouseholds, had children under one year of age juana in foods.
(P<O.OOl), and that all the case households, but
Table 1. Personal or household exposure to
only 41 percent of the control households, in- marijuana among cases and controls in Michigan,
cluded persons 15 to 35 years old (P<O.OOl).
January through February 1981..
Because of the association with young adults,
Controls
the lack of association with a food, restaurant,
Total
Both 1 Yes Both
or activity, and some anecdotal information, we
Yes
1 No
No
Triplets
asked about the use of illicit drugs in a second
interview with members of case households.
1
4
8
1
3
Cases
None of the household members admitted
0
1
3
4
Total triplets
1
5
11
17
using any drug except marijuana, but exposure
to marijuana appeared to be high. In Ohio and
a There was a significant difference between cases and conMichigan 49 patients (78 percent) had been ex- trols (P = 0.0006, relative risk = 20).
{F:
454
Laboratory Investigations
S . muenchen was isolated from marijuana samples from three of four case households in Ohio
and from one case household each in Michigan,
Alabama, and Georgia. The plate-count estimate of salmonella organisms per gram of
marijuana indicated extremely high levels of
contamination (Table 2). The microorganisms
isolated in the three other samples were similar.
Biochemical analysis of the suspected marijuana samples confirmed that they were marijuana, but the pattern did not suggest their
country of origin.
The S . muenchen strains isolated from the patients and the marijuana in Ohio and Michigan
were sensitive to all antibiotics tested and were
biochemically indistinguishable from control S.
muenchen (in a battery of 36 test reactions). Ini-
Table 2. Characteristics of salmonella isolated from marijuana samples obtained from case households,
February 1981.
~~
Sample
State
A
B
Ohio
C
D
E
F
Ohio
Ohio
Michigan
Georgia
Alabama
Serotype
S. muenchen
S . muenchen
s. Oslo
S . muenchen
S. muenchen
S . muenchen
S . muenchen
Colony count
per grama
8 x 105
4 x 107
4 x lo6
5x107
QNS
QNS
QNS
Other
microorganisms
foundb
1,2,3,4
1,2,3,4
NT
1,2,3,4,5
NT
NT
NT
Antibiotic
sensitivity
S
S
S
Tudor et ul.
Table 3. Salmonella muenchen plasmid profiles.
Origin of
organism
Characteristic
plasmidsa
None or
other
plasmids
No. of strains
Alabama
Arizona
Georgia
West Virginia
Ohio
Michigan
Massachusetts
California
Wisconsin
New Hampshire
Washington (state)
Pennsylvania
Texas
Vermont
C2 controls
Controls with
S. muenchen
11
12
1
4
3
7
3
1o c
16 c
75
a
455
456
Tudor el ul.
(19) Salmonellosis and Marijwma: Unexpected Transmission (Georgia Epidemiology Report). Atlanta:
Georgia Department of Human Resources, February
1981, p. 1 .
(20) The National Narcotics Intelligence Consumers Committee. Narcotics Intelligence Estimate: The Sup-
457
Purpose of Study. The purpose of the present study was to attempt to determine, so far as
possible by clinical investigations, statistical
Tobacco as a Possible Cause of Increase. The methods, and experimental studies, the imporsuggestion that smoking, and in particular ciga- tance of various exogenous factors that might
rette smoking, may be important in the produc- play a role in the induction of bronchiogenic
tion of bronchiogenic carcinoma has been made carcinoma. In this regard we intended to learn
by many writers on the subject even though well the relative importance of previous diseases of
controlled and large scale clinical studies are the lungs, rural and urban distribution of palacking. Adler (7) in 1912 was one of the first to tients, various occupations, and hereditary
think that tobacco might play some role in this background as well as smoking habits. By obregard. Tylecote (8),Hoffman (9),McNally ( l o ) , taining all this information, we hoped to deterLickint ( I I ) , Arkin and Wagner (IZ),Roffo (13), mine whether any of these factors, either singly
and Maier (14) were just a few of the workers or in combination, have had an effect in increaswho thought that there was some evidence that ing the incidence of bronchiogenic carcinoma.
tobacco was an important factor in the increase
In the present paper the chief emphasis will
of cancer of the lungs. Muller (15) in 1939, be placed on our findings in regard to smoking.
METHOD OF STUDY
Source: Journal of the American Medical Assocation
143(4):329-336, 1950.
This study has been aided by a grant from the American
Cancer Society. Other phases of it will be presented in subsequent publications.
From the Department of Surgery, Washington University
School of Medicine and Barnes Hospital.
458
459
of the interview. The reason for this is the wellknown existence of a time lag between the exposure to a carcinogenic substance and the appearance of cancer. Many patients coming into
the hospital with chronic disease of the lungs
had stopped smoking months, or even years,
previously. We therefore asked the patients to
estimate the average use of tobacco during the
last twenty years of their smoking period. The
control patients were questioned in an identical
manner; thus any possible error lying in this
method of estimating smoking habits was balanced.
In questioning patients about occupations, we
attempted to learn all the occupations of a given
patient, the years during which he had held
these jobs and to what type of dusts or fumes he
had been exposed. Similar details were obtained in regard to other possible exposures,
such as those a patient might have had in connection with certain hobbies.
Classification of Smoking. In order to facilitate a statistical analysis of the results, the arbitrary classification of smoking habits given in
Table 2 was established. If a patient smoked for
less than twenty years, his amount of smoking
was adjusted to a twenty-year period. Thus a
T h e questionnaires were sent to male and female patients
with cancer of the lungs from Dr. W. L. Watson's Thoracic
patient smoking 20 cigarettes for ten years only
Surgery Service at Memorial Hospital, New York.
was classified as smoking 10 cigarettes daily
CALIFORNIA:
Private patients of Drs. L. Brewer, Daniels,
F. Dolley, D. Dugan, H. Garland, E. Holman, J. Jones, W. (class 2). Such adjustments were rarely necesRogers, P. Samson, B. Stephens. Hospitals: Birmingham
sary, since only a few patients had smoked for
General, French, Good Samaritan, Letterman General, Los
less than twenty years.
Angeles County, Southern Pacific General, Stanford, United
If a man smoked habitually more than one
States Marine, United States Naval University of California,
Wadsworth General. COLORADO:
Private patients of Drs. A.
type of tobacco during the last twenty years, the
Brown, F. Condon, ,J. Grow, F. Harper, M. Peck. Hospitals:
various types were added together to make up
Colorado General, Denver General, Fitzsimmons, Fort Logan
Veterans, General Rose, St. Lukes. DISIKICTOF COLUMBIA: his classification. Thus a man who smoked one
Private patients of Drs. B. Blades, E. Davis. Hospitals:
package of cigarettes daily, as well as two cigars,
Georgetown, George Washington, Walter Reed. ILLINOIS:
was classified as a class 4, or an excessive,
Hospitals: Cook County, Veterans Administration, Hines, Ill.
MARYLAND:
Hospitals: Johns Hopkins, United States Naval.
smoker.
MASSACHUSETTS:
Private patients of Drs. D. Harken, R. Overholt. Hospitals: Boston City, Massachusetts General, New
England Deaconess. MICHIGAN:
Hospitals: Dearborn Veterans, M i s s o u ~ r Private
:
patients of Drs. J. Flance, A. Goldman, R. Smjth. Hospitals: Jefferson Barracks Veterans, Jewish, St. Louis City, St. Louis County. NEWJERSEY:Hospitals:
Berthold S. Pollack, Newark City. NEWYORK: Private patients
of Drs. W. Cahan, H. Maier, J. Pool, W. Watson. Hospitals:
Bellevue, Veterans Administration, Brooklyn Cancer Institute, Kings County Memorial, Montefiore, New York City
Cancer Institute, New York Hospital, Presbyterian, Roswell
Park Memorial Institute. OHIO:Hospitals: Veterans Administration. PENNSYLVANIA:
Private patients of Dr. J. Johnson.
Hospitals: Jefferson Medical College, Temple University,
University of Pennsylvania. UTAH: Private patients of Drs. W.
Rumel, Cutler. Hospitals: Holy Cross, St. Marks, Salt Lake
County General, Veterans Administration, Dr. W. H. Groves,
Latter-Day Saints.
460
types were the epidermoid or squamous carcinoma and its variant the undifferentiated carcinoma. These are the most common types found
in males. In females the adenocarcinoma has so
...........................................
Age: . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1. Have you ever had a lung disease? If so, state time, duration, and site of disease:
Tuberculosis
Chest Injuries
Asthma
Lung Abscess
Pneumonia
Influenza
Yes
Bronchiectasis
Others
No 0
Cigarettes .......................
Yes
Cigars .......................
Pipes .......................
No 0
7. Do you have a chronic cough which you attribute to your smoking, especially upon first smoking in the
morning? If so, for how long?
No
Yes 0
8. Do you smoke before or after breakfast?
Before 0 After
9. Name the brand or brands, and dates, if any given brand has been smoked exclusively for more than five
years.
Change frequently? 0
First brand-from
Second brand-from
19.
. . . to
19.
19.
...
. . . to 19. . . .
10. What kind ofjobs have you held? Have you been exposed to dust or fumes while working there? (Use back
of page for detailed description of possible exposure)
To
From
Dust or Fumes
Position
11. Have you ever been exposed to irritative dusts or fumes outside of your job? In particular have you ever
..............................
12. How much alcohol do you or have you averaged per day? State time and duration in years.
Whiskey
.......................
Beer .......................
Wine
.......................
13. Where were you born and where have you lived most of you life? State the approximate time span you have
14. State the cause of death of your parents, and of brothers and sisters if any
Microscopic Diagnosis
Papanicolaou C h s
Interuiewer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Etiologtcal Class
........................
Group 0
3
4
5
46 1
462
13.6 x 2.3
100
= 0.312, the
463
Figure 1. Percentages for amount of smoking among 605 male patients with cancer
of the lungs (solid bars) and 780 men in the general hospital population without
cancer (lined bars) with the same age and economic distribution.
LIGHT
MODERATELY
HEAVY
HEAVY
EXCESSIVE
CHAIN
Table 3. Control Study 11: amount of smoking in 83 cases of proved cancer of the lung as determined by
investigators using the same questionnaire as that used in the cases of this study.
Cases
Amount of Smoking:
None
Light
Moderately heavy
Heavy
Excessive
Chain
Bellevue
Hospital
(Turner)
Boston
City
Hospital
(Ware)
Crilc
Veterans
Hospital
(Surington)
Hines
Veterans
Hospital
(Shabart)
22
16
15
30
0
0
3
0
la
0
9
6
2
2
6
8
3
0
1
6
14
464
NONE
LIGHT
MODERATELY
HEAVY
HEAVY
EXCESSIVE
CHAIN
Table 4. Study 111: Amount of smoking among male patients with cancer of the lung in relation to area
or hospital where cases were observed (Wynder).
Smoking classification, %
Areas or hospitals
Barnes (76)
Los Angeles (50)
San Francisco (50)
Mountain states (50)
St. Louis (25)
Eastern and northern
states (50)
New York City (66)
Memorial Hospital (55)
Total (422)
a
V
18.4
12.0
20.0
14.0
20.0
IV
31.5
32.0
28.0
26.0
36.0
I11
43.4
12.0
36.0
36.0
32.0
11
5.3
12.0
14.0
18.0
12.0
I
0.0
0.0
2.0
4.0b
0.0
0
1.3
4.0
0.0
2.0
0.0
Average
Age
54.8
59.5
54.9
60.1
58.2
16.3
20.0
23.6
23.6
40.0
27.3
40.0
32.3
29.1
20.0
7.7
10.9
0.0
0.0
5.5b
0.0
0.0
3.6
56.1
55.8
57.6
18.2
30.8
36.7
11.4
1.4
1.4
56.7
This Table does not include any cases represented in control studies 1 and 11.
Includes one minimal smoker.
465
Percentage
of cases
30-39
40-49
50-59
60-69
70-79
2.3
17.4
42.6
30.9
6.8
Figure 3. Control study II: percentages for amount of smoking in 83 caaes of cancer
of the lungs collected independently by Dr. E. J. Sbabart (Chicago), Dr. C. T.
Surington (Cleveland),Dr. H. G. Turner (New York), and Dr. G. W. Ware (Boston).
31.3
3.7
2.4
LIGHT
MODERATELY
HEAVY
HEAVY
EXCESSIVE
CHAIN
Table 6. Percentage distribution of amount of smoking in respect to age groups among 780 men in the
general hospital population and 605 men with cancer of the lungs?
40-49
30-39
No. of cases
(146)
(14)
class: 0
1
2
3
4
5
13.6
5.5
17.1
41.0
14.3
8.2
7.1
7.1
14.3
42.9
28.6
0.0
Age Groups
50-59
(164) (105)
(210)
Amount of smoking,
9.7
0.0
14.8
9.7
1.9
7.1
18.9
3.8
17.6
37.1
29.5
43.3
14.0
28.6
10.5
10.3
36.2
6.7
60-69
70-79
(258)
(160)
(187)
(100)
(41)
1.6
1.6
7.4
36.0
34.1
19.4
14.3
18.7
20.6
28.7
10.6
6.8
1.1
1.1
13.6
38.0
30.5
15.5
25.0
13.0
21.0
16.0
15.0
10.0
2.4
12.2
24.4
29.3
17.1
14.6
The percentages for the general male hospital population are given in the left-hand columns
466
91.2
W w d e r and Graham
467
Figure 5. Percentages for duration of smoking in years, starting with the time when
the patient first began to smoke habitually, in 605 cases of cancer of the lungs.
20.0
17.2
18'0
17.2
7.9
6.4
=
1.3
1.8
0.3
0.0
and only 1.2 percent of the controls in the cancer age13 were excessive or chain smokers for at
least twenty years (Figure 6).
COMMENT
See Table 5 .
Figure 6. Amount of smoking in percentage among 780 male patients (iined bars) and
552 female patients (solid bars) of the general hospital population with the same age
and economic distribution as found among cases of cancer of the lungs.
35.6
11.5
LIGHT
MODERATELY
HEAVY
HEAVY
EXCESSIVE
CHAIN
468
Infiuence of Tobacco. Since in a small percentage of cases cancer of the lungs occurs in
nonsmokers and minimal smokers and since it
obviously does not develop in every person who
has been a heavy smoker for a long time, it is
apparent that smoking cannot be the only
etiologic factor in the induction of the disease.
From the evidence presented, however, the
temptation is strong to incriminate excessive
smoking, and in particular cigarette smoking,
over a long period as at least one important
factor in the striking increase of bronchiogenic
carcinoma for the following reasons: (1) it is
rare to find a case of epidermoid or undifferentiated carcinoma in a male patient who has not
been at least a moderately heavy smoker for
many years; ( 2 ) the use of cigarettes is much
greater among patients with cancer of the lungs
than among other patients of the same age and
economic groups; (3) the sex distribution of
cancer of the lungs roughly corresponds to the
ratio of long-term smoking habits of the two
sexes (see section on Duration of Smoking);
(4) the enormous increase in the sale of cigarettes in this country approximately parallels
the increase of bronchiogenic carcinoma.14
Amount of Smoking. The data have clearly
shown that the average patient with cancer of
the lungs smokes much more heavily than the
average patient of the same age and economic
group with some other disease. This contrast
becomes even greater if our observation of Mormons is considered, who as a group smoke far
less than the general hospital population; Mormons with cancer of the lung were, with one
exception, considerable smokers.
The fact that patients with bronchiogenic
cancer in their forties and fifties had smoked
more heavily than those in whom the cancer
developed at a later age may indicate that the
greater the irritation the sooner will cancer develop in a susceptible person. Such an observation obviously does not apply to the individual
case but rather to the age groups taken collectl4 It is taken for granted, of course, that by itself such
parallelism would mean little since similar curves could be
drawn for many other commodities.
469
is greater than among the general hospital population. It seems, therefore, that tobacco smoke
has also some influence on the induction of
adenocarcinoma in men, even though, as
judged from the incidence, the influence on
this type is much less marked than on the other
types of bronchiogenic carcinoma.
Cancer of the Lungs in Women. Many observers have commented on the fact that bronchiogenic carcinoma, while on the increase
among men and women, is increasing more
rapidly among men. In 100 consecutive cases
collected by Lindskog (19)in 1938 to 1943 the
ratio was 4.5 to 1, and in another series collected in 1947 and 1948 the ratio had reached
24 to 1 (20).At Barnes Hospital the ratio in our
last 150 cases has been 18.5 to 1. This shift in
ratio has been noted in varying proportions
throughout the country. Such a radical change
warrants a careful analysis.
The insufficient number of cases of cancer of
the lungs in women in our survey does not allow
definite conclusions at this time. So far, however, smoking seems to have had no apparent
effect on the incidence of adenocarcinoma in
women. It is of great interest that we have observed 10 cases of epidermoid and undifferentiated carcinoma of the lungs of women who were
nonsmokers with no history of occupational or
other irritative exposure. This percentage of
nonsmokers in women with cancer of the lung
is much higher than that found among men.
Proper explanations for this finding remain to
be advanced. At the same time it appears
strongly suggestive that heavy smoking plays a
significant role in the induction of epidermoid
and undifferentiated carcinoma of the lungs in
women, since the percentage of heavy smokers
is considerably higher in the cancer group than
in the general hospital control group.
CONCLUSION AND SUMMARY
470
47 1
472
Cancer
Noncancer
1045
605
Pb
a
All types
Cigarettes
Pipes
Cigars
84.8
77.8
0.01
56.0
46.0
0.01
30.3
24.3
0.01
22.4
20.8
0.47
Age standardized.
P denotes probability here and in Tables 2 and 3.
80.0
60.0
11
All Smokers
I-
z
W
U
a
W
40.0
v)
20.0
0
AGE IN YEARS
(ON ADMISSION)
473
Lwin et al.
Table 2. Prevalence of smokers among male patients by type of smoking and diagnostic group.
Percentage of smokers
Diagnosis
No. of
cases
Cigarettes (2)
%
P
Pipes (3)
P
Cigars (4)
%
P
236
666
124
48 1
84.7
82.9
81.1
78.3
0.53
0.39
0.03
66.1
48.0
53.1
44.1
0.01
0.02
0.01
13.5
25.8
25.5
25.3
0.01
0.09
0.01
11.2
20.3
13.4
22.7
0.01
0.64
0.01
143
666
51
554
84.5
82.9
74.0
78.1
0.58
0.09
0.23
45.3
48.0
43.0
46.4
0.48
0.78
0.81
48.1
25.8
30.7
23.8
0.01
0.02
0.01
26.5
20.3
34.9
19.6
0.11
0.22
0.08
Age standardized.
Table 3. Prevalencea of cigarette and pipe smokers among male patients by duration of smoking and
diagnostic group.
Duration of smokinE
Under 25 yr.
Diagnosis
No. of
cases
(1)
No.
(2)
%
236
666
25
74
11.7
13.0
124
48 1
19
92
6.3
4.3
(3)
No.
(4)
0.62
148
233
54.1
34.9
0.01
0.23
0.34
54
128
36.9
29.8
0.01
0.01
Cigarette smokers
Lung cancer compared with:
Other cancer (except lung and
lip)
Lung nontumors
Other noncancer
Pipe smokers
Lip cancer compared.with:
Other cancer (except lung and
lip)
Lip nontumors
OLher noncancer
143
666
8
26
7.8
3.9
0.02
60
162
35.7
22.9
0.01
51
554
5
35
0.6
5.8
0.35
0.24
11
87
21.5
17.9
0.04
0.01
Age standardized.
Table 4. Comparison of the proportiona of cases of lung and lip cancer among male nonsmokers and
smokers of 25 years duration or more at Roswell Park Memorial Institute, 1938-1948.
Lung Cancer
No. of
Lip Cancer
persons
Cases
Rate
Pb
Cases
Rate
Pb
293
600
353
263
23
148
33
22
8.6
20.7
8.6
8.5
0.01
1.00
0.97
20
37
60
39
6.9
5.9
13.4
12.5
0.55
0.01
0.03
474
Table 5. Comparison of the proportion. of cases of lung and lip cancer among male nonsmokers and
smokers of 25 years duration or more at Roswell Park Memorial Institute, 1938-1948.
Lip cancer
Lung cancer
No. of
persons
Cases
Rate
Pb
Cases
Ratea
Pb
293
76 1
479
166
116
23
148
115
19
14
8.6
17.1
20.9
10.6
12.8
0.01
0.01
0.48
0.18
20
73
22
34
17
6.9
8.9
5.1
15.7
11.6
0.29
0.29
0.01
0.13
In England and Wales the phenomenal increase in the number of deaths attributed to
cancer of the lung provides one of the most
striking changes in the pattern of mortality recorded by the Registrar-General. For example,
in the quarter of a century between 1922 and
1947 the annual number of deaths recorded
increased from 612 to 9287, or roughly fifteenfold. This remarkable increase is, of course, out
of all proportion to the increase of population-both in total and, particularly, in its older
age groups. Stocks ( I ) , using standardized
death rates to allow for these population
changes, shows the following trend: rate per
100 000 in 1901-1920, males 1.1, females 0.7;
rate per 100 000 in 1936-1939, males 10.6,
females 2.5. The rise seems to have been particularly rapid since the end of the First World
War: between 1921-1930 and 1940-1944 the
death rate of men at ages 45 and over increased
sixfold and of women of the same ages approximately threefold. This increase is still continuing. It has occurred, too, in Switzerland, Denmark, the U S A . , Canada, and Australia, and
has been reported from Turkey and Japan.
Many writers have studied these changes,
considering whether they denote a real increase
in the incidence of the disease or are due
merely to improved standards of diagnosis.
Some believe that the latter factor can be regarded as wholly, or at least mainly, responsible-for example, Willis (Z), Clemmesen and
Busk (3), and Steiner (4). On the other hand,
Kennaway and Kennaway ( 5 ) and Stocks (1)
have given good reasons for believing that the
rise is at least partly real. The latter, for instance, has pointed out that the increase of
certified respiratory cancer mortality during
Source: British Medical Jouml, September 30, 1950, pp.
739-74s.
Member of the Statistical Research Unit of the Medical
Research Council.
* Professor of Medical Statistics. London School of Hygiene
and Tropical Medicine; Honorary Director of the Statistical
Research Unit of the Medical Research Council.
475
476
THE DATA
477
478
Disease group
Group A.
Diagnosis confirmed
at necropsy,
etc.
Group B.
Other criteria
of
diagnosis
489
178
412
220
28
19
Carcinoma of lung
Carcinoma of stomach
Carcinoma of colon and rectum
Other malignant diseases
Diseases other than cancer (controls)
Other cases
Excluded
Total
709
206
43 I
81
709
335
4
2475
All cases
It will be seen that the lung-carcinoma patients and the control group of non-cancer patients are exactly comparable with regard to sex
and age, but that there are some differences
with regard to social class and place of residence. The difference in social class distribution is small and is no more than might easily be
due to chance ( X 2 = 1.61; n = 2 ; 0.30<P<0.50).
The difference in place of residence is, however,
large (X* = 3 1.49; n = 5 ; P<O.OO l), and Table 2
shows that a higher proportion of the lung patients were resident outside London at the time
Table 2. Comparison between lung-carcinomapatients and non-cancer patients selected as controls, with
regard to sex, age, social class, and place of residence.
Age
25303540455055606570-74
All ages
No. of
lungcarcinoma
patients
No. of
noncancer
control
patients
M
F
18
36
87
130
145
109
88
28
1
0
3
4
10
11
9
9
9
4
2
6
18
36
87
130
145
109
89a
27a
1
0
3
4
10
11
9
9
9
4
649
60
649
60
Social class
(RegistrarGeneral's
categories,
men only)
No. of
lungcarcinoma
patients
No. of
noncancer
patients
I and I1
111
IV and V
77
388
184
87
396
166
649
649
330
203
377
23 1
23
95
43
16
54
27
15
709
4
709
All classes
Place of residence
County of London
Outer London
Other county
borough
Urban $strict
Rural district
Abroad or in
services
Total (M
F)
One control patient was selected, in error, from the wrong age group.
London, and none elsewhere; of their non-cancer controls the corresponding numbers were
60, 38, and 0, clearly an insignificant difference.
It is evident, therefore, that the control group
of patients with diseases other than cancer is
strictly comparable with the group of lung-carcinoma patients in important respects but differs slightly with regard to the parts of England
from which the patients were drawn. It is unlikely that this difference will invalidate comparisons, but it must be kept in mind; fortunately, it can be eliminated, if necessary, by
confining comparisons to the smaller group of
patients seen in the district hospitals.
ASSESSMENT OF SMOKING HABITS
479
480
Table 3. Amount of tobacco smoked daily before present illness as recorded at two interviews with the
same patients at an interval of six months or more.
Second interview, no. of persons smoking
First
interview;
no. of
persons smoking
1 cig.-
1
4
1
1 cig.-
5 cigs.15 cigs.-
25 cigs.-
5 cigs.-
1
13
4
50 cigs. or more
Total
18
15 cigs.-
25 cigs.-
3
9
1
13
50 cigs.+
3
1
0
0
Total
9
5
17
14
50
Table 4. Proportion of smokers and non-smokers in lung-carcinoma patients and in control patients with
diseases other than cancer.
No. of
nonsmokers
No. of
smokers
Probability
test
Males:
Lung-carcinoma patients (649)
2 (0.3%)
647
P (exact method)
= (0.00000064
27 (4.2%)
622
Females:
Lung-carcinoma patients (60)
19 (31.7%)
41
32 (53.3%)
28
Disease group
X 2 = 5.76;n = 1
0.01<P<0.02
481
Table 5. Most recent amount of tobacco. consumed regularly by smokers before the onset of present
illness: lung-carcinoma patients and control patients with diseases other than cancer.
No. smoking daily
50 cigs. +
Probability
test
Disease group
1 cig.-.
5 cigs.-
15 cigs.-
25 cigs.-
Males:
Lung-carcinoma patients
(647)
33
(5.1%)
250
(38.6%)
196
(30.3%)
136
(2 1.0%)
32
X 2 = 36.95;
(5.0%) n = 4;
P<0.001
293
(47.1%)
190
(30.5%)
71
(11.4%)
13
(2.1%)
19
(46.3%)
9
(22.0%)
6
(14.6%)
10
(35.7%)
6
(21.4%)
0
(0.0%)
0
X2=5.72;
(0.0%) n = 2 ;
0.05<P<0. 10
(Women smoking
0
15 or more ciga(0.0%) rettes a day
grouped together)
12
(42.9%)
a Ounces of tobacco have been expressed as being equivalent to so many cigarettes. There is 1 oz of tobacco in 26.5 normalsize cigarettes, so that the conversion factor has been taken as: 1 oz of tobacco a week = 4 cigarettes a day.
482
WOMEN
32
10
1-4
5-14
15-24
25
Control patients
without cancer
Table 6. Maximum amount of tobacco ever consumed regularly by smokers: lung-carcinoma patients and
control patients with diseases other than cancer.
No. smoking as a daily maximum
Disease group
1 cig.-
5 cigs.-
Males:
Lung-carcinoma patients
(647)
24
(3.7%)
38
(6.1%)
Probability
15 cigs.-
25 cigs.-
50 cigs.+
test
208
(32.1%)
196
(30.3%)
174
(26.9%)
45
(7.0%)
X2=23.16;
n=4;
P<O.O01
242
(38.9%)
20 1
(32.3%)
118
(19.0%)
23
(3.7%)
6
15
(14.6%) (36.6%)
12
(29.3%)
8
(19.5%)
(0.0%)
12
9
(42.9%) (32.1%)
6
(21.4%)
0
(0.0%)
1
(3.6%)
X2=7.58;
n=2;
0.02<P<0.05
(Women smoking 15
or more cigarettes a
day grouped together)
483
Table 7. Estimate of total amount of tobacco ever consumed by smokers: lung-carcinoma patients and
control patients with diseases other than cancer.
No. who have smoked altogethet
Disease Group
365
cigs.
Males:
Lung-carcinoma patients
(647)
19
(2.9%)
50 000
cigs. -
150 000
cigs. -
250 000
cigs. -
500 000
cigs. +
145
(22.4%)
183
(28.3%)
225
(34.8%)
X 2 = 30.60;
75
(11.6%) n = 4 ;
P<O.OOl
190
(30.5%)
182
(29.3%)
179
(28.9%)
35
(5.6%)
19
(46.3%)
5
(12.2%)
7
(17.1%)
0
(0.0%)
5
(17.9%)
3
(10.7%)
1
(3.6%)
0
(0.0%)
19
(67.9%)
Probability
test
X2=12.97;
n=2;
0.0001<P<0.01
(Women smoking
15 or more cigarettes a day
grouped together)
Table 8. Age of starting to smoke, number of years smoked, and number of years stopped smoking in
lung-carcinomapatients and control patients with diseases other than cancer (male and female).
Age at
Under
20
20-
Lungcarcinoma
patients
No.
%
Control
patients
No.
%
541 78.6
118 17.2
488
129
75.1
19.8
650
No. of
years
smoking
Lungcarcinoma
patients
No.
%
it]
110 2040+
351
302
Total
688
X24.65;
Control
patients
No.
%
5.1
51.0
43.9
7.7
338
262
52.0
40.3
650
n = 2 ; 0.05<P<0.10
No. of
years
1-
10 20 +
Total
Lungcarcinoma
patients
No.
%
Control
patients
No.
%
649
30
590
37
5
688
94.3
4.4
1.3
90.8
5.7
%, 3.5
650
X2=8.59; n = 2 ; O.Ol<P<O.02
484
INHALING
The method by which the patients with carcinoma of the lung were obtained has been discussed earlier; there is no reason to suppose
that they were anything other than a representative sample of the lung-carcinoma patients
attending the selected London hospitals. The
control patients, as was shown in Table 2, were
exactly comparable so far as sex and age were
concerned and they were sufficiently comparable with regard to social class for the difference
between the two series to be ignored. They were
not wholly comparable from the point of view
485
Table 9. Most recent amount smoked by lung-carcinoma and control patients seen in district hospitals
(male and female).
No. smoking daily
Disease group
1 cig.-
5 cigs.-
15 cigs.-
25 cigs.+
12
36
27
21
9
X2=
50
19
11.68; n = 4 ; O.Ol<P<O.O2
11
characteristic of-persons suffering from all diseases other than carcinoma of the lung. Yet in
Table 10 the smoking habits of the patients in
five main groups of diseases are compared, allowing for their sex and age composition, and
no significant difference can be demonstrated
between them. (We have brought into this table
all the patients with diseases other than carcinoma of the lung.)
As in other tables where sex and age differences between groups have had to be taken into
account, the expected numbers have been obtained by taking the actual numbers of patients
with each type of disease in each age and sex
subgroup, and calculating what proportion of
them would fall in each smoking category if
they had had exactly the same habits as all the
patients included in the Table. In other words,
we have computed what ought to be the smoking habits of each disease group if it behaved in
each sex and at each age like the total popula-
Table 10. Most recent amount smoked by all patients other than those with carcinoma of lung, divided
according to type of disease (male and female).
No. smokinp dailv
Disease group
1 cig.-
5 cigs.-
15 cigs.-
25 cigs.t
2361
220.0
237
236.9
110
122.8
57
53.0
34
38.I
39
55.7
31
32.3
128
136.1
64
73.8
143
130.2
98
84.1
22
17.7
78
85.3
33
29.7
19
16.7
38
39.5
81
75.8
23
18.3
34
34.5
38
36.6
24
21.1
91
86.0
44
48.9
18
22.1
42
47.0
486
Table 11. Most recent amount smoked by all patients other than those with carcinoma of lung, divided
according to whether they were notified or selected for interview (male and female).
No. smoking daily
Method of selection of
patient
307a
301.8
70
75.2
1 cig.-
114
119.0
71
66.0
5 cigs.-
354
345.2
309
317.8
15 cigs.-
179
25 cigs.+
78
186.1
80.0
192
88
184.9
86.0
487
Table 12. Most recent amount smoked by patients with carcinoma of lung and by patients thought
incorrectly by the interviewers to be suffering from carcinoma of lung (male and female).
No. smoking daily
Disease group
Patients with carcinoma of lung (709)
Patients incorrectly t h o u g h t to have carcin o m a of l u n g (209)b
1 cig.-
2l a
40
48.0
25
17.0
31.7
35
24.3
5 cigs.-
269
276.0
83
76.0
15 cigs.-
205
201.0
50
54.0
25 cigs.+
174
152.7
16
37.3
X2=29.76; n = 4 ; P<O.OOl.
See footnote to Table 10.
There is a large number of cases in this group because one hospital notified all cases admitted for bronchoscopy; 147 out of
the 209 incorrectly thought to have carcinoma of the lung were interviewed at this hospital.
a
DISCUSSION
Table 13. Most recent amount smoked by patients incorrectly thought by the interviewers to be suffering
from carcinoma of lung and all other patients not suffering from carcinoma of lung (male and female).
No. smoking daily
Disease group
35a
36.8
340.2
342
1 cig.-
25
20.4
160
164.6
5 cigs.-
15 cigs.-
83
82.0
50
48.8
580
581.0
32 1
322.2
25 cigs. +
16
20.8
150
145.2
488
Table 14. Ratios of patients interviewed with carcinoma of lung and with a
given daily consumption of tobacco to the estimated populations in Greater
London smoking the same amounts (male and female combined; ratios per
million).
Daily consumption of tobacco
Age
25 35 45 55 65-74
a
OQ
2
12
14
21
1-4 cigs.
5-14 cigs.
15-24 cigs.
25-49 cigs.
I1
9
34
133
110
2
43
178
380
300
6
41
24 1
463
510
28
67
429
844
1063
50 cigs.+
77
667
600
2000
Ratios based on less than 5 cases of carcinoma of the lung are given in italics.
Total
4
29
147
244
186
Figure 2. Death rate from cancer of the lung and rate of consumption of tobacco
and cigarettes.
a:
W
n
250
(3
200
1-
489
6.0
v)-
a
Q
5.0
4.0
pK
00)2
2,:
Q:
O W
+n
2.0
1.0
D.5
3
THE RATES ARE EASED ON 3 YEAR AVERAGES FOR ALL
YEARS EXCEPT 1947.
L U
g
9.g
!.id
z
3
v)
z
0
O
490
SUMMARY
49 1
Social Factors
It is an old clinical experience that cancer of
the cervix uteri is most common in the lower
strata of society, whereas cancer of the body is
more common on a higher social level.
Theilhaber ( I ) was among the first to study
the social distribution of cervical cancer. His
series comprised 5848 women, 133 of whom
had cervical cancer. Patients with cancer of the
body were ruled out. After dividing this series
into 11 social groups, Theilhaber found 3.7 percent in the highest three groups and 96.3 percent in the remaining groups.
In 1910, Theilhaber and Greischer (2) studied the death certificates for women over 25
years of age dying in Munich during the period
1907-09. The number of women dying from
cervical carcinoma proved to increase down the
social scale, while the reverse applied to breast
cancer. They found also that poor patients with
cervical cancer died sooner than the well-to-do
ones.
An objection which may be raised against
Theilhaber and Greischers investigation is the
absence of any account of the social distribution
of the entire population in Munich. Moreover,
they do not give the age distribution in their
own material.
The relation of social factors to uterine carcinoma in England and Wales was first dealt with
in the Report of the Registrar General for
1930-32 (3).This series, comprising 7831 cases
of uterine carcinoma among married and 1294
among unmarried women, is divided into five
occupational groups (the husbands as far as the
married women are concerned). The mortality
from uterine cancer proved to be higher in the
lower classes of society. Hurdon ( 4 ) has reSource: Excerpted from J@rgen Rejel, The Interrelation Behueen Utoine Canrer a d Syphtlis: A Patho-Demoraphic Study.
Table 1. Social distribution of 1262 patients with cervical cancer, 176 patients with cancer of the body,
and 1392 controls. The incidence of syphilis is given for each social group.
~
Social
group
A. Absolute
number of
patients
with cervical
cancer
E. Absolute
number of F. Syphilitic
D % of C
controls
controls
F % of E
~
A
B
C
D
Total
101
263
278
620
14
28
43
82
1262
167
13.8
10.6
15.4
13.2
38
59
30
49
0
1
5
4
1.6
16.6
8.0
176
10
100
389
319
584
2
10
14
31
2.0
2.5
4.3
5.3
1392
57
494
moreover the incidence of syphilis in the various groups. In a more detailed investigation of
the social status-to which I shall revert-the
series were divided into age groups. In this
analysis, however, no regard was paid to age, as
it does not purport to be any more than a preliminary orientation to find whether the poorer
classes of society show an accumulation of syphilitics among the women with cervical carcinoma. If this had shown a massive accumulation of syphilitics in the lowest social groups, the
preponderance of syphilis found in the series
with cervical cancer might have been due to a
chance coincidence.
Such an accumulation, however, is not apparent. On the contrary, the incidence of syphilis
among the patients with cervical carcinoma is
higher in social group A than B. It is, therefore,
evident that syphilitics are not accumulated
among the patients with cervical carcinoma in
the lowest social strata, but are diffusely distributed in all four social groups. In the control
series the syphilitics are also diffusely distributed in all four social groups.
495
Rhel
Table 2. Number of persons among the patients with cervical carcinoma and the controls who were
found on the files of the Third Police Chamber.
Controls
Patients
Absolute
numbers of % of 1262
pts. with cer- patients with
vial
cervical
Number of
carcinoma
carcinoma
syphilitics
Absolute
number of
controls
% of 1392
controls
Number of
syphilitics
~~
Prostitutes
Semiprostitutes
Presumed prostitutes
(not proved)
Involved in infertion
proceedings
Delinquents
Total
0.8
0
5
3
1
5
0.07
0.03
26
17
40
6
3.1
0.4
15
0
11
0
0.3
11
15
0.8
1.1
77
2.
496
Table 3. Age distribution of 1262 patients with cervical cancer, 176 patients with cancer of the body, and
1392 controls; the incidence of syphilis is stated in percent for each age group.
A. Absolute B. Absolute
D. Absolute
number of number of
C. Absolute
no. of
patients
syphilitic
number of syphilitic
F. Absolute
with cerpts. with
pts. with
pts. with
E. Absolute number of
vical
cervical
cancer of cancer of
number of' syphilitic
B % o f A the body the body D % o f C controls
controls
F % of E
cancer
cancer
Age
~~
-40
41-50
51-60
61-70
71Total
129
36 1
362
266
144
7
61
51
36
12
1262
167
5.4
16.9
14.1
13.5
8.3
-
0
10
39
71
56
0
3
2
3
2
176
10
30
5.1
4.2
3.6
-
136
357
414
299
186
3
20
25
5
2.2
5.6
6.0
1.7
2.2
1392
57
Table 4. Age and social distribution in the patient and control series; incidence of syphilis stated in
percent for each age group.
Social
group
Age
-40
41-50
51-60
61-70
71Total:
C + D
less
prostitutes
-40
41-50
51-60
61-70
71Total:
prostitutes
-40
41 -50
5 1-60
61-70
71Total:
Absolute
Absolute number of
number of syphilitic
patients
patients
with
with
cervical
cervical
cancer
cancer
Absolute
number of
pts. with
cancer of
the body
Absolute
number of
syphilitic
patients
with cancer
of the body
3
-
25
73
129
186
76
0
4
5
3
0
5
4
2
1.0
489
12
2.5
Absolute
Absolute number of
number of syphilitic
controls
controls
19
65
52
183
45
2
10
14
11
5
10
15
26
6
11
24
32
37
0
0
0
1
0
364
42
11.5
97
105
277
296
75
99
5
45
30
23
7
4
16
10
30
7
0
6
15
39
19
0
3
2
2
2
50
13
5
10
109
282
280
111
110
3
15
17
2
4
2
5
6
852
110
12.9
79
11.4
892
41
4.5
5
19
14
8
0
0
6
7
2
0
2
2
5
2
0
0
1
3
0
0
46
15
11
33
36
497
Rdiel
~
Table 5. Annual assessable income of 1392 controls and 1438 patients with uterine cancer.
Controls
Assessahle income
Age
Income
not
known
~
21-25
26-30
3 1-35
36-40
4 1-45
46-50
51-55
56-60
61-65
66-70
71Total
<lo00
kr.
~~
1-4000
kr.
4-9000
kr.
>9000
kr.
~
11
26
22
31
31
22
42
0
2
11
27
53
87
64
61
47
25
20
14
9
10
8
9
34
196
697
397
68
2.4
14
50
28.5
4.8
0
0
4
4
6
5
3
1
4
5
<lo00
kr.
1-4000
kr.
4-9000
kr.
>go00
kr.
1
3
20
50
62
95
109
96
75
76
110
0
0
4
Income
not
known
0
0
1
4
8
5
25
60
1
1
16
40
75
110
101
91
104
75
95
0
0
12
31
41
62
64
65
40
24
18
0
0
1
1
7
10
8
5
11
8
12
68
241
709
357
63
4-7
16.7
49.3
24.8
4.3
0
0
0
5
10
18
9
2
5
4
15
0
1
7
13
19
19
24
32
41
% of
1392
controls
% of
1438
patients
with
uterine
cancer
498
Figure 1. Graphic repmaentation of the annual assearable income of 1392 controls and 1438 patients with
uterine cancer.
110
100
90
cn
80
70
-4000 kr (controls)
1-4000 kr (patients)
8a
60
UI
50
40
30
4-9000 kr (controls)
4-9000 kr (patients)
> 9000 kr (patients)
> 9000 kr (controls)
20
10
I
21-25
31-35
41-45
51-55
61-65
7126-30
36-40
46-50
56-60
66-70
AGE
R$jel
499
Table 6. Number of controls and patients who bad received poor relief, been convicted or had otherwise
forfeited their civil rights.
Controls
Patients
Absolute
number of
husbands
55
31
9
Absolute
number of
husbands
39
24
7
% of 902
husbands
Absolute
number of
women
% of 1392
controls
6.2
87
13
22
% of 780
husbands
Absolute
number of
women
% of 1438
patients with
uterine cancer
59
12
37
4.1
3.4
5
3
-
1.5
2.5
does not eliminate the preponderance of syphilitics among the patients with cervical carcinoma.
References
(I) Theilhaber, A. Zur lehre von der enststehung
der Uterustumoren. Miinch Med Wsch 58:1272, 1909.
(2) Theilhaber, A. and S. Greicher. Zur aetiology
der carcinome. Z Krebsforsch 9:530, 1910.
(3) Registrar Generals Decennial Supplement.
England and Wales, 1931. Part 11. London, 1938.
(4) Hurdon, E. Cancer ofthe Uterus. London, 1942.
(5) Smith, F.R. Nationality and carcinoma of the
cervix. A m J Obstet GynecoZ41:424, 1941.
(6) Kennaway, E.L. T h e racial and social incidence of cancer. Br 1 Cancer 2:177. 1948.
(7) Handley, W.S. The prevention of cancer. Lancet 987, 1936.
(8) Handley, W.S. Penile carcinoma. BY Med J
11:841, 1947.
(9)Plaut, A. and A.C. Kohn-Speyer. T h e carcinogenic action of smegma. Science vol. 105. April 11,
1947.
( l a ) Clemmesen, J. and A. Nielsen. The incidence
of malignant diseases in Denmark 1943 to 1947.
Copenhagen, 1952.
J
500
Lilien@ki el al.
obtaining this information from the hospital
record was not informed which birth was a case
or control.
The behavior problem children were those
who had been referred by public school teachers or principals in Baltimore to the Division of
Special Services of the Department of Education and who had been born during the period
1940-1945 (5). (The psychological consultation
service is located in this division.) A child of the
same sex in the same class as the child with the
behavior problem and who had not been referred to the Division of Special Services was
selected as a control. This selection of a control
resulted in automatic matching by race, economic status, and age due to segregation and
school districting. Controls were chosen in this
manner in order to be certain that they were
without behavior problems as defined in the
study. Birth certificates and hospital records
were reviewed and the necessary information
was obtained. The methods of selection of the
groups in these studies are summarized in
Table 1.
50 1
RESULTS
Table 1. Summary of sources of information for each type of neuropsychiatric disorder studied.
Type of
neuropsychiatric
disorder
No. of
cases
studied
Cerebral palsy
561
Epilepsy
564
Geographic
area of
study
Years of
birth of
cases
Upstate
New York
1940-1947 Crippled
Childrens
Register, New
York State
Department
of Health
Reporting on
birth certificate
Baltimore
Birth certificates
and hospital
obstetrical records
Baltimore
Child in same
classroom not
referred to Division of Specia1 Services
Birth certificates
and hospital
obstetrical records
424
Behavior prob1151
lems in children
Source of
cases
Source of
control
P u P
Source of record of
mothers pregnancy
experience
502
Type of
neuropsychiatric
disorder
Cerebral palsy
Abnormal
neonatal
Total
Prematurity conditions abnormalities
(percent)
(percent) (percent)
Not studied
Cases
Controls
38
21
22
5
White
Cases
Controls
27
19
13
4
17
6
34
25
Nonwhite
Cases
Controls
47
43
15
12
14
3
55
50
White
Cases
Controls
35
25
12
7
16
8
46
31
Nonwhite
Cases
Controls
59
55
18
12
7
6
65
60
White
Cases
Controls
33
25
6
2
10
39
31
Nonwhite
Cases
Controls
64
51
17
5
14
15
73
54
Cases
Controls
25
19
5
8
9
11
29
26
Epilepsy
Mental
deficiency
Behavior
disorders
Speech defects
White only
Maternal
age and
birth
order
+?a
Maternal
history of
previous
infant
loss
+b
Operative
procedures
--E
f ? = Suggestive association
b + = Definite association
c-
= No association
Lilienfld el al.
ropsychiatric disorders. The data presented in
Table 2 refer to cases with only one disorder.
The findings with regard to epilepsy and
mental deficiency are in general similar to those
observed in the case of cerebral palsy. There
exists an association of these two conditions
with complications of pregnancy and abnormal
neonatal conditions and prematurity. In addition, those complications which were actually
significant were those associated with nonmechanical fetal injury. In the case of mental
deficiency, there was an association with maternal age and birth order which was not present
in epilepsy. On the other hand, mothers of the
epileptics had had more previous infant loss
than the control group. This was not observed
among mothers of the mental defectives.
The frequencies of the abnormalities that
were studied were not significantly greater
among the nonwhite cases as compared to their
controls, although they were all in the same
direction as was found among the white cases. It
is not completely clear why the nonwhite differences were not significant. It is possible that the
much smaller number of cases was not sufficient to demonstrate a difference. In addition,
the nonwhites have a smaller percentage of hospitalized births leading to some selection for
abnormal pregnancies arid deliveries which
would tend to diminish differences between the
cases and controls. In neither of these disorders
was an association with operative procedures
found.
In the behavior disorder group, an increased
frequency of complications of pregnancy and
prematurity was observed. The specific complications that were excessive were those productive of nonmechanical fetal injury. N o association was found with regard to operative
procedures, maternal age, birth order, and
mothers history of previous infant loss. Of considerable interest was the fact that when the
various types of behavior disorders were studied individually only a specific type was found
to be associated with maternal and fetal factors.
The behavior disorders so incriminated were
those classified as the confused, disorganized
and/or hyperactive group.
In the case of speech defects where there
exists an impression that there is a relationship
with pregnancy experience, no associations with
the factors studied were found. Frankly, we
must admit that such a completely negative re-
503
sult was comforting since it eliminates the possible existence of a constant bias that may have
been inadvertently introduced into the routine
processing of the material.
Several considerations must be taken into account in interpreting these results. Of prime
importance is the fact that in cerebral palsy,
mental deficiency, and epilepsy the association
is not with a single set of factors. There appears
to exist a pattern of relationship which provides
additional support to the findings. This is not
true in the case of behavior disorders where the
association would appear to be more limited in
scope. However, in this instance, some weight
must be given to the fact that the association
appears to exist with a particular type of behavior disorder which has been suspected on clinical grounds as being the result of organic brain
damage.
DISCUSSION
504
Lilienfeld et al.
References
( 1 ) Lilienfeld, A.M. and E.A. Parkhurst. A study of
the association of factors of pregnancy and parturition with the development of cerebral palsy: Preliminary report. Am J Hyg 53:262, 1951.
505
Method
An attempt was made to trace all children in
England and Wales who had died of leukemia
or cancer before their tenth birthday during the
years 1953 to 1955 (case group) and to compare
their prenatal and postnatal experiences with
those of healthy children (control group). Details of control selection and method of recording data are given later, but the basic idea was to
obtain the necessary facts from the mothers by
sending a specially appointed survey doctor to
interview, first, the mother of a dead child, and,
secondly, the mother of a live child matched for
age, sex, and locality but otherwise picked at
random from the local birth register.
Available Cases. The total number of deaths in
the category required was 1694, of which 792
were ascribed to leukemia and 902 to other
cancers (2).By May, 1957, the mothers of 1416
of these children had been interviewed (677
leukemia and 739 other cancers). The lost cases
represented 16.4 percent of the total, and included 6.8 percent who belonged to families
that had moved abroad or to an unknown address and 3.1 percent where the interview could
not be arranged in time. The remainder represented refusals to cooperate, 4.5percent by parents and 2.0 percent by doctors.
Appointment of Survey Doctors. It was a basic
principle of the survey that the same doctor
Source: British Medicaljournal 1:1495-1508, 1958.
In receipt of a grant from the Medical Research Council.
All from the Department of Social Medicine, Oxford University.
u.
Collecting of Data
Each survey doctor was given a list of the
cases in his area. If the mothers or foster mothers were still living in the area he was to see
them and the corresponding control mothers: if
a mother had left the area he was to find and
interview a control, but return the case papers
to Oxford. These were eventually sent on to the
new area, but in this way a central record was
kept of all transfers. N o case/control pairs
seen by different doctors have been included in
the analyses making direct comparison between
cases and controls, but they may feature in other
analyses-for example, incidence of mongolism. Since the records obtained from foster
mothers contained no information about the
prenatal environment they too have been excluded from the casekontrol comparisons.
The questionaries for recording the interviews were the same for cases and controls, and
were distributed in pairs bearing the same serial
number and a so-called final date. This was the
date of death of the case, and was a reminder
that the medical history of the control child
should cease at the so-called onset date-that is,
the date when the corresponding child fell ill.
The first half of the questionary described the
childrens medical experiences before the onset
date-that is, illnesses, X-ray exposures, and
antibiotics-their feeding habits, and their ex-
506
Stewart et al.
507
posures to nonmedical ionizing radiations (television, luminous toys, and pedoscopes) during
this period. The second half described the
mother and other relatives. The mothers illness
and X-ray histories ceased at the date of birth of
the survey child, but the family histories continued to the date of the interview (this allowed
the total number of children in the family, and
the number of relatives dying from cancer, to be
the number up to the time of the interview).
The mothers illness and X-ray histories were
recorded separately for three periods of her
life: (1) the first period, before marriage; (2) the
middle period, between marriage and the relevant conception; and (3) the final period, during the relevant pregnancy. If the child was
illegitimate the first period extended to the relevant conception.
Control Selection
First-Born Children and Migrants. The cases included 510 first-born children and the 427 controls. By national standards the first figure is
508
Conurbations
Towns with over 100 000 inhabitants
Towns with 50 000 to 100 000 inhabitants
Towns with under 50 000 inhabitants
Rural areas
XY4)
Value of P
Leukemia
Other
malignant
disease
108
98
124
92
86
100
100
110
96
100
12.372
<0.02
1.451
>0.80
Stewart et al.
509
Figure 1. Approximate leukemia death rates of first- and later-born children, 19531955. These rates are based only on 663 leukemia cases of known parity. The inclusion of 14 adopted cases and 115 cases not followed up would give somewhat higher
rates per million but would he most unlikely to change the appearance of the graphs.
a
FIRST BORN
0
J
CC
0)
a
w
lo
kATERBORN
LATER BORN
- 0
AGE IN YEARS
AGE IN YEARS
5 10
____
Table 2. Number of mothers reporting X-ray examination of different sites (abdomen and
other) in three periods (see text).
X-ray examinations
~~
Period
Abdominal
Other
Any
Before marriage
The casekontrol ratio of 1.91 for abdominal x-ray exposures during :he relevant pregnancy (1) differs from
the expected ratio (1.00) at the level of P<lO-; (2) differs from :he contemporary ratio for other x-ray
exposures (1.17) at the level P=0.011 and from :he ratio for other X-ray exposures in any period (1.16) at the
level PCO.OO1; and (3) differs from the ratio for abdominal X-ray exposures in any period (1.38) at the level P0.012.
44126, etc., represent ratio of case mothers to control mothers.
Stewart et al.
and than the ratio for abdominal X-ray examinations as a whole. Hence there is prima facie
evidence that abdominal X-ray examinations
during pregnancy-the only type of examination involving direct exposure of the fetus-can
contribute to the etiology of malignant disease
in children.
Most of the remainder of this section is devoted to further analysis of the apparent association between fetal irradiation and malignant
disease in childhood, but it is necessary first to
say something about the figures for the other
maternal X-ray histories.
One group of maternal X-ray historiesthose relating to X-ray exposures of the abdomen during the middle period-shows a
small excess on the control side. This was at first
thought to be due to the fact that on the control
side there were more previous children, and
hence more occasions for obstetric X-ray exposures in the middle period. However, when
the obstetric X-ray examinations were separated
from the other abdominal X-ray examinations
belonging to this period and related to the
number of pregnancies at risk (see Table 3)
there was still a slight excess on the control side.
This strengthens the view that the control
mothers were as efficient as the mothers of
cases in recalling their abdominal X-ray examinations.
In the pre-marriage period there was a case
excess for abdominal X-ray exposures which
has to be accepted as technically significant
(P<0.05). The possibility has therefore to be
considered that damage to the maternal gonads
may increase the risk of childhood malignancy.
But even if this is the correct explanation such
damage is of minor interest, for, as judged by
the numbers involved, its importance can only
be one-quarter of that of direct fetal irradiation.
Unlike the case excess for abdominal X-ray
examinations, that for other X-ray examinations showed no tendency to concentrate in the
5 11
Table 3. Comparison between cases and controls in respect of obstetric X-ray examinations
in the middle period (i.e., between marriage and the relevant conception).
Cases
1623
79
84.41
Controls
1742
96
90.59
5 12
Table 4. Histories of direct fetal irradiation distinguishing three birth-rank groups of cases
and controls.
Position in
family of child
(birth rank)
First
Second
Later
All
Controls
Cases
No.
No.
Ratio
85/510
471393
461396
17811299
16.7
12.0
11.6
13.7
361427
281448
291424
9311299
8.4
6.3
6.8
7.2
1.99
1.90
1.71
1.91
than the obstetric X rays, they might be expected to concentrate on the case side if there
were a causal relationship between direct fetal
irradiation and childhood malignancies. Hence
the slight excess of other abdominal X-ray exposures on the case side in no way disturbs this
hypothesis.3
We are now in a position to judge the effect of
all the factors which might possibly have contributed a spurious element to the case/control
ratio for abdominal X-ray examinations in the
relevent pregnancy. To do this we have made
three assumptions which are unfavorable to an
explanation in terms of a causal link between
fetal irradiation and childhood malignancies.
These are: (1) that the excess of twins in the
case group is wholly spurious; (2) that ill health
during pregnancy is of such overruling importance that any associated X-ray exposure should
be ignored; and (3) that the case mothers were
16 percent more efficient than the control
mothers in reporting X-ray exposures (on the
grounds that the case/control ratio for other XJudging by the number of deaths among the children of
case and control mothers, the older and the younger children
of the case mothers were just as healthy as the older and
younger children of the control mothers, nor was there any
difference in the stillbirth rates for the case and control
groups (see Section VII).
Table 5. Frequency of direct fetal irradiation of cases belonging to migrant and static
families and of corresponding controls.
Migrant series
Fetal irradiation
Yes
No
Total
X2
Static series
Cases
Controls
Cases
Controls
57
325
23
359
121
796
70
847
382
382
917
917
15.205
<0.001
14.610
<0.001
Stewart et al.
5 13
Table 6. Numbers of mothers reporting obstetric and other abdominal X-ray examinations
during the relevant pregnancy.
Cases
X-ray category and reasons
for X-ray examinations
Total
Other abdominal:
Barium meal
Intravenous pyelogram
Injury
Other
Total
Total of all known indications
Indication for X-ray examination
unknown
No.
61
53
35
4
36.3
31.5
20.8
2.4
29
29
20
1
34.5
34.5
23.8
1.2
153
91.0
79
94.0
6
3
4
2
3.6
1.8
2.4
1.2
1
3
15
9.0
6.0
168
100.0
84
100.0
10
Controls
Total
Irradiated
141( 118.44)
81(103.56)
222
Not irradiated 1125(1147.56) 1204(1181.44) 2328
Total
1266
1284
No.
Obstetric:
Position
Size (? twins)
Routine pelvimetry
Diagnosis of pregnancy
Controls
2550
,$(I)
with continuity correction is 9.644, equivalent to a
normal deviate of 3-10; P<0.002.
1.2
1.2
3.6
Dose-Response Relationship
Court-Brown and Doll (7) have studied the
incidence of leukemia and aplastic anemia in
adults given deep X-ray therapy to the spine.
Although they had only 37 cases at their disposal they were able to demonstrate a strong
relation between dose and risk of disease. The
present series of children includes no fewer
than 178 in whom malignant disease had been
preceded by antenatal irradiation, and therefore appears to offer ample material for studying dosage effects. There is, however, an important difference which nullifies the advantage of
larger numbers. Court-Brown and Doll had an
almost pure series of radiation-leukemias,
since they gave evidence that over 90 percent of
them would not have occurred but for the preceding X-ray exposures. In the present series a
high proportion of the cases X-rayed in utero
developed the fatal disease for other reasons.
The existence of these cases is bound to hinder
recognition of a dose-response relationship.
Attempts to collect data for the study of dosage effects met with great difficulties. Over 200
5 14
0
1
2
3
4 or more
(Unknown number)
Total
Cases
Controls
Ratio
1121
37
60
23
32
(26)
1206
27
26
18
10
(12)
0.93
1.37
2.31
1.28
3.20
(2.17)
1299
1299
1.00
Tissue Sensitivity
Previous work suggests that, in adults, leukemia is more likely to be produced by ionizing
radiations than other malignant diseases (7, 8).
But our own data suggest that in a fetus there
may be other tissues which are at least as sensitive to X rays as the reticulo-endothelial system
(see Table 10). In this table the cases have been
divided into eight diagnostic groups, and the
number of irradiated cases in each of these is
compared with the number which would be expected if the X-ray exposures were evenly distributed among the cases. In calculating these
expected numbers, we have standardized them
in ten-yearly age groups, in order to allow for
changes in X-ray habits during the years in
which the survey children were born. On this
Table 9. Distribution of cases and controls irradiated in utero according to month of pregnancy.
Gestation period in months
1st to 5th
6th, 7th
8th
9th
10th
Total known
Unknown
Total
Controls
Gaselcontrol
ratio
18
15
20
55
58
2
9
19
30
25
(9.00)
1.67
1.05
1.83
2.32
166
12
85
8
1.95
(1.50)
178
93
1.91
Cases
A statistical comparison of the cases and controls in the five knowncategories above yields x2(,)= 9.656, for which P<0.05.
5 15
Stewart et al.
Table 10. Comparative incidence of direct fetal irradiation in eight diagnostic groups.
Irradiation in utero
No. of
cases
Diagnosis
~
Lymphatic leukemia
M yeloblastic
kidney
suprarenals
(incl. all neuroblastomas)
Malignant tumors of other sites
Total
Actual
no.
Expected
no.
292
124
203
109
212
120
14.4
7.3
13.8
7.3
12.7
15.8
42
9
28
8
27
19
40.31
16.24
27.93
14.24
28.75
16.68
87
152
18.4
19.1
16
29
12.68
21.16
1299
13.7
178
177.99
In an 8 x 2 table the comparison of actual and expected numbers irradiated and not irradiated x2(,) = 11.832,compared with
12.017 at the level P=O.lO.
The survey was deliberately restricted to children under 10 years of age because it was
thought that this was a convenient limit to the
time that mothers might be expected to recollect events. But by so doing we may have missed
some of the consequences of direct fetal irradiation. If the effects of prenatal irradiation are
exhausted by the age of 10 years one would
expect the casekontrol ratio for fetal irradiation
to be higher for children who were under 5
years of age at the onset date than for children
who were 5 or over. But, as Table 11 shows, this
ratio is 1.73 for the younger children and 2.50
for the older children.
Deaths at ages 0 to 4
5 to 9
0 to 9
Cases
Controls
Ratio
123
55
178
71
22
93
1.73
2.50
1.91
Estimates of Risk
The following estimates are based on the figures shown in Table 2 and necessarily represent
a rough appraisal of the situation. In the case
group 13.7 percent of the children were
X-rayed before birth and 86.3 percent were not.
The corresponding percentages for control
children were 7.2 and 92.8. On this showing
children who have been X-rayed in utero are
(137
5 16
exposure to X rays and the subsequent development of malignant disease, and there are indications that the risk is related both to the dose
of X rays and to the date of exposure. There is
nothing to suggest that irradiation in utero explains the early peak of leukemia mortality (see
Section IV for data on postnatal X-ray exposures) and it may even cause deaths from
malignant disease after the age of 10 years.
Thorough estimates of risk which are given
here apply only to one period of time and to the
X-ray doses in use at that time. Evidently radiotherapy of pregnant women is an extremely
rare event in this country, and contributed
nothing to the cases we have considered.
Addendum. After this report had been written
we received the results of an independent study
(9)which had been designed specifically to test
the conclusion announced in our preliminary
communication. This study concerned children
who had died under the age of 10 in the State
of Louisiana during the years 1951-5. Although conducted on a rather small scale it had
one important advantage over our own inquiry:
all the information was obtained direct from
professional sources, thus minimizing the risk
from emotional bias or selective memory in the
informants. The incidence of irradiation in
utero discovered by Paterson was as follows:
among 77 children who died of leukemia, 27.3
percent; among 70 children who died of other
Table 12. Illnesses of mothers arising before and during the relevant pregnancy.
Arising before
relevant pregnancy
International classification main category
~
Arising during
relevant pregnancy
Cases
Controls
86
9
24
32
53
219
74
62
32
33
15
5
48
82
11
22
37
44
207
91
44
32
43
20
7
57
1
0
12
12
0
113
5
3
0
692
697
162
109
565
554
155
104
~~
Cases
Controls
I. Infective
11. Neoplasms
13
2
1
2
3
0
4
13
4
0
77
5
0
5 17
Stewart et al.
Illnesses Originating Before the Relevant Pregnancy. The total number of illnesses in this category was 1389 (cases 692 and controls 697) and
the number of mothers affected was 1119 (cases
565 and controls 554). Unlike the series reported by Manning and Carroll (6) there were
no casekontrol differences for allergic conditions (category III), but compared with this
American series the incidence of these conditions was low. The only group of illnesses with a
substantial difference between cases and controls was category X (diseases of the genitourinary system) with 62 records for cases and 44 for
controls, The difference was confined to disease
of the renal tract as such (49 cases and 30 con-
Table 13. Illnesses of mothers specifically associated with the relevant pregnancy.
Other
cancers
Leukemia
I.C.D.
numbers
642
640-64 1
648.0
Total
Illness
Toxomias of pregnancy
Infections of genitourinary
tract during pregnancy
Threatened abortion
Total
Cases
Controls
Cases
31
38
26
67
57
3
4
38
11
13
62
7
6
39
19
27
113
10
10
77
Cases
Controls
29
8
14
51
Con-
trols
Corresponding children
No. of cases
Rubella
Mumps
Herpes zoster
Infective hepatitis
Leukemia at 4 years
Cerebral tumor at 6 years
5 18
infective diseases as the only ones with a significant and unambiguous excess on the case side.
The infective disease group included 10 virus
infections on the case side and one on the
control side. These undated infections were associated with the childhood diseases shown in
Table 14.
To summarize: there is nothing in the illness
records to suggest that the case mothers were,
before the relevant pregnancy, less healthy than
the control mothers. During the relevant pregnancy they appear to have suffered more from
threatened abortions and virus infections than
the control mothers, but the numbers involved
are small.
SECTION IV. CHILDRENS X-RAYHISTORIES
Nonmedical X-Ray Exposure. Three non-medical sources of ionizing radiations were considered-television sets, luminous clocks or toys,
and pedoscopes. Only one fact relating to these
(undated) exposures has been analyzed-
Total
Other cancers
Total
Cases
Controls
Cases
Controls
43
47
5
42
38
1
45
43
3
39
2
90
95
81
91
99
186
58
Includes some children for whom the number of examinations was not ascertained.
Cases
Controls
88
100
77
180
Stewart et al.
Time Relationship
In view of the evidence already presented
that a fetus may be specially radiosensitive, the
postnatal X-ray exposures (diagnostic and therapeutic combined) were analyzed by age at first
exposure (see Table 16) to discover whether
there were any time relationships. Included in
this Table are 67 children (27 cases, 40 controls)
who have been classified as age at first exposure unknown. These children were definitely X-rayed before the onset date, but the
possible dates covered a period of more than 12
months. The remaining children included 128
where the month and year of the first X-ray
exposure was known, and 171 where the date
was given to within 12 months. In the latter
group we have assumed that the actual date of
exposure was midway between the two possible
extremes and have placed the child accordingly.
The last two columns of the Table (where all
cases and all controls are compared) show a
significant excess on the case side of children
exposed before their third birthdays. This ex-
5 19
Table 16. Distribution of children with postnatal medical X-ray exposure according to age at
first X-ray exposure.
Age at
at first
first
Arre
exposure
exposure
Leukemia
Leukemia
Cases
Cases
Other Malignant
Malignant
Other
Controls
Controls
8
99
6
5
5
5
3
1
00
Not known
13
17
14
23
27
40
95
81
91
99
186
180
Total
25
4
14
9
3
2
1
00 .
9,
33
10
9
5
2
1
0
00
Controls
Controls
3
4
5
6
7
10
23
12
15
Cases
Cases
19
21
17
57
31
Controls
Controls
0
1
12
Cases
Cases
Total
Total
50
13
15
15
27
13
8
6
3
3
0
00
43
42
33
32
33
16
14
10
7
4
1
00
107
52
25
25
24
17
22
15
6
5
1
00
74
66
520
Table 17. Distribution of children with postnatal medical X-ray exposure according to age of the child at
the onset date.
Age at
onset date
(years)
Leukemia
Other malignant
Total
Cases
Controls
Cases
Controls
Cases
0, 1
2, 3
4, 5
6, 7
8, 9
6
35
23
17
14
7
11
25
22
15
14
23
21
17
15
10
16
32
27
13
20
58
44
34
29
57
49
28
Not known
Total
95
81
91
99
186
180
Controls
17
27
a The onset date is defined as either the date of onset of the fatal illness (case children) or the date when the corresponding
case developed its fatal illness (control children).
borne the brunt of the recent increase in childhood leukemia mortality both in this country
and in the U.S.A.
Thus far the evidence in favor of a link between childhood malignancies and postnatal Xray exposure has depended on casekontrol
comparisons of arguable validity. We have
therefore attempted to assess the evidence by
means of comparisons within the case and the
control series, using the following argument. In
a normal group of children, such as our own
controls, the percentage who have ever been Xrayed should steadily increase with age, for each
year adds to the number of children who are Xrayed for the first time. For the sake of simplicity we shall assume that the percentage of
first X-ray exposures is normally the same in
each of the first ten years of life. In order to test
this assumption we have calculated for the control children the expectedpercentages of children ever X-rayed in each of the five age groups
shown in Table 17. In the first two columns of
Table 18 these expected figures are shown in
italics together with the actual percentages and
numbers. It will be seen that the correspondence between the two sets of figures is not
exact but lies well within the limits of chance
fluctuation (P>O.l). It follows that the original
assumption is a reasonable one. Now, if the
postnatal X-ray exposure had not influenced the
risk of the children dying later of malignant
diseases, the correspondence between the actual and expected figures for the case series
should be much the same as in the control series. But the discrepancies are much greater in
the case series than in the control. Subdivision
of the cases into leukemias and other cancers
shows that in the second age group of the leukemia series (which includes children who developed leukemia between their second and fourth
birthdays) the proportion of X-rayed children
is actually higher than the proportion in the
third age group, although for these children the
period during which they had an opportunity
to be X-rayed was half as long again. For children with other cancers the largest discrepancy
is in the first age group, where the actual number of children X-rayed (14) is nearly twice the
expected number (7.43).
In earlier parts of this paper (and again in
Section IV) we discuss evidence which indicates
that the decisive factor in some of the cases,
particularly in the leukemia series, dates back to
a period well before birth. If this is so then there
must be, in the newborn population, a number
of children in what may be loosely called a premalignant state. T h e present analysis of
postnatal X-ray exposures suggests that if such
children are X-rayed during infancy overt signs
of the disease may appear relatively quickly.
This inference can be drawn from the tables
already given, but is perhaps better illustrated
by the distribution of cases and controls according to the interval between the first postnatal
X-ray exposure and the final date. For children
X-rayed in the first three years of life this was:
Cases
Controls
12
62
33
11
26
37
521
Stewart et al.
Table 18. Distribution (by age at onset date) of children with postnatal X-rayexposure compared with
the distribution expected on the basis of a regular increase with age of the percentage of children ever
X-rayed since birth.
Age of
children
at onset
date
(years)
Controls
All cases
Leukemia
Other cancers
Actual
Expected
%
No.
Actual
Expected
%
No.
Actual
Expected
%
No.
Actual
Expected
%
No.
0, 1
4.9
-
2, 3
7.2
4, 5
17.7
6, 7
24.8
8, 9
j1.9
Total
xY4)
Value of P
17
5.8
20
4.4
-
27
15.4
-
58
17.9
-
19.7
57
15.2
-
44
16.1
23
~19.0
27.20
14.4
21
~-
25.9
49
18.0
34
18.3
-
17
-
17.7
-
3.5
12.28
10.6
51.15
46.83
32.2
13.8
13.8
40.03
28
27.71
178
178.00
7.622
>o. 10
3.7
11.0
18.4
25.8
33.3
33.1
14.4
14.4
12.76
41.61
53.16
48.67
29
28.80
185
185.00
19.414
<0.001
This interpretation would also explain the deficiency on the case side of X-rayed children who
developed the fatal disease between their fourth
and eighth birthdays (see Table 17). For, if irradiation in infancy hastens overt signs of the
disease, then there must be some vulnerable
children who survive a comparatively long time
partly because they have not been X-rayed. If
this is so, we would expect the percentage of
X-rayed children in the older half of the case
series to be below the level for healthy children.
Pedoscopes
We must now reconsider the records for shoeshop X-ray exposures. The same rays are emitted from pedoscope machines as from diagnostic X-ray sets. If, therefore, the medical Xray examinations have had an effect, one would
expect the pedoscope records to show the same
signs. A possible reason for there being no such
excess has already been mentioned-namely,
inclusion on the control side of children whose
first pedoscopic examination was later than the
onset date. Another reason should now be apparent. The data on medical X-ray examina-
3.8
11.4
26.6
6
-
5.25
35
-
22.37
24.77
6.7
-
14
-
12.7
23
3.6
10.7
17.8
24.9
7.43
19.30
25.95
17
-
23.89
34.2
31.1
14
15.41
35.7
32.0
13.44
15.4_
15.4
95
_
95.00
13.4
90
12.478
<0.02
13.4
15
-
90.01
11.015
10.05
Comment
To sum up: there are two reasons for thinking
that X-ray films taken shortly after birth influence the distribution of childhood deaths from
malignant disease. In the first place, there is a
significant case excess for X-ray exposures dur-
522
ing infancy. Secondly, children who died of leukemia at the age of 2 and 3 years were concentrated within the group of cases exposed to
X rays in infancy. The effect of postnatal X-ray
exposure appears to be more marked in respect
of leukemia than other cancers and to be much
weaker than the effect of prenatal X-ray exposure. It may in fact be restricted to speeding
the date of death in children who are already
predisposed to leukemia.
The lack of any association between the parts
of the body exposed to X rays and subsequent
developments may be related to the fact that in
infancy the reticulo-endothelial system is so
widespread that it is likely to be involved in any
exposure.
SECTION V. CHILDRENS ILLNESSES AND
THEIR TREATMENT
Congenital Defects
These were mentioned in 75 of the case and
46 of the control records, but only two conditions, mongolism and naevi, were more common among the cases.
Mongols. There were 17 mongols in the leukemia series, one among the other malignant disease, and none in the control group, though
one child placed first on a control selection list
was said to be a mongol and to have been in an
institution since infancy, and was for this reason
not chosen (see Section I). Until two years ago
only four cases of leukemia associated with
mongolism had been reported. Since then
Krivit and Good (12), Merrit and Harris (Ijr),
Carter (14), and Paterson (9), have between
them reported 13 cases. With the present series
this makes a total of 34 reported cases. Since the
incidence of mongolism in the present leukemia series (2.6 percent) is nearly 20 times as
high as the incidence of mongolism in 14 000
consecutive births (13, the association is evidently not a fortuitous one. We have already
shown that excessive maternal age at the time of
the childs birth (which undoubtedly pre-
disposes to mongolism) also predisposes to leukemia (see Section I). It is therefore more likely
that the two diseases are influenced by a common factor than that the antecedent condition,
mongolism, predisposes to the later condition,
leukemia. If this is so, then for some of the cases
of leukemia the decisive event must date back at
least to the onset of the mongolism; that is to
say, at the latest to the second month of gestation. A number of facts relating to the mongols
in the survey are shown in Table 19. The
mongol who developed a glioma had been
X-rayed 71 days before birth, but none of the
mongols who died of leukemia had been irradiated in utero. Four mothers had had previous
abortions, but there were no threatened abortions in the relevant pregnancy and no previous
stillbirths. One mother was married to a second
cousin.
Nevi. These were reported 34 times on the
case side and 21 times on the control side. As
previously stated, eight of the cases and three of
the controls were treated with X rays. The overall difference between cases and controls is not
statistically significant, and might be due to underreporting on the control side of a condition
which is only cosmetically important. For girls
there were 18 cases and 14 controls, and for
boys 16 cases and 7 controls.
The other congenital defects took various
forms and were evenly distributed between the
two series, with 25 records on the case side and
25 on the control side.
523
Stewart et al.
Table 19. Features of the 18 mongols included in the survey.
Cause of
death
Cerebral glioma
leukemia
Cell type
(Leukemias)
Monocytic
Lymphatic
Myeloid
Lymphatic
M
M
M
F
F
M
F
M
M
M
F
F
F
Aleukemic
Lymphatic
Monocytic
Lymphatic
M
M
M
Myeloid
Lymphatic
Myeloid
NIR
Lymphatic
Sex
F
F
Birth
rank
2
8
4
5
2
3
2
2
1
4
1
3
1
1
1,
Age at death
3 years
2 years
1 year
1 year
5 years
2 years
2 years
1 year
4 years
5 years
2 years
2 years
2 years
1 year
8 years
6 years
4 years
6 years
5 months
11 months
2 months
6 months
0 months
1 month
8 months
11 months
9 months
11 months
0 months
6 months
0 months
5 months
4 months
5 months
6 months
11 months
Age of
mothep
40
44
44
43
42
40
39
38
36
34
34
34
32
32
29
26
26
19
524
Table 20. Childhood illnesses in 1299 pairs of survey (M) and control (C) children,
Illness more
than 2 Years
before
onset date
Illness
within 2
years of
onset date
Measles
Chickenpox
Whooping cough
Mumps
Rubella
Scarlet fever
Total
Bronchitis
Bronchopneumonia
Total
Acute tonsillitis
Acute otitis media
Total
Other infections
Other illnesses
Total
Grand total
208
131
105
57
47
20
235
173
136
61
46
24
236
125
154
45
34
19
568
675
38
39
Additional
illnesses of
uncertain
date
Total of
dated
illnesses
M
219
102
158
38
34
15
444
256
259
102
81
39
454
275
294
99
80
39
60
46
48
10
12
4
70
55
36
22
23
6
613
566
1181
1241
80
212
24
17
19
37
19
32
57
76
43
49
6
2
3
1
77
41
56
51
133
92
30
20
22
25
14
15
12
9
44
35
34
34
4
2
50
47
29
21
79
68
88
29
54
12
65
29
47
12
153
58
101
24
12
1
9
3
117
66
94
59
211
125
13
12
812
829
792
697
1604
1526
202
234
Stewart et al.
vaguely described to be considered separately.
In both periods there was a marked excess on
the case side, but since the risk of inflated
reporting by bereaved mothers is probably at a
maximum with minor or ill-defined conditions
(which form the bulk of this section), we prefer
to draw no conclusions from these figures. Trivial conditions not included in the Table are
feeding difficulties (254 cases, 198 controls),
allergic conditions (28 cases, 25 controls), worm
infections (89 cases, 82 controls), and frequent
colds (already mentioned). The even distribution of cases and controls for three of these four
conditions does not suggest underreporting on
the control side; it is therefore possible that the
figures in the fourth section of Table 20 and the
figures for recurrent attacks of bronchitis indicate a genuine excess on the case side.
Finally, some reference must be made to injuries and operations. A vast number of minor
injuries, including bruises, were recorded, but
only two types were regarded as important
enough to merit analysis, and then only if they
happened within two years of the onset date. In
this period there were 44 fractures (26 cases, 18
controls) and 28 burns or scalds (16 cases, 12
controls). The case excess lay entirely within the
leukemia series, and for the two items combined
these children produced 28 records, compared
with 14 for the children with other cancers. For
recent operations, other than tonsillectomy, the
figures were 16 cases and 17 controls.
In summary: Neither in their lifetime as a
whole, nor in the two years immediately preceding the fatal illness, did the children with leukemia and other cancers have an excessive num-
525
Other
cancers
(680 pairs)
Total
(1299 pairs)
Cases
Contmls
Cases
Controls
Cases
Controls
Case/
control
ratios
Sulphonamides:
Adequately dated
Not adequately dated
Total
94
30
124
96
18
114
84
26
110
61
34
95
178
56
234
157
52
209
1.13
1.08
1.12
Antibiotics:
Adequately dated
Not adequately dated
Total
109
36
145
83
36
119
95
29
124
77
41
118
204
65
269
160
1.28
0.84
1.14
Treatments
77
237
526
with sulphonamides, but the corresponding figures for antibiotics show a small case excess.
Since the high incidence of pulmonary infections on the case side was likely to have influenced these figures, Table 22 has been prepared in which the children who had such an
illness in the two years before the onset date are
shown separately. The effect of doing this is to
reduce the casekontrol ratio for antibiotics
from 1.28 (all treatments) to 1.18 (treatments
other than those probably but not certainly related to a recent pulmonary infection). At this
level it is within the conventional limits of
chance fluctuation ( P = approximately 1 in 15).
The relatively high casekontrol ratio for both
types of drugs in this Table probably reflects the
fact (not yet mentioned) that in the case series
these infections were usually more severe, as
well as more numerous, than in the control series, and for this reason were more likely to
receive antibiotics or sulphonamides.
A record was also kept of treatments with
ultraviolet light. These were reported for 50
cases and 40 controls. No conclusions can be
drawn from these small numbers.
We conclude that there is no evidence for a
direct relationship between childhood malignancies and the drugs considered here. There
may, however, be an indirect relationship between antibiotics and leukemia. In recent years
antibiotics have revolutionized the prognosis for
acute pulmonary infections and have undoubtedly kept alive children who would otherwise
have died. If these infections occasionally
provoke a latent leukemic process, then, by increasing the number of children who survive,
antibiotics may indirectly increase the prevalence of leukemia.
Concern is often expressed about the widespread use of new chemicals, particularly food
preservatives and detergents, on the grounds
that these may be carcinogenic. In the present
survey no attempt was made to collect information about exposure to specific chemicals, but
the mothers were asked a number of questions
about their own habits and what they allowed
their children to eat and drink.
Feeding Habits
The survey doctors were told to ask whether
the children had been given certain foods and
drinks every day, less than daily but at least once
a week, less often, or never. They were also
asked to state when these foodstuffs were first
given; and what the mother thought the child
had consumed in the way of colored sweets.
The completed records showed that a few control mothers had reported foods which were so
unlikely to have been given during the preonset period-for example, fish and chips before the age of one-that the analysis has been
restricted to caselcontrol pairs for whom the age
at the onset date was not less than an arbitrarily
chosen qualifying age for each foodstuff (see
Table 23).
The figures shown in this Table relate to children who had ever been given a food regardless
of the reported frequency. In aggregate the
caselcontrol differences for the eight foodstuffs
considered appear to be statistically significant
(P<0.05), but only two foods-tinned vegetables and colored drinks-showed a significant
difference, and both were reported more often
Table 22. Sulphonamide and antibiotic treatments, distinguishing children who had an acute pulmonary
infection within two years of the onset date.
All controls
All cases
Sulphonamide:
Children who had acute pulmonary infection within two years of onset date
Remainder
Antibiotics:
Children who had acute pulmonary infection within two years of onset date
Remainder
a
No.
No.
Ratio
22177
28.57
1014 1
24.39
1.17
12.77
147/1258
11.69
1.09
46.75
13141
31.71
1.48
13.75
147/1258
11.69
1.18
15611222
36177
16811222
Stewart et al.
527
Table 23. Feeding habits; percentage of cases and controls who had ever taken certain foodstuffs.
% of children
who had ever
taken this item
Qualifying
age
I tem
Dried rnilk
Fruit juice
Tinned sieved vegetables
Other tinned vegetables
Highly colored cakes
Highly colored fruit drinks
Colored sweetsa
Shop-fried fish and chips
Total
a
3 days
1 month
4 months
12 months
Cases
Controls
69.6
90.6
61.1
55.1
33.5
47.4
32.6
43.2
67.1
90.7
58.2
60.6
33.7
52.6
33.0
43.9
X2(d
P
approx.
1.718
0.002
1.976
6.797
<0.001
5.729
0.018
0.066
0.19
0.96
0.16
0.009
>0.98
16.306
0.05>P>0.02
0.011
0.90
0.79
For this item over = more than v4 Ib. (113 g.) per week.
528
Table 24. Miscarriages and stillbirths (up to the date of the survey childs death).
Other
Cancers
Leukemia
Cases
Controls
129
173
26
26
93
122
23
28
Total
Cases
Controls
Cases
Controls
119
153
27
32
118
156
23
25
248
326
53
58
211
278
46
53
529
Stewart et al.
~
~~
~~
Sibling
Age at
death
(years)
Sex
Diagnosis
122
289
290
347
F
F
F
M
Lymphosarcoma
Acute lymphatic leukemia
Acute recticular leukemia
Acute lymphatic leukemia
476
1293
1656
F
F
M
7,
7,
,,
Age at
6
1
0
5
1
1
death
(years)
Sex
Diagnosis
Acute anemia
Leukemia
Leukemia
Generalized lymphosarcoma
Retroperitoneal sarcoma
Brain tumor
Growth in abdomen
Lymphatic leukemia
F
F
M
M
M
F
M
1
1
0
3
2
4
2
5
Cases
Controls
Ratio
172
215
140
176
152
164
1.23
1.22
1.22
1.23
186
20 1
530
Discussion
eases which might have affected fetal development-virus infections and threatened aborAll scientific investigations are faced with the tions in the relevant pregnancy-were not
problem of eliminating unknown bias from test accompanied by a case excess for other materand control observations. Where, as in the pres- nal illnesses, either during this period or preent case, the findings are based on human data viously. Again, the case excess for pulmonary
and involve comparisons between facts which diseases in childhood must be viewed against a
have been obtained by different observers from background of almost equal numbers of all
inexpert witnesses the main sources of such bias childhood diseases in the two groups, and the
are: (1) unequal recording of events by differ- case excess for X-ray exposure in infancy
ent observers; ( 2 ) unequal reporting of events against almost equal numbers for postnatal Xby the individuals chosen to represent case and ray exposure as a whole.
control groups; and (3) faulty selection of cases
The third source of bias-namely, faulty selection of cases and controls-has been dealt
and/or controls.
The first source of error has been controlled with in the following ways. The cases were
by insisting on the same doctor seeing each drawn from the total number of cases for a
member of a given casekontrol pair. By so three-year period, and those actually included
doing, errors due to lack of skill in interviewing in the survey represent such a high proportion
and recording should be equally represented on of this total that, however atypical the lost cases
the case and control sides. The same device may be, the consequences would remain numermight be expected to control the second source ically unimportant. The controls were drawn
of error; but here we were on less certain from the general population of surviving chilground, and have therefore applied other tests dren, thus avoiding some of the confusion
of reliability. Thus for several items specified in which might have resulted from using more
the questionaries the proportions of completed accessible children-for example, other hospirecords in the two samples have been calculated tal patients, children attending welfare clinics,
and found to be alike. To illustrate this point we or siblings of cases. The controls actually inhave summarized the findings for six items in cluded in the survey have been shown to differ
which a low response rate might, with some in two important respects (first-born children
justification, have been expected (Table 27). By and twins) from the national population. But
these criteria it would seem that the doctors because the numbers in these classes can be
were equally pertinacious when questioning the calculated from official vital statistics, it has
cases and controls, and the mothers were been possible to show that the deficiencies in
equally helpful when replying to the doctors the control group had a negligible effect on the
questions.
main finding-namely, a large excess for direct
Other safeguards include repeated demon- fetal irradiation in the case series. We have also
stration that significant differences between shown that this finding was not affected either
cases and controls are not compatible with any by systematic checking of the mothers stategeneral tendency on the part of control mothers ments against hospital records, or by the pubto be unduly forgetful or uncooperative. For lication of a preliminary report half-way
instance, the case excess for two maternal dis- through the survey. Finally, we already know
Tale 27. Response of case and control mothers to certain questions.
Response rate (i.e., percentage
of complete answers)
Items
Family income
Contraception
Dated illnesses (children)
Cases
Controls
87
88
89
76
85
82
88
89
87
71
78
81
Stewart et al.
531
References
(I) Hewitt, D. B r J Prev SOCMed 9:81, 1955.
of England and Wales, 1953-5. Part 1 Tables, Medical. London, H.M.S.O., 1954-6.
(3) Stewart, A,, J. Webb, D. Giles, and D. Hewitt.
Lancet 2:447, 1956.
(4) Logan, W.P.D. and E.M. Brooke. The Suruey of
Sickness, 1943-1952. General Register Office Studies
on Medical and Population SubjGts, No. 12. London,
H.M.S.O., 1957.
532
SUMMARY
The role and limitations of retrospective investigations of factors possibly associated with the occurrence of a disease are discussed and their relationship to forward-type studies emphasized. Examples of situations in which misleading
associations could arise through the use of inappropriate control groups are presented. The possibility of misleading associations may be minimized by controlling
or matching on factors which could produce such associations; the statistical analysis
will then be modified. Statistical methodology is presented for analyzing retrospective study data, including chi-square measures of statistical significance of the
observed association between the disease and the factor under study, and measures
for interpreting the association in terms of an increased relative risk of disease. An
extension of the chi-square test to the situation where data are subclassified by
factors controlled in the analysis is given. A summary relative risk formula, R , is
presented and discussed in connection with the problem of weighting the individual
subcategory relative risks according to their importance or their precision. Alternative relative-risk formulas, R,, R,, R,, and R,, which require the calculation of
subcategory-adjusted proportions of the study factor among diseased persons and
controls for the computation of relative risks, are discussed. While these latter
formulas may be useful in many instances, they may be biased or inconsistent and
are not, in fact, averages of the relative risks observed in the separate subcategories.
Only the relative-risk formula, R , of those presented, can be viewed as such an
average. The relationship of the matched-sample method to the subclassification
approach is indicated. The statistical methodology presented is illustrated with
examples from a study of women with epidermoid and undifferentiated pulmonary
carcinoma.
INTRODUCTION
1959.
Biometry Branch, National Cancer Institute, Bethesda,
Maryland; National Institutes of Health, Public Health Service, U.S. Department of Health, Education, and Welfare.
'
533
534
OBJECTIVES
Restrospective studies are relatively inexpensive and can play a valuable role as scouting
forays to uncover leads on hitherto unknown
effects, which can then be explored further by
other techniques. The effects may be novel and
not suggested by existing data, as in the pioneer
work on the association of smoking and lung
cancer or the association of blood type and gastric cancer, or they may represent refinements
of current knowledge. T h e latter category
might include collection of lifetime residence
and/or work histories to elaborate differences in
incidence and mortality which appear when
some diseases are classified by last place of resi-
535
536
with the course of the illness, unrelated to increased or decreased risk of developing the disease. The theoretical point has been raised that
factors conducive to longer survival of patients
may be found in prevalence samples and interpreted erroneously as being associated with
excess liability to the disease (33).Loopholes of
this type are minimized when investigations are
restricted to samples of newly diagnosed patients (incidence).
A partial remedy for these uncertainties lies
in employing a conservative approach to interpretation of the associations observed. Recognizing the ease with which associations may be
influenced by extraneous factors, the investigator may require not only that the measure of
relative risk be significantly different from unity
but also that it be importantly different. He
may, for instance, require that the data indicate
an increased relative risk for a characteristic of
at least 50 percent, on the assumption that an
excess of this magnitude would not arise from
extraneous factors alone. However, the use of
such conservative procedures emphasizes a corresponding need to pinpoint the disease entity
under study. A strong relationship between a
factor and a disease entity might fail to be revealed, if the entity was included in a larger, less
well-defined, disease category. After the event
from data now at hand, we know that a study of
the association of cigarette smoking with epidermoid and undifferentiated pulmonary carcinoma is more revealing than an inquiry covering all histologic types of lung cancer.
MULTIPLE COMPARISON PROBLEM
individuals who both smoke and drink as opposed to all other persons, or as opposed to
those who neither smoke, nor drink coffee. In
addition, the relative risk associated with smoking might be obtained separately for drinkers
and nondrinkers of coffee, with a weighted
average of these two relative risks constituting
still another item. Conversely, risks associated
with coffee drinking, with adjustments for cigarette smoking, could be computed.
The potential comparisons arising from a
comprehensive retrospective study can be large.
Almost any reasonable level of statistical significance used to test a single contrast, when applied to a long series of contrasts, will, with a
high degree of probability, result in some contrasts testing significant, even in the absence of
any real associations. The usual prescription for
coping with this multiple comparison problem-requiring individual comparisons to test
significant at an extreme probability level to
reduce the number of associations incorrectly
asserted to be true-would result only in making real associations difficult to detect.
However, the multiple comparison problem
exists only when inferences are to be drawn
from a single set of data. If the purpose of the
retrospective study is to uncover leads for fuller
investigation, it becomes clear there is no real
multiple significance testing problem-a single
retrospective does not yield conclusions, only
leads. Also, the problem does not exist when
several retrospective and other type studies are
at hand, since the inferences will be based on a
collation of evidence, the degree of agreement
and reproducibility among studies, and their
consistency with other types of available evidence, and not on the findings of a single study.
Nevertheless, it would be wise to employ testing procedures which do not lead to a superabundance of potential clues from any one study.
This may be achieved by employing nominal
significance levels in testing factors of primary
interest incorporated into the design of an investigation and applying more stringent significance tests to comparisons of secondary interest
or to comparisons suggested by the data. For
the usual problem of multiple significance testing, this would be equivalent to allocating a
large part of the desired risk of erroneous acceptance of an association as real to a small
group of comparisons where fruitful results
were anticipated, and parceling out the re-
537
mainder of the available risk to the large may be advanced for any special groups used as
bulk of comparisons of a more secondary sources of cases and controls.
nature. This minimizes the risk of diluting,
However, a mere catalogue of biases arising
through inclusion of many secondary com- from the possibly unrepresentative nature of a
parisons, t h e chances for detecting a n sample of cases and controls should not ips0
important primary effect.
fact0 invalidate any study findings. This is a
substantive issue to be resolved on its merits for
a specific investigation. Collateral evidence may
provide information on the potential magniREPRESENTATIVE NATURE OF DATA
tude of bias and the size of spurious associations
The fundamental assumption underlying the which could result. In some situations the difanalysis of retrospective data is that the as- ference between cases and controls may be so
sembled cases and controls are representative of great that postulation of an unreasonably large
the universe defined for investigation. This ob- bias would be required. Whether he consciously
ligates the investigator not only to examine the recognizes it or not, the investigator must aldata which are the end product but also to go ways balance the risks confronting him and debehind the scenes and evaluate the forces which cide whether it is more important to detect an
have channeled the material to his attention, effect, when present, or to reject findings, when
including such items as local practices of refer- they may not reflect the true situation. If opporral to specialists and hospitals and the patients tunities for further testing exist, one should not
condition and the effect of these items on the be too hasty in rejecting an association as an
probability of diagnosis or hospital admission. artifact arising from the method of data colWe reemphasize that this requires the exercise lection, and in foreclosing exploration of a poof judgment on the potential magnitude of tentially fruitful lead.
biases and as to whether they could result in
Because of the important role retrospective
factors seeming to be related to a disease, in the studies play in studies of human genetics, menabsence of a real association of the factor with tion may be made of a bias frequently encounpresence or absence of the disease. The danger tered in studies dealing with the familial distriof bias may be greatest in working with material bution of diseases. A frequently used procedure
from a single diagnostic source or institution. takes a group of diagnosed cases for a disease in
Among the more important practical consid- question and a group of controls and compares
erations affecting retrospective studies is that the prevalence of this disease among relatives of
they are ordinarily designed to follow the line of the probands and controls. The bias arises from
least resistance in obtaining case and control the unrepresentative nature of the probands
histories. This means that cases and controls with respect to familial distribution and is
will often be hospital patients rather than per- known in other fields as the problem of the
sons in the general population outside hospi- index case or the effect of method of ascertals. As a result, any factor which increases the tainment. It has long been recognized that the
probability that a diseased individual will be characteristics for a random sample of families
hospitalized for the disease may mistakenly be will differ from those for families to whom the
found to be associated with the disease. For investigators attention has been directed beexample, Berkson (34) and White (35)have cause the family rosters include individuals sepointed out that positive association between lected for study on the basis of a specified attwo diseases, not present in the general popula- tribute. For example, data on family size
tion, may be produced when hospital admis- (number of children) obtained from siblings,
sions alone are studied, because persons with a rather than parents, are biased, since two or
combination of complaints are more likely to three potential index cases are present in the
require hospital treatment. I n theory, bias population for two- and three-child families as
might also be produced in reverse manner, if opposed to one for one-child families and none
the suspect factor diminished the probability of for childless couples. The analogy for disease
hospitalization for other diagnoses used as con- occurrence is apparent. Families with two or
trols. The difficulties are not unique for hospi- three cases of the disease under study may have
tal patients. Similar loopholes in interpretation double or triple the probability of being repre-
538
While easier accessibility to and lesser expense of hospital controls are important considerations, they should not deter one from
collecting control data for a sample representing a more general population, if the latter are
demonstrably superior. Some of the uncertainties about the superiority of hospital or general
population controls arise from the need to
maintain comparability in responses. The dependence of retrospective studies on comparability of responses from cases and controls
cannot be overemphasized. When more accurate answers can be obtained from controls in a
medical-care environment, the gain in comparability of responses for these controls could
outweigh the other advantages to be derived
from the more representative nature of general
population controls. The difficulties may be illustrated by the experience with smoking histories. Hospital controls invariably yield a higher
proportion of smokers for each sex than controls of comparable age drawn from the general
population (27). Does this mean more complete
smoking histories are collected in hospitals or
does it imply that smokers have higher hospital
admission rates? If the first alternative is correct, hospital controls are the appropriate
choice for measuring the association of smoking
history with a given disease. The second alternative calls for general population controls and
539
540
To progress further, questions on the representative nature of the case and control series
must have been resolved affirmatively. With this
condition in mind, let us suppose that a controlled retrospective study has been conducted
and that the number of diseased cases, N , , consists of A individuals with the factor being investigated and B free of the factor, while the number of controls, N,, consists of C individuals
with, and D individuals without, the factor. Let
M , = A + C , M , = B -t D,T = N , + N , = M ,
+ M , = A + B + C + D. What statistical
evidence is there for the presence of an association and what is an appropriate measure of the
strength of the association?
A commonly employed statistical test of association is the chi-square test on the difference
between the cases and controls in the proportion of individuals having the factor under test.
A corrected chi square may be calculated routinely as
((AD
-BCI
- %T)2T/N,MlN,M,
Free of
factor
PI
p3
P, + P ,
p2
p4
P3
+ P,
Total
p , + p2
p3 + p4
1
54 1
With disease
Free of disease
With factor
Free of
factor
P,=4%
P3 = 16%
P4= 72%
12%
88%
20%
80%
100%
Total
P2=8%
Total
542
Free of
factor
Total
A,
C,
4
D,
N,,
NZr
MI,
MB
T,
With factor
With disease
Free of disease
Total
543
Within this subgroup the approximate relative risk associated with the disease may be written as A,D,/B,C,.One may compare the observed
number of diseased persons having the factor,
A,, with its expectation under the hypothesis of
a relative risk of unity, E(A,) = N , , M , , / T , . The
discrepancy between A , and E(A,) (which is also
the discrepancy for any other cell within a 2 x 2
table) can be tested relative to its variance
which, subject to the fixed marginal totals-N,,,
N,,, M I , , a n d M,,-is
given by V ( A , ) =
N,,N,,M,,MZ,/T,2(T, - 1). The corrected chi
square with 1 degree of freedom (I A , - E(A,) I
- ? h ) W ( A , )reduces in this case to (1 A,D, - B,C,
I - %T,)Z(T, - l ) / N l l N Z , M l t M 2This
t . formula
for the variance of A, is obtained as the variance
of the binomial variable N , P Q ( P = M l / T , Q =
M , / T ) , multiplied by a finite population correction factor (T - N,)/(T - 1) = N , / ( T - 1). The
earlier chi-square formula, which is ordinarily
used, essentially employs a finite population
correction factor of N J T .
There is thus a difference between the two
chi-square formulas of a factor of (T - 1)/T
which, though trivial for any single significance
test with respectably large T , can become important in the overall significance test. It is with the
latter formula, just presented, that chi square is
computed as the ratio of the square of a deviation from its expected value to its variance.
The adjustment for control factors is at this
point resolved for the resulting separate subclassifications. The problem of overall measures
of relative risk and statistical significance still
remains. A reasonable overall significance test
which has power for alternative hypotheses,
where there is a consistent association in the
same direction over the various subclassifications between the disease and a study factor, is
provided by relating the summation of the discrepancy between observation and expectation
to its variance. The corrected chi square with 1
degree of freedom then becomes (I'CA, - 'CE(A,)
I - Yz)*/XV(A,)where E(A,) and V(A,) are defined as above.
The specification of a summary estimate of
544
545
yr
Q1
?P
1 + pack
cigarettes
daily
18
46
64
1 + pack
13
236
249
3
10
13
1 1 2 1 3
1 1 9 2 0
0
15
15
31
4
5
1
1
5
4
4
1
6
8
1 8 2 0
2 3 2 5
11
12
2
2
2
1
cigarettes
daily
0
3
3
0
Nonsmokers
0
7
7
1 2 4 2 5
0 4 9 4 9
0
42
42
Total
Total
5
6
4
0
1
6
3
Other
under age 45
occupa- 45-54
tions
55-64
65 and over
4
0
1
3
4
6
6
0
2
4
6
3
2
2
0
2
7
9
11
Whiteunder age 45
collar
45-54
workers 55-64
65 andover
>
0
2
3
0
Housewives
2
5
6
11
Total
under age45
45-54
55-64
65andover
Group
1 + pack
cigarettes
daily
0
0.156
0
0
0.714
2.667
0
0.790
1.201
0
0.056
0.231
0
1.636 0
1.500 0.167
1.484 0.258
0.333
0
0
1.500
2.534
0
313 12.825
14
18
26
19
11
24
31
18
9
32
58
53
9.286
9.389
19.231
14.211
4.364
16.667
21.774
11.333
7.000
22.656
46.466
42.000
0.204
0.767
0.178
0.166
0.595
0.483
0.562
0.222
0
0.480
0.380
0
0.286
1.389
0.231
0.211
1.364
0.667
0.774
0.333
0
0.656
0.466
0
,769 13.000 0
4.615 10.400 2.600
20.000
5.700 0
4.000 3.750 11.250
2.250 8.000 0
3.600 10.000 10.000
5.520 8.333 16.667
12.000
5.500 0
7.000
2.000 0
6.720 7.143 17.857
9.000 16.333 32.667
42.000
11.000 0
0.231
0.385
0.300
0
0.750
0.400
0.480
0.500
0
0.280
0
0
Derivative computations
1%
282
10
13
25
18
6
20
27
17
9
29
55
53
Nonsmokers
Controls
Total
Epidermoidundifferentiated
pulmonary
carcinoma
Nonsmokers
m
w
a
. I *
Table 1. Illustrative computations for chi square and for summary measures of relative risk (R, R , , R p ,R3, and R4)relating to the
association of epidermoid and undifferentiated pulmonary carcinoma in women with smoking history.
w (N0 Z
- Z
If any N,, or N,, should equal zero, the estimator R, would not be defined. R, is not defined for any zero-valued N 2 2 ,and R, is not
defined for any zero-valued N,,. In these instances it would be necessary to exclude the
zero-frequency categories to define the estimators. The estimator R, retains these catego-
547
The matched-sample study previously described can be considered a special case of the
classification procedure with the number of
classifications equal to the number of pairs of
individuals. The status of pairs of well and diseased individuals classified with respect to the
presence or absence of the suspect factor in
each individual will be represented as F, G, H, or
J in the following fourfold table. The meanings
attached to the marginal totals A, B, C, and D
are the same as those in the first schematic
representation.
548
Well
individuals
Diseased individuals
Free of
With factor factor
Total
549
u1
u1
16.60
=
14.97 =
V,
x<
% (ad.)
11.83 = 28.43
16.60
V2
103
11.83
11.5678
11.70
- 2.0670
X;,
16.60
4.88
12.2900
23.76
V3
Y3
232.85
16.60 =
22.15 = X&
14.0723
Controls
251
- 18.15 =
69.15
51
103
71
5.9163
36
32
Contmls
Epiderundiffermoid-
Total
302
302
339
441
4%)
1)
Ni,Nz, MpMh,
T;L(T, - 1)
-7
c.Ni&"tMp(Tt
T,2(T,
v, =
H N,,M,JT,
The variance of the total
case frequency is
102
339
441
For the general situation
the total expected case
frequency at the jth level
of a test factor is
102
entiated
pulmonary
ConTotal
carcinoma
tmls
Total
(Df3) (W,)
(ZN,)
(ZT)
Occasional or nonsmokers
Epidermoidundifferentiated
Total pulmonary
(SM,) carcinoma
+8.24 = Y2
26.91
9.09
36
Total
(XM,)
Epidermoidundifferentiated
pulmonary
carcinoma
+9.91 = Y I
17
19
Controls
Epidermoidundifferentiated
pulmonary
carcinoma
Table 2. Illustrative computation of summary chi square, when there are three levels for study factor. The data relate to the association of
epidermoid and undifferentiatedpulmonary carcinoma in women with smoking history:
551
for its own variance. The first deviate squared is squares with the number of degrees of freedom
simply the uncorrected chi square at the first being one less than the number of test levels.
level in line (6)-the variance of the deviate
Exactly orthogonal standardized deviates
remaining as initially calculated. The second would be obtained if, in the summary analysis,
deviate is the deviation at the second level ad- as each successive total deviation from expectajusted for its correlation with the first deviation tion were evaluated, it was adjusted for its multi[adjusted Y2 = Y2 - b,,Y,; bB1= covariance ( Y , , ple regression on the preceding deviations, and
Y,)lvariance Y,)]. The variance of the adjusted then standardized by the adjusted variance.
second deviate is the initial value reduced by, This, of course, would no longer be a simplified
that portion of the variation accounted for by approximate procedure. However, it can be
the first deviation [Var. (adjusted Y 2 ) = vari- shown that for a single classification, in the mulance Y,-covariance2 (Y,, YJvariance Y , ) ] .
tiple regression of any deviation from expectaIn the present instance the summary chi tion on any sqbset of deviations, the regression
square with 2 degrees of freedom is 28.43 [line coefficients wig all be, equal; the multiple re(1 l)]. This presumably is close to the chi square gression on the set of deviations will be the same
with 1 degree of freedom which would have as the simple regression on their sum. The
been obtained had only the two most extreme equality of regression coefficients, while holdsmoking classes been compared. If one exam- ing true exactly for deviations in the separate
ines the individual uncorrected chi squares [line subclassifications, will hold only approximately
(6)], their total is found to be 45.55, the max- for
total deviations from expectation (it
imum individual figure being 23.42. It will nec- would hold exactly if equal numbers of indiessady be true that the summary chi-sqwsre value will viduals were observed from level to level at each
lie between the largest of the three chi squares and subclassification). Nevertheless, this result sugtheir total. At almost any reasonable probability level gests that approximately orthogonal deviates
these limits would be sufficient to establish statistical would be obtained if, in evaluating each sucsignificance without further calculation. In our cessive total deviation, it were adjusted for the
companion paper (27) this rule sufficed in al- cumulative total of deviations already evalumost all instances to separate the significant ated. Computing procedures to accomplish this
from the nonsignificant results.
can readily be devised.
Both approximate chi-square procedures just
outlined,
which may have merit when more
COMMENTS ON EXTENSIONS TO MORE
than three groups are being compared simulTHAN THREE FACTORS
taneously, should, in theory, yield linear combiTwo procedures can be suggested for getting nations of independent chi squares. While testapproximate summary chi squares, when there ing the chi-square values obtained as though
are a large number of levels for the test factors, they were exact is not likely to be too inapwithout the burden of computation that the propriate, it may be more correct to obtain a
exact method would entail. Both methods cal- modified number of degrees of freedom, along
culate the approximate summary chi square as a the lines suggested by Satterthwaite (47) for
sum of squares of approximately orthogonal problems involving such linear combinations.
What the modified number of degrees of freestandardized deviates.
In the first method one computes an uncor- dom would be has not been investigated by us,
rected chi square with 1 degree of freedom for and it may prove as easy to apply the exact chithe difference of the first level from all the square procedure, indicated later, as to deterremaining levels combined (the same first step mine the appropriate degrees of freedom for
as in the illustration for the three-level case). the approximate chi square.
It is of interest that a somewhat similar task of
Discarding the data from the first level, a second chi square is computed for the difference obtaining an appropriate summary chi square
between the second test-factor level and the re- appears in the birth-order problems described
maining levels combined. This is done suc- by Halperin (48). There, it was necessary to
cessively up to and including the last two re- compare a set of total observations (across fammaining levels. The approximate summary chi ily sizes) with a set of total expectations, one for
square is then the sum of the separate chi each birth order. Halperin described a matrix-
552
inversion procedure for reducing the set of correlated deviations into a summary chi square. In
that problem it can be shown that all the regression coefficients are equal in the multiple regression of the deviation at a particular birth
order on the set of deviations at all succeeding
birth orders. The second approximate method
described previously for the present problem
could thus be used exactly for the birth-order
problem, permitting simplified computation of
chi square. T h e procedure indicated by
Halperin has the advantage of generality and
could be applied to the current and related
problems, if one obtained all the necessary variances and covariances and inverted the resulting matrix.
References
( 1 ) Snow, J. On the mode of communication of
cholera. In Snow on Cholera. New York, The Commonwealth Fund, 1936, pp. 1-139.
( 2 ) Holmes, O.W. The contagiousness of puerperal fever. In Medical Chsics. Baltimore, Williams &
Wilkins Co., 1936, vol. 1, pp. 21 1-243.
(?) Stern, R. Nota sulle ricerche del dottore Tanchon intorno la frequenza del cancro. Annali Universali di Medicina 110:484-503, 1844.
(4) Stocks, P. and J.M. Campbell. Lung cancer
death rates among non-smokers and pipe and cigarette smokers. Brit M J 2:923-929, 1955.
(5) Wynder, E.L. and J. Cornfield. Cancer of the
lung in physicians. New Engl J Med 248:441-444,
1953.
(6) Lane-Claypon, J.E. A further report on cancer
of the breast, with special reference to its associated
antecedent conditions. Rept Pub1 Health 6' M Subj no.
32, pp. 1-189, 1926.
(7) Clemmesen, J., K. Lockwood, and A. Nielsen.
Smoking habits of patients with papilloma of urinary
bladder. Danish M Bull 5:123-128, 1958.
(8) Denoix, P.R. and D. Schwartz. Tobacco and
cancer of the bladder. (Bulletin de L'Association Francake pour I'etude du Cancer.) Cancer 43:387-393,
1956.
( 9 ) Lilienfeld, A.M., M.L. Levin, and G.E. Moore.
The association of smoking with cancer of the urinary
bladder in humans. AMA Arch Int Med, 1956.
(10) Mustacchi, P. and M.B. Shimkin. Cancer of
the bladder and infestation with Schistosoma hematobium. J Nut Cancer Inst 20:825-842, 1958.
(11) Lilienfeld, A.M. The relationship of cancer of
the female breast to artificial menopause and marital
status. Cancer 9:927-934, 1956.
(12) Lilienfeld, A.M. and M.L. Levin. Some factors
involved in the incidence of breast cancer. In Proc
Third National Cancer Conference, Philadelphia, J.B.
Lippincott Co., 1957, pp. 105-112.
( 1 3 ) Segi, M., I . Fukushima, S. Fujisaku, M.
Kurihara, S. Saito, K. Asano, and M. Kamoi. An
553
( 4 2 ) Lilienfeld, A.M. Possible existence of predisposing factors in the etiology of selected cancers of
nonsexual sites in females. A preliminary inquiry.
Cancer 9:111-122, 1956.
( 4 3 ) Winkelstein, W., Jr., M.A. Stenchever, and
A.M. Lilienfeld. Occurrence of pregnancy, abortion
and artificial menopause among women with coronary artery disease: a preliminary study. J Chron Dis
7:273-286, 1958.
( 4 4 ) Billington, B.P. Gastric cancer-relationships
between ABO blood-groups, site, and epidemiology.
Lancet 2:859-862, 1956.
( 4 5 ) Schwartz, D. and G. Anguera. Une cause de
biais dans certaines enquCtes mkdicales: le temps de
skjour des malades a I'hBpital. Communication a I'Institut International de Statistique, 30eme Session.
Stockholm, 1957.
( 4 6 ) Cornfield, J. A statistical problem arising from
retrospective studies. Proc Third Berkeley Symposium on
Mathematical Statistics and Probability 4: 135-148, 1956.
(47) Satterthwaite, F.E. Synthesis of variance. Psychometrika 6:309-316, 1941.
( 4 8 ) Halperin, M. The use of X2 in testing effect of
birth order. Ann Eugenics 18:99-106, 1953.
CASE SERIES
CONTROL GROUP
554
555
RESULTS
MANH
BROOKLYN
-.,..-.. .
556
I1
111
IV
V
VI
Number of accidents
in categoryb
19
7
4c
1
Id
11
See text.
Only one driver was fatally injured in each accident contributing to the case series.
One of these resulted in the death of the driver of a
fleeing stolen car who went through a red light and hit
another car broadside. A second accident in this category
resulted when the vehicle driven by a case-series member
crossed into the opposite lane on a Brooklyn to Manhattan
bridge and hit an oncoming car head on. The third accident
in this category involved a driver seen to be slumped over his
wheel before hitting two pedestrians and another car. As the
car approached the pedestrians, its estimated speed was only
10-15 miles per hour. The fourth accident resulted when the
case-seriesdriver, a transit bus driver by occupation, was seen
by two independent witnesses to slump over his wheel and
lose control of the car he was driving. It is not known whether
or not the drivers in the last two of these accidents were
literally fatally injured, but for lack of further evidence
these accidents have been classified as listed. (See also footnotes, Table 5.)
This accident resulted when a car crossed a center island
and hit an oncoming car, the driver of which was killed. (See
also footnotes, Table 5.)
Number wet
Number dry
Total
Accident
Site visitb
Total
9
34
43
6
37
43
15
71
86
P = 0.38
a Includes all sites at which the road was wet fmm any
cause, for example, from concurrent or recent rain or from
the use of street-washing equipment.
See text.
I, 11, I11
rv, v, VI
Total I-VI
Cases
Controls
37.0
47.5
39.0
34.5
35.0
35.0
The ages of the vehicles, taken as the calendar year-model year difference, in the case and
control groups were not significantly different
(P > 0.05) in any of the groups studied (1-111,
IV-VI, and I-VI), allowance being made in the
I, 11, I11
IV,
v, VI
I-VI
Total
a
Male
557
Female
Total
Cases
30
30
Controls
160
20
180
Cases
13
13
Controls
69
78
Cases
43
43
Controls
229
29
258
272
29
30 1
Pb
0.04
0.23
0.01
558
Table 5. Alcohol concentration of fatally injured drivers dying within 6 hours of their accidents, by time
of day and accident category.
Accident
category
111
a.m.
pm.
12-2
3-5
6-8
9-11
12-2
3-5
6-8
9-11
+3b
+3
+3
+3
+3
+3
No test
0
0
Oe
+3
N o testf
+ 1s
+ Ih
+3
+3
+3
+I
0
No rep
01
0
+3
+3
+2
Od
+2
v, VI
OC
+3
+3
+3
0
Table 6. Blood alcohol concentration of fatally injured drivers dying within 6 hours compared with
those of noninvolved drivers at the same accident sites.a
Accident
categoryb
I, 11, 111
IV, V, VI
Total
Blood alcohol
concentration
in mg (%)
Casesd
(#)
(%)
Controls (#)
(%)
Casesd
(#)
(%)
Controls (#)
(%)
Cases
(#)
(%)
Controls (#)
(%)
00
7
27
115
74
7
64
50
76
14
38
165(195)
74(76)
< 20
20-99
0
0
4
3
0
0
5
8
0
0
9(14)
4(5)
3
12
27
17
0
0
7
11
3
8
34(34)
15(13)
100-249
2
7
8
5
0
0
1
2
2
5
9(9)
4(3)
250-399
12
46
0
0
3
27
0
0
15
41
.om
O(0)
Lab. loss
no test
no report
2
8
2
1
1
9
3
5
3
8
5(6)
2(2)
Total
26
100
156
100
11
100
66
102
37
100
222(258)
99(99)
a Cases in which postaccident survival was 6 hours or more have been omitted to avoid the artifacts introduced by
postaccident, antemortem metabolic lowering of the alcohol concentrations present at the times of the accidents. However, for
completeness with respect to the noninvolved drivers, the figures in parenthesis give the data for the drivers sampled at all
accident sites, regardless of the duration of survival of the corresponding, fatally injured drivers.
See text.
10 mg. = 0.01 %
by wt. = 0.1 per mille.
See footnotes, Table 5.
the corresponding control group, a highly significant difference. The case-control difference
in the numbers of drivers in the accident-responsible group (1-111) with no detectable alcohol (negative) was not significantly different
(P = 0.66) from that of drivers with relatively
low concentrations (> 0, < 100 mg %).
Within the married group there was a significant difference in blood alcohol concentration
(P < 0.001) between the accident-responsible
case and control subjects (Table 7). The married
care-series members more often had higher concentrations (" 100 mg %) than the married
control-group members. Within the nonmarTable 7. Marital status and alcohol concentration
of fatally injured, accident responsible drivers,
dying within 6 hours of their accidents, and of
their site-matched controls.
Alcohol concentration
0-99 mg 7% b 100 mg % Total
Marital statusd
Marriedb
Cases
Controls
Non-rnarriedc Cases
Controls
Totald
5
103
5
43
156
4
5
10
3
22
9
108
15
46
178
559
Controls
Percent
Number
Percent
35(25)=
81(83p
166(108)a
64(60)a
6(4)
14(13)
44(35)
17(19)
5(3)
29(20)
19(17)
258(180)
lI(11)
7(9)
99(99)
O(0)
43(30)
0)
lOO(99)
a The parentheses give the data from accidents in Categories, I, 11, 111, and fmm their site-matched controls.
P = 0.04 (x 2) for the entire (I-VI) case-control residence difference.
P = 0.03 (x 2, for the accident-responsible (1-111) case-control residence difference.
P = 0.03 (partial x 2) for entire (I-V) case-control residence difference comparing residence in New York City with residence
elsewhere.
P = 0.02 (partial x 2) for accident responsible (1-111)case-control residence difference comparing residence in New York City
with residence elsewhere.
560
tent with the distributions almost invariably observed elsewhere. T h e significant preponderance of evening, nighttime, and very early
morning accidents has also been often previously observed. It is worth noting, however,
that the periods with the highest numbers of
accidents, 12-2 and 3-5 a.m., have relatively
light traffic. In contrast, the smallest number of
accidents (one) in any three-hour period occurred in the three hours ending at 9 a.m., a
period during which the citys traffic greatly
increases. These data are very similar to those
obtained in the Manhattan pedestrian accident
study, and suggest, as stated previously, that
gross exposure per se is not the major determinant of such accidents, and that the factors of
importance are more active at other times of the
day (2). This also must lead to the questioning
of the appropriateness of using the unqualqied
total of miles driven as the demoninator of accident rates calculated for research and other
purposes.
The predominance of evening, nighttime,
and early morning accidents again underscores
the importance of emphasizing the same period
in the organization of emergency medical care,
a point of general relevance in the motor-vehicle accident picture(2). For example, only six of
the forty-three member case series survived six
hours or more and thirty-four were either dead
on arrival at the medical facilities to which they
were taken, or were dead within one hour of
their accidents. Further, among the twenty-nine
injured between 6 p.m. and 9 a.m., only three
were alive at 9 a.m. Consequently, prompt medical care during these off-hours is essential.
However, since death was immediate in many
cases, only the better packaging of the vehicle
occupants could have contributed, once their
accidents had been initiated, to the possible survival of many of those killed.
The lack of a significant association between
road condition and accident occurrence (Table
2) was also seen in the pedestrian-accident
study. However, as noted in that context, the
design employed does not correct for possible
weather-associated shifts in density (2).
It is generally accepted that young drivers
have higher fatal, injury-producing, and property-damage accident rates per license holder than
do those in following decades (11). This has
been widely interpreted to indicate that such
drivers drive more dangerously. However, such
561
562
563
564
case group. The fatally injured were also significantly less often married, and in the entire casecontrol group those not married had significantly higher alcohol concentrations than those
married.
The fatally injured drivers were not significantly different in age or socioeconomic status
from the similarly exposed but noninvolved
controls, and no association between accident
involvement and vehicle age was found.
Finally, it is suggested that alcoholism rather
than merely social drinking was involved in the
case of the drivers with very high alcohol concentrations.
SUMMARY
ACKNOWLEDGMENTS
This work would not have been possible without the superb cooperation of Dr. Milton Helpern, Chief Medical Examiner, City of New York,
and of the officers and men of the New York
City Police Department. The authors particularly wish to thank former Police Commissioner
Stephen P. Kennedy; Assistant Chief Inspector
John J. King, Commanding Officer, Safety Division; Captain Milton Zarchin, Commanding Officer, Accident Investigation Squad; Lieutenant
James Donnelly, Commanding Officer, Accident Records Bureau; and Lieutenant Hugh A.
Cleary and Sgt. Joseph Cea of the Accident
Investigation Squad.
In addition, Blaine A. Braniff, Robert B.
McFadden and Richard M. Sallick of Cornell
Medical College, and Robert E. Carroll, Albany
Medical College, carried a large share of the
burden imposed by the often inconvenient site
visits, and this, together with the high level of
professional competence with which they
worked, are greatly appreciated. The authors
also wish to thank Dr. Alan M. Gittelsohn, Director, Office of Biostatistics, New York State
Department of Health, and Mrs. Wilhemina
Calhoun for their very considerable assistance
in the evaluation of the data, and the staffs of
the Division of Epidemiologic Research and the
Epidemiology Residency Program for their continued support.
References
565
METHODS
566
567
MacMahon et al.
populations included exhibited a wide range of
incidence rates for breast cancer-from a high
of 55 per 100 000 persons per year in Boston,
U.S.A., to about 10 per 100 000 persons in
Tokyo, Japan, and Taipei, Taiwan. As far as
possible, the cases included all female residents
of the study areas who were hospitalized for a
first diagnosis of breast cancer during the study
period. The controls were patients hospitalized
in the same hospitals for conditions other than
breast cancer. The three eligible patients in the
beds closest to that of the index case were interviewed for each breast cancer patient interviewed. Eligibility required being a resident of
the study population, never having had cancer
of the breast, and being over 35 years of age
(unless the breast cancer patient was under 35
years of age, in which event an age-match
within two years of the breast cancer patients
age was required). The interview form and the
study protocol were the same for all centers.
Coding, data-processing, and analyses for all
study areas were carried out in a single co-ordinating center.
In all, more than 4000 breast cancer cases
and nearly 13 000 control patients were interviewed. In five of the centers, the breast cancer
cases included 80 percent or more of the cases
known to have occurred during the study period. In two centers total ascertainment was not
possible but the interviewed cases are believed
to have represented about 50 percent of all incident cases in one (Tokyo) and about 70 percent
in the other ( S ~ O
Paulo, Brazil).
For the purpose of the present analyses, her
age at the time of birth of each of her full-term
Table 1. Number of breast cancer cases and controls interviewed in the various study centers.
Numbers included in present analyses
Cases
Numbers
excludeda
Controls
Center
Nonparous
Parous
Total
Nonparous
Parous
Total
Case
Control
Boston, U.S.A.
Glamorgan, Wales
Athens, Greece
Slovenia, Yugoslavia
S%oPaulo, Brazil
Taipei, Taiwan
Tokyo, Japan
All centers
203
161
216
153
112
34
224
1103
374
446
579
60 1
420
177
623
3220
577
607
795
754
532
211
847
4323
467
32 1
554
419
229
55
409
2454
1262
1492
1910
1862
1298
589
1832
10 245
1729
1813
2464
2281
1527
644
2241
12 699
29
12
4
18
5
3
2
73
78
37
6
27
28
4
9
189
a Women whose interview was rated unreliable by the interviewer and those for whom parity or age at first birth was not
recorded are excluded.
568
Boston
Glamorgan
Athens
Slovenia
Sao Paul0
Taipei
Tokyo
76
68
76
93
78
74
82
81
60
93
89
87
48
84
100
100
100
100
100
100
100
25
64
59
54
61
63
42
68
77
58
90
83
84
57
60
62
47
48
41
34
60
59
Table 3. Percentage distribution of cases and controls, and estimates of relative risk of breast cancer,
by age at first birth.
Nulliparous
<20
20-24
25-29
30-34
235
Tom1
19.6
27.2
22.5
27.4
39.1
41.7
29.4
27.9
27.2
37.0
26.1
33.7
42.1
48.5
41.4
34.9
55
49
71
74
65
45
49
60
23.4
24.5
22.5
28.4
13.2
14.7
27.9
23.4
23.5
24.6
23.6
27.3
9.8
20.7
24.5
22.7
76
67
79
94
94
37
78
78
12.5
11.7
14.0
12.6
3.4
8.1
9.1
10.7
10.7
10.6
10.9
10.2
2.9
4.8
6.2
8.6
90
73
106
112
84
89
100
94
6.2
6.3
5.7
6.2
2.3
2.8
4.4
5.1
4.1
3.4
3.7
4.8
0.9
1.4
2.2
3.2
117
124
127
118
175
106
138
122
100.0
100.0
100.1
99.9
100.0
100.0
100.1
100.0
100.0
100.0
100.0
100.1
100.0
100.1
100.0
99.9
100
100
100
3.1
3.8
8.2
5.0
20.5
16.6
2.8
7.4
7.5
6.7
13.2
5.7
29.3
16.2
7.5
11.2
32
38
51
81
49
54
26
50
100
48
59
76
Group
Center
Cases
Boston
Glamorgan
Athens
Slovenia
S%oPaulo
Taipei
Tokyo
All centers
Boston
Glamorgan
Athens
Slovenia
S%oPaul0
Taipei
Tokyo
All centers
Boston
Glamorgan
Athens
Slovenia
SPo Paul0
Taipei
Tokyo
All centers
35.2
26.5
27.2
20.3
21.5
16.1
26.5
25.5
27.0
17.7
22.5
18.4
15.0
8.5
18.2
19.3
100
100
100
100
All centers
Controls
Relative
riska
Married
onlyb
____
a
100
~_______
91
119
__
MacMahon et al.
S5o Paulo, but only 7 percent of those in Glamorgan, Wales, had a birth before the age of 20
years, but the relative risks for women in this
group are similar in the two centers. It is also of
interest that the reduction in relative risk appears not to be dependent on the overall level
of breast cancer rates in a particular area.
The risks for women who had their first birth
between the ages of 30 and 34 years approach
those of nonparous women, and, in all centers,
women whose first birth was delayed until the
age of 35 years or over actually had higher risks
than nulliparous women.
In view of the general similarity of these
trends, it seemed reasonable to pool the data for
all centers. Table 3 gives an estimate, from the
pooled data, that the breast cancer risk for
women having their first birth under the age of
20 years is about half that for nulliparous
women and 40 percent of that for women whose
first birth is delayed until the age of 35 years or
over.
To evaluate the effect of the assumption
made in five centers, that single women were
non-parous, the bottom row of Table 3 shows
the pooled relative risks based on married
women only. The values are almost identical
with those based on all women, and the remaining analyses are therefore based on all women
regardless of marital status.
Pooling of the data from all centers enables
estimates of relative risk for individual years of
age at first birth to to be made. These values,
for ages of 14 years to 41 years, inclusive, are
plotted in the accompanying figure. The figure
suggests that, at least up to about 30 years of
age, breast cancer risk increases linearly with
increasing age at first birth. For women having
569
Table 4. Observed number of breast cancer cases with first births when under 20 years of age and
expected values computed from the control series.
Center
Boston
Glamorgan
Athens
Slovenia*
Sao Paul0
Taipei
Tokyo
Observed
variables
Parity
Age at
interview
Duration
of schooling
18
22
65
38
117
35
24
38.2
36.5
98.2
41.6
144.6
31.3
57.1
34.7
33.5
94.0
41.4
135.8
32.6
47.8
37.7
35.9
99.7
41.8
144.2
33.6
52.5
37.5
35.4
93.2
41.6
140.5
28.1
52.5
No.
The expected values are based on the distribution of the control series, specific for the stated variables.
570
1'41
-.
..
Table 5. Estimates of relative risk of breast cancer by age at delivery, for women of parity one only.
No. of:
<20
Boston
Glamorgan
Athens
Slovenia
S5o Paul0
Taipei
Tokyo
19
(5W
44
123
66
(92Ib
52
All centers
58
20-24
72
29
64
81
70
(61)b
61
62
25-29
30-34
235
Any
age
Cases
60
100
65
83
102
(121)b
67
107
55
120
126
74b
119
118
106
81
88
(175)b
(8l)b
152
76
68
76
93
78
74
82
77
117
129
136
63
22
135
233
345
433
399
165
48
302
77
98
104
78
679
1925
(50)b
Controls
~~
Relative risks are expressed relative to a risk of 100 for nonparous women. Estimates are based on direct comparison of cases
and controls, without adjustment.
Values for cells containing less than 20 controls are shown in parentheses.
571
MacMahon et al.
Table 6. Observed and expected numbers of breast cancer cases having specified numbers of births after
the first, adjusted for age at the first birth.
No.
of cases
having
one or
None
Center
births
Obs.
Exp."
mnw
~~~...
Obs.
Expa
Obs.
4-8
Exp."
Obs.
ExD.'
Obs.
29
ExD.'
Obs.
ExD.~
77
Boston
Glamorgan
Athens
Slovenia
S o Paul0
Taipei
Tokyo
374
446
579
60 1
420
177
623
117
129
136
63
22
135
All centers
3220
679 693.7
100
Relative riskb
a
75 77.4
91 85.4
109 105.5
110 114.6
59 68.9
29 34.0
120 121.6
42
51
64
75
47
29
95
46.2
42.7
69.1
77.8
50.8
29.3
94.2
58 59.4
45 53.8
62 74.2
106 100.9
118 122.9
62 62.6
86 123.3
33
9
1
11
3.3
4.1
3.7
8.4
39.4
7.1
5.0
945 841.2
593 607.4
403 410.1
537 597.1
63
71.0
116
101
102
93
0
4
5
92
Expected values are based on rates in the control series specific for individual years of age at first birth.
Relative to a risk of 100 for women having no births after the first; data for all centers combined.
Table 7. Observed and expected numbers of breast cancer cases with specified numbers of births prior
to the age of 25 years.
No. of births prior to 25 years
No.
23
of
Cases
Obs.
Exp."
Obs.
Exp."
Obs.
89
122
137
158
109
46
172
76.0
113.3
125.3
143.9
97.9
52.1
173.9
31
46
67
64
117
47
80
38.3
51.6
76.6
68.7
115.7
39.3
76.7
11
Athens
Slovenia
SBo Paul0
131
187
243
245
321
123
272
19
39
23
95
30
20
16.7
22.1
41.1
32.4
107.4
31.6
21.5
All centers
1522
833
782.4
452
466.9
237
272.8
Center
Boston
Glamorgan
Taipei
Tokyo
Relative riskb
68
61
EX^.^
55
a The table is based on women with at least one birth before the age of 25 years. Expected values are based on rates in the
control series specific for individual years of age at first birth.
Relative to a risk of 100 for women with no births before the age of 25 years.
572
hble 8. Observed and expected numbers of breast cancer cases having births other than the first at the
age of 35 years or older.
Women whose first
birth occurred
under 35 years
of age
Center
Boston
Glamorgan
Athens
Slovenia
S%oPaul0
Taipei
Tokyo
All centers
Relative riskb
No.
of
women
No. having
one or more
other births
No. having
one or more
births at 35
years of age
or older
No.
of
women
Obs.
EXp
Obs.
Exp.=
36
38
45
47
12
6
37
18
12
25
28
6
12
17.9
8.6
11.7
21.9
4.9
3.3
12.5
338
408
534
554
408
171
586
118
131
127
22 1
94
54
141
124.6
125.3
116.0
183.7
93.6
50.5
175.8
22 1
105
80.8
2999
886
869.5
157
103
Expected values are based on rates in the control series specific for individual years of age at first birth.
Relative to a risk of 100 for women in the same age-at-first-birth category who had no subsequent births when aged 35 years
or more.
a
Socioeconomic status
Socioeconomic status, being related to both
age at first birth and breast cancer risk, must
also be examined as a possible confounding variable. In our data, the duration of a patients
schooling was found to be the measure of socioeconomic status most closely related to breast
cancer risk. As shown in Table 4, adjustment for
this variable does reduce the expected values
for patients with births under the age of 20
years but, again, the reductions are relatively
small and substantial differences between expected and observed values persist after the
adjustment.
We have also examined the possibility that the
association of breast cancer risk with age at first
birth may explain the previously observed association of the disease with socioeconomic status.
However, in those centers where differences exist between cases and controls with respect to
socioeconomic status-in particular Athens,
S5o Paulo, and Tokyo-the extent of the differ-
573
MacMahon et al.
Table 9. Estimates. of risk of breast cancer for women with first births before the age of 25 years,
relative to risks for women with first births at 30 years or older, by age at diagnosis of cancer.
Age at diagnosis
~
<45
45-54
Boston
Glamorgan
Athens
Slovenia
SPo Paul0
Taipei
Tokyo
56
54
63
65
42
70
38
28
61
54
66
64
29
60
All centers
Ab
B'
54
55
55
55
Center
55-64
65-74
275
51
37
68
81
63
92
57
62
72
77
113
33
37
59
56
62
66
70
81
75
82
Estimates are not shown for cells containing less than five expected cases.
Estimates derived from actual sums of the data in each age group from all centers.
Weighted means of the values shown for individual centers. Weights are the number of controls included, as shown in
Table 1.
a
574
Table 10. Observed and expected numbers of breast cancer cases, by marital status and age at first
marriage; nonparous cases only.
Observed
or
expected
Never
married
Boston
Ohs.
Exp.a
114
108.0
Glamorgan
Obs.
EX^.^
Athens
Center
25-29
30-34
8
7.3
16
21.7
25
19.4
8
12.1
28
30.5
199
199.0
61
61.2
3
3.8
25
28.5
32
28.4
12
12.3
27
25.9
160.0
160.1
Obs.
Exp.a
80
74.5
16
27.5
27
32.1
29
31.3
25
20.4
34
25.2
21 1
211.0
Slovenia
Obs.
EX^.^
92
89.9
1
1.9
13
11.7
17
17.9
16
12.0
17
22.6
156
156.0
S%oPaul0
Obs.
Exp.a
67
55.2
18
26.0
17
20.1
3
7.8
3
3.1
8
3.9
116
116.1
Taipei
Ohs.
Exp.=
1
8.2
10
8.6
14
11.8
4
0.9
2
0.6
2
2.9
33
33.0
Tokyo
Obs.
Exp.=
65
58.3
20
18.6
66
75.7
40
46.4
17
13.6
15
10.5
223
223.1
All centers
Obs.
Exp.
480
455.3
76
93.7
178
201.6
150
152.1
83
74.1
131
121.5
1098
1098.3
105
81
88
99
112
108
102
104
90
104
109
96
<20
235
Total
Expected values are derived from the control group of the same center, adjusted for age at interview (5-year groups).
Excludes 1 case with unknown age at marriage.
Relative to a risk of 100 in all nonparous women
Again, numbers are too small for examination of trends in individual centers. The pooled
data for all centers do suggest lower risks for
nulliparous women married under the age of
25 years than for those married later. However,
relative to the trend in risks associated with age
at first birth (Table 3), that with age at marriage
is weak. In addition, the deficit of cases observed among nulliparous women first married
under the age of 20 years is confined to two
centers. If these are excluded (bottom line of
Table lo), the trend disappears. We have no
explanation for the appearance of this feature
in these two centers. In view of the relatively
small change in risk associated with it and its
limitation to two of the seven centers, we conclude that early marriage is not associated with
reduction in risk of cancer of the breast, unless
it is associated with early confinement.
DISCUSSION
MacMahon et al.
575
576
ACKNOWLEDGMENTS
This study was made possible by the cooperation of our colleagues who gave access to their
records and permission to interview their patients. More detailed recognition than can be
given here of our indebtedness to these individuals and to our staff who conducted the interviews has been given in the reports from the
individual centers.
We are grateful to Mrs. Hazel Coven, Mrs.
Joyce Berlin, and Mrs. Linda Desmond who
have been responsible for the coding, data processing, and computer programming, respectively, and to Dr. Dimitrios Trichopoulos for
assistance in the statistical analysis.
For encouragement and counsel at all stages
of the investigation, we are indebted to staff of
the World Health Organization.
References
( 1 ) Lowe, C.R. and B. MacMahon. Lancet
1:153-156, 1970.
( 2 ) Salber, E.J., D. Trichopoulos, and B. MacMahon. J Nat Cancerlnst 43:1013-1024, 1969.
(?) Valaoras, V.G., B. MacMahon, D. Trichopoulos, and A. Polychronopoulou. Znt J Cancer
4:350-363, 1969.
( 4 ) Yuasa, S. and B. MacMahon. Bull W H O
42:195-204, 1970.
( 5 ) MacMahon, B., T.M. Lin, C.R. Lowe, A.P.
Mirra, B. Ravnihar, E.J. Salber, D. Trichopoulos, V.G.
Valaoras. and S. Yuasa. Bull WHO 42:185-194, 1970.
(6) Lane-Claypon, J.R. A further report on cancer
of the breast with special reference to its associated
antecedent conditions. Report on Public Health and
Medical Subjects, No. 32. H.M. Stationery Office,
London, 1926.
(7) Wainwright, J.M. AmerJ Cancer 15:2610-2645,
1931.
(8) Gilliam, A.G. J Nut Cancer Znst 12:287-304,
1951.
(9) Stocks, P. Schweit Z Path 18:706-717, 1955.
( 1 0 ) Segi, M., I . Fukushima, S. Fujisaku, M.
Kurihara, S. Saito, K. Asano, and M. Kamoi. Gann 48,
Suppl.:l-63, 1957.
(11) Wynder, E.L., I.J. Bross, and T. Hirayama.
Cancer 13:559-601, 1960.
(12) Levin, M.L., P.R. Sheehe, S. Graham, and 0.
Glidewell. Amer/ Pub1 Hkh 54t580-587, 1964.
MacMahon et al.
(13) Stocks, P. Practitioner 179:233-240, 1957.
(14) Kaplan, S.D., and R.M. Acheson. J C h o n Dis
19:1221-1230, 1966.
(15) Cole, P. and B. MacMahon. Lancet 1:604-606,
1969.
577
( 1 6 ) MacMahon, B . , a n d P. C o l e . C a n c e r
24:1146-1150, 1969.
(17) Moon, R.C. ZntJ Cancer 4:312-317, 1969.
(18) Dao, T.L., and J. Sunderland.JNat Cancer Znst
23:567-585, 1959.
ABSTRACT
The possibility that the use of conjugated estrogens increases the risk of endometrial carcinoma was investigated in patients and a twofold age-matched control
series from the same population. Conjugated estrogens (principally sodium estrone
sulfate) use was recorded for 57 percent of 94 patients with endometrial carcinoma,
and for 15 percent of controls. The corresponding point estimate of the (instantaneous) risk ratio was 7.6 with a one-sided 95 percent lower confidence limit of
4.7. The risk-ratio estimate increased with duration of exposure: from 5.6 for 1 to
4.9 years exposure to 13.9 for seven or more years. The estimated proportion of
cases related to conjugated estrogens, the etiologic fraction, was 50 percent with a
one-sided 95 percent lower confidence limit of 41 percent. These data suggest that
conjugated estrogens have an etiologic role in endometrial carcinoma.
Between 1962 and 1973, dollar sales of estrogen quadrupled in the United States (1, 2).
Conjugated estrogens (Premarin, Ayerst Laboratories) containing principally sodium estrone
sulfate constituted the vast majority of the
quantity supplied. A recent series of articles by
Siiteri and his colleagues (3-5)has suggested
that the estrone form of estrogen might be associated with the development of endometrial
cancer. Siiteris theory is consistent with previous data from animal experiments indicating
carcinogenicity of estrogen. (6-9). In addition,
MacMahon (10) cites clinical and epidemiologic
evidence that exogenous estrogen increases the
risk of the development of endometrial cancer.
The present work addresses the relation of
estrogen to endometrial cancer using the casecontrol approach. Members of the Kaiser Foundation Health Plan with endometrial cancer
who were reported to the tumor registry of the
Kaiser Permanente Medical Center, Los Angeles, were compared with control subjects selected from the same Health Plan population
and matched for age, duration of Health Plan
membership, and area of residence.
Source: New England Journal of Medzczne 293(23): 1167-1 170,
1975.
Fmm the Department of Obstetrics and Gynecology, Kaiser Permanente Medical Center, Los Angeles and the Department of Medical Economics, Kaiser Foundation Health Plan,
Southern California Region.
Patients
Between July 1, 1970, and December 31,
1974, the diagnosis of endometrial cancer was
made in 94 patients at the Kaiser Permanente
Medical Center, Los Angeles, and reported to
its tumor registry. The criterion for the definition of endometrial cancer was a pathological
diagnosis of endometrial adenocarcinoma or
adenoacanthoma; mixed Miillerian sarcoma
and choriocarcinoma were excluded.
Control Subjects
Control subjects were selected in the following way. The membership files of the Southern
California Kaiser Foundation Health Plan population were reviewed, and all members in the
vicinity of the Los Angeles facility whose record
designations ended in arbitrarily selected numbers were identified and listed. From the list,
two control subjects were selected for each patient and matched for birth date within one
year, area of residence by postal zip code, duration of Health Plan membership (each control
subject had been a member at least as long as
the associated patient), and potential for the
development of endometrial cancer by the control subjects having an intact uterus. The patient and the two control subjects thus constituted a matched triple.
578
579
Record Review
RESULTS
Table 1. History of conjugated-estrogen use among 94 patients with endometrial cancer and 188
matchedP control subjects.
Distribution of triples by use of conjugated estrogens
Controls' use of
conjugated estrogens
Patients' use of
conjugated estrogens
Used
Did not use
Totals
both
one
neither
totals
1
0
16
37
29
66
54
11
27
1
Used
conjugated
estrogens
Exposure rates:
Patients
Controls
No.
54
29
40
94
%
(57)
(15)
No.
40
159
Totals
%
(43)
(85)
94
188
a The matching criteria were age, area of residence, duration of Health Plan membership, and potential for development of
uterine cancer.
580
(Rw,
Patients (No.)
Controls (No.)
Unknown
27
5.0-6.9
1.0-4.9
<1
14
14
9
5
14
3
4
40
159
RR
(RR)
X21
Nonexposed
9.3
4.2
22
<10-5
13.9
6.0
26
<10-5
7.2
2.8
12
<.01
10
5.6
2.7
Totals
188
94
( 1 .O)b
15
c.01
~
Duration of use for both patients and controls was defined by difference in year between date of most recent prescription
for conjugated estrogens and date of first such prescription given in record. If a first prescription date was given, hut
subsequent prescription dates were absent, a statement in the record, such as conjugated estrogens for five years was
acceptable as a statement of duration. If a first prescription date was given, hut subsequent dates were absent and no summary
of conjugated estrogens use was in the record, the duration of use was defined as unknown.
By definition.
a
Data were also recorded on the use of conjugated estrogens by control subjects who had been excluded from the
study because of hysterectomy or radiation therapy to the
uterus. If the restrictive criterion of an intact uterus in the
control subjects had not been applied, the estimated relative
risk and etiologic fraction would have been 4.9 and 46 percent respectively.
581
~~
tients and 80 percent of the controls; the confounding risk ratio, crude risk ratio, and confounding effect were estimated as I.I., 5.7, and
2 percent respectively.
The analogous calculations for confounding
by the risk factor age at menopause were as
follows. The risk ratio for endometrial cancer
for subjects with an age at menopause of 51
years or more versus those with an age at menopause below 51 was estimated to be 1.3, and the
rate of conjugated-estrogen use was slightly
lower in the former age group. On the basis of
data obtained from 90 percent of the case records and 70 percent of the control records, the
confounding risk ratio, crude risk ratio, and
confounding effect were estimated as 1.08, 5.6,
and 2 percent, respectively.
From these considerations, it is apparent that
the observed association between conjugatedestrogen use and the development of endometrial cancer cannot be explained to any appreciable extent by confounding due to age,
parity, excessive weight, or age at menopause. It
is possible, of course, that there could be a major confounding effect by other, unknown factors that lead to the development of endometrial cancer.
As an overall check on the validity of the
method, data were also collected on the use of
diazepam (Valium), reserpine, and thyroid
drugs by both patients and controls. The point
estimates of the risk ratio were 0.7,0.5, and 0.9,
respectively. All these estimates are small in
comparison with the estimated risk ratio of 7.6
that was found for conjugated estrogens. In addition, data were collected on the indications
given in the record for prescribing conjugated
estrogens. Where an indication was recorded
(for 54 percent of patients and 72 percent of
controls), it was hot flashes for 72 percent of
patients and 71 percent of controls, indicating
that the reasons for prescribing this drug were
similar for patients and controls. Chance is an
extremely unlikely explanation, considering the
magnitude of the P value.
Causal explanation of the association involves
the difficulty of explaining why an association
of this magnitude has remained undetected until now. Estrogens have been used extensively
only during the last decade or two (1,2), and if
the results of this study are generalizable to all
post-menopausal women, there should have
been an appreciable increase in the occurrence
582
ularly before 1971. However, t h e d a t a do indicate a significant increase (uncorrected for hysterectomy) in 1972 to 1974.
Causal interpretation of t h e association between conjugated-estrogen use and t h e development of endometrial cancer has some biologic credibility. Carcinogencity of estrogens has
been demonstrated i n laboratory animals a t
various sites (6-9), including t h e e n d o m e t r i u m
(9). In addition, t h e cases of endometrial cancer
observed in surgically castrated women or in
girls with ovarian dysgenesis exceed t h e expected n u m b e r (28-33),and these women receive replacement estrogen therapy. Moreover,
endometrial cancer has been f o u n d t o be associated with high estrogen-producing granulosatheca ovarian tumors (34-36).
Recently, Siiteri e t al. (3-5) have suggested a
theory of h o r m o n e conversion that implies a
higher level of estrone i n women i n whom endometrial cancer develops t h a n i n those in
whom it does not. They f o u n d rates o f conversion of androstenedione to estrone two t o three
times as high in women with endometrial cancer or hyperplasia (a precursor of endometrial
carcinoma) (37,38) as in women without such
cancer. Confirming Siiteri's findings, Schindler
and his associates (36) discovered t h a t adipose
tissues of patients with endometrial cancer converted androstenedione to estrone nearly f o u r
times as fast as those of subjects without cancer.
T h e evidence for a connection between t h e
use of conjugated estrogens and t h e development of endometrial cancer seems r a t h e r persuasive. Caution is u r g e d , however, i n view of
the absence of d a t a b o t h f r o m similar epidemiologic studies in o t h e r populations and
from follow-up studies. Such information is necessary before policy conclusions c a n be drawn.
Further studies are necessary t o evaluate t h e
possible relation between t h e use of o t h e r estrogens and endometrial cancer.
References
(1) Shipments of Pharmaceutical Preparations, Except Biologicals, 1962, U.S. Bureau of the Census
Current Industrial Reports, Series M28G[62]- 1.
Washington, D.C., Government Printing Office,
1963.
( 2 ) Pharmaceutical Preparations, Except Biologicals, 1973, U.S. Bureau of the Census Current Industrial Reports, Series Ma-28G[73]-1. Washington D.C.,
Government Printing Office, 1975.
(3) Siiteri, P.K., B.E. Schwartz, I. Moriyama, et al.
Estrogen binding in the rat and human. Adv Exp Med
Biol 36:97-112. 1973.
( 4 ) Siiteri, P.K. and P.C. MacDonald. The role of
extraglandular estrogen in human endocrinology.
Handbook of Physiology. Section 7, Endocrinology, Vol.
2, Part 1. Edited by R.O. Creep and E.B. Astwood.
Washington, D.C., American Physiological Society,
1973, pp. 615-629.
( 5 ) Siiteri, P.K., B.E. Schwartz, and P.C. MacDonald. Estrogen receptors and the estrone hypothesis in relation to endometrial and breast cancer.
Gyneco Oncol 2:228-238, 1974.
(6) Cook, J.W. and E.C. Dodds. Sex hormones and
cancer-producing compounds. Nature 131:205-206,
1933.
(7) Perry, I.H. and L.L. Gintzon. The development
of tumors in female mice treated with 1:2:5:6-dibenzanthracene and theelin. A m J Cancer 29:680-704,
1937.
(8) Gardner, W.U. Tumors in experimental animals receiving steroid hormones. Surgery 16:s-32,
1944.
(9) Meissner, W.A., S.C. Sommers, and G. Sherman. Endometrial hyperplasia, endometrial carcinoma, and endometriosis produced experimentally
by estrogen. Cancer 10:500-509, 1957.
(10) MacMahon, B. Risk factors for endometrial
cancer. Gynecol Oncol 2:122-129, 1974.
( 1 1 ) Miettinen, O.S. Estimation of relative risk
from individually matched series. Biometrics 26:75-86,
1970.
(12) -.
Individual matching with multiple
controls in the case of all-or-none responses. Biometrics
25:339-355, 1969.
(13) -.
Simple interval-estimation of risk
ratios. Am J Epidemzol 100:515-516, 1974.
(14) -.
Proportion of disease caused or prevented by a given exposure, trait, or intervention. Am
J Epidemiol 99:325-332, 1974.
(15) -.
Estimability and estimation in casereferent studies. Am J Epidemiol (in press).
(16) ___. Components of the crude risk ratio.
Am J Epidemiol 96:168-172, 1972.
(17) Cramer, D.W., S.J. Cutler, and B. Christine.
Trends in the incidence of endometrial cancer in the
United States. Gynecol Oncol 2:130-143, 1974.
(18) Kinlen, L.J. and R. Doll. Trends in mortality
from cancer of the uterus in Canada and in England
and Wales. B r J Preu SOCMed 27:146-149, 1973.
(19) McMahon, B. and M. Feinleib. Breast cancer
in relation to nursing and menopausal history. J Natl
Cancer lmt 24:733-753, 1960.
583
I E . C O H ~ RSTUDIES
T
I
_____~
584
Goldberger et al.
merce and a faulty system of arrangement between landlords and the cultivators of the soil,
all tending to depress agriculture and to reduce
the peasantry at large to a state of much misery
and privation. Continuing this discussion, Holland remarks further (p. 333): Animal food
rarely forms a part of their diet, and although
living on a soil which produces wine their poverty almost precludes the use of it, even when
sickness and debility render it most needful.
The same condition of poverty is evident in
their clothing, in their habitations, and in the
want of all the minor necessaries and comforts
of life. The immediate effect of these privations
is obvious in the aspect of squalid wretchedness
and emaciation which forms so striking a spectacle at the present time throughout the greater
part of Lombardy. I say particularly at thepresent
time (italics in original), because whatever may
have been the progress of misery among the
peasants of this country during the last half
century it appears to have increased in a tenfold
ratio during the last two years, the effect of bad
harvests added to the preceding wars and political changes which have distressed this part of
Italy.
Hameau (lo),in the first recorded observations of pellagra in France, reported that this
disease attacks individuals of both sexes and all
ages, but I have not yet seen it in any but the
poor and uncleanly who subsist on coarse
food.
Lalesque ( I I ) , in his account of pellagra of
the Landes, cites a number of instances illustrating the conditions of misery under which
pellagra occurred, finally exclaiming (p. 42 1):
These are the individuals attacked by pellagra,
for it attaches itself to poverty as the shadow to
the body.
In a discussion of pellagra in Gorz-Gi adisca,
Berger (12), very significantly observes: The
appearance during the last decennium of diseases of the vine, the reduction in value of the
product of the soil because of foreign competition, crop failures, increase in taxes, increasing
living costs, all operated to undermine economic conditions, particularly of the poorer
country folk, and thus prepared favorable conditions for the spread of the disease.
Discussing the therapy and prophylaxis of
pellagra in Bessarabia, V. Rosen (13)bewails the
attendant difficulties in that, on the one hand,
the alimentation with cornmeal porridge is a
585
586
Goldberger et al.
PLAN AND METHODS OF PRESENT STUDY
Locality
The study was made in seven representative
cotton-mill villages situated in the northwestern
part of South Carolina.
Population
The villages were of about average size; none
had over 800 or less than 500 inhabitants. Each
constituted a distinct, more or less isolated community in close proximity to a cotton-cloth
manufacturing plant and was composed practically exclusively of the mill employees and
their families. The few Negro families present
and living somewhat apart were not considered,
so that our study deals with an exclusively white
population, which, with hardly a single exception, was of Anglo-Saxon stock born in this
country of American-born parents. Besides the
Negroes, there were also excluded from this
study the mill executives, store managers,
clerks, and their households, so that we had left
for study an exceptionally homogeneous group
with respect to racial stock, occupation, and
general standard of living, including dietary
custom. An enumeration of the population was
made in May and June in connection with the
collection of our dietary and economic data,
and totaled about 4160 people, included in
about 750 households.
Pellagra Incidence
The procedure adopted for determining the
incidence of pellagra in this population has
been described at length in a previous paper of
this series (2).
Briefly, in order to ascertain the incidence of
the disease as completely as possible, the expedient of a systematic biweekly house-to-house
search for cases was employed and practically
exclusively depended on.
Orily cases with a clearly defined, bilaterally
symmetrical dermatitis were recorded as pellagra; cases with poorly defined eruptions, or
those with more or less suggestive manifestations but without clearly marked eruption, were
recorded at most as suspectsand are excluded
from present consideration.
Just as in our study of pellagra incidence in
relation to diet, so here, in relating pellagra
incidence to economic conditions, no distinc-
587
Season
It would seem reasonable to expect, if diet,
economic status, or other factor has any influence in relation to the seasonal rise in incidence
of the disease, that this influence is most effective during a period immediately anterior to the
sharp rise and peak of incidence. Such statistics
of pellagra morbidity as were available to us at
the beginning of our study indicated that the
rise of the seasonal curve of pellagra incidence
in the southern states began in the late spring
and reached its peak in June. It was assumed,
therefore, that the factors favoring the production of pellagra were most effective during the
season beginning some time in the late winter
or early spring and continuing u p to or possibly
somewhat into June. The period actually selected by us as representative of this season extended from April 16 to June 15, 1916. Information relating to family income, household
food supply, and the composition of the households, etc., for sample sections of this period
was secured by trained enumerators who canvassed the village in successive 15-day periods
under the immediate direction and supervision
of one of us (E. S.).
Dietary Data
The methods adopted for securing data relating to diet have been described fully in a pre-
588
Family Income
The data relating to family income were secured by inquiries of the housewife or of some
other responsible member or members of each
family, supplemented by data from the mill
payrolls. For the latter we are greatly indebted
to the willing cooperation of the administrative
officials of the mills.
The information obtained from the families
covered ( a ) the rate of daily earnings of each
member earning wages during the half month
preceding the week of the canvass and the various rates of daily earnings of all members who
had been employed during the 12 preceding
months; (b) the days not at work for all members
who had worked for wages during the 12 preceding months; (6) the income from all other
sources during the preceding half month as
well as during the preceding 12 months, this
information being secured in detail for each
source of income. On the basis of this information it was possible to approximate the total
income of each family for the half month preceding the visit of the enumerator, and,
roughly, for any part or all of the preceding
year.
Goldberger et al.
sources of information and the nature of the
information sought were as follows:
(1) Statements were obtained from households as to the immediate source of every article
of food entering into their half-months supplies. Thus it was ascertained, for example,
whether the fresh milk used by the household
was produced at home, purchased from another mill workers household in the village, or
from some specific farmer, dairy, or store, or
donated by a relative, neighbor, or other person. In the event that a household had a source
of supply not common generally to households
in the village, inquiries were directed with a
view of ascertaining the length of time the
household had had such a supply, particularly,
with respect to the period after January 1, 1916.
(2) From farmers, hucksters, or peddlers
selling from house to house, statements were
secured relating to the quantities sold, prices,
frequency of selling, and character of produce
sold since January 1, 1916.
( 3 ) From m a n a g e r s a n d clerks in t h e
stores, markets, and other retail establishments at which mill workers households
largely dealt, data were secured relating to
( a ) prices during the 15-day period and price
changes during 1916; (b) sources of each food
sold, whether direct from nearby farms o r
through middlemen from local agricultural territory or from other sections of the United
States; ( c ) names of brands and quantities of the
foods sold; (d) practices with respect to credit to
mill workers households, especially as affected
by the amount of earnings by the mill workers.
Economic Classification
Method of Classifcation According to Economic
status
As has already been mentioned, the great
majority of the individuals composing the population studied were members of families who
subsisted on the income of families composed
of several persons; the small proportion not
subsisting on such family income were boarders
living under substantially the same conditions
as the families with which they boarded. It
would seem permissible, therefore, to classify
these economically with the members of the
family with which they boarded, although it is
fully recognized that in so doing a certain,
589
590
Male
Female
1 .O
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.8
0.8
0.7
0.6
0.6
0.5
0.4
0.3
~~
Results of Classification
The 747 families for which income data were
sufficiently accurate and complete for consideration have been classified by this method and
grouped into four convenient classes, each containing a fair proportion of the total number.
Table 1 presents this classification and also the
resulting distribution of individuals and their
equivalent adult male units.
Families
Personsa
Equivalent
adult male
unitsb
Number
Less than $6.00
$6.00-$7.99
$8.00-$9.99
$10.00 and over
217
183
139
208
1289
972
704
800
866.2
675.9
529.2
607.1
All incomes
747
3765
2678.2
Percent
100.0
Percent
100.0
Percent
100.0
29.1
24.5
18.6
27.9
34.2
25.8
18.7
21.3
32.4
25.2
19.8
22.6
All incomes
Goldberger et al.
59 1
Table 2. Average half-month family income, computed in terms of per family, per person, and
per adult male unit, for various income classes of the population in seven cotton-mill villages
in South Carolina.
Average income during
a half month
All family
income during
a half month
Per family
Per person
Per adult
male unitb
$3990.45
4780.85
4642.29
7777.99
$18.38
26.12
33.40
37.39
$3.09
4.92
6.55
9.72
$4.61
7.07
8.77
12.81
All incomes
21 191.58
28.36
5.63
7.92
of the wage earners, including chiefly the factors of natural ability, length of training, and
state of health. In the classification of this population according to family income per adult
male unit, those persons in the higher income
classes appeared distinctly to have the advantage in each of these respects over those in the
lower income classes.
PELLAGRA INCIDENCE ACCORDING TO
ECONOMIC STATUS
marked degree of regularity as income increases. This inverse correlation is even more
clearly shown when weight is given to households with more than one case of the disease,
as is done in Table 4, in which the incidence of
pellagra is expressed as a rate per 1000 persons
in each income class.
The occurrence of multiple-case families, especially from the point of view of difference in
income, invites special comment. The 97 cases
of pellagra occurred in 61 families. In each of
24 of these families, two or more cases occurred,
while in each of 8, three or more cases developed. Taking into consideration the size of the
families and assuming that all individuals were
equally susceptible to the disease,3 a computation of the probability of the occurrence of multiple-case families according to purely chance
distribution indicated that in the 747 families
we should expect about 90 Pamilies with one
case each, about 8 families with two or more
cases, while the probability of households each
with three or more cases would be less than 2 in
10 000. The actual occurrence of 24 families
with two cases each and of 8 families with three
or more cases would thus seem to be far in
~
* Upon
592
Table 3. Number and percent of households of different income classes affected with pellagra in seven
cotton-mill villages of South Carolina in 1916.
Pellagrous households
in which were:
Half-month family income
per adult male unit
All
households
One or
more cases
of pellagra
Two or
more cases
of pellagra
Three or
more cases
of pellagra
Number
Less than $6.00
$6.00-$7.99
$8.00-$9.99
$10.00-$13.99
$14.00 and over
217
183
139
144
64
28
21
8
3
1
17
3
4
0
0
All incomes
747
61
24
Percent
~~
100.00
100.00
100.00
100.00
100.00
12.9
11.5
5.8
2.1
1.5
7.8
1.6
2.9
0.0
0.0
3.2
.5
0.0
0.0
0.0
All incomes
100.00
8.2
3.2
1.1
'
Goldberger et al.
593
Table 4. Number of definite cases of pellagra and rate per 1000. among persons of different income
classes in seven cotton-mill villages of South Carolina in 1916.
Males
Total
Halfmonth family
income per adult
male unit
Number
of
persons
Number
of
cases
Rate
Per
1312
1037
784
736
29 1
56
27
10
3
1
42.7
26.0
12.8
4.1
3.4
All incomes
4160
97
23.3
1000
Number
of
persons
Females
Number
of
cases
Rate
Per
1000
Number
of
persons
Number
of
cases
Rate
Per
1000
650
521
376
363
161
20
6
4
0
1
30.8
11.5
10.7
0.0
6.2
662
516
408
373
130
36
21
6
3
0
54.4
40.7
14.7
8.0
0.0
2071
31
14.9
2089
66
31.6
~~
Comparison of crude pellagra rates and of rates after adjustment for age
to a standard population for each income class
(Standard population-total population, all incomes)
Family income per adult
male unit
Less than $6.00
$6.00-$7.99
$8.00-$9.99
$10.00-$13.99
$14.00 and over
Adjusted
42.7
26.0
12.8
4.1
3.4
41.0
24.8
14.2
5.2
2.5
a Since a marked variation in the pellagra rate according to age and sex was found for the population studied (3),and since,
ordinarily, differences in the distribution of persons according to age occur in different economic groups, computation of rates
adjusted to a standard population was made. The influence of differences in the sex distribution in any age group was
insignificant, and practically the same incidence rates were obtained after making adjustments to a standard age distribution, as
is shown in the above table.
594
Table 5. Number and percent of persons in each income class, classified according to age,in seven
cotton-mill villages of South Carolina in 1916.
(The classes being divided from each other at those ages at which the pellagra incidence rate for the
whole population varies most sharp1y.a)
Age group
~~~
All
ages
Under
5 years
5-9
10-19
20-29
30-44
45-54
55 years
andover
Number
~~
1312
1037
784
736
29 1
260
162
104
95
27
251
166
108
69
15
317
270
229
173
71
162
172
149
215
91
217
166
114
102
63
49
60
48
46
9
56
41
32
36
15
All incomes
4160
648
609
1060
789
662
212
180
24.2
26.0
29.2
23.5
24.4
25.5
12.4
16.6
19.0
29.2
31.3
19.0
16.5
16.0
14.5
13.9
21.6
15.9
3.7
5.8
6.1
4.3
3.9
4.1
4.9
5.2
4.3
Percent
Less than $6.00
$6.00-$7.99
$8.00-$9.99
$10.00-$13.99
$14.00 and over
All incomes
a
100
100
100
100
100
100
19.8
15.7
13.3
12.9
9.3
15.6
19.1
16.0
13.8
9.4
5.2
14.6
6.2
3.1
5.1
See (2).
crr
u1
100
126
138
144
138
167
$14.00andmore
$10.00-$13.99
$8.00-$9.99
$6.00-$7.99
Less than $6.00
Pellagrous
households
100
100
113
107
100
34
31
31
35
33
31
119 110
100
121
120
138
134
150
126
152
151
174
169
Dried fruits
10
10
9
9
9
10
361
447
434
410
460
399
127
319
302
342
317
282
Sugar syrup
48
59
61
67
55
50
Butter
11
30
35
14
19
19
5
4
5
Rice
Canned vegetables
20
58
36
37
46
35
Preserved milk
4
4
3
1
2
35
63
54
49
49
37
Canned meats
16
20
12
16
14
12
Cheese
3
2
3
2
2
Eggs
31
59
57
44
38
33
61
61
49
60
105
63
Green vegetablesb
85 100
100 100
151 90
137 90
124 90
103 100
81
40
81
37
75
50
100 100
62 100
64 75
79 25
60 50
55
100
86
78
78
59
80
..
100 100
60 67
80 100
70 67
60 67
53
100
97
75
64
56
Fresh fruits
10
41
33
31
30
19
9
10
8
6
4
Canned corn
58
24
67
Relative numbers. Base: Supply per adult male unit per day in households with highest incomes
60
71
107
97
88
73
Potatoes
65
Pellagrous
households
39
49
54
56
54
Salt pork
$14.00and more
$10.00-$13.99
$8.00-$9.99
$6.00-$7.99
Less than$6.00
Half-month family
income per adult
male unit
Canned fruits
59
100
96
89
52
44
16
27
26
24
14
12
Fresh meats
16
47
32
30
21
19
9
6
4
5
2
55
34
22
11
10
8
9
5
Table 6. Average supply (per adult male unit) during a fifteen-day period between April 15 and June 16, 1916, of various
purchased artides of food. in households of different income classes and in the group of households in each of which two or
more cases of pellagra OCCURHI prior to August 1, 1916.
15
100
53
45
38
23
12
20
24
28
53
596
This is confirmed by the quite definite differences in food supply above actually shown to be
associated with differences in family income,
and further by the fact that when comparison is
made, such as Table 6 permits, it is found that
in a general, but quite definite, way the food
supply of the households of the lowest-income
class tends to be similar to that of the group of
pellagrous households in each of which at least
two cases of pellagra occurred prior to August 1 , 1916; that is, similar to that of the
group whose food supply more closely approximates a representative sample of a pellagra-producing diet than does any other
afforded by our study. [Graphic comparison
deleted. ED.]
1. Dqfirences in Diet Consumed Among Individuals of the Household. Although all members
of a household presumably have the same diet
available, as the result of individual likes and
dislikes, observable at almost any table, slight
differences in diet actually consumed are common and marked differences, amounting in
some instances to outstanding individual eccen-
2. Dq$rences in Individual Susceptibility or Resistance. Assuming identity of diet actually consumed, differences in incidence among individuals of the same household or other dietary
group may result from individual variation in
resistance or susceptibility, which may conceivably be related to ( a ) an inherent individual
characteristic, (b) the age or sex of the individual, (c) the existence of some exhausting underlying disease or condition (hookworm, dysentery, duodenal fistula), o r ( d ) to unlike
physical strain or exertion.
3. Combinations of Factors I and 2. Thus, in the
village of I n , where the highest of the incidence
rates observed by us in 1916 occurred and
where the rate among persons constituting the
households with incomes under the average was
90 per 1000, over 65 percent of these poorer
households appeared not to be affected, and, in
varying degree, this was true of each of the
seven villages studied. That the exemption of
these families from pellagra was not due to a
lack of subjects of susceptible sex is evident
from what has already been said on this point;
and that it could not be attributed to lack of
human material of susceptible age appears
very clearly when the distribution of the population according to age is compared for the pellagrous and for the poorer nonpellagrous
households in a representative village, as is
done in Table 7. Manifestly, therefore, the
amount of family income-that is, money income (in the sense here used), such as wages,
cash payments from boarders, cash receipts
from sales of supplies, and other sources-was
not the sole factor determining the character of
the household diet.
597
Goldberger et al.
Table 7. Age distribution of population constituting the nonpellagrous households with low family
income"and the pellagrous households of the mill village of In.
~
Age groups
Households
A11 ages
Under 5
5-9
10-14
20-29
30-44
45-54
33
19
52
45
31
76
14
5
19
12.5
11.3
12.0
17.0
18.5
17.5
55 and over
Number of persons
Nonpellagrous
Pellagrous
All households
265
168
433
52
31
83
53
32
85
61
49
110
7
1
Percent
Nonpellagrous
Pellagrous
All households
a
100.0
100.0
100.0
19.6
18.5
19.2
20.0
19.0
19.6
23.0
29.2
25.4
5.3
3.0
4.4
2.6
.6
1.8
That is, under $8 per adult male unit during a half-month period in the late spring of 1916.
This is quite in accord with common experience, which teaches that there are many factors
that, singly or in varying combinations, may
have an important influence on the character of
the diet and that may vary among and thus may
distinguish different households of the same
income. In illustration of this, reference may be
made to the group of factors that tend to determine the amount and proportion of family income available for the purchase of food, an
example of which is the occurrence of sickness
or injury, making an unusual draft on the family income. Related to such factors are the general spirit of the household with respect to thrift
(which, when unwisely directed, may be harmful) and the intelligence and ability of the
housewife in utilizing the available family income.
More tangible than these, and perhaps of
more immediate practical importance in its effect on the household diet, is the difference
among households with respect to the availability of food supplies. We found that, among
households with similar incomes and of the
same village and thus with access to the same
markets, there were some more favorably situated in having sources of food supplies that
others either did not possess or possessed in a
lesser degree. Such sources frequently were gardens, home-owned cows, swine, poultry, and
the like.
598
____
disclosed. Clearly the differences in pellagra incidence among these villages cannot be accounted for by differences in the economic status of the populations concerned.
(d) As family income is simply an index of the
power to buy, and as this power is obviously
Table 8. Comparison of the relation of rate of
pellagra incidence to proportion of population of
low family income in seven mill villages of South
Carolina in 1916.
Percent of population
whose half-month family
income per adult male unit
was less than:
Village
All villages
At
In
NY
Rc
Sn
Sa
WY
$6.00
$8.00
Pellagra rate
per 1000
population
(all incomes in
1916)
31.5
37.0
40.9
26.2
13.2
38.3
28.3
31.0
56.5
64.3
66.6
45.7
23.7
58.1
57.4
64.0
23.4
20.7
64.6
0.0
24.9
10.9
25.7
18.7
Table 9. Comparison of the age distribution of the population constituting the households with low
family incomes. of seven cotton-mill villages of South Carolina.
Classified by age periods (years)
Villages
All ages
Under 5
years
5-9
10-19
20-29
30-44
45-54
55 and over
Number of persons
At
In
NY
Rc
Sn
Sa
WY
All villages
367
433
331
206
338
268
407
65
83
60
37
65
51
62
65
85
56
42
46
51
72
82
110
87
50
69
68
120
63
52
45
34
61
40
39
59
76
57
32
52
34
73
18
19
15
5
14
14
24
15
8
11
6
31
10
17
2350
423
417
586
334
383
109
98
Percent
At
In
NY
Rc
Sn
Sa
WY
All villages
a
100.0
100.0
100.0
100.0
100.0
100.0
100.0
17.5
19.2
18.1
18.0
19.2
19.0
15.2
17.5
19.6
16.9
20.4
13.6
19.0
17.4
22.3
25.4
26.3
24.3
20.4
25.4
29.5
17.2
12.0
13.6
16.5
18.0
14.9
9.6
16.1
17.5
17.2
15.5
15.4
12.7
17.9
4.9
4.4
4.5
2.4
4.1
5.2
5.9
4.1
1.8
3.3
2.9
9.2
3.7
4.2
100.0
18.0
17.7
24.9
14.2
16.3
4.6
4.2
That is, under $8 per adult male unit during a half month in the late spring of 1916.
Goldberger et al.
599
600
Table 10. Comparison of availability of fresh meat as shown by the number of purchases and the average
daily supply of this food during the period May 16-30, 1916, in households, with family incomes less
than the average, of two mill villages of South Carolina.
~
Number of purchases
during 15-day period
Percent of total
households
Percent of total
households
None
1
2
17
6
7
4
5
More than 5
6
6
6
31.0
10.9
12.7
12.7
10.9
10.9
10.9
46
18
4
1
1
0
0
65.8
25.7
5.7
1.4
1.4
0.0
0.0
Goldberger et al.
60 1
Table 11. Comparison of availability of certain foods in two cotton-mill villages of South Carolina, as
indicated by the proportion of the households with family incomes under the average of the contrasted
villages purchasing the specified articles from nearby farms during the period May 16-30, 1916.
NY
Average quantity
per household
Article purchased
purchasing
Fresh milk
Butter
Eggs
Fresh vegetables
Fresh fruit
Poultry
Any of the
above articles
None
22.5 qts.
3.4 lbs.
2.9 doz.
4.0 lbs.
In
Households purchasing
Number
Percent of total
households
24
23
19
31
8
1
51.0
49.0
40.5
66.0
17.0
2.1
40
8
83.3
16.7
Average quantity
per household
purchasing
29.3 qts.
4.0 lbs.
6.0 doz.
3.0 lbs.
Households purchasing
Number
Percent of total
households
3
1
1
1
0
1
4.5
1.5
1.5
1.5
0.0
1.5
6
61
91.0
9.0
with the farming sections in states where onecrop agriculture has not been the rule. Cotton
was a relatively less important crop, and beef
cattle, swine, poultry, and milch cows seemed
much more abundant than in the In section.
Apparently greater emphasis was given to gardens, and the amount of truck produce was
noticeably larger. The physical character of the
section apparently was one cause of this difference in products. The land around In is almost
level, lies well below the foothills of the Blue
Ridge Mountains, and is well suited for the
growing of cotton. The land around Ny is quite
rolling and even hilly, being, in fact, in the
foothills of the mountains and thus not so well
suited to cotton growing. Land not suitable for
the cultivation of cotton and, hence, available
and used for corn and truck products was consequently far more abundant near Ny than near
In.
( 6 ) Contrast in market conditions. Conditions affecting the market for farm produce
from the two sections were quite different in
some important respects. The village of Ny is
itself more isolated than the village of In and is
not near any important community. The nearest railway station is a mile away and is surrounded by only about a dozen houses, including three small stores. Seneca, the nearest town
of any size (population 1313 in 1910), is some 4
miles from Ny, and Greenville, the nearest city
(population 15 741 in 1910), is about 40 miles
distant. Seneca exports comparatively little produce and hence its market is limited to local
602
Table 12. Percentages of cotton mill operatives households having supplies of various articles of food in
different quantities per adult male unit per day, compared for the mill villages of Ny and In, South
Carolina.
(All households considered have incomes of less than the average for the two villages.)
Percent of households whose average daily
supply per adult male unit was:
Article of food
Village
Fresh meats
NY
C u r e d lean meats
NY
In
Fresh milk
In
NY
In
NY
In
NY
Preserved milk
Nv
C a n n e d meats
Eggs
In
Average daily
supply per adult
male unit
(grams)
None
34
7
24
20
19
17
34
50
426
457
1
3
31.2
67.2
37.5
46.3
22.9
35.8
31.2
7.5
8.3
33.3
87.5
73.6
6.2
10.4
4.2
6.0
10.4
3.0
4.2
6.0
10.4
0.0
2.1
1.5
16.7
13.4
27.1
14.9
37.5
31.3
31.2
26.9
45.8
30.2
2.1
1.5
45.8
9.0
31.2
32.8
29.2
29.9
33.3
59.7
35.4
36.5
8.3
22.4
Goldberger et al.
603
Article of food
Butter
Cheese
Dried peas and beans
Canned peas and beans
Wheat flour
Wheat bread, cakes,
and crackers
Cornmeal
Grits
Rice
Salt pork
Dried fruits
Canned fruits
Irish potatoes
Village
Average daily
supply per adult
male unit
(grams)
None
NY
In
32
25
2
4
358
358
16.7
14.9
87.5
97.0
25.0
32.8
83.3
85.1
6.2
18.5
10.4
16.4
2.1
0.0
14.6
7.5
0.0
0.0
0.0
3.1
33.3
21.4
0.0
0.0
29.8
29.9
0.0
0.0
43.7
32.3
39.6
46.3
10.4
3.0
39.6
29.9
16.7
14.9
29.2
46.2
NY
In
NY
In
NY
In
NY
In
NY
In
13
18
139
180
4
2
4
5
54
53
18.7
25.4
29.8
20.9
87.5
95.6
75.0
70.2
4.3
10.4
12.5
6.0
4.3
0.0
4.3
0.0
33.3
22.4
29.8
17.9
0.0
0.0
0.0
0.0
57.2
41.8
35.4
46.3
36.2
61.2
12.5
4.5
25.0
29.9
34.0
47.8
NY
In
NY
In
NY
In
41
40
11
1
1
4
6.2
10.4
68.7
100.0
97.9
89.5
4.2
3.0
0.0
0.0
0.0
0.0
52.1
37.3
0.0
0.0
0.0
0.0
37.5
49.3
31.2
0.0
2.1
10.5
NY
In
88
46
14.6
22.7
12.5
16.7
39.6
37.9
33.3
22.7
NY
In
NY
In
NY
In
*Y
In
*Y
In
NY
In
NY
In
NY
In
NY
In
NY
In
36
36
40
20
12
8
10
20
34
60
0
0
5
3
46
39
17
17
3
9
29.2
26.9
25.0
43.9
53.2
70.2
66.7
56.7
45.8
53.7
0.0
0.0
81.2
88.1
10.4
9.0
68.7
64.2
70.2
40.3
2.1
7.5
10.4
9.1
0.0
1.5
0.0
0.0
4.2
3.0
0.0
0.0
0.0
22.9
28.4
20.8
28.8
17.0
6.0
2.1
1.5
20.8
3.0
0.0
0.0
0.0
0.0
45.8
43.3
2.1
0.0
4.3
0.0
45.8
37.3
43.7
18.2
29.8
22.4
31.2
41.8
29.2
40.3
0.0
0.0
18.7
11.9
39.6
38.8
29.2
35.8
25.5
58.2
NY
In
NY
In
NY
In
NY
In
26
30
3
0.0
0.0
0.0
0.0
0.0
4.2
9.0
0.0
0.0
0.0
2.5
Tabulation of approximate average dady supply of various foods in these same households deleted. ED.
Less than 0.5 gram.
604
It was quite evident, however, that home gardens contributed but very slightly, if at all, to
the food supply of households in either village
during the spring of 1916. With the exception
of an occasional (rare is perhaps a more accurate term) messor dish of greens, a very little
lettuce, and a few young onions, the gardens
had yielded no supplies during 1916 up to
about June 1. Not until after June 15 did garden produce become abundant, a condition
that was somewhat contrary to the expectation
of the authors, who had anticipated finding
considerably earlier garden production in this
section. The principal reason for this tardiness
appears to be the fact that gardens in mill villages are usually planted later than gardens
elsewhere in this section. Difficulty in getting
the ground prepared early enough, owing in
part to the fact that the long hours of work in
the mill leave no available daylight for gardening until well along in the spring, lack of initiative in making other preparations, and possibly other causes, apparently almost preclude
good early spring gardens in most of the mill
villages studied, including Ny and In, although
climatic conditions ordinarily are such that gardens can be made to yield supplies of early
varieties of vegetables during May and even in
April. Aside from a half dozen households reporting that they had had radishes, lettuce, or
English peas, only about one-third of the I n
households reported that they had had greens
or young onions even occasionally and in small
quantities before this date. In Ny the proportion
was even less.
Summing up the principal differences in
availability of food supplies during the spring
of 1916 as between Ny and In, it may be said
that (1) supplies of fresh milk, butter, green
vegetables, and fresh fruit were available to a
greater degree (better distributed among the
households) in Ny than in In, because, in the
farm territory adjacent to Ny, there was a larger
production of these articles of food and because
Ny occupied a more advantageous location as a
market for such products, and (2) that a supply
of fresh meat was available to a greater degree
in Ny than in I n because of the existence of a
fresh-meat market in Ny all the year round. In
practically all other respects the availability of
food supplies appeared to be generally similar
in the two villages.
The conditions outlined above are reflected
Goldberger et al.
in a comparison of the total food supplies during the 15-day period May 16-30, 1916, of
households in Ny and In. In this comparison
(Table 12) in order to eliminate as far as practicable the influence of differences in economic
status, only those households with less than the
average of incomes7 have been considered.
In Table 12 is shown the average quantity of
each article of food for all the households considered, as well as the percentages of the households in each village which had various quantities of each article of food, such quantities
being expressed in terms of the average for all
households in order to shorten the statistical
presentation.
This comparison indicates that during the
15-day period, May 16-30, 1916, (1) supplies of
fresh meat, fresh milk, green vegetables, and
fresh fruit were more abundant (i.e., better distributed) in Ny than in I n households; (2) supplies of cured and canned meats, salt pork, butter, flour, lard, and lard substitutes, and dried
peas and beans in Ny households were quite
similar to those in In households; and (3) supplies of eggs, corn meal, Irish potatoes, and
most canned goods were more abundant in I n
than in Ny households. Other differences in the
supplies of articles of food occurring either
rarely or in small quantities are indicated.
From the foregoing considerations it clearly
appears that the character of the household
food supply in the two villages was considerably
influenced by the availability of certain foods,
notably fresh meats, fresh milk, green vegetables, and fresh fruits, all of which were relatively
less abundant or less equally distributed in I n
than in Ny. It is clear also that these differences
in the food supply of Ny and In households are
quite similar to the differences which, as already
reported, we found to exist in the food supply
of nonpellqgrous and of pellagrous households
(1, 2).
We have here, therefore, a striking and significant correspondence between the differences
in the availability of certain foods (and thus, it is
permissible to assume, in the character of the
diet) in the two villages, on the one hand, and
the difference with respect to the incidence of
pellagra among their households on the other.
605
Since between these two villages no other differences to which significance could properly be
attached were disclosed by our study, the conclusion would seem to be warranted that the
difference in the availability of food supplies
above summarized was the outstanding determining factor in relation to the marked difference in the incidence of the disease.
Thus, of all the factors we have studied in
relation to differences in pellagra incidence
among our villages, the factor of food availability is the only one in connection with which
significant evidence of such relationship was
found. The conclusion would, therefore, seem
to be warranted that in this factor we have the
explanation for the differences among the villages studied in the incidence of the disease, so
far as this incidence was a reflection of community conditions.8
DISCUSSION
606
Goldberger et al.
607
The results of the present study clearly suggest fundamental lines along which efforts
looking to the eradication of the disease should
be directed, namely, (1) economic, by improvement of economic status (income), and (2) food
availability, by improvement in availability of
food supplies.
Measures for improving the economic status
of those people most subject to the disease, are
in the main, outside of the sanitarians sphere
and but little subject to his influence. While
much the same may be said to apply to the
conditions of food availability, this field is more
easily accessible, both directly and indirectly, to
his activities and influence. Thus, for instance,
by avoiding ill-considered regulations governing milk production he can, negatively at least,
favor an adequate supply of this invaluable
food. Furthermore, he can and should aid in
improving the conditions of food availability by
lending his powerful influence in support of
and, by cooperating with, the agencies at work
in this field, in their efforts to stimulate milk
production (particularly through cow ownership)
and to induce the farmer to adopt a suitable
system of crop diversification.
And in this connection it may perhaps be
remarked that certain preliminary observations
have created in our minds a rather strong suspicion that the single-crop systems practiced in at
least some parts of our southern states, by reason of apparently unfavorable conditions of
food supply and of other conditions of an economic character bound u p therein, will be
found indirectly responsible for much of the
pellagra morbidity and mortality with which
local agricultural labor is annually afflicted.
Although considerable study will be required
to determine definitely the factors responsible
for the high incidence of the disease in the rural
areas in question, it would, nevertheless, seem
to be the better part of wisdom to make an
earnest effort to improve conditions in the ways
suggested above.
SUMMARY AND CONCLUSIONS
lo
ting the market, may affect family income (of the farmer) as
disastrously as may crop failure.
I We hope to consider the relation of economic status to
the course of the disease from year to year in a separate
paper.
1. In the present paper are reported the results of the part of the pellagra study of cottonmill villages, during 1916, dealing with the relation of conditions of an economic nature to the
incidence of pellagra. It is the first reported
study in which the degree of the long-recog-
608
8. The most potent factors influencing pelnized association between poverty and pellagra
incidence is measured in a definite, purely ob- lagra incidence in the villages studied were: ( a )
low family income, and (b) unfavorable condijective manner.
2. The study was made among the white mill tions regarding the availability of food supplies,
operatives households in seven typical cotton- suggesting that under the conditions obtaining
mill villages of South Carolina. Pellagra inci- in some of these villages in the spring of 1916
dence was determined by a systematic, biweekly, many families were without sufficient income to
house-to-house canvass and search for cases, enable them to procure an adequate diet, and
only active cases being considered. Information that improvement in food availability (particurelating to household food supply, family in- larly of milk and fresh meat) is urgently needed
come, etc., was secured by enumerators for a in such localities.
sample section of the period April 16 to June
15, assumed to be representative of the season References
during which the factors favoring the produc( I ) Goldberger, J., G.A. Wheeler, and E. Sydtion of pellagra were assumed to be most effec- enstricker. A study of the diet of nonpellagrous and
tive.
of pellagrous households, etc. JAMA, 71:944-949,
3. Family income was made the basis of classi- 1918.
( 2 ) Goldberger,J. A study of the relation of diet to
fication according to economic status, the Atwater scale for food requirements being used pellagra incidence in seven textile-mill communities
of South Carolina in 1916. Public Health Rep
for computing the size of families in comparing 35:648-713, 1920.
their incomes.
( 3 ) Goldberger, J. Pellagra incidence in relation to
4. In general, pellagra incidence was found sex, age, season, occupation, and disabling sickness
to vary inversely according to family income. As in seven cotton-mill villages of South Carolina during
Public Health Rep 35:1650-1664, 1920.
the income fell, the incidence of the disease rose 1916.
( 4 ) Sydenstricker E., G.A. Wheeler, and J. Goldand showed an increasing tendency to affect berger. Disabling sickness among the population of
members of the same family; as the income seven cotton-mill villages of South Carolina in relation
rose, incidence fell, being reduced almost to the to family income. Public Health Rep 33:2038-2051 ,
point of practical disappearance in the highest 1918.
(5) Goldberger, J. A study of the relation of facof our income classes, although the income en- tors of a sanitary character to pellagra incidence in
joyed by this class was comparatively quite low. seven cotton-mill villages of South Carolina in 1916.
5 . The inverse correlation between pellagra Public Health Rep 35:1701-1714, 1920.
(6) Casal, G. Obra postuma del Dr. Casal pubincidence and family income depended on the
licada en 1762. Cowesp Med (Madrid) 5:78, 1870.
unfavorable effect of low income on the char(7) Strambio, G. Abhandlungen ueber das Pellagra.
acter of the diet; but family income was not the Leipzig, 1796.
(8) Roussel, T. La Pellagre. Paris, 1845.
sole factor determining the character of the
( 9 ) Holland, H. On the pellagra, a disease prevailhousehold diet.
in Lombardy. Med Chir Trans (London) 8:313-346,
6. Differences in incidence among house- ing
1820.
holds of the same income class are attributable
(10) Hameau. Note sur une maladie peu connue
to the operation of such factors as tend to deter- observke dans les environs de la teste (Gironde). Jour
mine the amount and proportion of family in- de Med Prat (etc.) de la Soc Roy de Med de Bordeaux
come available for the purchase of food, the 1:310-314, 1829.
( 1 1 ) Lalesque, fils. Actes de IAcad Roy d Sc (etc.) de
intelligence and ability of the housewife in uti- Bordeaux. p. 421, 1846.
lizing the available family income, and to the
(12) Berger, L. Pellagra. Wiener Klinik Wien
differences among households with respect to 16:161-179, 1890.
(13) Rosen, H.V. Ueber die pellagra in Russland,
availability of food supplies from such sources
Petersburg. Med Wchnschft n. F. 11:21-23, 1894.
as home-owned cows, poultry, gardens, etc.
( 1 4 ) Sofer, T. Die pellagra in Oesterreich und ihre
7. Differences in incidence among villages Bekampfung als Volkskrankheit. Therap Monatshefte
whose constituent households are economically 23:216, 219, 1909.
(15) Probizer, yon. Die Pellagra. Die Heilkunde
similar, are attributable to differences among
them in availability of food supplies resulting (Wien) 4:139-142, 1899.
(16) Babes, V. Ueber Pellagra in Rumanien. Wien
from differences (a)in the character of the local Med Presse 44:1184-1239, 1903.
markets, (b) in the produce from adjacent farm
( 1 7 ) Calmarza, J.B. Memoria Sobre L a Pelagra.
territory, and (c) in marketing conditions.
Madrid, 1870.
Goldbewer et al.
(18) Roussel, T. TraitB de la Pellagre, . . . Paris,
1866.
(19) Huertas, F. La pelagra en Espafia. Arch Latin
de Med y de Biol (Madrzd) 1:9-15, 1903.
(20) Sandwith, F.M. How to prevent the spread of
pellagra in Egypt. Lancet 1:723, 1903.
(21) Gaumer, G.F. Pellagra in Yucatan. Trans Natl
Conf on Pellagra (Columbia, S.C., 1910), pp. 101-107.
(22) Manning, C.J. Report on Certain Cases of
Psilosis Pigmentosa Which Have Recently Occurred at
the Lunatic Asylum. Barbados, 1907.
(23) Siler, J. and P. E. Garrison. An intensive study
of the epidemiology of pellagra. Am J M e d Sci
146:July and August, 1913.
(24) Jobling, J.W. and W. Petersen. T h e epidemiology of pellagra in Nashville, Tennessee, 11. J
In& Dis 21:109-131, 1917.
(25) Atwater, W. 0.Principles of Nutrition and Nutritive Value of Food. Farmkrs Bull (US Dept of Agnc
142:33, 1915.
(26) White, R.G. Report on an Outbreak of Pellagra
Among Armenian Refugees at Port Said, 1916-1917 ,
Cairo, Egypt, 1919.
609
(27) Boyd, F. D. and P. S. Lelean. Report of a Committee of Enquiry Regarding the Prevalence of Pellagra
Among Turkish Prisoners of War Alexandria, Egypt,
1919. Also J Roy Army Med Corps 33:426 et al.,
1919.
(28) Sydenstricker, E. T h e prevalence of pellagra-Its possible relation to the rise in the cost of
food. Public Health Rep October 22, 1915.
(29) Goldberger, J. The cause and prevention of
pellagra. Public Health Rep 29:2354-2357, 1914.
(30) Goldberger, J. Pellagra-Causation
and a
method of preventionJAMA 66:471-476, 1916.
(32) Vedder, E. B. Dietary deficiency as the etiological factor in pellagra. Arch IntMed 18:137-172, 1916.
(32) Roberts, S.R. Types and treatment of pellagra
JAMA 75:21-25, 1920.
(33) Vedder, E. B. Beriberi. New York, 1913.
(34) Weiss, E. Die pellagra in Sudtirol und die
staatliche bekampfungsaktion. Das Osterreichische Sanztiitrwesen (Wzen) 26:309-331, 1914.
(35) Niederman, J., E. Konrad, and E. Farkas. A
report on pellagra in Transylvania (abstract). Lancet
2:164, 1898.
Sixty-one cases of lung cancer have been recorded in persons with asbestosis (1, 2) since
Lynch and Smith (3) reported the first case. In
view of the infrequency of asbestosis, this large
number of cases suggests-but does not
prove-that lung cancer is an occupational hazard of asbestos workers. The strongest evidence
that it may be a hazard has been produced by
Merewether and by Gloyne. Merewether ( 4 )
found that lung cancer was reported at necropsy in 13.2 percent of cases of asbestosis (31
out of 235) but in only 1.3 percent of cases of
silicosis (91 out of 6884) and Gloyne (5), on
personal examination, found lung cancer in 14
percent of necropsies on subjects with asbestosis
(17 out of 121) against 6.9 percent in silicotics
(55 out of 796). Neither author gave full details
of the sex composition of the groups examined,
but since women form a higher proportion of
asbestos workers than of persons employed in
occupations liable to give rise to silicosis (coalminers, stonemasons, pottery workers, foundrymen, metal grinders) and since lung cancer is
less common among women, the differences in
the proportions of cancer cases cannot be accounted for by differences in sex distribution.
In fact the proportions which are more properly
comparable with the findings in silicotic subjects are the proportions of lung cancer found
among men with asbestosis, 17.2 percent in
Merewethers series and 19.6 percent in
Gloynes.
Animal experiments are inconclusive. A positive result was reported by Nordmann and
Sorge (6) who found that of 10 mice which had
been exposed to asbestos dust and survived for
240 days, two developed lung carcinoma. Smith
(7), however, considers that one of the carcinomas was, in fact, an example of squamous
metaplasia and that the other, an adenocarcinoma, may have developed spontaneously
from the common mouse adenoma. A negative
result has been reported by Vorwald and Karr
Source: Britzsh Journal of Zndustnal Meduzne 12231-86, 1955.
From the Statistical Research Unit, Medical Research
Council, London.
(8).The majority of workers (cited in 2) consider that a causal relationship between asbestosis and lung cancer is either proved or is
highly probable and the reality of the relationship was agreed at the recent International
Symposium on the Endemiology of Lung Cancer (9). A minority, however, remains sceptical
(10, 11), and, according to Hueper (Z), Lanza
and Vorwald, so that it was thought desirable to
undertake a fresh investigation.
NECROPSY DATA
Asbestosis Asbestosis
present
absent
All
cases
Heart failure
Pulmonary tuberculosis
Lung cancer
Other diseases of the
respiratory system
Other diseases
34
11
45
12
21
18
10
14
All causes
75
30
105
15
* Necropsies on asbestos workers are ordered by the coroner when, in his opinion, there may be a question of asbestosis being a contributory cause of death.
610
M.52 Disintegrater
M.52 Weaver
1940
1941
1942
M.51 Weaver
1952
M.36 Weaver
M.43 Fiberizer
Weaver
Weaver
Spinner
Maintenance man
Spinner
Fiberizer
Weaver
1944
1951
1948 M.59
1948 M.53
1948 M.48
1948 M.65
1950 F.51
1951 M.74
1951 M.60
M.49 Distintegrater
1939
Weaver
Weaver
Fiberizer
Weaver
M.62
M.54
M.65
M.47
1935
1935
1936
1938
Occupation
Sex
and
age
Year
of
death
1942-44
1942-44
1939-48
1941 (3/12)
1945-52
1919-32
19-32
19
1909-32
1909-32
1913-36
1913-36
1910-12
1920-37
1910-14
1919-39
1911-15
1-15
191
1919-21
1919-21
1923-39
1923-39
1913-19
1913-19
1924-38
1924-38
191 3-4 1
1912-14
1918-48
1922-35
1922-35
1922-48
1922-48
1919-48
19
19-48
1915-42
1915-42
1917-43
1919-25
1929-50
1929-50
Period
of
exposure
0
0
2
12
11
32
27
2
9
26
26
29
35
34
10
10
13
17
15
13
26
29
27
26
29
36
28
29
29
16
26
23
28
19
16
14
15
17
14
13
23
19
Years
from
first
exposure
to death
28
32
20
22
24
13
23
23
19
Years of
exposure
Years
before
of
Jan. 1,
exposure
1933
Asbestosis
Histological type
of primary
lung cancer
Pathological report
Carcinoma
Carcinoma
Carcinoma
Oat-celled carcinoma
Carcinoma
Adenocarcinoma
Oat-celled carcinoma
Anaplastic carcinoma
Oat-celled carcinoma
Carcinoma
Carcinoma
Less than 1
Carcinoma
Anaplastic carcinoma
13
Less than 1
Less than 1
8
8
1
Less than 1
Less than 1
Less than 1
3
Present Carcinoma
3
Epithelial carcinoma
Less than 1
Endothelioma of pleura
1
Carcinoma
Years
from last
exposure
to death
Table 2. Occupational history and necropsy data of asbestos workers with primary lung cancer.
Doll
61 1
6 12
recorded among the whole population of England and Wales over the same period. It might,
therefore, have been anticipated that a larger
number of cases in which the two conditions
were associated would have been found in the
last 10 years. National regulations for the control of asbestos dust were, however, introduced
in 1931 (12) and the precautions taken to prevent dust dissemination in the works had become effective by the end of the following year.
All the subjects in whom the two diseases were
found together had been employed for at least
nine years under the old conditions, and although 11 of the 15 men and women died
within 30 years of their first exposure, the association of the two conditions has not yet been
found in any person taken into employment
during the last 31 years (1923-1953).It is, therefore, possible that the reason more cases were
not found in the second half of the period is
that reduced exposure to dust has already begun to lessen the incidence and severity of asbestosis.
METHOD OF ESTIMATION OF RISK
Doll
used for the period 1922-1933, rather than
those for the mid-years, since disproportionately few men were under observation during
the early part of the period. As an example of
the method, the mortality rate for all neoplasms
other than lung cancer among men in England
and Wales aged 55 to 59 in 1951 was 2.778 per
1000. The numbers of years lived in this age
group in the five years 1949-1953 were respectively 15 years, 15 years, 17% years, 19 years,
and 19 years. The number of deaths expected
in the period was, therefore, estimated to be (15
+ 15 + 17% + 19 19) X 2.778/1000 =
0.238. The total number of deaths expected
from each category of diseases was obtained by
adding the numbers thus calculated for each
age group for each of the five periods.
The great majority of the men lived and,
when they died, died in the town in which the
works was situated, so that it would have been
preferable to have based the calculation of the
expected deaths on the death rates observed in
that town rather than on the rates for all England and Wales. These, however, were not
known in sufficient detail. Little error in the
expected number of deaths from lung cancer is
likely to have been introduced on this account
since, according to Stocks (I3),the age-adjusted
death rate for lung cancer among men in the
town concerned was 96 percent of the rate for
England and Wales. Stocks figure was calculated only for the period 1946-1949, but the
proportion is unlikely to have varied greatly
over the longer period of the investigation. The
expected number of deaths from all causes is,
however, likely to be somewhat underestimated
6 13
RESULTS
Table 3. Number of man-years lived by men with 20 or more years of work in a scheduled area.
Period
Age
(years)
30354045505560657075-79
All ages
All
1922-33
1934-38
1939-43
1944-48
1949-53
periods
0
4.5
9.5
9.5
6.5
12
15
0.5
2
16
19.5
25.5
6
3
13.5
2
0
1.5
11
33.5
50
39.5
30
5.25
3
9
0
17.5
48
78.5
85
52
25.5
10
3
1.5
0
9
55
84
96.5
85.5
36
21.5
3.5
0.5
2
44
162
241.5
253
185.5
84.75
49
17.5
3
391.5
1042.25
0
0
58
88
183.75
32 1
6 14
Table 4. Causes of death among male asbestos workers compared with mortality experience of all men in
England and Wales.
No. of deaths
Cause of death
Lung cancer?
No.
observed
Test of
significance
of difference
between
observed
and
expected
(value of P)
Expected
on
England
and
Wales
rates
<0.000001
7.6
<0.001
4
4
2.3
4.7
<0.1
39
15.4
11
0.8
14
6
<0.000001
It certainly cannot be claimed that the records of the Personnel Office were necessarily
complete, but they were believed to be complete
and no deficiency on this score would account
for the total excess of deaths unless it were so
gross that more than half the defined population had been omitted. Moreover, the number
of deaths due to conditions unrelated to asbestosis was close to the estimated number and
this is unlikely to have happened unless the
population had been estimated approximately
correctly and the deaths from all causes fully
reported.
All the 11 deaths attributed to lung cancer
were confirmed by necropsy and histological
examination so that the excess number cannot
be attributed to incorrect diagnosis among the
group of asbestos workers. Some of the excess
may well be due to an underestimation of the
expected deaths since part of the increase in
mortality attributed to lung cancer over the past
30 years is certainly due to improvements in
diagnosis and in therapy (14). Even, however, if
it were postulated that the whole of the recorded increase between 1931 and 1951 was
spurious and that the real mortality from the
disease throughout was that ascribed to it in
1951, the expected number of deaths is increased to only 1.1 and the observed excess is
still grossly significant. For the actual number of
lung cancer cases to be so little in excess of the
expected as to be reasonably attributable to
chance, it would be necessary for the expected
cases to be 6.2, that is 5.6 times the number
estimated on 1951 rates. In other words, it
would be necessary to postulate that in 1951
(and throughout the previous 20 years) there
was 5.6 times as much cancer of the lung as was
recognized in 1931, which would mean that the
condition would have to have been present and
capable of detection in over 20 percent of all
men at death. Moreover, even if this were so, it
would still not account for the fact that all the
cases of lung cancer were found in association
with asbestosis.
It is, therefore, concluded that the fourth explanation is the most reasonable one and that
the asbestos workers who had worked for 20 or
more years in the scheduled areas suffered a
notably higher risk from lung cancer than the
rest of the population.
To test if the risk has altered since the 1931
regulations were introduced, it is not only necessary to make allowance for duration of em-
Doll
ployment before the end of 1932, but also to
allow for the mens ages and for the total durations of their employment in the scheduled
areas, since the men employed in the earlier
periods can also have been employed longer
and lived to be older. On the other hand, there
is no need to consider the changing incidence
of lung cancer in the total population of England and Wales since the non-industrial risk has
been shown to be small in comparison with the
industrial one. The data required or coniparing the risks among men employed for under
10 years, for 10 to 14 years, and for 15 years and
over in the pre-1933 conditions are shown in
Table 5. The ages shown are the ages at death of
the men who have died and the ages in
mid-1953 for the men who are still alive. The
expected numbers of men in each pre-1933 employment group found to have asbestosis or asbestosis and lung cancer are estimated by multiplying the numbers in each age, total employment, and pre-1933 employment subgroup by
the proportions of men with asbestosis or with
asbestosis and lung cancer in the same age and
total employment group for all lengths of
pre-1933 employment combined. For example,
three out of the nine men aged 50 to 54 years
who had been employed for 20 to 24 years in
the areas in which they might be exposed to
asbestos dust were found to have asbestosis and
lung cancer. Since three men had worked for
under 10 years in the pre-1933 conditions,
three had worked for 10 to 14 years, and three
had worked for 15 or more years, the expected
number of cases in each of the pre-1933 employment groups would have been the same,
i.e., 3 x 3/9, or 1. In fact, the numbers of cases
found were 0, 1, and 2. The total numbers
expected in each pre-1933 employment group
are obtained by adding the numbers calculated
for each of the age and total employment
groups within it. The results are as follows:
615
>$21td
10- 14
yrs. 15 yrs.
14
13
21.9
16
10.3
10.8
1
5.5
3
2.4
7
3.1
+
The expected numbers of lung cancer are small and the
probability that the differences could arise by chance has
consequently been somewhat, but not seriously, underestimated. If all men with more than 10 years pre-1933 employment are grouped together and Yales correction made for
small numbers, x2=5.82, n = 1 , P=O.O2.
35 +
30-34
25-29
20-24
Total
length of
employment
in
scheduled
area
(years)
1
0
0
0
0
0
556065-9
2
2
3
10
6
6
3
1
0
0
0
No.
of
men
455055606570-4
556065707580-4
50-
4045-
354045505560657075-9
Age at
June 30,
1953,
or at
death
(years)
2
2
1
0
0
0
of
men
1
0
4
1
1
-
1
1
0
3
1
1
No. of
men with
asbestosis
No.
No. of
men with
asbestosis
No. of
men with
cancer
of lung
10-14 Years
0-9 Years
1
0
0
0
No. of
men with
cancer
of lung
1
1
2
0
2
3
2
0
1
1
1
0
0
0
1
3
0
0
0
1
3
0
1
1
1
0
No.
of
men
2
1
0
1
3
0
0
1
0
1
0
0
1
0
0
1
-
1
3
0
1
1
1
No. of
men with
cancer
of lung
No. of
men with
asbestosis
15 + Years
1
1
2
2
1
1
1
3
12
6
15
No.
of
Men
0
0
0
0
0
2
1
0
1
0
3
2
6
4
1
0
1
0
1
1
2
No. of
men with
asbestosis
1
0
1
1
1
0
1
3
0
0
1
-
No. of
men with
cancer
of lung
Table 5. Numbers of men employed for different periods before 1933 and numbers known to have asbestosis and lung cancer in association
with asbestosis divided by total duration of employment in a scheduled area and by age.
All
lengths of
employment
in a
scheduled
area
(20 yrs.)
Total
length of
employment
in
scheduled
area
(years)
26
13
61
50-
354045-
All ages
No.
of
men
1
0
0
0
No. of
men with
cancer
of lung
556065707580-84
No. of
men with
asbestosis
1
1
2
8
11
2
of
men
No.
0-9 Years
14
1
3
5
2
1
-
No. of
men with
asbestosis
10-14 Years
I
0
0
1
No. of
men with
cancer
of lung
26
0
0
1
5
5
7
3
3
No.
of
men
1
7
19
11
12
7
3
0
1
0
Age at
June 30,
1953,
or at
death
(years)
Table 5. (Continued.)
1
2
1
3
3
5
1
2
1
0
16
43
3
2
1
113
No. of
men with
asbestosis
1
1
5
10
12
8
2
2
2
0
No.
of
men
2
8
22
24
28
16
1
1
1
0
1
No. of
men with
cancer
of lung
No. of
men with
asbestosis
15+ Years
11
4
1
1
2
0
1
No. of
men with
cancer
of lung
618
The cause of death, as determined at necropsy, is reported for 105 persons who had been
employed at one asbestos works. Lung cancer
was found in 18 instances, 15 times in association with asbestosis. All the subjects in whom
both conditions were found had started employment in the industry before 1923 and had
worked in the industry at least nine years before
the regulations for the control of dust had become effective.
One hundred and thirteen men who had
worked for at least 20 years in places where they
were liable to be exposed to asbestos dust were
followed up and the mortality among them
compared with that which would have been expected on the basis of the mortality experience
of the whole male population. Thirty-nine
deaths occurred in the group whereas 15.4 were
expected. The excess was entirely due to excess
deaths from lung cancer (11 against 0.8 expected) and from other respiratory and cardiovascular diseases (22 against 7.6 expected). All
the cases of lung cancer were confirmed histologically and all were associated with the presence of asbestosis.
From the data it can be concluded that lung
cancer was a specific industrial hazard of certain asbestos workers and that the average risk
among men employed for 20 or more years has
been of the order of 10 times that experienced
by the general population. The risk has become
progressively less as the duration of employment under the old dusty conditions has decreased.
I would like to offer my thanks to the management of the firm concerned for permission
to carry out this work and to the Medical Officer and members of the staff of the works
where the men were employed, who carried out
the greater part of the work on which this report is based.
References
( I ) Boemke, F. Med Mschr 7:77, 1953.
( 2 ) Hueper, W. C. Proceedings of the Seventh Saranac
Symposium. To be published.
( 3 ) Lynch, K.M. and W.A. Smith. A m J Cancer
24:56, 1935.
( 4 ) Merewether, E.R.A. Annual Report of the Chref
Inspector of Factorzes for the Year 1947. London,
H.M.S.O., 1949.
(5) Gloyne, S.R. Lancet 1:810, 1951.
( 6 ) Nordmann, M. and A. Sorge. 2 Krebsforsch
51:168, 1941.
( 7 ) Smith, W.E. Arch Industr Hyg 5:209, 1952.
(8) Vorwald, A.J. and J.W. Karr A m J Path 14:49,
1938.
6 19
620
For the past 13 years the National Heart Institute of the Public Health Service, acting
through the Massachusetts Department of Pub-
Dawber et al.
lic Health and the Framingham Health Department, has been conducting a long-term study of
coronary heart disease in the town of Frarningham, Massachusetts. A number of reports
of the progress of this study have been made
(1 -4). The authors have been engaged in the
conduct of this study from its inception, and
their experience gained in this undertaking
serves as the basis for a discussion of the methodology of longitudinal community studies of
which the Framingham Study will serve as a
prototype.
The Framingham Study was set up as a community-wide study of the general adult population of a town. It is a prospective study concerned with the incidence of coronary heart
disease and with the study of those factors, both
host and environmental, which may contribute
to its development.
It is safe to say that up to the present time this
study has been considered reasonably successful. It is customary to attribute success to
such factors as careful planning and hard work.
Such ingredients are admittedly necessary. It is
not customary to point out that success in
epidemiologic studies usually depends on finding significant associations between some factors and disease development, i.e., having the
good fortune to select those factors for study
which turn out to be related to the disease. If all
of the factors studied had turned out not to be
associated with coronary heart disease, this
study would have been considered a failure, and
it would probably have been abandoned. Nevertheless, such a study might have been almost
as valuable, and the methodology just as worthwhile, even if it only showed that large numbers
of factors did not appear to be related to disease
development, particularly if the factors studied
were widely accepted as being causally related.
The methodology of the Framingham Study
should not, therefor?, he evaluated in terms of
the successor failure of this project to turn
up positive findings, but rather on whether it
was, in fact, able to study a population adequately for a reasonable length of time in order
to determine whether or not the factors studied
could be demonstrated to be associated with
coronary heart disease development.
OBJECTIVES OF THE STUDY
The first aspect of any study to warrant attention is that of its objectives. As previously indi-
62 1
622
Dawber et al.
based on absolute estimates of disease incidence, may be inadequate for the demonstration of significant differences in relative disease
incidence in making comparisons between subgroups of the population. If the factors to be
investigated are strongly related to the disease,
this estimate of the population size will probably
be adequate. If, however, the factor bears only a
weak relationship to the disease, a much larger
population, and/or a longer period of ohservation will be needed. It cannot be hoped to study
a population large enough to provide answers
as to the effects of factors, regardless of the
potency of their effect, or their distribution in
the population. The major determining factor
must be the expected incidence of the disease
and the assumption that at least some of the
factors operating are strongly enough associated so that within-group comparisons will show
significant differences. Since it is usually not
possible to estimate reliably the potency of the
characteristic under investigation in relation to
development of disease before the study is undertaken, it is necessary, arbitrarily, to assign
some potency to the characteristic in estimating
the size of the population sample required to
demonstrate statistically significant differences.
In making this decision, one is motivated by
how large a difference would be biologically
meaningful or important.
AGE OF POPULATION
623
Having decided on the size, age, and sex distribution, and type of population desired, the
next task was that of finding the population.
Inquiry had been made through several state
health departments as to the possible commu-
624
less the cooperation of the population is essentially complete, it is difficult, if not impossible,
to consider it a random sample and representative of a larger population. If the characteristics to be studied influence the cooperation
of the population and in effect determine the
response rate, it is conceivable that a systematic
bias may be introduced, thus making random
sampling highly desirable for such studies. If,
however, the personal traits under consideration are apparently unrelated to population response, the use of volunteers would seem to be
adequate. If the objective in random sampling
is to insure that the sample be representative of
the entire community, steps must also be taken
to ascertain the degree to which the population
sample represents larger population segments
such as the county, the state, the country, etc.
At the inception of the Framingham Study it
was believed that the selection of a sample
should be made by random sampling methods,
to provide a study population representative of
the town. If a truly random sample could have
been obtained, it would then have been possible
to have made observations on the absolute prevalence and incidence of coronary heart disease
which could then be compared with estimates of
prevalence and incidence in other locations.
Since the major effort of the Framingham
Study involved within-group comparisons, it
would appear that a great deal of unnecessary
effort may have been expended in attempting
to get a random sample of the Framingham
population.
The possibility that studies of other population groups may obtain data on coronary heart
disease in a similar manner may yet justify the
effort expended in trying to get a representative sample. To date, however, this effort
seems to have been unwarranted.
A two-thirds random sample of the adult
population age 30-59 of the town of Framingham was selected, and an attempt made to
bring in those selected for examination and
continued follow-up.
A systematic sampling scheme was used, arranging persons by precincts of residence and
by size of family grouping; and two out of every
three family groupings were selected.
This method was chosen over a geographic
sampling, presumably on the basis that the sample obtained would be more representative of
the entire town. The major difficulty encoun-
Dawber et al.
625
Men
Women
6507
4469
3074
2024
3433
2445
740
312
428
4393
734
1975
307
2418
427
5127
2282
2845
Random sample
Respondents
Volunteers
626
Lost
Outside Info.
[IIIID Dead
0Examined
EXAMINATION NUMBER
Dawber et al.
nial clinic examinations, follow-up also has consisted of surveillance of the study population
with respect to hospital admissions, records of
interim illnesses obtained from subjects, physicians, relatives and friends, medical examiner
reports, and death certificates.
On the basis of all information available for
each subject, conclusions have been reached as
to the state of health with particular reference
to cardiovascular disease.
The need to consider changes in concepts of
disease, in laboratory procedure, and in observers is peculiar to longitudinal studies. It is necessary, therefore, to consider modifying criteria
and laboratory methodology when better information concerning these develops. In so doing,
it is necessary to consider the effect of such
changes on the continuity of information.
Changes in the subjects must be distinguished
from changes in method. Unless previous observations are proved invalid by more recent
knowledge, it is not desirable to make frequent
or drastic changes in study techniques.
Since the inception of the Framingham
Study, enzyme studies for the detection of myocardial infarction have been introduced. These
have been incorporated in the diagnostic criteria to supplement electrocardiographic information.
CRITERIA FOR DISEASE
627
628
147
CHOLESTEROL LEVEL
ANALYSIS OF DATA
The nature of the analytical problems in longitudinal prospective studies, such as Framingham, are such that machine processing of
data is imperative. Since the incidence of disease has been relatively low and a large population has been required, the difficulty in handling longitudinal data collected on such a large
population under continuous surveillance for
so many years has made the use of machines
mandatory. Huge quantities of data on personal
characteristics and environmental influences
collected continuously throughout a long period of observation must be tabulated and crossclassified. While this is a counting and sorting
operation, the large numbers of observations
make use of machines necessary not only for
tabulation, but also for storage and retrieving of
information.
Up to the present time, analysis of data has
been based on an assumed uniform period of
observation of all persons initially examined.
n
I I
186
aMorbidily ratio =
Observed cases
Expected cases
x 100
I-
Q
m
a
118
Dawber et al.
Serum cholesterol level, blood pressure, electrocardiographic abnormalities not considered
diagnostic of coronary heart disease, weight,
cigarette smoking habits, and vital capacity are
factors which, by simple analysis, have been
demonstrated to be significantly associated with
the development of coronary heart disease. Assessment of the independent contribution of
each of these often interrelated factors to the
development of coronary heart disease has required detailed, rather cumbersome, analysis
which would have been almost impossible if
done by hand.
Because of the many observations on such a
large population, even relatively simple operations of counting, sorting, classifying, and crossclassifying of information would have been tedious and prohibitively time-consuming if done
by hand. Calculations of means, standard
deviations, and correlation coefficients, while
perfectly feasible with the use of a desk type
calculating machine, are extremely time-consuming when done on such a large number of
observations.
As the data from the repeated biennial examinations have accumulated, the number of ways
in which these data can be examined seems
infinite. To date, analysis has been based on
observations made at the initial examination in
relation to the subsequent development of disease. Since these initial characteristics may have
been lost, acquired, or otherwise changed
through the years on repeated observation, it
will be desirable to know whether some parameter of these continuous observations, e.g.,
trends, means, variability, etc., will improve the
ability to predict disease or give further insight
into the mechanism of disease production. Accumulation of the additional persons in the
population at risk, by making use of all the
longitudinal observations, allows more detailed
analysis of the relationship of a characteristic
under observation to development of disease.
Data analysis, accounting for changes in characteristics or personal attributes with the passage
of time in relation to development of disease,
requires use of computers so that variable periods of follow-up from the onset, loss, or change
of some characteristic to development of disease can be taken into account.
From the beginning of the Framingham
Study, plans were made to analyze data by
means of machine tabulators and computers.
629
630
References
(I) Dawber, T.R., G.F. Meadors, and F.E. Moore,Jr.
Epidemiologic approaches to heart disease: the Framingham study. A m J Pub Health 41: 279-286, 1951.
(2) Dawber, T.R., F.E. Moore, and G.V. Mann. Coronary heart disease in the Framingham study. Am J
Pub Health 47: 4-24, 1957.
( 3 ) Dawber, T.R., W.B. Kannel, N. Revotskie, J.
Stokes 111, A. Kagan, and T. Gordon. Some factors
associated with the development of coronary heart
disease: six years follow-up experience in the Framingham study. Am J Pub Health 49: 1349-1356,
1959.
(4) Kannel, W.B., T.R. Dawber, A. Kagan, N. Revotskie, and J. Stokes 111. Factors of risk in the development of coronary heart disease-six-year follow-up
experience: the Framingham study. Ann Internal Med
55: 33-50, 1961.
(5) Doyle,J.T., A.E. Heslin, H.E. Hilleboe, P.F. Formel, and R. F. Korns. A prospective study of degenerative cardiovascular disease in Albany: report of three
years experience-I. Ischemic heart disease. Am J
Pub Health 47: 25-32, 1957.
(6) Drake, R.M., R.W. Buechley, and L. Breslow.
An epidemiological investigation of coronary heart
disease in the California health survey population. Am
J Pub Health 47: 43-57, 1957.
(7) Report of the Conference on Longitudinal Cardiovascular Studies at the Hotel Beaconsfield,
Brookline, Massachusetts,June 17-18, 1957.
(8) Epidemiology of cardiovascular diseases methodology-hypertension and atherosclerosis. Report
of Conference, Princeton, N.J., April 24-26, 1959.
Supp. to Am J Pub Health 50: 1-124, 1960.
DATA
To this request we had 40 637 replies sufficiently complete to be used-34 445 from men
and 6192 from women. From a 1 in 10 random
sample of the register that we subsequently
drew and analyzed we estimate that these figures represent answers from 69 percent of the
men and 60 percent of the women alive at the
63 1
632
answerers and 179 (84 percent) of the nonanswerers responded. Comparison of these two
groups shows 2 1 percent (answerers) and 6 percent (nonanswerers) as nonsmokers and 15 percent (answerers)and 28 percent (nonanswerers)
as moderate or heavy cigarette smokers (15 or
more daily). While these differences are large
and must contribute measurably to the continuing favorable mortality of the group that replied
in 1951, they are unlikely to account for it
wholly. As a further factor we suspect (but obviously cannot prove) that there may be some
more general association between mortality and
the tendency not to reply to such an inquirywhether the tendency be due to a deliberate
refusal (which is rare) or a mere neglect of these
things (which is frequent) In this respect it is
perhaps not too fanciful to note that one nonreplier died of smallpox and another of diabetic
coma.
Second Questionary
According to the doctors replies in 1951 we
allocated them to the appropriate nonsmoking
or smoking groups (subdivided by manner and
amount, continuing or stopped). Our previous
calculations of their subsequent death rates
were based upon the number of persons in each
of those groups at that time. We knew nothing
about any subsequent change of smoking habits, either in the dead or in the living, and the
further we moved from 1 November 1951 the
more likely it was that changes in habits had
occurred. In particular, it was probable that a
large number of doctors had given up cigarette
smoking. We therefore decided to approach
again all the survivors of those who had previously replied, and, taking advantage of this
opportunity, we added questions on: (1) the
past use of cigarettes by pipe and cigar smokers,
and (2)4nhaling (a factor that had since become
prominent in argument). Allowing for repeated
inquiries, we sought answers from men between
1 November 1957 and 31 October 1958, and
from women between 1 November 1960 and 3 1
October 1961.
Between the issue of our first and second
questionaries 2579 of the men and 320 of the
women had died; seven men had been struck
off the Register, and these we omitted. Of the
remaining 31 859 men and 5872 women, all but
2 percent replied to our second approach. Of
Deaths
Through the courtesy of the Registrars-Genera1 in the United Kingdom, a form with particulars of the cause of death has been provided to
us for every death since November 1951 identified as referring to a medically qualified person. We have also obtained lists of deaths notified to the General Medical Council and of
those recorded by the British Medical Association, and we have sought information from the
records of the fighting Services and from other
sources at home and abroad. A few deaths came
to light from our second questionary. As a result
of these several approaches we believe that very
few deaths can have been missed. In fact, combining their total numbers appropriately with
our 10 percent sample of non-repliers gives us a
mortality rate for all British doctors which is 93
percent of the corresponding mortality of all
males in England and Wales. This figure compares well with the standardized mortality ratio
of 89 percent that the Registrar-General gives
for doctors aged 20-64 and 65-74 years in his
Occupational Mortality Supplement for
1949-1953.
In total, in the ten years to which this paper
relates (1 November 1951 to 31 October 1961)
there have been 4597 deaths of men and 366
deaths of women. (Preliminary data for the eleventh year give another 472 deaths of men and
48 deaths of women, and these will also be used
for analyses where the numbers would otherwise be too few.) Except for deaths attributed to
cancer of the lung, we have accepted without
further inquiry the certified cause of death, and
(unless otherwise mentioned) have classified
these deaths according to the specified underlying causes. In only one case have we failed to
obtain any evidence of the cause of death.
633
634
Habits in
1951
Nonsmokers
Ex-smokers
Current smokers:
Pipe or cigar
Cigarette and
other
Cigarette only
Total
a
Nonsmokers
Exsmokers
5272
Pipe or
cigar
Cigarette
and other
Cigarette
only
Total
54
13
100
5439
4247
207
65
293
4812
707
2575
213
118
36132
520
629
2083
570
3802
2840
64 1
787
9274
13 542
5272
8314
4106b
3161
10 355
31 208
Table 2. Amount of smoking in 1951 and 1958, men, all forms of smoking.
Habits in 1958
Some tobacco (g/day)
Reduced by
Habits in
1951
No tobacco
15 or
more
10-14
Nonsmokers
5272
Ex-smokers
4247
Smokers of (glday):
1-4
59 1
5-9
808
10-14
854
15-19
629
20-24
677
508
25 +
27
131
392
60
190
290
336
13 586
550
876
Total
Increased by
5-9
1-4
No change
139
487
569
394
300
162
473
938
1439
1504
1991
1832
1804 2051
8177
115
436
427
453
373
15 or
more
Total
47
24
144 122
29
164
5439
4812
21
68
154 53
289 98
244 67
216 128
183 116
35
32
38
31
39
78
1569
2909
4159
3802
4385
4133
446
31 208
1-4
5-9
67
135
242
322
376
289
160
153
10-14
Table 3. Pure cigarette smokers in 1951 and their smoking habits in 1958, men.
Habits in 1958
Habits in 1951
(No. smoked
per day)
1-4
5-9
10-14
15-19
20-24
25 +
Total
Changed
to pipes
and/or cigars
1-4
5-9
328
48 1
550
438
557
397
31
72
120
101
163
130
301
171
125
59
54
31
275 1
617
741
Stopped
smoking
10-14
15-19 20-24
25+
Total
Total
65
495
280
156
87
43
18
101
920
240
268
90
8
11
25
19
188
88
1056 191
213 1753
149 222
3
5
23
79
316
2027
406
816
1624
1781
2691
2562
765
1369
2294
2320
341 1
3089
1126
1637
1642 2281
2453
9880
13 24@
This total is less than the total of men smoking only cigarettes shown in Table 1 since it excludes 294 men who were known
635
636
Mortality from cancer of the lung is examined in Tables 4 to 11, where we have included
the five deaths for which cancer of the lung was
mentioned on the death certificate as contributory with the 207 in which it was given as the
underlying cause.6 Many of the rates, however,
are based on small numbers, and although they
may contribute usefully to the general picture
they cannot be relied on individually. In spite of
this there is a steadily rising death rate with
increasing consumption of cigarettes at every
age above 45 years (Table 4). For all ages a more
detailed analysis in Figure 1 indicates a linear
relationship, the death rate rising step by step
from the 0.07 per 1000 in nonsmokers to 3.15
per 1000 per annum in men smoking 35 or
more cigarettes daily.
In examining the effect of the method of
smoking (Table 5), we are dependent upon the
Ten certified deaths were excluded because additional
evidence suggested that the diagnosis was incorrect (see p.
633).
Table 4. Age and number of cigarettes smoked death rates per 1000 per annum from
cancer of the lung (numbers of deaths in parentheses).
Cigarettes per day
Age
Nonsmokers
35-44
45-54
55-64
65-74
75 and over
All ages
a
1-14
0.05 (1)
0.00
0.00
0.00
1.11 (2)
0.07
0.31
0.48
2.69
2.68
(1)
(3)
(3)
(9)
(6)
0.07 (3)
0.57 (22)
15-24
0.00
0.62
2.31
5.16
7.27
25 +
(9)
(20)
(17)
(8)
0.11 (1)
0.75 (8)
3.88 (26)
6.48 (14)
16.33 (8)
1.39 (54)
2.27 (57)
Number reported at 1 November 1951 for men continuing to smoke cigarettes only at that time.
Method of smoking
0.96 (143)
0.53 (42)
0.43 (24)
0.42 (12)
0.45 (6)
0.96 (3)
At 1 November 1951.
637
mixed, pipe and/or cigar alone) the rate of continuing smokers is higher than that of those
who had stopped, and the rate in ex-cigarette
smokers (0.24) is notably less than that in current pipe smokers (0.47).
The difference between the ex-smokers and
the continuing smokers is most marked for
those who smoked cigarettes, where the rate for
those who had given up is only 19 percent of
the rate for those who continued. This difference, moreover, may well underestimate the
true difference, since in this analysis we take no
account of the changes in habits in the years
following 1951. (See p. 648, where allowance
has been made for some of the changes in habits which took place and where comparison is
made between men who have given up smoking
for different lengths of time.)
In Table 7 we consider the nature of the
evidence available to the doctors who certified
cancer of the lung as the cause of death (excluding the one case for which this information was
refused to us). In more than half of the deaths
(56 percent) there was histological, cytological,
or necropsy evidence together with X-ray or
bronchoscopic confirmation of the site of the
primary growth. In another 38 percent an X-
Continuing at 1111/51
Stopped before 1/11/5 1
Cigarette
smokers
Mixed
smokers
Pipe and/or
cigar smokers
1.25 (133)
0.24 (10)
0.59 (36)
0.48 (6)
0.47 (21)
0.23 (3)
Tablu 7. Standard of diagnosis: standardized death rates from cancer of the lung.
Death rate per 1000
Standard
of
diagnosisd
(No. of deaths)
Cigarette smokersb
Nonsmokers
All
smokers
Continuing
Given
UP
1-14
daily
15-24
daily
25 +
daily
Grade 1 (118)
2 (81)
3 (12)
0.01
0.03
0.03
0.41
0.28
0.04
0.68
0.52
0.04
0.17
0.05
0.03
0.37
0.15
0.02
0.76
0.60
0.03
1.02
1.21
0.05
0.07
0.73
1.24
0.24
0.54
1.39
2.27
a Grade 1= Necropsy evidence or histological or cytological together with evidence of primary from x-ray picture or
bronchoscopy. Grad 2 =Clinical evidence together with evidence of primary from x-ray picture or bronchoscopy. Grade
3 =Evidence from hfsrory and physical examination alone.
One case for which information was refused has been excluded (a man aged 74 smoking 14 cigarettes a day).
At 1 November 1951.
638
ray picture or bronchoscopy supported theclinical evidence. In both these groups the various
associations of mortality with smoking are quite
distinct. On the other hand, it is of interest that
for the 12 cases in which the diagnosis rested on
history and physical examination there is no
clear evidence of association. In other words,
these figures show that doctors are not diagnosing cancer of the lung in their colleagues without adequate evidence, and that in the few less
definite cases they are not making such a diagnosis merely because the sick person was a
heavy cigarette smoker.
Similarly Table 8 shows no marked associations with smoking in the 13 cases of adenocarcinoma, whereas they are distinct with both the
squamous and the oat and anaplastic growths.
Taking the death rate of the continuing cigarette smokers in each histological group as standard (loo), we have ratios for those smoking 1
to 14, 15 to 24, and 25 or more daily, of 68, 104,
and 139 percent in the squamous group, and
44,91, and 169 percent with the oat celled and
anaplastic group.
The rise in lung cancer mortality with increasing number of cigarettes smoked occurred
in all types of area, and the rates in Table 9
provide no reason to suppose that the association was any closer in big towns than in small
towns or in the countryside. It is, however, possible that the rural mortality is affected by the
retirement there of men who had passed their
active lives in towns. We therefore consider in
Table 10 doctors under 65 years of age. The
number of deaths in the different areas is then
small, but they do suggest that there is a lower
lung cancer mortality in the rural, and possibly
the small town, areas which is not due to differences in amounts smoked.
Table 11 shows the mortality in the principal
smoking categories for two periods of time-1
November 1951 to 31 October 1956 and 1 November 1956 to 31 October 1961. The first of
these quinquennia followed immediately after
the questionaries were sent out. Although the
majority were returned within a few weeks,
there were some which were not returned for
several months, so that the first year of observation was biased by the inclusion of some months
of observation for which there could not be any
corresponding mortality. A much more important type of bias is the differential selection of
healthy men among the repliers, and we have
shown that the total mortality rate was lower
during the early years of follow-up than in any
subsequent period. Clearly, therefore, we would
expect the mortality from lung cancer to be
higher in the second quinquennium and that
Table 8. Histological type: standardized death rates from cancer of the lung.
Death rate per 1000
Histological
Y Pe
(No. of
deaths)
Cigarette smokers
Nonsmokers
All
smokers
Continuing
Given
UP
1-14
daily
15-24
daily
25 +
daily
0.00
0.19
0.32
0.09
0.22
0.33
0.45
0.01
0. I4
0.22
0.05
0.10
0.20
0.38
0.00
0.05
0.07
0.03
0.03
0.12
0.07
Squamous (55)
Oat and anaplastic (40)
Adenocarcinoma
(13)
Table 9. Place of residence: standardized death rates from cancer of the lung.
Death rate per 1000
Place
of
residence
Conurbation
Large towns
Small
Rural areas
a
Nonsmokers
1-14
15-24
25 +
0.03
0.00
0.11
0.12
0.48
0.32
0.87
0.52
1.31
1.88
1.06
1.15
1.90
4.43
2.20
1.17
No.
of
deaths
49
34
32
18
639
Table 10. Place of residence: standardized death ratesa at ages 25-64 years.
Death rate per 1000
Place of
residence
Conurbation
Large towns
Smalil
Rural area
a
Nonsmokers
Cigarette
smokersb
No. of
deaths
0.04
0.00
0.00
0.00
0.62
0.65
0.52
0.40
31
17
16
8
Table 11, Period after start of inquiry: standardized death rates from cancer of the lung.
Death rate per 1000
Period
(No. of
deaths)
1951-6 (102)
1956-1 (110)
All
men
Nonsmokers
All
cigarette
smokers
0.69
0.64
0.05
0.08
0.98
0.95
Cigarette
smokers
Current cigarette
smokers
Current
Ex-
1-14/d
15-24/d
25+/d
1.22
1.28
0.41
0.09
0.64
0.50
1.20
1.53
2.25
2.32
640
Site
(1)
All
smokers
(5)
Cigarette
smokers
(6)
Mixed
smokers
(3)
Nonsmokers
(4)
(7)
(8)
19 (2)
0.06
0.00
0.06
0.05
0.10
0.04
16 (5)
29 (1)
0.05
0.09
0.00
0.04
0.06
0.10
0.05
0.06
0.03
0.19
0.10
0.08
64 (8)
0.20
0.04
0.22
0.16
0.32
0.22
No. of
deathsa
All
men
(2)
Mouth,
pharynx,
or nose
Larynx or
trachea . .
Esophagus
Total
Pipe or
cigar
smokers
a The numbers in parentheses are of deaths (included in the total figure) in which cancer of the specified site was certified as
being associated with the death but not its direct or underlying cause.
Currently smoking or past smokers in 1951.
Table 13. Standardized death rates from cancer of the upper respiratory and
digestive tracts in relation to the amount smoked.
Death rate per 1000
Amount of tobacco smoked daily (g.)
Site
Mouth, pharynx, or
nose
Larynx or trachea
Esophagus
Total
1-14
15-24
25+
All
amounts
Given up
smoking
0.04
0.02
0.08
0.01
0.02
0.14
0.21
0.15
0.20
0.07
0.06
0.12
0.06
0.05
0.02
0.13
0.17
0.56
0.24
0.13
and 17. For chronic bronchitis there is a closeand statistically highly significant-association
with smoking. While this was most marked
when chronic bronchitis was described as the
underlying cause of death, it was also clearly
present when chronic bronchitis served as a
contributory cause in cardiovascular mortality.
For pulmonary tuberculosis the evidence is
much less clear. The mortality was higher in
cigarette smokers than in nonsmokers, in continuing cigarette smokers compared with those
who had stopped, and in heavy cigarette
smokers compared with light. The deaths were,
however, few, and none of these differences was
statistically significant. Moreover, the few additional deaths for which pulmonary tuberculosis
was described as a contributory cause tend to
diminish the relationship rather than reinforce
it. In other words, the data suggest that death
from pulmonary tuberculosis may be associated
with smoking, while the disease itself is not.
64 1
Site
No. of
deaths
All
men
Nonsmokers
All
smokers
Cigarette
smokers
Mixed
smokers
Pipe or
cigar
smokers
381
84
97
35
69
194
0.1 1
0.26
0.30
0.11
0.21
0.61
0.17
0.27
0.31
0.05
0.33
0.64
0.11
0.25
0.30
0.11
0.20
0.61
0.13
0.30
0.26
0.12
0.18
0.68
0.12
0.14
0.32
0.12
0.16
0.52
0.07
0.30
0.39
0.08
0.26
0.51
27
0.08
0.09
0.08
0.05
0.09
0.15
542
1.67
1.87
1.65
1.72
1.45
1.77
Bladder
Stomach
Bowel
Rectum
Prostate
Other sites
Primary site
unknown
Total
a Two deaths from other causes in which cancer of the bladder was mentioned as contributory are included in this category,
but are excluded from the total cancer deaths.
Table 15. Standardized death rates from cancer of other sites in relation to the
amount smoked.
Death rate per 1000
Amount of tobacco smoked daily (g.p
Site
1-14
15-24
25+
All
amounts
Given u p
smoking
Bladder
Stomach
Bowel
Rectum
Prostate
Other sites
Primary site unknown
0.10
0.28
0.37
0.10
0.16
0.57
0.10
0.11
0.28
0.22
0.06
0.19
0.59
0.07
0.13
0.26
0.44
0.22
0.12
0.83
0.06
0.12
0.28
0.32
0.1 1
0.17
0.63
0.09
0.08
0.18
0.23
0.12
0.28
0.54
0.05
1.67
1.52
2.03b
1.71
I .49
Total
a
Chronic bronchitis
Chronic bronchitis
as associated
causea
Pulmonary tubercu 1osis
Pulmonary tuberculosis as associated cause
Other respiratory
diseases
a
No. of
deaths
All
men
Nonsmokers
All
smokers
Cigarette
smokers
Mixed
smokers
Pipe or
cigar
"
smokers
111
0.34
0.05
0.37
0.51
0.33
0.15
55
0.17
0.03
0.18
0.20
0.21
0.12
42
0.13
0.06
0.14
0.15
0.1 1
0.11
14
0.04
0.05
0.04
0.05
0.02
0.04
181
0.54
0.63
0.53
0.55
0.49
0.47
642
Chronic bronchitis
Chronic bronchitis as
associated causea
Pulmonary tuberculosis
Pulmonary tuberculosis
as associated cause
Other respiratory diseases
Given
1-14
15-24
25+
All
amounts
smoking
cigarettes
0.34
0.64
1.06
0.58
0.38
0.15
0.10
0.20
0.16
0.30
0.26
0.2 1
0.17
0.16
0.12
0.05
0.09
0.02
0.06
0.11
0.41
0.68
0.40
0.51
0.61
Cardiovascular Disease
In Tables 18 and 19 we set out the data relating to mortality from cardiovascular diseasesdue to nephritis. With the deaths attributed to
cardiovascular accidents or to coronary disease
we have ,separately considered those in which
reference was made on the death certificate to
the presence of hypertension, for these deaths
might be related to the causes of hypertension
more closely than to any other factor.
For the 138 deaths in which hypertension was
given as the primary cause there is no association with smoking habits-neither with method
(Table 18) nor with the amount of cigarette
smoking (Table 19). This is equally true for the
143 cerebrovascular accidents and for the 89
coronary disease deaths in which hypertension
was also mentioned. Adding the three hyper-
tensive groups together gives, with one exception, remarkably similar rates in all the smoking
categories, varying only between 1.26 per 1000
in nonsmokers and 1.10 per 1000 in men who
had given up smoking (the final lines of Tables
18 and 19). The exception lies in the rather low
death rate of 0.8 1 per 1000 in pipe and/or cigar
smokers, a figure which is based upon only 53
deaths, and has 95 percent confidence limits of
0.69 and 0.93.
We can also conclude from these tables that
there is no association with smoking habits in
the 135 deaths from other heart diseases
(rheumatic heart 35, other valvular disease 22,
other diseases 78) nor in the 135 from other
cardiovascular diseases (general arteriosclerosis 64, dissecting and atherosclerotic aneurysm
of the aorta 47, and other vascular diseases 24).
With the larger group of cerebrovascular accidents without reference to hypertension, there
also seems to be no association. Though the
death rate of all smokers at 1.41 per 1000 is a
trifle above that of the nonsmokers (1.24), this
small excess is not related specifically to one
method of smoking (Table 18), and there is no
clear gradient with number of cigarettes
smoked nor fall in the death rate on giving up
smoking (Table 19).
With the 337 deaths attributed to myocardial
degeneration and the 43 deaths attributed to
nephritis there is some suggestion of an association, but the evidence is slight. With myocardial
degeneration there is a substantial-and statistically significant (P<O.Ol)-difference between smokers (1.02) and non-smokers (0.59),
but there is no difference between the different
643
No of
deaths
All
men
Nonsmokers
All
smokers
Cigarette
smokers
Mixed
smokers
Pipe or
cigar
smokers
Cerebrovascular
accidents
(a).
(b)b
All
462
143
605
1.39
0.45
1.84
1.24
0.51
1.76
1.41
0.44
1.85
1.48
0.50
1.98
1.27
0.44
1.71
1.44
0.23
1.67
Coronary
disease
(a).
(b)b
All
1287
89
1376
3.99
0.28
4.26
3.31
0.30
3.61
4.08
0.28
4.36
4.39
0.26
4.65
3.87
0.37
4.25
3.18
0.21
3.39
337
138
135
0.97
0.42
0.41
0.59
0.45
0.41
1.02
0.42
0.4 1
1.01
0.45
0.42
0.98
0.36
0.43
1.oo
0.37
0.33
135
43
0.4 1
0.13
0.41
0.09
0.4 1
0.14
0.46
0.14
0.38
0.13
0.35
0.12
2769
370
8.44
1.15
7.32
1.26
8.61
1.14
9.11
1.22
8.24
1.17
7.23
0.81
Myocardial degeneration
Hypertension
Other heart diseases
Other cardiovascular diseases
Nephritis
Total
Total hypertensive
Cause of
death
Given
"P.
smoking
cigarettes
1-14
15-24
25+
All
amounts
Cerebrov ascu1ar
accidents
(a).
(b)b
All
1.46
0.46
1.93
1.43
0.48
1.91
1.69
0.54
2.23
1.49
0.51
2.00
1.42
0.51
1.93
Coronary
disease
(a)a
(b)b
All
4.35
0.29
4.65
4.28
0.29
4.57
4.97
0.18
5.16
4.57
0.28
4.86
3.73
0.19
3.92
Myocardial degeneration
H ypqrtension
Othet heart diseases
Othet cardiovascular
diseases
NepHritis
1.10
0.44
0.32
0.94
0.45
0.59
0.97
0.45
0.21
1.07
0.46
0.41
0.87
0.40
0.41
0.41
0.16
0.45
0.14
0.62
0.29
0.54
0.17
0.25
0.08
Total
Total, hypertensive
9.01
1.20
9.05
1.22
9.93
1.17
9.51
1.25
7.86
1.10
644
Age
in
years
No.
of
deaths
Nonsmokers
Current
cigarette smokers
smoking daily
All
cigarette
smokers,
ex and
current
1-14
15-24
25 +
35-44
45-54
55-64
65-74
75-84
85 +
38
149
319
389
314
78
0.1 1
1.12
4.90
10.83
21.20
32.35
0.41
1.66
6.8 1
16.44
21.22
33.06
0.49
3.40
7.03
13.04
15.04
58.54
1.50
2.73
8.81
17.59
17.30
-b
0.61
2.40
7.20
14.69
19.18
35.93
All ages
1287
3.31
4.36
4.28
4.97
4.39
a Includes all deaths attributed to coronary thrombosis, except those for which hypertension was
mentioned as a contributory cause (89).
Very small total number of man-years at risk (2?/z)-no deaths due to coronary thrombosis.
myocardial degeneration or other heart disease-and that these constitute another cardiovascular group which is also associated with
smoking. On the evidence available, however,
this group would not appear to be large enough
to have materially affected the results. Possibly
most such deaths among doctors are attributed
directly to chronic bronchitis.
Other Diseases
645
Peotic ulcer
Ciirhosis of liver
and alcoholism
Other digestive
diseases
Genito-urinary
diseasesb
Indefinite causes
Violence
All remaining
causes of death
No. of
deaths
All
men
Nonsmokers
All
smokers
Cigarette
smokers
Mixed
smokers
Pipe or
cigar
smokers
541
0.17
0.03
0.18
0.21
0.16
0.12
33
0.10
0.00
0.11
0.12
0.11
0.05
87
0.26
0.07
0.28
0.32
0.20
0.25
82
50
248
0.15
0.24
0.77
0.33
0.17
0.94
0.24
0.14
0.75
0.27
0.13
0.79
0.21
0.12
0.68
0.22
0.20
0.64
150
0.46
0.50
0.47
0.49
0.48
0.45
a Including 15 deaths in which peptic ulcer was certified as being associated with the death but not its direct or underlying
cause.
Excluding nephritis.
mortality rates for smokers (0.18) and nonsmokers (0.03) is sufficiently great to be unlikely to be due to chance (P = 0.05).
With alcoholism (6 deaths) and cirrhosis of
the liver (27 deaths) the association is strong.
No deaths from these causes occurred among
nonsmokers, and, like cancers of the mouth,
pharynx, and larynx, the mortality fell almost
wholly on the heaviest smokers.
The evidence relating to other digestive diseases is inconcludive. This is a heterogeneous
group, the major tomponents of which were 11
deaths from appendicitis, 12 from hernia, 18
from obstruction, 9 from diverticulitis, and 17
from gallstones or cholecystitis. Several of these
Table 22. Standardized death rates from other diseases in relation to the
number of cigarettes smoked.
Death rate per 1000
No. of cigarettes smoked daily
Cause of
death!
Given
UP
1-14
15-24
25+
All
amounts
smoking
cigarettes
Peptic ulcer
Cirrhosis of liver and
0.07
0.31
0.22
0.21
0.16
0.05
0.26
0.08
0.33
0.43
0.36
0.15
0.32
0.03
0.32
0.28
0.21
0.85
0.29
0.12
0.57
0.29
0.12
1.15
0.29
0.16
0.83
0.21
0.08
0.65
0.53
0.34
0.53
0.45
0.55
alcdholism
646
Table 23. Standardized death rates from causes related to smoking and from causes unrelated
to smoking.
Death rate per 1000
Cause of death
Related causes:
Cancer of lung
Other upper respiratory and
digestive cancers
Chronic bronchitis
Coronary disease
without hypertension
Peptic ulcer
(including
hematemesis)
Cirrhosis of liver
and alcoholism
Pulmonary
tuberculosis
No. of
deaths
All
men
Non-
smokers
All
smokers
Cigarette
smokers
Mixed
smokers
Pipe or
cigar
smokers
207
0.65
0.07
0.71
0.93
0.52
0.43
56
111
0.17
0.34
0.04
0.05
0.20
0.37
0.15
0.51
0.28
0.33
0.16
0.15
1287
3.99
3.31
4.08
4.39
3.87
3.18
39
0.12
0.00
0.13
0.13
0.12
0.10
33
0.10
0.00
0.11
0.12
0.11
0.05
42
0.13
0.06
0.14
0.15
0.11
0.11
1775
5.49
3.53
5.74
6.39
5.33
4.17
542
1.67
1.87
1.65
1.72
1.45
1.77
181
0.54
0.63
0.53
0.55
0.49
0.47
462
1.39
1.24
1.41
1.48
1.27
1.44
337
370
0.97
1.15
0.59
1.26
1.02
1.14
1.01
1.22
0.98
1.17
1.oo
0.81
135
43
0.41
0.13
0.41
0.09
0.41
0.14
0.42
0.14
0.43
0.13
0.33
0.12
135
0.41
0.41
0.41
0.46
0.38
0.35
87
248
282
0.26
0.77
0.85
0.07
0.94
1.00
0.28
0.75
0.85
0.32
0.79
0.89
0.20
0.68
0.81
0.25
0.64
0.87
Total
2822
8.55
8.53
8.58
8.99
8.00
8.06
All causes
4597
14.05
12.06
14.32
15.38
13.34
12.23
Total
Unrelated causes:
Other cancers
Other respiratory disease
Cerebrovascular
accidents without hypertension
Myocardial
degeneration
All hypertension
Other heart
disease
Nephritis
Other cardiovascular disease
Other digestive
disease
Violence
Remainder
647
Table 24. Standardized death rates from causes related to smoking and from
cau$es unrelated to smoking in relation to the number of cigarettes smoked.
Death rate per 1000
No. of cigarettes smoked daily
Cause of
death
Given
15-24
25 or
more
All
amounts
0.57
1.29
2.23
1.20
0.24
0.04
0.34
0.18
0.64
0.43
1.06
0.20
0.58
0.05
0.38
4.35
4.28
4.97
4.57
3.73
0.02
0.18
0.19
0.13
0.12
0.05
0.08
0.43
0.15
0.03
0.10
0.16
0.26
0.17
0.12
5.48
6.81
9.56
7.01
4.67
1.77
1.56
2.31
1.82
1.47
0.41
0.69
0.40
0.51
0.61
1.46
1.43
1.69
1.49
1.42
1.10
1.20
0.32
0.16
0.94
1.22
0.59
0.14
0.97
1.17
0.21
0.29
1.07
1.25
0.41
0.17
0.87
1.10
0.41
0.08
0.41
0.45
0.62
0.54
0.25
0.26
0.85
1.02
0.33
0.57
0.75
0.36
1.15
0.94
0.32
0.83
0.90
0.32
0.65
0.84
Tatal
8.96
8.66
10.11
9.31
8.02
All causes
14.44
15.47
19.67
16.32
12.68
Related causes:
Cancer of lung
Other upper respiratory and digestive
cancers
Chronic bronchitis
Coronary disease
without hypertension
Peptic ulcer (including hematemesis)
Cirrhosis of liver
and alcoholism
Pulmonary tuberculosis
Total
Unrelated causes:
Other cancers
Other respiratory
disease
Cerebrovascular accidents without
hypertension
Myocardial degeneration
All hypertension
Other heart disease
Nephritis
Other cardiovascular
disease
Other digestive
disease
Violence
Remainder
1-14
smoking
cigarettes
648
difference (0.3<P<0.4) nor that between nonsmokers and cigarette smokers (P= 0.09) is statistically significant.
MORTALITY OF MEN WHO HAVE GIVEN UP
SMOKING
Cigarette Smokers
Table 25. Mortality among cigarette smokers at different times after stopping smoking (numbers of
deaths in parentheses).
Death rate per 1000
Cause of death
Lung cancer
Chronic bronchitis
Coronary disease without hypertension
Other related causes
Unrelated causes
All causes
Continuing
cigarette
smokersa
5-9
20 or more
Nonsmokers
1.28 (124)b
0.58 (48)
0.67 (5)
0.71 (5)
0.49 (7)c
0.81 (11)
0.18 (3)
0.06 (1)
0.19 (2)
0.30 (4)
0.07 (3)
0.05 (2)
4.72 (464)
0.65 (69)c
9.43 (865)
3.52 (28)
0.50 (4)
6.26 (49)
4.17 (61)
0.40 (6)
8.49 (120)
3.87 (59)
0.33 (5)
9.27 (136)
3.74 (40)
0.22 (2)
8.80 (105)
3.34 (113)
0.10 (3)
8.52 (315)
16.62 (1566)
10-19
11.62 (91) 14.25 (204) 13.60 (204) 13.38 (153) 12.09 (436)
Calculate by method B (see text) and therefore not in all cases exactly the same as the rates in previous Tables.
Including three deaths, also included under their primary cause, in which the specified disease was a contributory cause.
Including one death, also included under its primary cause, in which the specified disease was a contributory cause.
a
649
Figure 2. Death raw from lung cancer, standardized for disease will remain higher in the ex-smokers
age and amount smoked, among men continuing to compared with the smokers until, in the long
smoke cigarettes and men who had given up smoking
for different periods (men who had regularly smoked run, the change in habits can exert a beneficial
pipes or cigars as well as cigarettes excluded). The cor- effect on the disease in the survivors.
responding rate fgr nonsmokers was 0.07 per 1000.
For the unrelated diseases selective factors
1.5
0,
ea
1.0
bLu
5
k
n
J
4
3
Z
Z
0.5
6
1
I
10
15
20
25
30
650
Table 26. Mortality among cigarette smokers: effect of grouping men who had stopped smoking for less
than five years with continuing smokers.
Death rate per 1000
Cause
of
death
Lung cancer
Chronic bronchitis
Coronary disease without hypertension
Other related
causes
Unrelated causes
All causes
a
Ex-
smokers
of 5 years
or more
(C)
All other
smokers
(D)
C
as
percenttage of
(D)
All exsmokers
(A)
Continuing
smokers
(B)
0.35b
0.42
1.28.
0.58
27
72
0.2gb
0.37
1.24a
0.59
23
63
3.88
4.72
82
3.95
4.63
85
0.36
8.44
0.65b
9.44
55
89
0.33
8.86
0.64b
9.18
52
97
13.42
16.62
81
13.77
16.23
85
Table 27. Average age at starting to smoke among cigarette smokers divided by
continuity and amount smoked (standardized for age at inquiry).
~
~~
1-14
15-24
25 or more
Smoking at 1/11/51
Stopped less than 10 years
10 years or more
20.3
20.7
19.6
19.7
19.9
19.3
19.1
19.8
19.3
a The ages are somewhat higher than would be anticipated from present experience: they
refer, however, to a section of the population composed of university graduates, some of whom
completed their education before the first world war. The average is heavily weighted by the few
doctors who began to smoke after 25 years of age; in each group the commonest age at starting
was one to two years less than the average.
65 1
Table 28. Percentage of inhalers among cigarette smokers, divided by continuity and amount smoked (standardized for age at inquiry).
Percentage of inhalers when daily number
of cigarettes last smoked was:
Group
1-14
15-24
25 or more
Smoking at 1/11/51
Stopped less than 10 years
10 years or more
68
59
34
80
77
72
81
80
73
Table 29. Mortality from various causes in first and second quinquennia of observation:
continuing and ex-cigarette smokers
Death rate per 1000
Continuing smokers
Cause of
death
Lung cancera
Chronic bronchitis
Coronary disease without
hypertension
Other related causes
Unrelated causes
All causes
Ex-smokers
1st
5 years
2nd
5 years
1st
5 years
2nd
5 years
1.22
0.55
1.28
0.61
0.41
0.45
0.09
0.32
4.50
0.67
9.14
4.64
0.64
9.37
3.19
0.36
7.51
4.17
0.30
8.63
16.02
16.48
11.93
13.51
a Including four deaths among continuing cigarette smokers in which lung cancer was certified as a
contributory cause (method A).
652
Table 30. Inhaling habits by age, method of smoking, and amount smoked; (Male doctors aged 25 years
and over).
Percentage of inhalers among
Continuing cigarette
smokers smoking daily
Age
(years)
25 or
more
All
continuing
cigarette
smokers
cigarette
smokers
Ex-
Continuing
smokers,
cigarette
and
other
Continuing
Pipe
andor
cigara
smokers
82
76
62
57
49
74
60
47
36
30
12
10
7
5
4
1-14
15-24
25-34
35-44
45-54
55-64
65-74
75 or
more
85
78
65
54
49
93
89
78
70
66
95
81
73
59
90
85
75
66
58
38
47
46
41
41
26
All ages
67
81
81
76
63
47
89
Figures for the few cigar smokers are almost identical with those for pipe smokers.
Table 31. Mortality among continuing cigarette smokers according to inhaling habits,
compared with non-smokers (numbers of deaths in parentheses).
Death rate Der 1000 men Der veal
Continuing cigarette smokers
Cause of death
Lung cancer
Chronic bronchitis
Coronary disease without
hypertension
Other related causes
Unrelated causes
All causes
Inhalers
Non inhalers
Nonsmokers
1.88 (33)a
0.72 (12)
1.10 (11)
0.51 (7)
0.03 (1)
0.07 (1)
5.22 (113)
0.70 (17)
9.27 (171)
5.09 (60)
0.48 (4)
8.65 (99)
3.29 (50)
0.12 (2)
7.32 ( 1 12)
17.73 (345)
15.83 (181)
10.82 (166)
a Including one death, also included under its primary cause, in which lung cancer was certified as a
contributory cause.
1-14
15-24
25 or more
a
Noninhalers
1.59 (1 1)
1.95 (13)
2.16 (9p
0.00
0.86 (4)
2.98 (7)
The women doctors available for study number less than a fifth of the male doctors (18
percent) and fewer of them are in the age
groups of high mortality rate (40 percent were
under 35 years of age against 29 percent of
men, and 5 percent were 65 years or over
Other upper respiratory and upper digestive diseases,
pulmonary tuberculosis, peptic ulcer, and cirrhosis of the
liver and alcoholism.
653
654
Table 33. Standardized death rates from various causes among women doctors in relation to
smoking habits.
Death rate per 1000 women per year
Cause of
death
No.
deaths
Lung cancer
7
Other respiratory and upper
digestive cancer
2
Chronic bronchitis
2
Pulmonary tuberculosis
1
Cirrhosis of liver
4
and alcoholism
Coronary disease without
56
hypertension
Other causes (unrelated)
342
All causes
414
All
NonExwomen smokers Smokers smokers
Continuing
smokers
1-14
15-24 25 or more
0.08
0.03
0.13
0.08
0.15
0.41
0.22
0.04
0.04
0.03
0.04
0.07
0.08
0.11
0.11
0.27
0.27
0.22
0.05
0.06
0.04
0.07
0.16
0.03
0.10
1.14
6.72
1.25
6.60
0.99
6.60
1.26
5.94
0.84
6.93
0.41
6.30
1.25
8.60
2.62
5.70
8.10
7.95
7.94
7.44
8.17
6.76
10.89
8.76
a Only one woman said she smoked a pipe as well as cigarettes and no women smoked only pipes or cigars. For the purpose of
this analysis the small additional amount of pipe tobacco has been ignored.
each level of smoking the proportion of inhalers was approximately twice as great in men as
in women.
We have not sufficient data to take these differences into account, but it appears (from
Table 32) that the difference in inhaling will
lower the female mortality rate in relation to the
male, and it is reasonable to suppose that the
later average age at starting to smoke would
have a similar effect. These differences, too,
may affect not only the relative lung cancer
mortality rate in the two sexes, but also, to a
Table 34. Inhaling habits by age and amount smoked; women doctors aged 25
years and over.
Percentage of inhalers among continuing
smokers
Number of cigarettes smoked daily
~
Age (years)
1-14
15-24
25 or more
25-34
35-44
45-54
55-64
65-74
75 or more
73
63
I!?
26
17
17
91
71
59
34
26
18
54
47
All ages
44
55
58
22
50
29
65
36
64
25a
Age groups combined because number of women in one age group was less than 10.
Standardized for age.
GENERAL DISCUSSION
655
656
657
1-4
5-9
10-14
1
20-24
15-19 I
Control
patients
50
Cancer of
lung patients
Control
patients
50
29
59
46
61
46
60
72
81
72
62
82
71
658
Table 36. Death rate from related and unrelated cancers by smoking habits and age (numbers of deaths
in parentheses).
Death rate per 1000
Cancer of lung and upper
respiratory and digestive tracts
Age
(years)
25-29
30-34
35-39
40-44
45-49
50-54
55-59
60-64
65-69
70-74
75-79
80-84
85 and over
25 and overa
a
Standardized rates.
Other
cancers
Nonsmokers
Smokers
Nonsmokers
Smokers
0.03 (1)
0.08 (3)
0.18 (6)
0.69 (21)
1.32 (36)
2.52 (51)
2.95 (44)
2.94 (34)
4.06 (33)
4.52 (22)
3.35 (8)
0.19 (1)
0.20 (2)
0.57 (6)
0.24 (2)
0.82 ( 5 )
0.66 (3)
1.16 (4)
1.76 (4)
6.02 (9)
2.75 (3)
12.87 (11)
16.46 (10)
32.35 (11)
0.49 (5)
0.23 (6)
0.27 (10)
0.39 (15)
0.56 (19)
1.18 (36)
1.11 (30)
2.82 (57)
3.89 (58)
6.41 (74)
8.00 (65)
12.92 (63)
13.82 (33)
0.91 (259)
1.87 (71)
0.10 (1)
-
0.92 (1)
2.34 (2)
0.11 (4)
1.65 (471)
The association of chronic bronchitis and emphysema with smoking that we observe in doctors has been a conspicuous feature of other
prospective as well as retrospective studies and
also of carefully conducted surveys of various
populations. Thus, bringing together the results of seven prospective studies (including our
own), the Surgeon Generals Advisory Committee calculates that cigarette smokers died of
chronic bronchitis and emphysema six times
more frequently than nonsmokers (546 observed deaths to only 89 expected). In all but
one of these studies the excess is marked.
As an example of a retrospective study we
may cite the comparison made by Oswald,
Harold, and Martin (27) of 300 chronic bron-
659
660
In 89 instances (6 percent of the total) hyperThat being so, and seeing that chronic bronchitis is a common disease in this country, we tension was also mentioned on the certificate,
must naturally anticipate the clinical finding and these cases, examined separately, showed
that chronic bronchitis and cancer of the lung no relationship with smoking (Tables 18 and
will quite frequently occur together in the same 19).
For the remaining 1287 deaths the results are
person. It is, of course, possible that the chronic
bronchitis itself plays a part in producing cancer very similar to those we previously reported
of the lung (38), and figures published in a with less than half the number of cases (2), and,
retrospective inquiry by Campbell and Lee (39) although the differences in mortality between
somewhat support that view. For the prospec- cigarette smokers and nonsmokers (33 percent)
tive inquiry, however, what we need is to observe and between heavy and light cigarette smokers
the incidence of cancer of the lung in non- (14percent) are not great, the numbers are so
smokers suffering from chronic bronchitis; and large that there can be no doubt that the relawith its rarity in the absence of smoking the tionship with cigarette smoking is not due to
chance (P= 0.001 and P=O.O4, respectively).
problem is to obtain enough of them.
The results are, moreover, similar to those obtained in the six other large prospective studies
Pulmonary Tuberculosis
reviewed by the Surgeon Generals Advisory
The relationship between smoking and morCommittee, all of which show that the observed
tality from pulmonary tuberculosis is quite dismortality from coronary disease among cigatinct, though it is based, we should observe,
rette smokers is from 1.7 to 2.0 times the morupon only 42 deaths. The further 14 deaths in
tality expected from the experience of the nonwhich the presence of the disease was mensmokers. Similar results have also been obtioned as a contributory and not as the primary
tained in Albany and Framingham by Doyle,
cause of death reveal no such association. It is
Dawber, Kannel, Heslin, and Kahn (42), who
therefore possible that smoking may hinder refollowed 4 120 initially healthy men and found
covery from the infection or, as Lowe (40) has
that the incidence of myocardial infarction was
concluded from his retrospective survey, it may
2.4 times higher in cigarette smokers than in
conceivably re-aggravate a quiescent lesion. We
nonsmokers and 1.3 times higher in men smoksee nothing unlikely in that. On the other hand,
ing more than 20 cigarettes a day than in men
Kissen (41) suggests that anxiety is one reason
smoking less. In all these studies as well as in
for a poor prognosis in tuberculosis and also
our data the relationship is specific for cigarette
one reason for smoking. Without, therefore,
smokers, and no excess mortality has been obmore-and more precise-data on a disease so
served among pipe or cigar smokers.
influenced by social factors, we hesitate to conAccording to Hammond and Horn (4) and to
clude that a direct causal relationship of morBuechley, Drake, and Breslow (43),the relationtality with smoking does exist.
ship is more marked at younger ages than at
old. In our data this feature is distinct (Table 20)
Coronary Thrombosis
and the 38 deaths that occurred at the early age
Deaths attributed to coronary disease of 35-44 years show a relationship with smoking
amount, in our data, to 30 percent of the total that is only a little less close than that for chronic
deaths (1376 out of 4597). The great majority bronchitis. At these ages the mortality among
are described as due to coronary thrombosis or nonsmokers is extremely low (0.11 per 1000 per
myocardial infarction, but there is a small pro- year), so that although the rate of 0.61 among
portion described less specifically as due, for cigarette smokers is only 0.50 per 1000 higher
example, to coronary-atheroma, coronary is- the ratio of the rates is 5.5 to 1. Over the next
chemia, or angina pectoris with myocardial four decades the mortality of nonsmokers indegeneration. With so many deaths it was not creases to 10.83 and of smokers to 14.69; in
practicable to check the diagnoses as we did other words, though the ratio of the rates has
with cancer of the lung, and it is probable that become quite small, the excess mortality of 3.86
many of them, particularly at ages over 75 is seven times the excess at ages 35-44 years. At
years, were not, in fact, associated with an acute the oldest ages (75 years and over) the differepisode of thrombosis.
ence between smokers and nonsmokers disap-
661
662
Peptic Ulcer
Even including 15 deaths in which peptic ulcer was certified as being contributory though
not the underlying cause, the total number (54)
is too small to allow us to examine gastric and
duodenal ulcers separately. Classed together as
peptic ulcer, the results agree with those of 12
other studies in which the relationship between
smoking and peptic ulcer has been specially
examined (five retrospective, six prospective,
and one cross-sectional survey, Surgeon General's Advisory Committee). In all the retrospective studies the proportion of nonsmokers
was higher in the control group than in the
ulcer group, and in all the prospective studies
the mortality was higher in smokers than in
nonsmokers. In all studies in which gastric and
duodenal ulcers have been separated the relationship has been stronger with gastric ulcer
than with duodenal, and, when the type of
smoking has been examined, the relationship
has been closer with cigarette smoking than
with the smoking of pipes or cigars.
An unusual feature of the prospective studies
has been th'at the mortality has been maximum
in the moderate or light smokers rather than in
the heavy cigarette smokers. This is shown in
the present data (Table 22), and a comparable
result was obtained in the one large-scale retrospective study in which this aspect was examined in detail (53).
It is difficult to believe that these findings
mean that smoking is a direct cause of gastric
ulcer, since gastric ulcers have been decreasing
in incidence over the long period of time in
which cigarette smoking has been increasing.
Duodenal ulcers, admittedly, have increased in
With the large group of other and unrelated diseases the death rates of all the smoking
subgroups lie, with one exception, in the narrow range of 8 to 9 per 1000 (nonsmokers 8.53,
cigarette smokers 8.99, mixed smokers 8.00,
pipe and/or cigar smokers 8.06, ex-cigarette
smokers 8.02, 1-14 cigarettes daily 8.96, 15-24
cigarettes daily 8.66,25 + cigarettes daily 10.11;
Tables 23 and 24). This close similarity makes it
unlikely that we have overlooked any major relationship between smoking and mortality. The
only departure from the prevailing pattern is
the rather high death rate of 10.11 per 1000 in
the heaviest cigarette smokers, an excess of 19
percent over the rate for nonsmokers. This excess comes from only 7 of the 11 causes of death
shown in Tables 23 and 24, and could, we believe, be partly due not to smoking per se but to
other features of life, both environmental and
constitutional, which are correlated with it-for
example, the psychological and possibly physical characteristics of heavy smokers and their
habits of eating and drinking, etc. It is also
likely that some of the excess is due to errors in
the certification of the cause of death. Unless
100 percent accuracy was achieved in death certification-and we know that it is never so-the
unrelatedcauses must inevitably include some
deaths that should in fact have been classified as
related. The unrelated causes will thereby
share some of the excess mortality of heavy
cigarette smokers that is characteristic of the
related causes. In short, we conclude that smoking bears no direct or causal relationship to
some 60 percent of the total death rate.
663
That is not to say that our related causes necessarily include all the diseases in which smoking is of direct etiological importance. We have
no evidence on nonfatal diseases such as tobacco amblyopia, and very little for many rare
diseases or diseases with a low fatality rate.
There is, for example, strong evidence to inculpate smoking as a cause of thromboangiitis
obliterans (Buergers disease), but in our data
no deaths were attributed to it.
Paucity of cases may also be the explanation
of our failure to find any association of smoking
with cancer of the bladder. Four retrospective
studies have shown an increased mortality for
this site among cigarette smokers-of the order
of two to three times the rate in nonsmokersand a still higher rate of mortality in heavy
cigarette smokers (44,58-60).Moreover, similar
results have been obtained in all the other six
prospective studies (Surgeon Generals Advisory Committee). Our data are based on only
38 cases-six of which occurred in nonsmokers-and with this small number of deaths
it would not be surprising if by chance we had
failed to demonstrate a weak association with
cigarette smoking. But it does not appear that
there can be a strong relationship. (Since the
conclusion of the ten-year period of observation
a further 14 deaths with bladder cancer have
been reported-six among cigarette smokers,
eight among other smokers, and none in nonsmokers.)
In the studies of Hammond and Horn (4)
and Dorn (5) several other broad groups of
diseases have also been reported to show an
excess mortality with smoking. The excesses
have mostly been quite small, and, with the exception of a heterogeneous group of other circulatory diseases (Surgeon Generals Advisory
Committee), none have consistently shown a
mortality among cigarette smokers as much as
double that among nonsmokers. Such differences, we believe, could easily be due to the
same causes as we have suggested above may
account for the excess mortality from unrelated diseases in doctors who are heavy cigarette smokers. We may also note that the very
large populations studied by Hammond and
Horn (4) and by Dorn (5) have necessarily been
heterogeneous in many respects. Their results
may thereby be influenced to some extent by
the social distribution of smoking. Doctors, on
the other hand, are a more homogeneous
664
was 29 percent greater than that of nonsmokers; in men who were continuing to smoke
cigarettes at the start of the inquiry it was 35
percent greater. These figures, we pointed out,
do not necessarily mean that cigarette smoking
increases the death rate by such amounts, since
other explanations are possible. However, after
our separate and detailed examination of the
principal causes of death we have concluded
that cigarette smoking is likely to be responsible
for at least the greater part of the excess.
The results of our assessment are summarized in Table 37, where we divide the excess
death rate in smokers of different ages into
three categories. ( I ) The first category comprises the excess deaths from the unrelated
diseases and from cirrhosis of the liver and alcoholism. Conceivably some small part of this excess may be due to cigarette smoking, insofar as
it is due to errors in the diagnosis of the cause of
death; the greater part, however, is likely to be
either an artifact (due to the selective failure of
doctors to stop smoking when they develop a
serious illness; Table 25 and p. 649), or a secondary effect of the association of cigarette
smoking with some other factor which itself
predisposes to disease and death-for example,
heavy drinking or excessive anxiety. (2) The
second category includes deaths from cancers
of the upper respiratory and upper digestive
tracts, from peptic ulcer, and from pulmonary
tuberculosis. Some of these deaths are, we suggest, directly attributable to smoking, but it is
impossible to say how many, since the effect of
smoking is complicated by other factors, en-
Nonsmokers
Causes not
attributable
to smoking
Causes partly
attributable
to smoking
Causes
attributable
to smoking
Cigarette
smokers
all causes
25-44
45-54
55-64
65-74
75 +
1.12
4.12
12.08
30.56
114.29
0.48
1.02
2.37
6.72
- 11.64
0.04
0.22
0.92
0.82
2.09
0.39
2.25
5.33
10.87
6.64
2.04
7.62
20.70
48.96
111.37
12.06
0.93
0.40
2.93
16.32
All ages
standardized
Percentage of
nonsmokers
100
24
135
665
666
667
~~
~~
6 19
620
For the past 13 years the National Heart Institute of the Public Health Service, acting
through the Massachusetts Department of Pub-
Dawber et al.
lic Health and the Framingham Health Department, has been conducting a long-term study of
coronary heart disease in the town of Frarningham, Massachusetts. A number of reports
of the progress of this study have been made
(1 -4). The authors have been engaged in the
conduct of this study from its inception, and
their experience gained in this undertaking
serves as the basis for a discussion of the methodology of longitudinal community studies of
which the Framingham Study will serve as a
prototype.
The Framingham Study was set up as a community-wide study of the general adult population of a town. It is a prospective study concerned with the incidence of coronary heart
disease and with the study of those factors, both
host and environmental, which may contribute
to its development.
It is safe to say that up to the present time this
study has been considered reasonably successful. It is customary to attribute success to
such factors as careful planning and hard work.
Such ingredients are admittedly necessary. It is
not customary to point out that success in
epidemiologic studies usually depends on finding significant associations between some factors and disease development, i.e., having the
good fortune to select those factors for study
which turn out to be related to the disease. If all
of the factors studied had turned out not to be
associated with coronary heart disease, this
study would have been considered a failure, and
it would probably have been abandoned. Nevertheless, such a study might have been almost
as valuable, and the methodology just as worthwhile, even if it only showed that large numbers
of factors did not appear to be related to disease
development, particularly if the factors studied
were widely accepted as being causally related.
The methodology of the Framingham Study
should not, therefor?, he evaluated in terms of
the successor failure of this project to turn
up positive findings, but rather on whether it
was, in fact, able to study a population adequately for a reasonable length of time in order
to determine whether or not the factors studied
could be demonstrated to be associated with
coronary heart disease development.
OBJECTIVES OF THE STUDY
The first aspect of any study to warrant attention is that of its objectives. As previously indi-
62 1
622
Dawber et al.
based on absolute estimates of disease incidence, may be inadequate for the demonstration of significant differences in relative disease
incidence in making comparisons between subgroups of the population. If the factors to be
investigated are strongly related to the disease,
this estimate of the population size will probably
be adequate. If, however, the factor bears only a
weak relationship to the disease, a much larger
population, and/or a longer period of ohservation will be needed. It cannot be hoped to study
a population large enough to provide answers
as to the effects of factors, regardless of the
potency of their effect, or their distribution in
the population. The major determining factor
must be the expected incidence of the disease
and the assumption that at least some of the
factors operating are strongly enough associated so that within-group comparisons will show
significant differences. Since it is usually not
possible to estimate reliably the potency of the
characteristic under investigation in relation to
development of disease before the study is undertaken, it is necessary, arbitrarily, to assign
some potency to the characteristic in estimating
the size of the population sample required to
demonstrate statistically significant differences.
In making this decision, one is motivated by
how large a difference would be biologically
meaningful or important.
AGE OF POPULATION
623
Having decided on the size, age, and sex distribution, and type of population desired, the
next task was that of finding the population.
Inquiry had been made through several state
health departments as to the possible commu-
624
less the cooperation of the population is essentially complete, it is difficult, if not impossible,
to consider it a random sample and representative of a larger population. If the characteristics to be studied influence the cooperation
of the population and in effect determine the
response rate, it is conceivable that a systematic
bias may be introduced, thus making random
sampling highly desirable for such studies. If,
however, the personal traits under consideration are apparently unrelated to population response, the use of volunteers would seem to be
adequate. If the objective in random sampling
is to insure that the sample be representative of
the entire community, steps must also be taken
to ascertain the degree to which the population
sample represents larger population segments
such as the county, the state, the country, etc.
At the inception of the Framingham Study it
was believed that the selection of a sample
should be made by random sampling methods,
to provide a study population representative of
the town. If a truly random sample could have
been obtained, it would then have been possible
to have made observations on the absolute prevalence and incidence of coronary heart disease
which could then be compared with estimates of
prevalence and incidence in other locations.
Since the major effort of the Framingham
Study involved within-group comparisons, it
would appear that a great deal of unnecessary
effort may have been expended in attempting
to get a random sample of the Framingham
population.
The possibility that studies of other population groups may obtain data on coronary heart
disease in a similar manner may yet justify the
effort expended in trying to get a representative sample. To date, however, this effort
seems to have been unwarranted.
A two-thirds random sample of the adult
population age 30-59 of the town of Framingham was selected, and an attempt made to
bring in those selected for examination and
continued follow-up.
A systematic sampling scheme was used, arranging persons by precincts of residence and
by size of family grouping; and two out of every
three family groupings were selected.
This method was chosen over a geographic
sampling, presumably on the basis that the sample obtained would be more representative of
the entire town. The major difficulty encoun-
Dawber et al.
625
Men
Women
6507
4469
3074
2024
3433
2445
740
312
428
4393
734
1975
307
2418
427
5127
2282
2845
Random sample
Respondents
Volunteers
626
Lost
Outside Info.
[IIIID Dead
0Examined
EXAMINATION NUMBER
Dawber et al.
nial clinic examinations, follow-up also has consisted of surveillance of the study population
with respect to hospital admissions, records of
interim illnesses obtained from subjects, physicians, relatives and friends, medical examiner
reports, and death certificates.
On the basis of all information available for
each subject, conclusions have been reached as
to the state of health with particular reference
to cardiovascular disease.
The need to consider changes in concepts of
disease, in laboratory procedure, and in observers is peculiar to longitudinal studies. It is necessary, therefore, to consider modifying criteria
and laboratory methodology when better information concerning these develops. In so doing,
it is necessary to consider the effect of such
changes on the continuity of information.
Changes in the subjects must be distinguished
from changes in method. Unless previous observations are proved invalid by more recent
knowledge, it is not desirable to make frequent
or drastic changes in study techniques.
Since the inception of the Framingham
Study, enzyme studies for the detection of myocardial infarction have been introduced. These
have been incorporated in the diagnostic criteria to supplement electrocardiographic information.
CRITERIA FOR DISEASE
627
628
147
CHOLESTEROL LEVEL
ANALYSIS OF DATA
The nature of the analytical problems in longitudinal prospective studies, such as Framingham, are such that machine processing of
data is imperative. Since the incidence of disease has been relatively low and a large population has been required, the difficulty in handling longitudinal data collected on such a large
population under continuous surveillance for
so many years has made the use of machines
mandatory. Huge quantities of data on personal
characteristics and environmental influences
collected continuously throughout a long period of observation must be tabulated and crossclassified. While this is a counting and sorting
operation, the large numbers of observations
make use of machines necessary not only for
tabulation, but also for storage and retrieving of
information.
Up to the present time, analysis of data has
been based on an assumed uniform period of
observation of all persons initially examined.
n
I I
186
aMorbidily ratio =
Observed cases
Expected cases
x 100
I-
Q
m
a
118
Dawber et al.
Serum cholesterol level, blood pressure, electrocardiographic abnormalities not considered
diagnostic of coronary heart disease, weight,
cigarette smoking habits, and vital capacity are
factors which, by simple analysis, have been
demonstrated to be significantly associated with
the development of coronary heart disease. Assessment of the independent contribution of
each of these often interrelated factors to the
development of coronary heart disease has required detailed, rather cumbersome, analysis
which would have been almost impossible if
done by hand.
Because of the many observations on such a
large population, even relatively simple operations of counting, sorting, classifying, and crossclassifying of information would have been tedious and prohibitively time-consuming if done
by hand. Calculations of means, standard
deviations, and correlation coefficients, while
perfectly feasible with the use of a desk type
calculating machine, are extremely time-consuming when done on such a large number of
observations.
As the data from the repeated biennial examinations have accumulated, the number of ways
in which these data can be examined seems
infinite. To date, analysis has been based on
observations made at the initial examination in
relation to the subsequent development of disease. Since these initial characteristics may have
been lost, acquired, or otherwise changed
through the years on repeated observation, it
will be desirable to know whether some parameter of these continuous observations, e.g.,
trends, means, variability, etc., will improve the
ability to predict disease or give further insight
into the mechanism of disease production. Accumulation of the additional persons in the
population at risk, by making use of all the
longitudinal observations, allows more detailed
analysis of the relationship of a characteristic
under observation to development of disease.
Data analysis, accounting for changes in characteristics or personal attributes with the passage
of time in relation to development of disease,
requires use of computers so that variable periods of follow-up from the onset, loss, or change
of some characteristic to development of disease can be taken into account.
From the beginning of the Framingham
Study, plans were made to analyze data by
means of machine tabulators and computers.
629
630
References
(I) Dawber, T.R., G.F. Meadors, and F.E. Moore,Jr.
Epidemiologic approaches to heart disease: the Framingham study. A m J Pub Health 41: 279-286, 1951.
(2) Dawber, T.R., F.E. Moore, and G.V. Mann. Coronary heart disease in the Framingham study. Am J
Pub Health 47: 4-24, 1957.
( 3 ) Dawber, T.R., W.B. Kannel, N. Revotskie, J.
Stokes 111, A. Kagan, and T. Gordon. Some factors
associated with the development of coronary heart
disease: six years follow-up experience in the Framingham study. Am J Pub Health 49: 1349-1356,
1959.
(4) Kannel, W.B., T.R. Dawber, A. Kagan, N. Revotskie, and J. Stokes 111. Factors of risk in the development of coronary heart disease-six-year follow-up
experience: the Framingham study. Ann Internal Med
55: 33-50, 1961.
(5) Doyle,J.T., A.E. Heslin, H.E. Hilleboe, P.F. Formel, and R. F. Korns. A prospective study of degenerative cardiovascular disease in Albany: report of three
years experience-I. Ischemic heart disease. Am J
Pub Health 47: 25-32, 1957.
(6) Drake, R.M., R.W. Buechley, and L. Breslow.
An epidemiological investigation of coronary heart
disease in the California health survey population. Am
J Pub Health 47: 43-57, 1957.
(7) Report of the Conference on Longitudinal Cardiovascular Studies at the Hotel Beaconsfield,
Brookline, Massachusetts,June 17-18, 1957.
(8) Epidemiology of cardiovascular diseases methodology-hypertension and atherosclerosis. Report
of Conference, Princeton, N.J., April 24-26, 1959.
Supp. to Am J Pub Health 50: 1-124, 1960.
DATA
To this request we had 40 637 replies sufficiently complete to be used-34 445 from men
and 6192 from women. From a 1 in 10 random
sample of the register that we subsequently
drew and analyzed we estimate that these figures represent answers from 69 percent of the
men and 60 percent of the women alive at the
63 1
632
answerers and 179 (84 percent) of the nonanswerers responded. Comparison of these two
groups shows 2 1 percent (answerers) and 6 percent (nonanswerers) as nonsmokers and 15 percent (answerers)and 28 percent (nonanswerers)
as moderate or heavy cigarette smokers (15 or
more daily). While these differences are large
and must contribute measurably to the continuing favorable mortality of the group that replied
in 1951, they are unlikely to account for it
wholly. As a further factor we suspect (but obviously cannot prove) that there may be some
more general association between mortality and
the tendency not to reply to such an inquirywhether the tendency be due to a deliberate
refusal (which is rare) or a mere neglect of these
things (which is frequent) In this respect it is
perhaps not too fanciful to note that one nonreplier died of smallpox and another of diabetic
coma.
Second Questionary
According to the doctors replies in 1951 we
allocated them to the appropriate nonsmoking
or smoking groups (subdivided by manner and
amount, continuing or stopped). Our previous
calculations of their subsequent death rates
were based upon the number of persons in each
of those groups at that time. We knew nothing
about any subsequent change of smoking habits, either in the dead or in the living, and the
further we moved from 1 November 1951 the
more likely it was that changes in habits had
occurred. In particular, it was probable that a
large number of doctors had given up cigarette
smoking. We therefore decided to approach
again all the survivors of those who had previously replied, and, taking advantage of this
opportunity, we added questions on: (1) the
past use of cigarettes by pipe and cigar smokers,
and (2)4nhaling (a factor that had since become
prominent in argument). Allowing for repeated
inquiries, we sought answers from men between
1 November 1957 and 31 October 1958, and
from women between 1 November 1960 and 3 1
October 1961.
Between the issue of our first and second
questionaries 2579 of the men and 320 of the
women had died; seven men had been struck
off the Register, and these we omitted. Of the
remaining 31 859 men and 5872 women, all but
2 percent replied to our second approach. Of
Deaths
Through the courtesy of the Registrars-Genera1 in the United Kingdom, a form with particulars of the cause of death has been provided to
us for every death since November 1951 identified as referring to a medically qualified person. We have also obtained lists of deaths notified to the General Medical Council and of
those recorded by the British Medical Association, and we have sought information from the
records of the fighting Services and from other
sources at home and abroad. A few deaths came
to light from our second questionary. As a result
of these several approaches we believe that very
few deaths can have been missed. In fact, combining their total numbers appropriately with
our 10 percent sample of non-repliers gives us a
mortality rate for all British doctors which is 93
percent of the corresponding mortality of all
males in England and Wales. This figure compares well with the standardized mortality ratio
of 89 percent that the Registrar-General gives
for doctors aged 20-64 and 65-74 years in his
Occupational Mortality Supplement for
1949-1953.
In total, in the ten years to which this paper
relates (1 November 1951 to 31 October 1961)
there have been 4597 deaths of men and 366
deaths of women. (Preliminary data for the eleventh year give another 472 deaths of men and
48 deaths of women, and these will also be used
for analyses where the numbers would otherwise be too few.) Except for deaths attributed to
cancer of the lung, we have accepted without
further inquiry the certified cause of death, and
(unless otherwise mentioned) have classified
these deaths according to the specified underlying causes. In only one case have we failed to
obtain any evidence of the cause of death.
633
634
Habits in
1951
Nonsmokers
Ex-smokers
Current smokers:
Pipe or cigar
Cigarette and
other
Cigarette only
Total
a
Nonsmokers
Exsmokers
5272
Pipe or
cigar
Cigarette
and other
Cigarette
only
Total
54
13
100
5439
4247
207
65
293
4812
707
2575
213
118
36132
520
629
2083
570
3802
2840
64 1
787
9274
13 542
5272
8314
4106b
3161
10 355
31 208
Table 2. Amount of smoking in 1951 and 1958, men, all forms of smoking.
Habits in 1958
Some tobacco (g/day)
Reduced by
Habits in
1951
No tobacco
15 or
more
10-14
Nonsmokers
5272
Ex-smokers
4247
Smokers of (glday):
1-4
59 1
5-9
808
10-14
854
15-19
629
20-24
677
508
25 +
27
131
392
60
190
290
336
13 586
550
876
Total
Increased by
5-9
1-4
No change
139
487
569
394
300
162
473
938
1439
1504
1991
1832
1804 2051
8177
115
436
427
453
373
15 or
more
Total
47
24
144 122
29
164
5439
4812
21
68
154 53
289 98
244 67
216 128
183 116
35
32
38
31
39
78
1569
2909
4159
3802
4385
4133
446
31 208
1-4
5-9
67
135
242
322
376
289
160
153
10-14
Table 3. Pure cigarette smokers in 1951 and their smoking habits in 1958, men.
Habits in 1958
Habits in 1951
(No. smoked
per day)
1-4
5-9
10-14
15-19
20-24
25 +
Total
Changed
to pipes
and/or cigars
1-4
5-9
328
48 1
550
438
557
397
31
72
120
101
163
130
301
171
125
59
54
31
275 1
617
741
Stopped
smoking
10-14
15-19 20-24
25+
Total
Total
65
495
280
156
87
43
18
101
920
240
268
90
8
11
25
19
188
88
1056 191
213 1753
149 222
3
5
23
79
316
2027
406
816
1624
1781
2691
2562
765
1369
2294
2320
341 1
3089
1126
1637
1642 2281
2453
9880
13 24@
This total is less than the total of men smoking only cigarettes shown in Table 1 since it excludes 294 men who were known
635
636
Mortality from cancer of the lung is examined in Tables 4 to 11, where we have included
the five deaths for which cancer of the lung was
mentioned on the death certificate as contributory with the 207 in which it was given as the
underlying cause.6 Many of the rates, however,
are based on small numbers, and although they
may contribute usefully to the general picture
they cannot be relied on individually. In spite of
this there is a steadily rising death rate with
increasing consumption of cigarettes at every
age above 45 years (Table 4). For all ages a more
detailed analysis in Figure 1 indicates a linear
relationship, the death rate rising step by step
from the 0.07 per 1000 in nonsmokers to 3.15
per 1000 per annum in men smoking 35 or
more cigarettes daily.
In examining the effect of the method of
smoking (Table 5), we are dependent upon the
Ten certified deaths were excluded because additional
evidence suggested that the diagnosis was incorrect (see p.
633).
Table 4. Age and number of cigarettes smoked death rates per 1000 per annum from
cancer of the lung (numbers of deaths in parentheses).
Cigarettes per day
Age
Nonsmokers
35-44
45-54
55-64
65-74
75 and over
All ages
a
1-14
0.05 (1)
0.00
0.00
0.00
1.11 (2)
0.07
0.31
0.48
2.69
2.68
(1)
(3)
(3)
(9)
(6)
0.07 (3)
0.57 (22)
15-24
0.00
0.62
2.31
5.16
7.27
25 +
(9)
(20)
(17)
(8)
0.11 (1)
0.75 (8)
3.88 (26)
6.48 (14)
16.33 (8)
1.39 (54)
2.27 (57)
Number reported at 1 November 1951 for men continuing to smoke cigarettes only at that time.
Method of smoking
0.96 (143)
0.53 (42)
0.43 (24)
0.42 (12)
0.45 (6)
0.96 (3)
At 1 November 1951.
637
mixed, pipe and/or cigar alone) the rate of continuing smokers is higher than that of those
who had stopped, and the rate in ex-cigarette
smokers (0.24) is notably less than that in current pipe smokers (0.47).
The difference between the ex-smokers and
the continuing smokers is most marked for
those who smoked cigarettes, where the rate for
those who had given up is only 19 percent of
the rate for those who continued. This difference, moreover, may well underestimate the
true difference, since in this analysis we take no
account of the changes in habits in the years
following 1951. (See p. 648, where allowance
has been made for some of the changes in habits which took place and where comparison is
made between men who have given up smoking
for different lengths of time.)
In Table 7 we consider the nature of the
evidence available to the doctors who certified
cancer of the lung as the cause of death (excluding the one case for which this information was
refused to us). In more than half of the deaths
(56 percent) there was histological, cytological,
or necropsy evidence together with X-ray or
bronchoscopic confirmation of the site of the
primary growth. In another 38 percent an X-
Continuing at 1111/51
Stopped before 1/11/5 1
Cigarette
smokers
Mixed
smokers
Pipe and/or
cigar smokers
1.25 (133)
0.24 (10)
0.59 (36)
0.48 (6)
0.47 (21)
0.23 (3)
Tablu 7. Standard of diagnosis: standardized death rates from cancer of the lung.
Death rate per 1000
Standard
of
diagnosisd
(No. of deaths)
Cigarette smokersb
Nonsmokers
All
smokers
Continuing
Given
UP
1-14
daily
15-24
daily
25 +
daily
Grade 1 (118)
2 (81)
3 (12)
0.01
0.03
0.03
0.41
0.28
0.04
0.68
0.52
0.04
0.17
0.05
0.03
0.37
0.15
0.02
0.76
0.60
0.03
1.02
1.21
0.05
0.07
0.73
1.24
0.24
0.54
1.39
2.27
a Grade 1= Necropsy evidence or histological or cytological together with evidence of primary from x-ray picture or
bronchoscopy. Grad 2 =Clinical evidence together with evidence of primary from x-ray picture or bronchoscopy. Grade
3 =Evidence from hfsrory and physical examination alone.
One case for which information was refused has been excluded (a man aged 74 smoking 14 cigarettes a day).
At 1 November 1951.
638
ray picture or bronchoscopy supported theclinical evidence. In both these groups the various
associations of mortality with smoking are quite
distinct. On the other hand, it is of interest that
for the 12 cases in which the diagnosis rested on
history and physical examination there is no
clear evidence of association. In other words,
these figures show that doctors are not diagnosing cancer of the lung in their colleagues without adequate evidence, and that in the few less
definite cases they are not making such a diagnosis merely because the sick person was a
heavy cigarette smoker.
Similarly Table 8 shows no marked associations with smoking in the 13 cases of adenocarcinoma, whereas they are distinct with both the
squamous and the oat and anaplastic growths.
Taking the death rate of the continuing cigarette smokers in each histological group as standard (loo), we have ratios for those smoking 1
to 14, 15 to 24, and 25 or more daily, of 68, 104,
and 139 percent in the squamous group, and
44,91, and 169 percent with the oat celled and
anaplastic group.
The rise in lung cancer mortality with increasing number of cigarettes smoked occurred
in all types of area, and the rates in Table 9
provide no reason to suppose that the association was any closer in big towns than in small
towns or in the countryside. It is, however, possible that the rural mortality is affected by the
retirement there of men who had passed their
active lives in towns. We therefore consider in
Table 10 doctors under 65 years of age. The
number of deaths in the different areas is then
small, but they do suggest that there is a lower
lung cancer mortality in the rural, and possibly
the small town, areas which is not due to differences in amounts smoked.
Table 11 shows the mortality in the principal
smoking categories for two periods of time-1
November 1951 to 31 October 1956 and 1 November 1956 to 31 October 1961. The first of
these quinquennia followed immediately after
the questionaries were sent out. Although the
majority were returned within a few weeks,
there were some which were not returned for
several months, so that the first year of observation was biased by the inclusion of some months
of observation for which there could not be any
corresponding mortality. A much more important type of bias is the differential selection of
healthy men among the repliers, and we have
shown that the total mortality rate was lower
during the early years of follow-up than in any
subsequent period. Clearly, therefore, we would
expect the mortality from lung cancer to be
higher in the second quinquennium and that
Table 8. Histological type: standardized death rates from cancer of the lung.
Death rate per 1000
Histological
Y Pe
(No. of
deaths)
Cigarette smokers
Nonsmokers
All
smokers
Continuing
Given
UP
1-14
daily
15-24
daily
25 +
daily
0.00
0.19
0.32
0.09
0.22
0.33
0.45
0.01
0. I4
0.22
0.05
0.10
0.20
0.38
0.00
0.05
0.07
0.03
0.03
0.12
0.07
Squamous (55)
Oat and anaplastic (40)
Adenocarcinoma
(13)
Table 9. Place of residence: standardized death rates from cancer of the lung.
Death rate per 1000
Place
of
residence
Conurbation
Large towns
Small
Rural areas
a
Nonsmokers
1-14
15-24
25 +
0.03
0.00
0.11
0.12
0.48
0.32
0.87
0.52
1.31
1.88
1.06
1.15
1.90
4.43
2.20
1.17
No.
of
deaths
49
34
32
18
639
Table 10. Place of residence: standardized death ratesa at ages 25-64 years.
Death rate per 1000
Place of
residence
Conurbation
Large towns
Smalil
Rural area
a
Nonsmokers
Cigarette
smokersb
No. of
deaths
0.04
0.00
0.00
0.00
0.62
0.65
0.52
0.40
31
17
16
8
Table 11, Period after start of inquiry: standardized death rates from cancer of the lung.
Death rate per 1000
Period
(No. of
deaths)
1951-6 (102)
1956-1 (110)
All
men
Nonsmokers
All
cigarette
smokers
0.69
0.64
0.05
0.08
0.98
0.95
Cigarette
smokers
Current cigarette
smokers
Current
Ex-
1-14/d
15-24/d
25+/d
1.22
1.28
0.41
0.09
0.64
0.50
1.20
1.53
2.25
2.32
640
Site
(1)
All
smokers
(5)
Cigarette
smokers
(6)
Mixed
smokers
(3)
Nonsmokers
(4)
(7)
(8)
19 (2)
0.06
0.00
0.06
0.05
0.10
0.04
16 (5)
29 (1)
0.05
0.09
0.00
0.04
0.06
0.10
0.05
0.06
0.03
0.19
0.10
0.08
64 (8)
0.20
0.04
0.22
0.16
0.32
0.22
No. of
deathsa
All
men
(2)
Mouth,
pharynx,
or nose
Larynx or
trachea . .
Esophagus
Total
Pipe or
cigar
smokers
a The numbers in parentheses are of deaths (included in the total figure) in which cancer of the specified site was certified as
being associated with the death but not its direct or underlying cause.
Currently smoking or past smokers in 1951.
Table 13. Standardized death rates from cancer of the upper respiratory and
digestive tracts in relation to the amount smoked.
Death rate per 1000
Amount of tobacco smoked daily (g.)
Site
Mouth, pharynx, or
nose
Larynx or trachea
Esophagus
Total
1-14
15-24
25+
All
amounts
Given up
smoking
0.04
0.02
0.08
0.01
0.02
0.14
0.21
0.15
0.20
0.07
0.06
0.12
0.06
0.05
0.02
0.13
0.17
0.56
0.24
0.13
and 17. For chronic bronchitis there is a closeand statistically highly significant-association
with smoking. While this was most marked
when chronic bronchitis was described as the
underlying cause of death, it was also clearly
present when chronic bronchitis served as a
contributory cause in cardiovascular mortality.
For pulmonary tuberculosis the evidence is
much less clear. The mortality was higher in
cigarette smokers than in nonsmokers, in continuing cigarette smokers compared with those
who had stopped, and in heavy cigarette
smokers compared with light. The deaths were,
however, few, and none of these differences was
statistically significant. Moreover, the few additional deaths for which pulmonary tuberculosis
was described as a contributory cause tend to
diminish the relationship rather than reinforce
it. In other words, the data suggest that death
from pulmonary tuberculosis may be associated
with smoking, while the disease itself is not.
64 1
Site
No. of
deaths
All
men
Nonsmokers
All
smokers
Cigarette
smokers
Mixed
smokers
Pipe or
cigar
smokers
381
84
97
35
69
194
0.1 1
0.26
0.30
0.11
0.21
0.61
0.17
0.27
0.31
0.05
0.33
0.64
0.11
0.25
0.30
0.11
0.20
0.61
0.13
0.30
0.26
0.12
0.18
0.68
0.12
0.14
0.32
0.12
0.16
0.52
0.07
0.30
0.39
0.08
0.26
0.51
27
0.08
0.09
0.08
0.05
0.09
0.15
542
1.67
1.87
1.65
1.72
1.45
1.77
Bladder
Stomach
Bowel
Rectum
Prostate
Other sites
Primary site
unknown
Total
a Two deaths from other causes in which cancer of the bladder was mentioned as contributory are included in this category,
but are excluded from the total cancer deaths.
Table 15. Standardized death rates from cancer of other sites in relation to the
amount smoked.
Death rate per 1000
Amount of tobacco smoked daily (g.p
Site
1-14
15-24
25+
All
amounts
Given u p
smoking
Bladder
Stomach
Bowel
Rectum
Prostate
Other sites
Primary site unknown
0.10
0.28
0.37
0.10
0.16
0.57
0.10
0.11
0.28
0.22
0.06
0.19
0.59
0.07
0.13
0.26
0.44
0.22
0.12
0.83
0.06
0.12
0.28
0.32
0.1 1
0.17
0.63
0.09
0.08
0.18
0.23
0.12
0.28
0.54
0.05
1.67
1.52
2.03b
1.71
I .49
Total
a
Chronic bronchitis
Chronic bronchitis
as associated
causea
Pulmonary tubercu 1osis
Pulmonary tuberculosis as associated cause
Other respiratory
diseases
a
No. of
deaths
All
men
Nonsmokers
All
smokers
Cigarette
smokers
Mixed
smokers
Pipe or
cigar
"
smokers
111
0.34
0.05
0.37
0.51
0.33
0.15
55
0.17
0.03
0.18
0.20
0.21
0.12
42
0.13
0.06
0.14
0.15
0.1 1
0.11
14
0.04
0.05
0.04
0.05
0.02
0.04
181
0.54
0.63
0.53
0.55
0.49
0.47
642
Chronic bronchitis
Chronic bronchitis as
associated causea
Pulmonary tuberculosis
Pulmonary tuberculosis
as associated cause
Other respiratory diseases
Given
1-14
15-24
25+
All
amounts
smoking
cigarettes
0.34
0.64
1.06
0.58
0.38
0.15
0.10
0.20
0.16
0.30
0.26
0.2 1
0.17
0.16
0.12
0.05
0.09
0.02
0.06
0.11
0.41
0.68
0.40
0.51
0.61
Cardiovascular Disease
In Tables 18 and 19 we set out the data relating to mortality from cardiovascular diseasesdue to nephritis. With the deaths attributed to
cardiovascular accidents or to coronary disease
we have ,separately considered those in which
reference was made on the death certificate to
the presence of hypertension, for these deaths
might be related to the causes of hypertension
more closely than to any other factor.
For the 138 deaths in which hypertension was
given as the primary cause there is no association with smoking habits-neither with method
(Table 18) nor with the amount of cigarette
smoking (Table 19). This is equally true for the
143 cerebrovascular accidents and for the 89
coronary disease deaths in which hypertension
was also mentioned. Adding the three hyper-
tensive groups together gives, with one exception, remarkably similar rates in all the smoking
categories, varying only between 1.26 per 1000
in nonsmokers and 1.10 per 1000 in men who
had given up smoking (the final lines of Tables
18 and 19). The exception lies in the rather low
death rate of 0.8 1 per 1000 in pipe and/or cigar
smokers, a figure which is based upon only 53
deaths, and has 95 percent confidence limits of
0.69 and 0.93.
We can also conclude from these tables that
there is no association with smoking habits in
the 135 deaths from other heart diseases
(rheumatic heart 35, other valvular disease 22,
other diseases 78) nor in the 135 from other
cardiovascular diseases (general arteriosclerosis 64, dissecting and atherosclerotic aneurysm
of the aorta 47, and other vascular diseases 24).
With the larger group of cerebrovascular accidents without reference to hypertension, there
also seems to be no association. Though the
death rate of all smokers at 1.41 per 1000 is a
trifle above that of the nonsmokers (1.24), this
small excess is not related specifically to one
method of smoking (Table 18), and there is no
clear gradient with number of cigarettes
smoked nor fall in the death rate on giving up
smoking (Table 19).
With the 337 deaths attributed to myocardial
degeneration and the 43 deaths attributed to
nephritis there is some suggestion of an association, but the evidence is slight. With myocardial
degeneration there is a substantial-and statistically significant (P<O.Ol)-difference between smokers (1.02) and non-smokers (0.59),
but there is no difference between the different
643
No of
deaths
All
men
Nonsmokers
All
smokers
Cigarette
smokers
Mixed
smokers
Pipe or
cigar
smokers
Cerebrovascular
accidents
(a).
(b)b
All
462
143
605
1.39
0.45
1.84
1.24
0.51
1.76
1.41
0.44
1.85
1.48
0.50
1.98
1.27
0.44
1.71
1.44
0.23
1.67
Coronary
disease
(a).
(b)b
All
1287
89
1376
3.99
0.28
4.26
3.31
0.30
3.61
4.08
0.28
4.36
4.39
0.26
4.65
3.87
0.37
4.25
3.18
0.21
3.39
337
138
135
0.97
0.42
0.41
0.59
0.45
0.41
1.02
0.42
0.4 1
1.01
0.45
0.42
0.98
0.36
0.43
1.oo
0.37
0.33
135
43
0.4 1
0.13
0.41
0.09
0.4 1
0.14
0.46
0.14
0.38
0.13
0.35
0.12
2769
370
8.44
1.15
7.32
1.26
8.61
1.14
9.11
1.22
8.24
1.17
7.23
0.81
Myocardial degeneration
Hypertension
Other heart diseases
Other cardiovascular diseases
Nephritis
Total
Total hypertensive
Cause of
death
Given
"P.
smoking
cigarettes
1-14
15-24
25+
All
amounts
Cerebrov ascu1ar
accidents
(a).
(b)b
All
1.46
0.46
1.93
1.43
0.48
1.91
1.69
0.54
2.23
1.49
0.51
2.00
1.42
0.51
1.93
Coronary
disease
(a)a
(b)b
All
4.35
0.29
4.65
4.28
0.29
4.57
4.97
0.18
5.16
4.57
0.28
4.86
3.73
0.19
3.92
Myocardial degeneration
H ypqrtension
Othet heart diseases
Othet cardiovascular
diseases
NepHritis
1.10
0.44
0.32
0.94
0.45
0.59
0.97
0.45
0.21
1.07
0.46
0.41
0.87
0.40
0.41
0.41
0.16
0.45
0.14
0.62
0.29
0.54
0.17
0.25
0.08
Total
Total, hypertensive
9.01
1.20
9.05
1.22
9.93
1.17
9.51
1.25
7.86
1.10
644
Age
in
years
No.
of
deaths
Nonsmokers
Current
cigarette smokers
smoking daily
All
cigarette
smokers,
ex and
current
1-14
15-24
25 +
35-44
45-54
55-64
65-74
75-84
85 +
38
149
319
389
314
78
0.1 1
1.12
4.90
10.83
21.20
32.35
0.41
1.66
6.8 1
16.44
21.22
33.06
0.49
3.40
7.03
13.04
15.04
58.54
1.50
2.73
8.81
17.59
17.30
-b
0.61
2.40
7.20
14.69
19.18
35.93
All ages
1287
3.31
4.36
4.28
4.97
4.39
a Includes all deaths attributed to coronary thrombosis, except those for which hypertension was
mentioned as a contributory cause (89).
Very small total number of man-years at risk (2?/z)-no deaths due to coronary thrombosis.
myocardial degeneration or other heart disease-and that these constitute another cardiovascular group which is also associated with
smoking. On the evidence available, however,
this group would not appear to be large enough
to have materially affected the results. Possibly
most such deaths among doctors are attributed
directly to chronic bronchitis.
Other Diseases
645
Peotic ulcer
Ciirhosis of liver
and alcoholism
Other digestive
diseases
Genito-urinary
diseasesb
Indefinite causes
Violence
All remaining
causes of death
No. of
deaths
All
men
Nonsmokers
All
smokers
Cigarette
smokers
Mixed
smokers
Pipe or
cigar
smokers
541
0.17
0.03
0.18
0.21
0.16
0.12
33
0.10
0.00
0.11
0.12
0.11
0.05
87
0.26
0.07
0.28
0.32
0.20
0.25
82
50
248
0.15
0.24
0.77
0.33
0.17
0.94
0.24
0.14
0.75
0.27
0.13
0.79
0.21
0.12
0.68
0.22
0.20
0.64
150
0.46
0.50
0.47
0.49
0.48
0.45
a Including 15 deaths in which peptic ulcer was certified as being associated with the death but not its direct or underlying
cause.
Excluding nephritis.
mortality rates for smokers (0.18) and nonsmokers (0.03) is sufficiently great to be unlikely to be due to chance (P = 0.05).
With alcoholism (6 deaths) and cirrhosis of
the liver (27 deaths) the association is strong.
No deaths from these causes occurred among
nonsmokers, and, like cancers of the mouth,
pharynx, and larynx, the mortality fell almost
wholly on the heaviest smokers.
The evidence relating to other digestive diseases is inconcludive. This is a heterogeneous
group, the major tomponents of which were 11
deaths from appendicitis, 12 from hernia, 18
from obstruction, 9 from diverticulitis, and 17
from gallstones or cholecystitis. Several of these
Table 22. Standardized death rates from other diseases in relation to the
number of cigarettes smoked.
Death rate per 1000
No. of cigarettes smoked daily
Cause of
death!
Given
UP
1-14
15-24
25+
All
amounts
smoking
cigarettes
Peptic ulcer
Cirrhosis of liver and
0.07
0.31
0.22
0.21
0.16
0.05
0.26
0.08
0.33
0.43
0.36
0.15
0.32
0.03
0.32
0.28
0.21
0.85
0.29
0.12
0.57
0.29
0.12
1.15
0.29
0.16
0.83
0.21
0.08
0.65
0.53
0.34
0.53
0.45
0.55
alcdholism
646
Table 23. Standardized death rates from causes related to smoking and from causes unrelated
to smoking.
Death rate per 1000
Cause of death
Related causes:
Cancer of lung
Other upper respiratory and
digestive cancers
Chronic bronchitis
Coronary disease
without hypertension
Peptic ulcer
(including
hematemesis)
Cirrhosis of liver
and alcoholism
Pulmonary
tuberculosis
No. of
deaths
All
men
Non-
smokers
All
smokers
Cigarette
smokers
Mixed
smokers
Pipe or
cigar
smokers
207
0.65
0.07
0.71
0.93
0.52
0.43
56
111
0.17
0.34
0.04
0.05
0.20
0.37
0.15
0.51
0.28
0.33
0.16
0.15
1287
3.99
3.31
4.08
4.39
3.87
3.18
39
0.12
0.00
0.13
0.13
0.12
0.10
33
0.10
0.00
0.11
0.12
0.11
0.05
42
0.13
0.06
0.14
0.15
0.11
0.11
1775
5.49
3.53
5.74
6.39
5.33
4.17
542
1.67
1.87
1.65
1.72
1.45
1.77
181
0.54
0.63
0.53
0.55
0.49
0.47
462
1.39
1.24
1.41
1.48
1.27
1.44
337
370
0.97
1.15
0.59
1.26
1.02
1.14
1.01
1.22
0.98
1.17
1.oo
0.81
135
43
0.41
0.13
0.41
0.09
0.41
0.14
0.42
0.14
0.43
0.13
0.33
0.12
135
0.41
0.41
0.41
0.46
0.38
0.35
87
248
282
0.26
0.77
0.85
0.07
0.94
1.00
0.28
0.75
0.85
0.32
0.79
0.89
0.20
0.68
0.81
0.25
0.64
0.87
Total
2822
8.55
8.53
8.58
8.99
8.00
8.06
All causes
4597
14.05
12.06
14.32
15.38
13.34
12.23
Total
Unrelated causes:
Other cancers
Other respiratory disease
Cerebrovascular
accidents without hypertension
Myocardial
degeneration
All hypertension
Other heart
disease
Nephritis
Other cardiovascular disease
Other digestive
disease
Violence
Remainder
647
Table 24. Standardized death rates from causes related to smoking and from
cau$es unrelated to smoking in relation to the number of cigarettes smoked.
Death rate per 1000
No. of cigarettes smoked daily
Cause of
death
Given
15-24
25 or
more
All
amounts
0.57
1.29
2.23
1.20
0.24
0.04
0.34
0.18
0.64
0.43
1.06
0.20
0.58
0.05
0.38
4.35
4.28
4.97
4.57
3.73
0.02
0.18
0.19
0.13
0.12
0.05
0.08
0.43
0.15
0.03
0.10
0.16
0.26
0.17
0.12
5.48
6.81
9.56
7.01
4.67
1.77
1.56
2.31
1.82
1.47
0.41
0.69
0.40
0.51
0.61
1.46
1.43
1.69
1.49
1.42
1.10
1.20
0.32
0.16
0.94
1.22
0.59
0.14
0.97
1.17
0.21
0.29
1.07
1.25
0.41
0.17
0.87
1.10
0.41
0.08
0.41
0.45
0.62
0.54
0.25
0.26
0.85
1.02
0.33
0.57
0.75
0.36
1.15
0.94
0.32
0.83
0.90
0.32
0.65
0.84
Tatal
8.96
8.66
10.11
9.31
8.02
All causes
14.44
15.47
19.67
16.32
12.68
Related causes:
Cancer of lung
Other upper respiratory and digestive
cancers
Chronic bronchitis
Coronary disease
without hypertension
Peptic ulcer (including hematemesis)
Cirrhosis of liver
and alcoholism
Pulmonary tuberculosis
Total
Unrelated causes:
Other cancers
Other respiratory
disease
Cerebrovascular accidents without
hypertension
Myocardial degeneration
All hypertension
Other heart disease
Nephritis
Other cardiovascular
disease
Other digestive
disease
Violence
Remainder
1-14
smoking
cigarettes
648
difference (0.3<P<0.4) nor that between nonsmokers and cigarette smokers (P= 0.09) is statistically significant.
MORTALITY OF MEN WHO HAVE GIVEN UP
SMOKING
Cigarette Smokers
Table 25. Mortality among cigarette smokers at different times after stopping smoking (numbers of
deaths in parentheses).
Death rate per 1000
Cause of death
Lung cancer
Chronic bronchitis
Coronary disease without hypertension
Other related causes
Unrelated causes
All causes
Continuing
cigarette
smokersa
5-9
20 or more
Nonsmokers
1.28 (124)b
0.58 (48)
0.67 (5)
0.71 (5)
0.49 (7)c
0.81 (11)
0.18 (3)
0.06 (1)
0.19 (2)
0.30 (4)
0.07 (3)
0.05 (2)
4.72 (464)
0.65 (69)c
9.43 (865)
3.52 (28)
0.50 (4)
6.26 (49)
4.17 (61)
0.40 (6)
8.49 (120)
3.87 (59)
0.33 (5)
9.27 (136)
3.74 (40)
0.22 (2)
8.80 (105)
3.34 (113)
0.10 (3)
8.52 (315)
16.62 (1566)
10-19
11.62 (91) 14.25 (204) 13.60 (204) 13.38 (153) 12.09 (436)
Calculate by method B (see text) and therefore not in all cases exactly the same as the rates in previous Tables.
Including three deaths, also included under their primary cause, in which the specified disease was a contributory cause.
Including one death, also included under its primary cause, in which the specified disease was a contributory cause.
a
649
Figure 2. Death raw from lung cancer, standardized for disease will remain higher in the ex-smokers
age and amount smoked, among men continuing to compared with the smokers until, in the long
smoke cigarettes and men who had given up smoking
for different periods (men who had regularly smoked run, the change in habits can exert a beneficial
pipes or cigars as well as cigarettes excluded). The cor- effect on the disease in the survivors.
responding rate fgr nonsmokers was 0.07 per 1000.
For the unrelated diseases selective factors
1.5
0,
ea
1.0
bLu
5
k
n
J
4
3
Z
Z
0.5
6
1
I
10
15
20
25
30
650
Table 26. Mortality among cigarette smokers: effect of grouping men who had stopped smoking for less
than five years with continuing smokers.
Death rate per 1000
Cause
of
death
Lung cancer
Chronic bronchitis
Coronary disease without hypertension
Other related
causes
Unrelated causes
All causes
a
Ex-
smokers
of 5 years
or more
(C)
All other
smokers
(D)
C
as
percenttage of
(D)
All exsmokers
(A)
Continuing
smokers
(B)
0.35b
0.42
1.28.
0.58
27
72
0.2gb
0.37
1.24a
0.59
23
63
3.88
4.72
82
3.95
4.63
85
0.36
8.44
0.65b
9.44
55
89
0.33
8.86
0.64b
9.18
52
97
13.42
16.62
81
13.77
16.23
85
Table 27. Average age at starting to smoke among cigarette smokers divided by
continuity and amount smoked (standardized for age at inquiry).
~
~~
1-14
15-24
25 or more
Smoking at 1/11/51
Stopped less than 10 years
10 years or more
20.3
20.7
19.6
19.7
19.9
19.3
19.1
19.8
19.3
a The ages are somewhat higher than would be anticipated from present experience: they
refer, however, to a section of the population composed of university graduates, some of whom
completed their education before the first world war. The average is heavily weighted by the few
doctors who began to smoke after 25 years of age; in each group the commonest age at starting
was one to two years less than the average.
65 1
Table 28. Percentage of inhalers among cigarette smokers, divided by continuity and amount smoked (standardized for age at inquiry).
Percentage of inhalers when daily number
of cigarettes last smoked was:
Group
1-14
15-24
25 or more
Smoking at 1/11/51
Stopped less than 10 years
10 years or more
68
59
34
80
77
72
81
80
73
Table 29. Mortality from various causes in first and second quinquennia of observation:
continuing and ex-cigarette smokers
Death rate per 1000
Continuing smokers
Cause of
death
Lung cancera
Chronic bronchitis
Coronary disease without
hypertension
Other related causes
Unrelated causes
All causes
Ex-smokers
1st
5 years
2nd
5 years
1st
5 years
2nd
5 years
1.22
0.55
1.28
0.61
0.41
0.45
0.09
0.32
4.50
0.67
9.14
4.64
0.64
9.37
3.19
0.36
7.51
4.17
0.30
8.63
16.02
16.48
11.93
13.51
a Including four deaths among continuing cigarette smokers in which lung cancer was certified as a
contributory cause (method A).
652
Table 30. Inhaling habits by age, method of smoking, and amount smoked; (Male doctors aged 25 years
and over).
Percentage of inhalers among
Continuing cigarette
smokers smoking daily
Age
(years)
25 or
more
All
continuing
cigarette
smokers
cigarette
smokers
Ex-
Continuing
smokers,
cigarette
and
other
Continuing
Pipe
andor
cigara
smokers
82
76
62
57
49
74
60
47
36
30
12
10
7
5
4
1-14
15-24
25-34
35-44
45-54
55-64
65-74
75 or
more
85
78
65
54
49
93
89
78
70
66
95
81
73
59
90
85
75
66
58
38
47
46
41
41
26
All ages
67
81
81
76
63
47
89
Figures for the few cigar smokers are almost identical with those for pipe smokers.
Table 31. Mortality among continuing cigarette smokers according to inhaling habits,
compared with non-smokers (numbers of deaths in parentheses).
Death rate Der 1000 men Der veal
Continuing cigarette smokers
Cause of death
Lung cancer
Chronic bronchitis
Coronary disease without
hypertension
Other related causes
Unrelated causes
All causes
Inhalers
Non inhalers
Nonsmokers
1.88 (33)a
0.72 (12)
1.10 (11)
0.51 (7)
0.03 (1)
0.07 (1)
5.22 (113)
0.70 (17)
9.27 (171)
5.09 (60)
0.48 (4)
8.65 (99)
3.29 (50)
0.12 (2)
7.32 ( 1 12)
17.73 (345)
15.83 (181)
10.82 (166)
a Including one death, also included under its primary cause, in which lung cancer was certified as a
contributory cause.
1-14
15-24
25 or more
a
Noninhalers
1.59 (1 1)
1.95 (13)
2.16 (9p
0.00
0.86 (4)
2.98 (7)
The women doctors available for study number less than a fifth of the male doctors (18
percent) and fewer of them are in the age
groups of high mortality rate (40 percent were
under 35 years of age against 29 percent of
men, and 5 percent were 65 years or over
Other upper respiratory and upper digestive diseases,
pulmonary tuberculosis, peptic ulcer, and cirrhosis of the
liver and alcoholism.
653
654
Table 33. Standardized death rates from various causes among women doctors in relation to
smoking habits.
Death rate per 1000 women per year
Cause of
death
No.
deaths
Lung cancer
7
Other respiratory and upper
digestive cancer
2
Chronic bronchitis
2
Pulmonary tuberculosis
1
Cirrhosis of liver
4
and alcoholism
Coronary disease without
56
hypertension
Other causes (unrelated)
342
All causes
414
All
NonExwomen smokers Smokers smokers
Continuing
smokers
1-14
15-24 25 or more
0.08
0.03
0.13
0.08
0.15
0.41
0.22
0.04
0.04
0.03
0.04
0.07
0.08
0.11
0.11
0.27
0.27
0.22
0.05
0.06
0.04
0.07
0.16
0.03
0.10
1.14
6.72
1.25
6.60
0.99
6.60
1.26
5.94
0.84
6.93
0.41
6.30
1.25
8.60
2.62
5.70
8.10
7.95
7.94
7.44
8.17
6.76
10.89
8.76
a Only one woman said she smoked a pipe as well as cigarettes and no women smoked only pipes or cigars. For the purpose of
this analysis the small additional amount of pipe tobacco has been ignored.
each level of smoking the proportion of inhalers was approximately twice as great in men as
in women.
We have not sufficient data to take these differences into account, but it appears (from
Table 32) that the difference in inhaling will
lower the female mortality rate in relation to the
male, and it is reasonable to suppose that the
later average age at starting to smoke would
have a similar effect. These differences, too,
may affect not only the relative lung cancer
mortality rate in the two sexes, but also, to a
Table 34. Inhaling habits by age and amount smoked; women doctors aged 25
years and over.
Percentage of inhalers among continuing
smokers
Number of cigarettes smoked daily
~
Age (years)
1-14
15-24
25 or more
25-34
35-44
45-54
55-64
65-74
75 or more
73
63
I!?
26
17
17
91
71
59
34
26
18
54
47
All ages
44
55
58
22
50
29
65
36
64
25a
Age groups combined because number of women in one age group was less than 10.
Standardized for age.
GENERAL DISCUSSION
655
656
657
1-4
5-9
10-14
1
20-24
15-19 I
Control
patients
50
Cancer of
lung patients
Control
patients
50
29
59
46
61
46
60
72
81
72
62
82
71
658
Table 36. Death rate from related and unrelated cancers by smoking habits and age (numbers of deaths
in parentheses).
Death rate per 1000
Cancer of lung and upper
respiratory and digestive tracts
Age
(years)
25-29
30-34
35-39
40-44
45-49
50-54
55-59
60-64
65-69
70-74
75-79
80-84
85 and over
25 and overa
a
Standardized rates.
Other
cancers
Nonsmokers
Smokers
Nonsmokers
Smokers
0.03 (1)
0.08 (3)
0.18 (6)
0.69 (21)
1.32 (36)
2.52 (51)
2.95 (44)
2.94 (34)
4.06 (33)
4.52 (22)
3.35 (8)
0.19 (1)
0.20 (2)
0.57 (6)
0.24 (2)
0.82 ( 5 )
0.66 (3)
1.16 (4)
1.76 (4)
6.02 (9)
2.75 (3)
12.87 (11)
16.46 (10)
32.35 (11)
0.49 (5)
0.23 (6)
0.27 (10)
0.39 (15)
0.56 (19)
1.18 (36)
1.11 (30)
2.82 (57)
3.89 (58)
6.41 (74)
8.00 (65)
12.92 (63)
13.82 (33)
0.91 (259)
1.87 (71)
0.10 (1)
-
0.92 (1)
2.34 (2)
0.11 (4)
1.65 (471)
The association of chronic bronchitis and emphysema with smoking that we observe in doctors has been a conspicuous feature of other
prospective as well as retrospective studies and
also of carefully conducted surveys of various
populations. Thus, bringing together the results of seven prospective studies (including our
own), the Surgeon Generals Advisory Committee calculates that cigarette smokers died of
chronic bronchitis and emphysema six times
more frequently than nonsmokers (546 observed deaths to only 89 expected). In all but
one of these studies the excess is marked.
As an example of a retrospective study we
may cite the comparison made by Oswald,
Harold, and Martin (27) of 300 chronic bron-
659
660
In 89 instances (6 percent of the total) hyperThat being so, and seeing that chronic bronchitis is a common disease in this country, we tension was also mentioned on the certificate,
must naturally anticipate the clinical finding and these cases, examined separately, showed
that chronic bronchitis and cancer of the lung no relationship with smoking (Tables 18 and
will quite frequently occur together in the same 19).
For the remaining 1287 deaths the results are
person. It is, of course, possible that the chronic
bronchitis itself plays a part in producing cancer very similar to those we previously reported
of the lung (38), and figures published in a with less than half the number of cases (2), and,
retrospective inquiry by Campbell and Lee (39) although the differences in mortality between
somewhat support that view. For the prospec- cigarette smokers and nonsmokers (33 percent)
tive inquiry, however, what we need is to observe and between heavy and light cigarette smokers
the incidence of cancer of the lung in non- (14percent) are not great, the numbers are so
smokers suffering from chronic bronchitis; and large that there can be no doubt that the relawith its rarity in the absence of smoking the tionship with cigarette smoking is not due to
chance (P= 0.001 and P=O.O4, respectively).
problem is to obtain enough of them.
The results are, moreover, similar to those obtained in the six other large prospective studies
Pulmonary Tuberculosis
reviewed by the Surgeon Generals Advisory
The relationship between smoking and morCommittee, all of which show that the observed
tality from pulmonary tuberculosis is quite dismortality from coronary disease among cigatinct, though it is based, we should observe,
rette smokers is from 1.7 to 2.0 times the morupon only 42 deaths. The further 14 deaths in
tality expected from the experience of the nonwhich the presence of the disease was mensmokers. Similar results have also been obtioned as a contributory and not as the primary
tained in Albany and Framingham by Doyle,
cause of death reveal no such association. It is
Dawber, Kannel, Heslin, and Kahn (42), who
therefore possible that smoking may hinder refollowed 4 120 initially healthy men and found
covery from the infection or, as Lowe (40) has
that the incidence of myocardial infarction was
concluded from his retrospective survey, it may
2.4 times higher in cigarette smokers than in
conceivably re-aggravate a quiescent lesion. We
nonsmokers and 1.3 times higher in men smoksee nothing unlikely in that. On the other hand,
ing more than 20 cigarettes a day than in men
Kissen (41) suggests that anxiety is one reason
smoking less. In all these studies as well as in
for a poor prognosis in tuberculosis and also
our data the relationship is specific for cigarette
one reason for smoking. Without, therefore,
smokers, and no excess mortality has been obmore-and more precise-data on a disease so
served among pipe or cigar smokers.
influenced by social factors, we hesitate to conAccording to Hammond and Horn (4) and to
clude that a direct causal relationship of morBuechley, Drake, and Breslow (43),the relationtality with smoking does exist.
ship is more marked at younger ages than at
old. In our data this feature is distinct (Table 20)
Coronary Thrombosis
and the 38 deaths that occurred at the early age
Deaths attributed to coronary disease of 35-44 years show a relationship with smoking
amount, in our data, to 30 percent of the total that is only a little less close than that for chronic
deaths (1376 out of 4597). The great majority bronchitis. At these ages the mortality among
are described as due to coronary thrombosis or nonsmokers is extremely low (0.11 per 1000 per
myocardial infarction, but there is a small pro- year), so that although the rate of 0.61 among
portion described less specifically as due, for cigarette smokers is only 0.50 per 1000 higher
example, to coronary-atheroma, coronary is- the ratio of the rates is 5.5 to 1. Over the next
chemia, or angina pectoris with myocardial four decades the mortality of nonsmokers indegeneration. With so many deaths it was not creases to 10.83 and of smokers to 14.69; in
practicable to check the diagnoses as we did other words, though the ratio of the rates has
with cancer of the lung, and it is probable that become quite small, the excess mortality of 3.86
many of them, particularly at ages over 75 is seven times the excess at ages 35-44 years. At
years, were not, in fact, associated with an acute the oldest ages (75 years and over) the differepisode of thrombosis.
ence between smokers and nonsmokers disap-
661
662
Peptic Ulcer
Even including 15 deaths in which peptic ulcer was certified as being contributory though
not the underlying cause, the total number (54)
is too small to allow us to examine gastric and
duodenal ulcers separately. Classed together as
peptic ulcer, the results agree with those of 12
other studies in which the relationship between
smoking and peptic ulcer has been specially
examined (five retrospective, six prospective,
and one cross-sectional survey, Surgeon General's Advisory Committee). In all the retrospective studies the proportion of nonsmokers
was higher in the control group than in the
ulcer group, and in all the prospective studies
the mortality was higher in smokers than in
nonsmokers. In all studies in which gastric and
duodenal ulcers have been separated the relationship has been stronger with gastric ulcer
than with duodenal, and, when the type of
smoking has been examined, the relationship
has been closer with cigarette smoking than
with the smoking of pipes or cigars.
An unusual feature of the prospective studies
has been th'at the mortality has been maximum
in the moderate or light smokers rather than in
the heavy cigarette smokers. This is shown in
the present data (Table 22), and a comparable
result was obtained in the one large-scale retrospective study in which this aspect was examined in detail (53).
It is difficult to believe that these findings
mean that smoking is a direct cause of gastric
ulcer, since gastric ulcers have been decreasing
in incidence over the long period of time in
which cigarette smoking has been increasing.
Duodenal ulcers, admittedly, have increased in
With the large group of other and unrelated diseases the death rates of all the smoking
subgroups lie, with one exception, in the narrow range of 8 to 9 per 1000 (nonsmokers 8.53,
cigarette smokers 8.99, mixed smokers 8.00,
pipe and/or cigar smokers 8.06, ex-cigarette
smokers 8.02, 1-14 cigarettes daily 8.96, 15-24
cigarettes daily 8.66,25 + cigarettes daily 10.11;
Tables 23 and 24). This close similarity makes it
unlikely that we have overlooked any major relationship between smoking and mortality. The
only departure from the prevailing pattern is
the rather high death rate of 10.11 per 1000 in
the heaviest cigarette smokers, an excess of 19
percent over the rate for nonsmokers. This excess comes from only 7 of the 11 causes of death
shown in Tables 23 and 24, and could, we believe, be partly due not to smoking per se but to
other features of life, both environmental and
constitutional, which are correlated with it-for
example, the psychological and possibly physical characteristics of heavy smokers and their
habits of eating and drinking, etc. It is also
likely that some of the excess is due to errors in
the certification of the cause of death. Unless
100 percent accuracy was achieved in death certification-and we know that it is never so-the
unrelatedcauses must inevitably include some
deaths that should in fact have been classified as
related. The unrelated causes will thereby
share some of the excess mortality of heavy
cigarette smokers that is characteristic of the
related causes. In short, we conclude that smoking bears no direct or causal relationship to
some 60 percent of the total death rate.
663
That is not to say that our related causes necessarily include all the diseases in which smoking is of direct etiological importance. We have
no evidence on nonfatal diseases such as tobacco amblyopia, and very little for many rare
diseases or diseases with a low fatality rate.
There is, for example, strong evidence to inculpate smoking as a cause of thromboangiitis
obliterans (Buergers disease), but in our data
no deaths were attributed to it.
Paucity of cases may also be the explanation
of our failure to find any association of smoking
with cancer of the bladder. Four retrospective
studies have shown an increased mortality for
this site among cigarette smokers-of the order
of two to three times the rate in nonsmokersand a still higher rate of mortality in heavy
cigarette smokers (44,58-60).Moreover, similar
results have been obtained in all the other six
prospective studies (Surgeon Generals Advisory Committee). Our data are based on only
38 cases-six of which occurred in nonsmokers-and with this small number of deaths
it would not be surprising if by chance we had
failed to demonstrate a weak association with
cigarette smoking. But it does not appear that
there can be a strong relationship. (Since the
conclusion of the ten-year period of observation
a further 14 deaths with bladder cancer have
been reported-six among cigarette smokers,
eight among other smokers, and none in nonsmokers.)
In the studies of Hammond and Horn (4)
and Dorn (5) several other broad groups of
diseases have also been reported to show an
excess mortality with smoking. The excesses
have mostly been quite small, and, with the exception of a heterogeneous group of other circulatory diseases (Surgeon Generals Advisory
Committee), none have consistently shown a
mortality among cigarette smokers as much as
double that among nonsmokers. Such differences, we believe, could easily be due to the
same causes as we have suggested above may
account for the excess mortality from unrelated diseases in doctors who are heavy cigarette smokers. We may also note that the very
large populations studied by Hammond and
Horn (4) and by Dorn (5) have necessarily been
heterogeneous in many respects. Their results
may thereby be influenced to some extent by
the social distribution of smoking. Doctors, on
the other hand, are a more homogeneous
664
was 29 percent greater than that of nonsmokers; in men who were continuing to smoke
cigarettes at the start of the inquiry it was 35
percent greater. These figures, we pointed out,
do not necessarily mean that cigarette smoking
increases the death rate by such amounts, since
other explanations are possible. However, after
our separate and detailed examination of the
principal causes of death we have concluded
that cigarette smoking is likely to be responsible
for at least the greater part of the excess.
The results of our assessment are summarized in Table 37, where we divide the excess
death rate in smokers of different ages into
three categories. ( I ) The first category comprises the excess deaths from the unrelated
diseases and from cirrhosis of the liver and alcoholism. Conceivably some small part of this excess may be due to cigarette smoking, insofar as
it is due to errors in the diagnosis of the cause of
death; the greater part, however, is likely to be
either an artifact (due to the selective failure of
doctors to stop smoking when they develop a
serious illness; Table 25 and p. 649), or a secondary effect of the association of cigarette
smoking with some other factor which itself
predisposes to disease and death-for example,
heavy drinking or excessive anxiety. (2) The
second category includes deaths from cancers
of the upper respiratory and upper digestive
tracts, from peptic ulcer, and from pulmonary
tuberculosis. Some of these deaths are, we suggest, directly attributable to smoking, but it is
impossible to say how many, since the effect of
smoking is complicated by other factors, en-
Nonsmokers
Causes not
attributable
to smoking
Causes partly
attributable
to smoking
Causes
attributable
to smoking
Cigarette
smokers
all causes
25-44
45-54
55-64
65-74
75 +
1.12
4.12
12.08
30.56
114.29
0.48
1.02
2.37
6.72
- 11.64
0.04
0.22
0.92
0.82
2.09
0.39
2.25
5.33
10.87
6.64
2.04
7.62
20.70
48.96
111.37
12.06
0.93
0.40
2.93
16.32
All ages
standardized
Percentage of
nonsmokers
100
24
135
665
666
667
~~
~~
the proportion of children suffering from significant hearing loss had been underestimated
at the original examination. It was therefore
considered advisable to extend the inquiry to
early rubella children living in the rest of the
country. Reports for 237 children were received
and the results were included in the report of
Manson et al. ( I ) . This examination is designated No. 2 in Table I .
Finally, in order to discover how they developed in later childhood a third inquiry was carried out in 1962, when the children were between the ages of 8 and l l years. The present
paper reports the results of this examination,
which is designated No. 3 in Table 1, in relation
to findings of the previous examinations.
The medical officers of health were requested to provide the following particulars:
( 1 ) Any abnormality of the eyes: visual acuity
right and left, distant and near, without and
with spectacles if worn. (2) Any abnormality of
the ears: hearing right and left for quiet conversational voice without lip-reading at 3 and 10 ft.
(0.9 and 3 m.). Also a full pure-tone audiogram.
(3) Condition of the heart as reported by a cardiologist or pediatrician. (4) Intelligence quotient, naming testing scale used. (5) Assessment
of the childs emotional development and social
behavior. (6) Any other pathological condition
present. (7) Type of school attended.
A total of 227 completed reports were received. Of the remaining 32, one child had emigrated, the parents of five children refused to
participate, and 26 were untraced.
Owing to the wide geographical distribution
of cases and the large number of medical examiners concerned, it was inevitable that the reports received were not equally informative, but
the general standard of recording was high.
Assessment of an abnormality as major o r
minor on the evidence available sometimes re-
668
Sheridan
669
Table 1.a
Week
1
No. of
abnormalities
per child
No. of
children
No. with
abnormalities
11
Major 1
Minor 1
2
1
1
16
14
Major 4
3221
Minor 1
Major 1
Minor 3
3
221
11
Major 3
411
22
Minor 2
Major 5
33221
Minor 3
221
Major 4
4421
14
10
16
11
17
12
20
13
16
14
15
15
24
16
Totals
12
227
Major
Minor
Major
Minor
Major
Minor
Minor
Major
Minor
Major
Minor
Major
0
1
1
2
2
4
1 1
21
2111
1 1
0
5
4
2
1 1 1 1
11
32211
Minor 7
1111111
Major Minor 1
Major 0
Minor 3
Major 2
42
Minor 6
211111
Major Minor 1
Major 33
Minor 37
211
1
Major 15%
Minor 16%
Unilateral deafness 3
Unilateral deafness 3; undescended testicles 3. D.
vision 3. Deafness 3
D. vision 3; deafness 1; E.S.N. 1; undescended testicles
3. Unilateral amblyopia 3; asthma 3
Deafness 3; heart murmur 3. Unilateral amblyopia 3.
Deafness 2. Deafness 2. Heart murmur 3. Heart
murmur 3
Heart murmur 3
The figures 1, 2 and 3 in the last column refer to the number of the examination (see text).
670
though in compiling the tables only those children whose abnormalities were considered from
the information available to be certainly or possibly due to rubella have been included. Hence
1 1 children showing single abnormalities which
were either known or thought unlikely to be
connected with the maternal rubella infection
have been omitted from the tables-that is, two
cases of myopia developing in middle childhood, two cases of hearing loss associated with
active otitis media, six cases of uncomplicated
educational subnormality, and one case of paralytic poliomyelitis.
Table 1 summarizes according to the pregnancy week of rubella infection the number of
children at risk, the number showing abnormalities, and the number and nature of the
abnormalities occurring in each affected child.
Major Abnormalities.-The number and associations according to week of infection are
shown in Table 2.
Minor Abnormalities.-The associations and
number of cases are shown in Table 3.
Table 2.
No. of
abnormalities
Nature
Single
Week
9
9
15
5
6
8
8
11
12
4
8
5, 8
5, 12
9
12
15
4,5, 7, 8, 10,
10, 10, 12
3, 7, 12
9
10
Sheridan
Table 3.
No. of
abnormalities
Single
Nature
No. of
cases
1
4
1
1
1
1
2
7
5
1
10
EYE DEFECTS
671
672
HEART DEFECTS
The amount of information available regarding cardiac abnormalities was particularly satisfactory. T h e report of a cardiologist or
consultant pediatrician was available for all except about a dozen children whose attendance
at hospital or special clinic had been too difficult to arrange. In these cases the school medical officer had examined the child. Fourteen
cases were definitely diagnosed as congenital
lesions, the weeks of infection being from 1 to
12. In 10 cases the cardiac lesion was associated
with another abnormality; in four it was the sole
abnormality noted. Eighteen heart murmurs
described as innocent or of no significance
were recorded (16 of them for the first time at
the third examination), but it is noteworthy that
nine of these (seven of the 16) were associated
with another abnormality.
OTHER CONDITIONS
INTELLIGENCE
The educational placement of the 227 children is shown in Table 5. Thus 206 (92 percent)
of the children were attending ordinary schools.
Of these seven (3 percent) are noted as having
special provision such as hearing-aids, speech
therapy, remedial teaching, etc. Twenty chilSOCIAL ADJUSTMENT
dren were attending special schools or classes,
It has been suggested that rubella children including the child whose handicap was the reoften show emotional instability and difficult sult of paralytic poliomyelitis. The majority of
behavior, but although the information was spe- those in special schools were severely deaf. The
cifically requested there was little supportive ev- child receiving home tuition was blind and malidence in the reports. Twelve children were adjusted. He had also been operated upon for a
variously noted as shy, immature, lacking congenital heart lesion, a residual murmur
in concentration, or liableto outbursts of tem- being present. The child who was in a hospital
per,but only one, a blind child, was reported as for the mentally subnormal had multiple handicaps.
psychologically difficult.
standardization group.
25
20
15
15'
[r
10
50
I.Q.
Week
I.Q.
No. of
cases
Given
Mean
Range
~
10
11
12
13
14
15
16
a
9
4
8
11
13
5
10
19
13
11
17
15
14
14
18
10
11
6
8
16
14
5
11
22
14
16
17
20
16
15
24
12
106.8.
102
116
110
103
101
112
103
108
95
107
107
107
110
103
109
116
Additional assessments
~~~
88-141
83-131
96- 139
80- 128
'71-150
80-130
80-142
84- 134
63-1 10
82-160
83-140
70-143
80-133
83-134
70-145
93-143
673
5 average
1 average
1 average
2 average, 1 borderline
1 average
4 average, 1 below average
2 average, 3 below average
1 above average, 1 average
1 average
5 average, 1 below average
2 average
674
Ordinary school
Ordinary school with special help
Special schools or classes, etc.:
For deaf or partially deaf
For partially sighted
For speech defectives
For E.S.N.
Home tuition
Hospital for mentally subnormal
[P.H. school-paralytic polio (not rubella)
199
7
15
1
1
1
1
1
1
SEASONAL DISTRIBUTION
It has often been suggested that a larger proportion of handicapped children are born in
the months of winter and early spring than in
the summer. The months of birth of the 259
early rubella children, with and without abnormalities, are given in Table 7.
Although the larger number of early
rubella children were born in the months October to February inclusive, reflecting the usual
springtime incidence of rubella epidemics,
there was no significant difference in the proportion of handicapped and non-handicapped
children born at any season of the year.
BIRTH WEIGHTS
MISSING CASES
Month
Normal
Abnormal
Total
January
February
March
April
May
June
July
August
September
October
November
December
21
18
6
5
1
2
4
6
21
34
35
32
15
3
4
1
1
1
2
5
6
6
18
12
36
21
10
6
2
3
6
11
27
40
53
44
Total
185
74
259
Ib.
Range
g.
02.
~
Normal (157)
With minor abnormalities (37)
With major abnormalities (3P)
With multiple major abnormalities (19)
a
7 2
6 10
6 4%
5 14
3230
3005
2850
2665
Ib.
02.
Ib.
4 10 3 13 3 134 6-
02.
~~
10
10
8
8
2
6
8
4
R.
2100-4590
1730-4705
1730-3855
1985-3740
Sheridan
SUMMARY
675
~~
Since its establishment in 1946 the Commission has maintained a continuous surveillance
of cases of leukemia occurring in Hiroshima
and Nagasaki. In addition, parallel efforts are
made to obtain adequate data for evaluating
cases of leukemia occurring in survivors of the
bombings living outside these cities. Each case
of leukemia is carefully investigated and reviewed at frequent intervals by the current investigative and consultant staff. The type of
leukemia is decided after review of all available
data and material; the certainty of diagnosis,
expressed as definite, probable, or possible, is
based on established criteria. This program of
case finding and review is described elsewhere
(30).Definite leukemia is defined as follows: cases
with good clinical information and a history of
well studied and convincing morphologic documentation of the disease by earlier investigators
of the Commission even though the material
currently is not at hand; cases with good clinical
and hematologic material that provide convincing evidence for the diagnosis of leukemia; and
cases with morphologic confirmation and a clinical history not inconsistent with leukemia, even
though the clinical information is scanty.
The definition of probable leukemia is as follows: cases with convincing clinical information
for the diagnosis of leukemia but with little or
no morphologic material; cases with inadequate
clinical information but with good morphologic
material consistent with leukemia; cases clinically consistent with the diagnosis and with adequate documentation of morphologic material
studied by earlier physicians of the Commission, and cases with adequate clinical and morphologic material in which it is only remotely
possible that some other clinical syndrome
caused the hematologic abnormality.
The following characteristics are included in
the definition of possible leukemia: cases with
Bazzozero et al.
677
~~
Hiroshima no.
of cases
Nagasaki no.
of cases
Total no.
of cases
Percentage
562
63
397
66
959
129
87.3
11.7
19
1.o
Acute
Chronic
Total
City
Hiroshima
Nagasaki
Hiroshima
Nagasaki
1501-10 000
0-1500 meters
meters
no. of cases
no. of cases
74
30
46
10
66
62
25
13
160
166
>10 000
meters
Patients born
after
Aug. 1945
no. of cases
no. of cases
171
154
52
35
122
89
6
4
412
22 1
678
40
10000m
- - - 0-10000m
__-----0-1500m
412
326
YEAR OF ONSET
RESULTS
Figure 2. Definite and probable cases of leukemia among persons within 1500
meters of the hypocenter, Hiroshima and Nagasaki.
24
Total
___-.
Acute
---
YEAR OF ONSET
Chronic
160
104
56
679
Bizzozero et al.
~~
Figure 3. Definite and probable casea of leukemia among persons 1500 to 10 000
meters from the hypocenter, Hiroshima and Nagasnki.
15
- --___
U 12
---
Total
166
Acute
128
Chronic
38
z 93
1945
1950
1955
YEAR OF ONSET
1960
%-
1965
1945
1950
1955
YEAR OF ONSET
*--.----..
1960
1965
680
The Life Span Study (36)sample of the Commission (referred to by Heyssel et al.-24-as
The chief value in studying a defined population is that rates for diseases causing death can
be more accurately estimated. Any effort to calculate rates for leukemia for both cities immediately founders on the formidable problem of
establishing a population denominator for all
Type of leukemia
0-1500meters
no. of cases
Acute:
Granulocytic
Lymphocytic
Monocytic
M yelomonocytic
Stem cell
Unknown
Other
Chronic:
Granulocytic
Ly mphocytic
Other
Total
Total
32
35
6
6
10
9
6
30.7
33.7
5.8
5.8
9.6
8.7
5.8
104
55
0
1
56
no. of c a m
44
22
14
17
8
18
5
%
34.4
17.2
10.9
13.3
6.2
14.1
3.9
128
98.2
-
1.8
34
4
0
38
89.5
10.5
-
Bizzozero et al.
681
Type of leukemia
Acute:
Granulocytic
Lyrnphocytic
Monocytic
Myelomonocytic
Stem cell
Unknown
Other
Total
Chronic:
Granulocytic
Lyrnphocytic
Other
Total
0-1500 meters
no. of cases
13
12
4
3
4
1
1
38
22
0
1
23
10-
a.
6-
34.2
31.6
10.5
7.9
10.5
2.6
2.6
4
5
1
4
2
1
22.2
27.8
5.5
22.2
11.2
5.5
5.5
18
95.7
4.3
100.0
9
0
0
9
-0-1500m
- - Chronlc
. _ _ _Acute
_
61
38
23
4.
2-
$ 0
2
I> a
0-1500m
61
682
1946-49
no. of
cases
1950-54
ratel
100 000
no. of
cases
no. of
ratel
100 000
3
6.22
29
48.12
Exposure, 1501-10 000 meters from hypocenter
0
4
1.32
Acute
Chronic
3
1.24
2
0.66
Total
3
1.24
6
1.98
Total
time of the bombings (Figure 6) invites particular attention since it is clearly excessive and contrasts with a null incidence in the other exposure group (Figure 6). The contrast between
groups is not altered appreciably if the group
up to 1500 meters is extended to less than 30
years at the time of the bombings. No typespecific leukemia tables are available for Japan,
1960.1964
1955-1959
no. of
ratel
ratel
I00 000
cases
100 000
15
6
26.24
10.50
5
3
9.24
9.54
21
36.74
14.78
10
2
3.46
0.69
4
2
1.46
0.73
12
4.15
2.20
cases
Figure 6. Average annual rate, per 100 000 population, of cases of acute and
chronic leukemia.
Age ATB
8.m.
All Ages
I
-
0-29
30+
. . .,, *
I
Age A T 8
I..-
I -
All Ages
0-14
15+
! I . 1 .I
A B C
0-1500 m
1501-10 000 rn
0-1500 m
ACUTE
A 1946-49
6:1950-54
C: 1955-59
I.
1501-10 000 m
CHRONIC
D: 1960-64
Bizzozero et al.
Appearance Time
This term is used to express in years the interval between the age at the time of the bombings and the age at onset-that is, between exposure to radiation and the onset of leukemia
(Table 6).
Acute Leukemia. All persons in whom acute
leukemia developed who were exposed up to
1500 meters from the hypocenter and who were
less than 30 years of age at the time of the
bombings had a significantly shorter appearance time (8.6 years for those under 15 and 9.4
for those 15 to 29 years of age) than persons
with the same type of leukemia in the sample
age group, but exposed between 1501 and
10 000 meters (11.6 years under 15 and 11.6
years 15 to 29 years of age). These differences
are statistically significant ( p less than 0.025) but
are not seen in persons who were 30 years of
age and over at the time of the bombings (13
years for those exposed up to 1500 meters and
11.6 years for those 1501 to 10 000 meters from
the hypocenter).
Chronic Leukemia. All persons exposed up to
1500 meters in whom chronic leukemia developed experienced a shorter appearance time in
all age groups than those with chronic leukemia
exposed at greater distances. The probability
values for both these groups are statistically significant ( p less than 0.005).
The length of appearance time in the group
exposed 1501 to 10 000 meters from the hypocenter for both types of leukemia and for all age
groups (mean, 11.4 years) is remarkably consistent.
DISCUSSION
0-1500 meters
p Value
no.
uartance
yr.
no.
uanance
Acute leukemia:
0- 14
8.6
15-29
9.4
13.0
30 +
38
40
26
19.9
22.1
14.3
11.6
11.6
11.6
36
29
63
23.7
17.1
16.4
<0.005
<0.025
N.S.
Chronic leukemia:
0-29
8.1
7.2
50 +
30
26
17.3
10.5
11.7
10.3
16
22
18.3
17.8
<0.005
<0.005
YV.
YT.
683
684
Bizzozero et al.
counterpart of this observation of a possible
genetic effect has been provided by MacMahon
and Levy (49) in their demonstration of an excessive rate of leukemia in the twins of patients
with leukemia.
The rates of leukemia in the Life Span Study
for the two exposure categories and for the periods 1946-1949, 1950-1954, 1955-1959 and
1960-1964, indicating an absolute increase in
incidence for those persons exposed within
1500 meters of the hypocenter and the demonstration of a shorter appearance time and, subsequently, earlier development of leukemia for
this group, can be construed to mean that radiation accelerated the development of leukemia.
The declining rates over time suggest either
that the subgroup of persons with the genetic
and environmental mosaicism that prefigured
the development of leukemia has been consumed, or that a radiation-produced loss of
specificity at a cellular level promoting an undirected cellular proliferation has been repaired.
SUMMARY
The observation of the occurrence of leukemia in Hiroshima and Nagasaki for the period
1946-1964 are presented.
During this period leukemia developed excessively in persons exposed within 1500 meters
from the hypocenter.
A decreased appearance time, defined as the
interval between exposure to radiation and the
clinical or symptomatic onset of leukemia, was
noted in patients exposed within 1500 meters.
Attention is invited to the role of age as a
covariable to distance in the appearance time of
acute leukemia developing in such persons.
In this group chronic granulocytic leukemia
developed in persons up to 29 years of age at
the time of the bombings in the period
1946-1955 at a disproportionate rate as compared to the distally exposed and the pattern of
spontaneous leukemia for this age group.
We are indebted to o u r predecessors at the
Atomic Bomb Casualty Commission, whose
careful labor made this study possible, and especially to the present consultants, Dr. S.C.
Finch, of Yale University, Professor M. TOmonaga, of Nagasaki University, and Professor
G. Wakisaka, of Kyoto University, and the pres-
685
686
Director, Medical Research Councils Social Medicine Research Unit, The London Hospital, London E.l.
Member, Medical Research Councils Social Medicine Research Unit, The London Hospital, London E.l.
Deputy Chief Medical Officer, London Transport Board,
London, S.W.I.
Age
Six of the 128 men who were in their forties
when first examined newly developed the disease (an incidence rate during the period of
687
688
Table 1. Incidence of ischemic heart disease in sample of London busmen during five years,
by age at initial examination.8
30-39
40-49
50-59
60-64
65-69
Total
a
No. of men
examined
Incidence
rate per
100 men
in 5 years
No. of manyears of
observation
Incidence
rate per 100
man-years of
observation
1
6
24
13
3
32
128
300
170
37
(3.1)
4.7
8.0
7.6
(8.1)
175
689
1461
917
207
(0.6)
0.9
1.6
1.4
(1.4)
47
667
7.0
3449
1.4
No. of
new cases
in 5 years
Figures in parentheses are incidence rates calculated from very small numbers of cases. All men were born in the British
Isles.
Occupation
Blood Pressure
Table 2. Incidence of ischemic heart disease in conductors and drivers during five years,
by age at initial examination.
Conductors
Age
(Yr.)
30-39
40-49
50-59
60-69
Total
Drivers
No. of
new cases
in 5 years
No. of
men
examined
Incidence
rate per
100 men
in 5 years
No. of
new cases
in 5 years
No. of
men
examined
Incidence
rate per
100 men
in 5 years
0
1
6
5
7
62
117
68
(1.6)
5.1
7.4
1
5
18
11
25
66
183
139
(4.0)
7.6
9.8
7.9
12
254
4.7
35
413
8.5a
a When standardized on the age structure of the conductors this was 8.6 per 100 men in 5 years.
, occupational difference.
Allowing for consistency by age, ~ < 0 . 0 5 for
The rates per 100 man-years of observation were 0.9 in the conductors and 1.7 in the drivers, the drivers rate again being
standardized for age.
689
Morris et al.
the number of new cases from high to low quarters being: 19, 16,5,and 7. B.P. was also taken at
rest and after mild exercise; simple combinations of these various pressures, for example by
adding the three systolic or diastolic readings
together, have not been more powerful in identifying future ischemic heart disease than the
casual systolic (or diastolic) reading alone.
Very few of the men in the high quarters of
S.B.P. distributions initially had any symptoms
referable (even with hindsight) to the cardiovascular system and, of course, none showed evidence of ischemic heart disease. No systematic
search was made for secondary hypertension
other than analysis for proteinuria.
Blood Pressure and Occupation
There are two connections between s.n.P., occupation, and ischemic heart disease. Drivers
aged over 50 have higher S.B.P. than conductors-higher average pressures and more men
with high pressures (Table 3). This is a factor,
therefore, in the higher incidence of the disease
among drivers. Secondly, there is an indication
that at the same level of S.B.P. drivers have a
higher incidence than conductors (Table 4).
Incidence in Relation to Initial Hood-lipid Levels
Table 3. Incidence of ischemic heart disease in conductors and drivers during five years,
by age and casual systolic blood-pressure level at initial examination.
No. of new cases among conductors
aged:
Quarter
40-49
50-59
60-69
40-69
40-49
50-59
60-69
All busmen
40-69
No.
~
High
Second
Third
Low
1
0
0
0
2
2
Total
2
0
1
2
5
2
3
2
12
No. of
cases
~
1
1
1
10
3
3
7
2
2
18
6
6
23
8
168
165
164
167
18
11
34
664b
47
1Oa
The quartile points (mm. Hg) for each column were, from high to low: (1) 153, 137, 125; (2) 161, 143, 131; (3) 165, 153, 143;
(4) 153, 135, 125; (5) 171, 153, 137; and (6) 175, 159, 145.
a
1 case was aged 39, a driver, systolic B.P. 128 mm. Hg.
The casual pressure of 3 men was not recorded.
690
Figure 1. Incidence of ischemic heart disease in sample of London busmen during 5 years by findings at initial
examination: ( 0 )By casual systolic B.P. (664 men); P < 0,001. (b)By total plasma cholesterol (607 men); P < 0.001.
(c) BYstature (663 men); P < 0.1. (d) By skim-foldthickness at right suprailiac crest (658 men); P < 0.2. Probability
levels in figures 1, 3, and 4 have been calculated using a test for trend in proportions (4).
Lu
LL
150-166
18.5-60.0
........
238-269
.....
171-174
[
I
Q
3
84-132
168-170
......
151-167
Systolic
blood pressure
(mmHg)
Plasma
cholesterol
(mg per 100 ml)
B.P.
(mm. Hg)
168-254
150-166
134-148
84-132
Conductors
8.5
3.9
Drivers
15
5.6
Total busmen
:\!}
13b
4.9b
664 men.
Standardization by indirect method.
Morbidity ratios are 190 2 40 and 692 14.
There was 1 case among the 7 conductors whose systolic
B.P. was 200 mrn. Hg and over, and 2 among the 16 such
drivers.
nal plasma was determined by the LiebermannBurchard reaction as described by Dodds and
Mills (6).Total plasma-cholesterol in 607 menmeasured regardless of season of year, time of
day, or time of last meal-is a good predictor
(Figure lb), with 19 new cases in the high
quarter and 5 in the low. p-lipoprotein cholesterol seems to be the effective fraction-19, 11,
6, and 3 new cases occurring in the quarters
from high to low-but the number of estima-
..........
..........
6.5-1 1.9
......
..........
..........
.........
Suprailiac
Skin fold
(mm)
tions was less again (544 men). With a-lipoprotein cholesterol the figures are 6 cases in the
high quarter, then 10, 14, and 9 in the other
quarters. The very low density lipoproteins,
S,20-400, that carry much of the plasma-triglyceride (7) were also estimated. They showed
an inconsistent trend with incidence of ischemic
heart disease; we will report when more cases
have accumulated.
Drivers have substantially higher blood-lipid
levels than do conductors of the same age, and
this is another factor in the higher incidence of
ischemic heart disease in the drivers.
These connections between age, occupation,
s.B.P., and blood-lipid levels and the subsequent
incidence of ischemic heart disease are the main
findings; we report briefly on five others, none
of them statistically significant.
Molwis et al.
Stature. The incidence of ischemic heart disease is greatest among the shortest men (Figure
Ic), and the same is true measuring sittingheight. Gertler and White (9) have noted this
association. There is no excess of high s.n.P.
among the 160 shortest men, but the 40 hypertensive (as defined) in this group have a particularly high incidence-with 10 new cases, compared to 7 in the other 120 men of similar
height. The short hypertensive men who developed the disease were not particularly obese,
nor did they have especially high blood-cholesterol levels.
691
Other Factors. Many other factors were examined and found to have no or negligible relationship, taken singly or in various combinations, to the development of ischemic heart
disease. These include body-weight, relative
weight, ponderal index, arm circumference,
waist, biacromial diameter, chest expansion, hemoglobin level, civil state, number of siblings
and children, social class of father, and alcohol
consumption. We will not mention these further.
692
..........
5 .........
.........
....
No. of new
cases
in 5 yr.
No. of
men
examined
Proportion
of all
of all
new
men
C
z examined
Individual
risk
552
h
.GZ
53 -2
L 5
0)
34
240
77 %
40%
1 in 7
10
364
23 %
60 O/O
1 in 36
5 - rt
I
each man. In this instance, the score is a measure in terms of the factors studied of an individuals risk of developing ischemic heart disease, and is calculated as a weighted addition of
their values. The weights are determined to
produce as small a range of scores as possible
within each of the two groups (with or without
the disease) and as little overlap as possible between the groups. The efficiency of the discriminant score for selecting men at greater or lesser
risk will be evident to the degree that the new
cases have higher values than the remainder; its
efficiency diminishes the more the scores for
the two groups overlap. The more efficient the
score ixi this sense, the better will it be as a
predictor of the small group of men (a),who are
at a particularly high risk (b),and among whom
the majority of the future cases in the population will occur (c). At present in the U.K., about
20 percent of men, we may estimate, develop
clinical ischemic heart disease during middle
age; in these terms the group is small. (a), (b),
and ( c ) are related, each is a function of the
other two (the argument is briefly set out in an
appendix to this article).
Theoretical requirements of the mathematical model underlying the technique of dis-
Morris et al.
mode of life;
(2) the precursor pathology, as causes begin to
show effects in disturbed function and
structure, without as yet any evidence of
the disease; then comes the
(3) early ancidence as disease, possibly reversible, appears; and so on to established
and advanced disease, etc.
Several postulated causes of ischemic heart
disease are included in our study, and two evident precursors-high levels of B.P. and plasmacholesterol-the minor electrocardiographic
abnormalities may be regarded as earZy disease.
So we organized these factors to give some picture of causes (Xl-X,), precursors (X, and XJ,
and early disease (Xg).
We consider the causes first. By obtaining a
weighted score (as described) from factors X I X,, ere sought to combine their individual predictive powers in an effective way; at the same
time the method indicates the extent to which
the factors are contributing independentlythis is essential in view of known interrelationships, for example of age, obesity, and
occupation. We calculated a score for each of
the 593 men, divided them into quarters of
about 150 and identified the number of new
cases occurring in each. Figure 3a shows that
these causes by themselves go some way to building a predictive device: the individual risk of a
man with a high score is several times greater
than that of a man with a low score. Occupation
is the most powerful of these causes; stature and
obesity contribute little to the prediction in the
presence of the other factors.
Next we produced a score, using only the
measurements of S.B.P. and plasma-cholesterol.
The equation shows the weights attaching to the
two factors and how the score is obtained:
..*......*
.......... .....*.*..
.......*. *.*.*.
.
b
..*.*
Z
w
0
U
W
I-
11I
a
3
LOW
1 ....
*..*.**.
I
CAUSES
693
PRECURSORS
CAUSES,
PRECURSORS,
AND EARLY
DISEASE
694
them developed ischemic heart disease. Dividing the men into five groups (instead of four)
for all these factors, yields 21, 10, 6, 6, and 0
new cases in each. If we extend this process to
six groups of about a hundred men, there are
18 new cases in the group with highest scores,
then 9, 7, 5, 4, and 0 in the remainder. Early
disease (X,) contributes very little.
One point is clear: the levels of S.B.P. and
plasma-cholesterol predominate, and adding
the other factors achieves little net improvement
in prediction. Comparing Figures 3b and 3c, it
seems that these causes (occupation, obesity, and
the rest), operate largely through these precursors. Moreover, comparing Figures 3a and 3b,
and in light of their many known and demonstrated connections, it seems that much of the
predictive power of these precursors is due to
these causes.
Finally, for the moment, we contrast in profile the sizable groups of men placed (Figure
3c) in opposite quarters of risk by scores based
on all the factors. In terms of these factors, and
of the occurrence of new cases of ischemic heart
disease, the men in the low and high quarters are very different (Table 5). In detail, for
instance, 37 of the 46 busmen having a casual
systolic B.P. reading less than 120 mm Hg are in
the low group as are 63 of the 88 men with
cholesterol levels under 200 mg per 100 ml.
Table 5. Profile of low and high quarters of the sample of London busmen for risk of developing
ischemic heart disease during five years.
Quarter of men
Finding at initial
examination
X1
x5 {ot;:rs
Non-cigarette smokers
x6 {s mokers
X7
X8
X,
At low risk
( n = 150)
52
71%
29%
173
11
59%
41%
37%
63%
131
205
96%
4%
2
(1.3)
1 in 75
At high risk
(n = 147)
58
37%
63%
170
17
17%
83%
15%
85%
176
272
77%
23%
25
17
1 in 6
M o w k et al.
There are many older drivers, smokers, men
who are fat and have a bad family history in the
high group; and there are many thin conductors with a good family history and an unblemished electrocardiograph in the low
group. Individually, there are interesting placements of men with high S.B.P. or plasma-cholesterol in the low group (none of these men
have developed ischemic heart disease) through
the compensating advantages in other factors.
In this way we can move towards a picture of the
individual who is susceptible to ischemic heart
disease and, equally interesting though not
elaborated here, the individual least likely to
develop the disease.
DISCUSSION
695
696
and its incidence is strongly connected with levels of S.B.P. if not of blood-lipids. Figure 4 shows
the five-year incidence of recognized new major
vascular disease in these busmen relative to initial systolic B.P. Since clinical ischemic heart disease is so common-affecting perhaps 20 percent of middle-aged men-it
has to be
recognized that individual approaches to its
prevention are likely to have only a limited
place. The approach to all who are clearly susceptible would involve more than 20 percent of
middle-aged men and so, with foreseeable
methods of prevention, would be impracticable.
It seems likely that the eventual approach will
be to combine mass campaigns with personal
treatment for special individuals-e.g., those
who can be identified as early candidates for the
disease, or the most severe disease, for example.
So the grand question is whether reduction
of symptomless hypertension and hypercholesterolemia from early middle age will re-
Figure 4. Incidence of e a r d l o v d n r disease in London busmen during 5 years by casual systolic B.P. level
at initial examination. (664 men; P < 0.oooOl.) Agestandardized incidence rates per 100 men in 5 years in
the four quartera are 22.3, 7.6, 9.2, and 5.2 from
high to 6610w. Two men developed two conditions,
but are counted only once.
SYSTOLIC
BLOOD-PRESSURE
OnmHgi
..........
.
.......
o.........
168-254
..0000000
00
O.OO0000
.O
150-166
0
W
U
0.0
+
a
4
134- 148
84- 132
0=
0=
=
U =
Ischemic heart-disease
Cerebrovasculardisease
Peripheral vascular disease
Hypertensive heart-disease
Morris et al.
duce the subsequent incidence of ischemic
heart disease. There are signs of acceptable
means of lowering high blood-cholesterol, and a
trial is under way to discover whether there is
any advantage for an otherwise healthy population in doing so. To assess the prophylactic
value of long-term lowering of mild and moderate hypertension by large-scale experiment is
now an urgent task for preventive medicine,
though it is doubtful if any of the available
agents are suitable. The tantalizing situation today in ischemic heart disease is that the main
causes probably are known, and the main mechanisms also. But how to prove it, and benefit
from the knowledge? Preventive studies on an
adequate scale if only they can be organizedand they give rise to quite new issues for medicine and the public-could give the answer. In a
situation where animal experiment is progressing so slowly, and where multiple interrelated
factors render human observation so difficult to
interpret, no other method is in sight for getting an answer. Such studies in the free-living
population could have tremendous theoretical
importance in determining whether the evident
precursors are in fact important mechanisms of
the disease, in distinguishing causes from mere
association, and in assessing the roles of inheritance, experience, and mode of life in the modern epidemic of ischemic heart disease.
SUMMARY
A sample of 667 middle-aged London busmen who showed no evidence of ischemic heart
disease when first examined were followed u p
for five years. During this period 47 of them
developed the disease-an incidence rate of
seven percent. The incidence is higher in later
than in early middle age, in men with a bad
family history of parental death, in drivers than
conductors, in cigarette smokers, in the more
obese, and in the shortest men.
Levels of casual systolic blood-pressure (s.B.P.)
and of plasma-cholesterol are, however, the predominant predictive factors. Three-quarters of
the new cases were among the men who were in
the top quarter of the distribution of either
S.B.P. or plasma-cholesterol. The risk, during
five years, for such an individual of developing
the disease was about one in seven-several
times the risk for the remainder.
697
APPENDIX
Also, the individual risk for a man in the highrisk group of developing the disease is 1 in Wr.
At any point of time during the follow-up
period the incidence of new disease in the total
population will be known. Thus the incidence
698
The concept of risk factors evolved from prospective epidemiological studies in the 1950s
relating personal characteristics of participants
to subsequent incidence of CHD. Risk factors
are based only on associations demonstrated in
epidemiological studies; hence they may be directly causative, secondary manifestations of
more basic underlying metabolic abnormalities
or early symptoms of the disease. Also, risk can
be expressed in terms of absolute risk, relative
risk ratios, or attributable risks. A high relative
699
700
Table 1. Coefficients for regression. of specified cardiovascular events on cardiovascular risk factors, for
men and women, ages 45 to 74, in the Framingham Study, over 20-year follow-up.
~
Risk factors
~~
Cigarettes
ECG-LVH
Serum cholesterol
Diabetes
Hypertension
Heart rate
Relative weight
Vital capacity
Proteinuria
Cardiovascular
disease
Cardiac failure
Men
Men
Women
Women
Intermittent
claudication
Men
Women
Brain infarction
Men
Women
Coronary
disease
Men
Women
.198
.02gh
.087b
.219
.394
,222
.255
.161b
.236
,160
.192
.I96
.414
.509
616
,144
.05Sb
,293
,219
.156
.227
-.179 -.330
-- ,392
,090 .062b
.195
.372
.217b
,316
,158
,287
.133b
.271
.244
,325
,316
.358
,362
.529
.504
.17Od -.060b
,201
.193b
.387 -.192
- ,667
-.264 -.469
.02P
.190b .042b
.154b .104b
,244
.682
,717
,305
.105b
.130b ,359
-.214b -.264
-
,168 -.023b
,212
,174
.255
,314
,117
.201
,338
.483
,125
,053'~
,206
,223
- ,116 -.292
,094
,120
a These numbers represent the strength of the relationships between the risk factor and the cardiovascular event. T h e higher
the number, the stronger is the relationship.
b Not statistically significant at P = <0.05.
- = Insufficient data.
Atherogenic Traits
The atherogenic risk attributes include the
blood lipids, blood pressure (BP), and clinical
diabetes. These factors jointly play a large role
in determining the pace of atherogenesis (Figure 1). Each has been repeatedly shown to be
independently related to the rate of development of clinical CHD and to the extent of occlusive coronary atherosclerosis as shown on angiography. Gout or hyperuricemia may be an
additional factor.
Blood Lipids
The evidence incriminating the serum total
cholesterol in the evolution of CHD is extensive
and unequivocal (3,4). Atherosclerosis has been
produced in animals by hypercholesterolemiainducing diets; the lesions have been shown to
contain the lipid, derived from plasma; CHD
cases have been shown to have higher choles-
Kannel
70 1
Figure 1. The risk of coronary heart disease according to the number of risk
factors. at initial exam In men aged 30 to 49 over 20-year follow-up, in the
Framingham Study.
158
81
one
Any
two
Three or
more
a Risk Factors:
Hypertension
Cholesterol F 250
Diabetes
Cigarettes F Pkg Day
ECG-ABN
702
SBP:
ChoP
185
210
235
260
285
310
335
2 3 3 4 5 6
3 3 4 5 7 8
4 4 6 7 9 1 1
5 6 7 9 11 14
6 8 9 12 14 17
8 10 12 15 18 22
10 13 15 19 23 27
2 3 3 4 5 7 Glucose intolerance
3 4 4 6 7 9 present
4 5 6 7 911
5 6 8 91214
6 8 10 12 15 18
8 10 13 15 19 22
11 13 16 19 23 28
185
210
235
260
285
310
335
4
5
6
8
11
13
17
2
3
4
5
6
235
260
285
3 10
335
Chol
185
210
235
260
285
310
335
185
SBP:
2 2 3 4 Glucose intolerance
2 3 4 5 absent
3 4 5 6
4 5 7 8
6 7 911
7 91114
9 12 14 17
2 10
2
2
3
3
4
6
8
5
6
8
10
13
16
21
6
7
10
12
16
20
25
7
9
12
15
19
24
30
9 1 1
11 14
15 18
18 22
23 28
29 34
35 40
8
1 0
1 3
17
21
26
32
1 3
17
21
27
33
39
46
SBP = 195mmHQ
LDL = 220 mgfdl
0
0
120 mmH(
100 rngldl
0
25
45
65
HDL CHOLESTEROL
(mg/dl)
85
Kannel
703
Figure 3. The 4-year risk of coronary heart disease according to the ratio of
cholesterol lipoprotein fractions, in the Framingham Study, exam 11.
MEN 50-79
300
300
-
250
2
+
d
Y
u,
a
>
t
E
rn
a
CLDUGHDL
GTOTAUCHDL
1.o
3.4
1
3
5.0
.6
understood and are under intensive investigation, although it is known that excessive calories
definitely lower this lipoprotein. A high intake
of complex carbohydrates is definitely associated with a low CHD mortality, particularly
when they are eaten in place of saturated fats.
Calorie balance has been clearly shown to improve the LDL/HDL cholesterol ratio. Conclusive demonstration of the relationship of diet
to CHD or hyperlipidemia within typical American population samples is difficult because of
the generally high consumption of the incriminated nutrients and inadequate methods of
quantifying both the intakes and energy expenditures of individuals (11-13).
Serum triglyceride values and the very lowdensity lipoproteins (VLDL), which transport
the triglycerides, are also positively associated
with CHD risk, but most prospective studies
indicate that this excess risk is dependent on
coexisting low HDL values, high LDL cholesterol, obesity, and impaired glucose tolerance.
O*>6.3
9.6
23.4
8.0
Blood Pressure
704
Diabetes
Prospective epidemiological studies have confirmed the clinical observation that diabetes
predisposes individuals to CVD (16, 17). Although some have reported that asymptomatic
Figure 4. The &year probability per 1000 men aged 40 of cardiovascular disease, according to systolic blood
pnssures from 105 to 195 mmHg at specined levels of other risk factors,in the Framingham Study.
(Source: Monograph No. 28)
700
GOO
0
r
500
400
>
t_J
m
a
m
a
a
a
Q
300
>
OD
200
100
SYSTOLIC B.P.
CHOLESTEROL.
GLUCOSE INTOL.
CIGARElTES:
ECG-LVH:
105-195
185
0
0
0
105-195
335
0
0
0
105-195
105-195
105-195
335
335
0
0
+
+
335
+
+
+
Kannel
hyperglycemia is not hazardous, some studies
have indicated that hyperglycemia, an abnormal glucose tolerance test, or even a high casual
blood sugar, is associated with an increased risk
of CVD (18). I n western industrial society,
asymptomatic hyperglycemia may well be an
independent risk factor for atherosclerotic disease.
In the Framingham cohort, diabetes was
found to double cardiovascular mortality (17,
18). Its relative impact is substantially greater in
women than in men, tends to diminish with
advancing age, and varies widely depending on
the level of coexisting risk factors (Figure 5).
(II
$
2-
03
zw
(II
(I,
43
::
4
>
P
n
a
a
0
8
0
Fa
a
W
2-
k
2
z
a
m
GLUCOSE
n
INTOLERANCE
SYSTOLIC B.P.
CHOLESTEROL
CIGARElTES
ECG-LVH
105
185
195
185
705
195
335
195
335
195
335
+
+
706
myocardium directly, predisposing the individual to cardiac failure (21). Diabetes appears
to be less dangerous in some parts of the world,
which suggests that cofactors, such as exercise,
obesity, cigarette smoking, diet, and form of
treatment, may be of great importance.
Gout
Gout or hyperuricemia appears to be indirectly atherogenic, an effect most likely mediated through associated hypertension, hyperlipidemia, and obesity (22, 23). There is little
net contribution of gout or hyperuricemia to
development of CHD in men when the associated atherogenic traits are taken into account.
Nonetheless, gouty arthritis has been found to
be associated with a doubled risk of CHD (22),
and there seems to be an association between
asymptomatic hyperuricemia and the development of CHD. The relation of hyperuricemia
induced by diuretics to the development of
CHD is not established.
Living Habits
The life style that predisposes individuals to
CVD is characterized by a diet too rich in calories, saturated fat, cholesterol, and salt and by
physical indolence, unrestrained weight gain,
cigarette smoking, and alcohol abuse.
Cigarette Smoking
Prospective epidemiological studies covering
Obesity
Obese persons in the Framingham Study developed twice as much cardiac failure and brain
infarction and a more modest excess of CHD
(18). The impact of obesity upon CHD was
greater for men than for women and was most
impressive for angina and coronary attacks (Figure 7). The incidence of sudden death was related to relative weight, and the fraction of
CHD deaths that were sudden also increased
progressively with the degree of overweight,
suggesting a specific relationship.
Kannel
707
Although it is generally agreed that obesity is ing these risk factors into account; this evidence
associated with CHD (18),its independent con- is corroborated by studies of Hawaiian Japanese
tribution to the disease has been questioned and cohorts of young subjects (28).
(26).Several studies have found that the association with CHD is accounted for by the other
cardiovascular risk factors that tend to accom- Physical Activity
pany obesity (16, 24). Overweight has been
Epidemiological evidence strongly suggests
found to be strongly associated with adverse that endurance exercise protects against CHD
lipid profiles, hypertension, and glucose intol- (29).Overall mortality, cardiovascular mortality,
erance (27). Change in weight is mirrored by and CHD mortality, in particular, all have been
corresponding changes in these cardiovascular found to be inversely related to the level of
physical activity in the Framingham cohort (30).
risk factors (27).
Although the contribution of obesity to CHD The protection seems to be confined to men
incidence may be largely mediated through ef- and is modest compared with the effects of mafects on the major cardiovascular risk factors, jor cardiovascular risk factors, but it persists
obesity must nevertheless be considered an even when the risk factors are taken into acimportant modifiable risk factor for the disease. count. The protection in men can be demonThere is also some evidence from the Fra- strated into advanced age (Table 3).
mingham Study that there is a net contribution
The amount of exercise required to achieve
of obesity to CHD incidence in men, even tak- the benefits of physical activity has not been
Figure 6. The risk of coronary heart disease (12 years) according to cigarette smoking habit and presence of predisposing factors in men aged 30 to 59 at entry in
the Framingham Study. OBS refers to the number of coronary heart disease events
observed, EXP to the number of events expected for this portion of the population.
The OBS divided hy the EXP provides the morbidity ratio.
307b
CIGARElTE SMOKING:
0NONE
>I
PKG.1DAY
196b
87
NONE
aPredisposing factors: cholesterol
ANY ONE
F
1
123
103b
< .05.
ANY TWO
708
d;
i
CORONARY INSUFFICIENCY
0
0
0
1501-
s1
aw:
a
w
z
w
0
CORONARY ATTACK
MYOCARDIAL INFARCTION
........
....
....
50
4
3
z
z
4
I?d
>
a
....
....
....
....
....
....
....
....
....
....
....
........
....
....
....
....
....
....
....
MEN
59
.......
....
....
....
...
....
....
...
....
.......
...
....
.......
....
....
....
....
....
....
.....
....
...
.......
.a.
?l
**
18 20
.......
MEN
WOMEN
WOMEN
UNCOMPLICATED
ANGINA PECTORIS
n
w
k-3)
aQ
w
0
l 50
oot
....
....
....
....
....
....
....
....
n
w
L
60
MEN
<loo
quantified. Epidemiological data based on general population samples indicate that moderate
amounts of exercise will suffice. Yet exercise
Table 3. Fourteen-year incidence of cardiovascular
mortality according to physical activity, for men
aged 35 to 64.
Physical
activity
index
24-29
30-33
34-37
38-83
t-Value
All ages
35-44
45-64
55-64
3.5
12.6
10.5
10.2
9.6
22.5
19.1
12.0
9.1
12.1
10.6
-0.55
-2.85
-2.86
4.0
1.8
1.4
- 1.74
7.7
6.5
....
MEN
WOMEN
....
....
....
....
....
....
WOMEN
Kannel
Societally conditioned psychosocial factors influence various important aspects of life style,
such as eating, drinking, smoking, and inactivity, and hence must be important in the evolution of atherosclerotic CVD. The exact role of
various emotional and social factors in the evolution of CHD remains speculative, perhaps because definition and quantification of these factors remain imprecise. Reviews of the role of
psychosocial factors have indicated some inconsistent associations with CHD. It is suspected
that sustained overstimulation of the central
nervous system may lead to CHD, and depression, sleep disturbance, prolonged anxiety, and
Type A behavior have also been incriminated.
Less consistently associated factors are work
overload, life dissatisfactions, social mobility,
and status incongruity. Social deprivation involving low income and low education have
been implicated, though social status, when
considered in isolation, appears to bear no consistent relationship to CHD incidence (31).
An overdeveloped sense of time-urgency,
preoccupation with deadlines, competitive
drive, and work-oriented behavior, all characterized as Type A, have been found to be related
to the development of CHD (32).The Western
Collaborative Study Group found that Type A
men had a doubled risk of CHD compared with
those characterized as antithesis, Type B (32).
The Framingham Study confirmed this finding
in men and demonstrated similar excess risks
for Type A women (33).Two prospective studies
have shown that the Type A behavior precedes
709
CHD, that its effect is independent of associated risk factors, and that risk is proportional to
the degree of Type A behavior exhibited (32,
33). However, both the Multiple Risk Factor Intervention Trial and the Aspirin Myocardial Infarction Study recently reported no relationship
between type and risk of CHD death (34).
Among women in the Framingham Study a
number of psychosocial factors have been
found to be related to CHD (Table 4). Working
women in the Framingham Study were more
likely than housewives to report Type A behavior, marital disagreements, job mobility, daily
stress, or marital dissatisfaction. Although employment itself is not associated with CHD in
women, the working environment (a nonsupportive boss, decreased job mobility), personality (suppressed hostility), and economic stress
appear to be determinants of vulnerability to
CHD.
Diet
Table 4. Somatic strain scores among CHD cases and non-cases in the Framingham Study, for women
ages 45 to 64.
Any clinical
manifestation of
CHD
Tension state
Daily stress
Anxiety symptoms
Anger symptoms
a
<O.Ol.
= <0.05.
Uncomplicated
angina pectoris
Cases
Non-cases
Cases
Non-cases
0.58
0.37
0.53
0.49
0.3@
0.30
0.20a
O.3Oa
0.67
0.46
0.59
0.54
0.3@
0.306
0.21a
0.31
7 10
sorption of lipids and carbohydrates (35). Recent evidence now connects cholesterol, polyunsaturated and saturated fats in the diet, and
serum cholesterol to coronary mortality prospectively in men (13). Positive calorie balance
produces an unfavorable LDUHDL cholesterol
ratio, impaired glucose tolerance, hyperuricemia, and hypertension, promoting accelerated atherogenesis.
Sodium
A high sodium content in the diet can produce hypertension in genetically predisposed
animals ( 3 6 ) . Some cross-population epidemiological studies have suggested a strong
association between sodium intake and the
prevalence of hypertension (36). Restriction of
sodium intake has been shown to be helpful in
patients being treated for hypertension (36).
The connection between salt intake and hypertension is hard to establish in humans because
of the difficulty in assessing their salt intake and
in accounting for such confounding influences
as weight, potassium intake, and alcohol usage.
There is also no good index of individual susceptibility to the effect of sodium aside from a
family history of hypertension. Perhaps measurement of intracellular sodium concentration
and fluxes may prove helpful in this regard, but
the evidence currently available is too sparse to
be definitive.
Common Beverages
Figure 8. The risk of coronary heart disease and angina according to alcohol intake
in men aged 50 to 62 over 18-year follow-up, in the Framingham Study.
TOTAL CHD
ANGINA
191
0
i-
131
n
105
NONE
15-69
(15
70 +
OZ I
NONE
15-69
(15
MONTH
70 -t
Kunnel
711
The use of oral contraceptives has been associated both retrospectively and prospectively
with an increased risk of CHD (41). Adverse
effects are more likely in women who are cigarette smokers, are over age 35, have hypertension, hypercholesterolemia, or diabetes, and are
long-term users of oral contraceptives. The
hazards associated with oral contraceptive use
include the adverse effects of the major cardiovascular risk factors, alteration in blood clotting
and platelet factors, histochemical vascular alterations, and thromboembolic events. In a few
susceptible women, oral contraceptives may
provoke severe hypertension, marked hyperlipidemia, and overt diabetes. Even in those not
so predisposed, oral contraceptives tend to shift
the distribution of these cardiovascular risk factors upward, which could have serious consequences once the women are beyond child-bearing age. The adverse effects are more common
in the oral contraceptives containing larger
amounts of estrogen. Although progestin-containing pills have less overall impact on cardiovascular risk factors, they tend to lower HDL
cholesterol, which estrogen raises.
Host Susceptibility
It was formerly believed that the atherosclerotic diseases were largely an inevitable
consequence of aging and genetic make-up. But
the progressive rise in CHD mortality in the
1940s and 1950s and its decline since 1968 have
been too rapid to be explained by genetic factors. Also the rates of CHD mortality vary
widely among populations with similar genetic
make-ups (1,2).Migrants from low to high incidence geographical areas tend to acquire the
7 12
Mortality from each of the atherosclerotic diseases is strikingly related to age in each sex and
in all races. Although uncommon in young
white women, CHD is already a major cause of
death for men aged 35 to 44 years. By ages 55 to
Table 5. Predictive value of CHD in older brothers on CHD risk in younger brothers in association with
other variables, in the Framingham Study.a
Risk factor variables
CHD in brother
Age
Systolic blood pressure
Serum total cholesterol
Metropolitan relative weight
Smoking (yesho)
Total CHD
,156
.185
.325b
,157
.383b
,165
.351b
- .056
,555
,422~
- ,102
,008
Myocardial
infarction
.47lC
-.loo
.45F
.349b
-.017
,154
CHD mortality
.5 1F
,412
,720~
,415
-.212
.245
a These are standardized multivariate logistic coefficients and represent the strength of the relationships between risk factors
and the occurrence of various aspects of coronary heart disease.
+ P = <0.05.
c
P = <0.01.
Kannel
7 13
Table 6. Age-adjusted CHD incidence rates for women having surgical menopause in the Framingham
Study, over a 24-year follow-up.
CHD
No.
Person years
~
Premenopausal
Postmenopausal
Ovaries removed
0
1 or 2
Unspecified
~~
Rate/1000
~
8384
6848
6
26
0.95
2.95
1396
4544
908
7
17
2
3.24
2.87
1.42
Indirect method
Preclinical Signs
unfavorable risk profile, it is reasonable to assume that the patient has a compromised coronary circulation. Such persons have been shown
to have an increased risk of coronary events
(18).
In coronary candidates with a normal resting
ECG, a compromised coronary circulation may
be revealed by exercise testing to elicit acute
coronary insufficiency, impaired ventricular
performance, or transient subendocardial ischemia. Positive reactors have been shown to be
at greatly increased risk of symptomatic coronary disease. Other noninvasive procedures, including nuclear imaging techniques and echocardiography, may also be useful in detecting a
compromised coronary circulation and impaired myocardial function.
7 14
PREVENTION
Kannel
risk factor level into account, the rate of recurrences and mortality is better shown to be related to conventional risk factors.
A variety of cardiovascular risk factors have
been found to be related to mortality and reinfarction following the occurrence of coronary
events. These factors include hypertension, cigarette smoking, physical inactivity, diabetes, and
ECG abnormalities following onset of angina.
By combining these risk variables it is possible
to identify a tenth of the postmyocardial infarction population from which 31 percent of the
deaths will arise. The Type A score has been
found to be a good predictor of recurrent MI,
even taking other risk factors into account.
The utility of correcting many of the risk
factors following the onset of CHD remains to
be demonstrated but seems rational and without great hazard. Comprehensive multifactorial
management along with careful surveillance
and prompt management of complications has
been found to lower the rate of new coronary
events (47). Smokers should be advised to quit,
since stopping smoking has been found to reduce mortality to half that of those who continue to smoke (47). Hypertension should be
controlled and weight reduction should increase exercise tolerance, reduce postinfarction
angina, and help control risk factors, although
evidence of benefit from these measures is still
lacking. Supervised exercise and lipid control
may be helpful in reducing the rate of new
coronary events, but evidence of efficacy is still
limited.
Primary Preventive Implications
7 15
7 16
atherogenic effect. Women taking oral contraceptives usually develop a rise in triglycerides
which may influence their HDL, depending on
the estrogen-progestin composition of the pill.
Oral contraceptives of predominantly progestin
composition may reduce HDL. In high-risk coronary candidates unresponsive to hygienic
measures, an improvement in the LDL/HDL
ratio may be achieved by lipid-lowering drugs.
Of these, clofibrate and nicotinic acid also raise
HDL. Trials involving clofibrate have not been
encouraging. Diet, cholestyramine, and nicotinic acid have been shown to reduce cholesterol by 30 to 40 percent when used in combination.
The atherogenic effects of blood lipids require decades to produce clinical disease, and
hence, clinical benefits cannot be expected from
corrective measures in a short period of time.
The improvement in the lipid profile should be
discernible in a month. Although cellular fatty
deposits may shrink in weeks, extracellular fatty
deposits require one or more years to change
significantly. Even the mass of fibrous lesions
can be reduced but only after four or more
years of vigorous treatment.
Obesity
The control of obesity, particularly on a fatmodified diet, is one of the chief hygienic measures available to control the major cardiovascular risk factors, including hypertension, lipid
abnormalities, and hyperglycemia. It is easier to
avoid than to correct long-standing obesity. A
greater sense of urgency is needed concerning
correction of modest weight gain, since this
often leads insidiously to intractable obesity.
Diabetes
Because even moderate degrees of hypertension have been shown to double the risk of a
coronary event and the Hypertension Detection
and Follow-up Program has demonstrated the
efficacy of treating this mild hypertension, it is
especially important to treat this group. It can
be shown, however, that the bulk of the cardiovascular sequelae over the first decade in this
mild hypertension is concentrated in a small
percentage who have other accompanying cardiovascular risk factors and hence a poor cardiovascular risk profile. It would seem best to
reserve drug treatment for these and use
hygienic treatment with weight reduction and
salt and alcohol restriction for the rest. The
regimen must be followed, since hypertension
tends to progress and drug treatment may be
required at a later stage.
Cigarette Smoking
Kannel
conscientious efforts in dispensing this advice
and checking on compliance are needed. Physicians should always raise the question of smoking in connection with the findings of CVD and
even in general health exams. They should
more often solicit help from the family in getting the patient to quit.
Because of the tenaciousness of the cigarette
habit and powerful vested interests in protecting the tobacco industry, an attempt to develop
a safercigarette has been undertaken. The tar
and nicotine content has been reduced, largely
through the introduction of filter cigarettes.
Preliminary evidence from the Framingham
Study indicates that this has been of little benefit in reducing the hazard of CVD. Men under
55 in the Framingham Study were classified as
to whether they smoked filter cigarettes or nonfilter cigarettes. The 58 percent who smoked
filter cigarettes tended to have slightly less prior
smoking exposure, but despite this, they did not
have lower CHD incidence than nonfilter
smokers (48). Filter cigarette smokers tend to
smoke more vigorously, getting as much nicotine and carbon monoxide as nonfilter cigarette smokers.
CONCLUSION
Effective risk factor control will require mobilization of community resources to assist in the
endeavor. Such measures should be multifactorial and should be begun as early in life as
possible, when the faulty habits are conditioned
and lesions are still only in the formative stages.
The entire family should be involved in the risk
factor modification of the high-risk candidate.
Physicians must develop the preventive skills
needed to encourage the behavior modification
required. Although the rewards may lie decades
in the future, the physicians must recognize
that such an endeavor will have a greater impact
on his patients ultimate well-being than almost
anything else he might do for them.
References
(1) Intersociety Commission for Heart Disease
Resources: Primary prevention of the atherosclerotic
diseases. Circulation 42:A-55, 1970.
(2) Keys, A. Coronary heart disease-the global
picture. Atherosclerosis 22: 149-192, 1975.
(3) Stamler, J. Lifestyles, major risk factors, proof
and public policy. Circulation 58:3-19, 1978.
( 4 ) Kannel, W.B., W.P. Castelli, and T. Gordon.
7 17
Cholesterol in the prediction of atherosclerotic disease. New perspectives based on the Framingham
Study. Ann Intern Med 90:85-91, 1979.
(5) Hegsted, D.M., R.B. McCandy, M.L. Nyers,
and F.J. Stare. Quantitative effects of dietary fat on
serum cholesterol in man. Am] Clin Nutr 17:281-295,
1965.
( 6 ) Keys, A., J.T. Anderson, and F. Grande.
Serum cholesterol response to changes in the diet. IV.
Particular saturated fatty acids in the diet. Metabollsrn
14:776-787, 1965.
(7) Mattson, F.H., B.A. Erickson, and A.M. Kligman. Effect of dietary cholesterol on serum cholesterol in man. A m J Clin Nutr 25:589-594, 1972.
(8) Vesselinovitch, D., R.W. Wissler, R. Highes,
and J. Borensztajn. Reversal of advanced atherosclerosis in rhesus monkeys. I. Light-microscope studies.
Atherosclerosis 23: 155-176, 1976.
(9) Glueck, C.J. Dietary fat and atherosclerosis.
Am J Clin Nutr 32:2703-2711, 1979.
(10) McGill, H.C., Jr. The relationship of dietary
cholesterol to serum cholesterol concentration and to
atherosclerosis in man. Am J Clin Nutr 32:2664-2702,
1979.
( 1 1 ) Beaton, G.H., J. Midner, P. Corey, et al.
Sources of variance in 24 hour dietary recall data:
implications for nutrition study design and interpretation. Am J Clin Nutr 32:2546-2559, 1979.
(12) Liu, K., J. Stamler, A. Dyer, and P. McKeever.
Statistical methods to assess and minimize the role of
intra-individual variability in obscuring the relationship between dietary lipids and serum cholesterol. J
Chron Dis 31:399-418, 1978.
(13) Shekelle, R.B., A.M. Shryock, 0. Paul, et al.
Diet, serum cholesterol and death from coronary
heart disease. The Western Electric Study. N Engl J
Med 304:65-70, 1981.
( 1 4 ) Kannel, W.B., T.R. Dawber, and D.L. McGee.
Perspectives on systolic hypertension. T h e Framingham Study. Circulation 61:1179-1182, 1980.
(15) Kannel, W.B., P. Sorlie, and T. Gordon. Labile
hypertension: a faulty concept? T h e Framingham
Study. Circulation 61:1183-1187, 1980.
(16) Report of the National Commission on Diabetes.
Washington, DC, DHEW Publ. No. (NIH) 76.1022,
1975.
(17) Kannel, W.B. and D.L. McGee. Diabetes and
cardiovascular risk factors: the Framingham Study.
Circulation 59:8-13, 1979.
(18) Kannel, W.B., and T. Gordon. T h e Framingham Study: an epidemiological investigation of
cardiovascular disease. Section 30. Some characteristics related to the incidence of cardiovascular disease and death: the Framingham Study. 18 year follow-up. US Dept. of Health, Education, and Welfare.
Public Health Service. National Institutes of Health.
DHEW Publ No. (NIH) 74-599, 1974.
(19) Garcia, M.J., P.M. McNamara, T. Gordon, and
W.B. Kannel. Morbidity and mortality of diabetes in
the Framingham population. 16 year follow-up. Diabetes 23:105-111, 1976.
(20) Knowles, H.C., Jr. Magnitude of the renal failure problem in diabetic patients. Kidney International
Vol. 6, No. 4, Suppl. I. New York, Springer Verlag,
1974.
7 18
INTRODUCTION
The Experimental subjects were born between 1915 and 1945 to schizophrenic mothers
confined to an Oregon State psychiatric hospi-
tal. Most of the subjects were born in the psychiatric hospital; however, hospital authorities
encouraged confinement in a neighboring general hospital whenever possible, in which case
the children were delivered during brief
furloughs. All apparently normal children born
of such mothers during the above time span
were included in the study if the mothers hospital record (1) specified a diagnosis of schizophrenia, dementia precox, or psychosis;
(2) contained sufficient descriptions of a thinking disorder or bizarre regressed behavior to
substantiate the diagnosis; (3) recorded a negative serologic test for syphilis and contained no
evidence of coincident disease with known psychiatric manifestations; and (4) contained presumptive evidence that mother and child had
been separated from birth. Such evidence typically consisted of a statement that the mother
had yielded the child for adoption, a note that
the father was divorcing the mother, the continued hospitalization of the mother for several
years, or the death of the mother. In practice
these requirements meant that the mothers as a
group were biased in the direction of severe,
chronic disease. No attempt was made to assess
the psychiatric status of the father; however,
none were known to be hospital patients. The
74 children ascertained as above were retained
in the study if subsequent record searches or
interviews confirmed that the child had had no
contact with its natural mother and never lived
with maternal relatives. (The latter restriction
was intended to preclude significant exposure
to the environment which might have produced
the mothers schizophrenia.)
All of the children were discharged from the
State hospital within three days of birth (in accordance with a strictly applied hospital policy)
to the care of family members or to foundling
homes. The records of the child care institutions made it possible to follow many subjects
through their early life, including, for some,
adoption. The early iife of those subjects discharged to relatives was less completely known,
719
720
or even adoptive homes. However, these disruptions occurred with equal frequency and intensity in the two groups and are considered random.
Table 1 gives the sex distribution of the subjects and the causes of further losses. Fifteen of
the 74 Experimental subjects died before
achieving school age. This rate is higher than
that experienced by the general population for
the ages and years involved, but not significantly so.
Table 1.
Experimental
Male Female
Number
Died, infancy or childhood
Lost to follow-up
33
25
6
2
Final Groups
30
17
Control
Male Female
33
25
5
3
33
17
FOLLOW-UP METHOD
Heston
721
were contacted by letter and asked to partici- approached because of risk of exposure of the
pate in a personal interview. The interview was subjects adoption. It did not seem reasonable to
standardized, although all promising leads were drop all of these 25 persons from the study,
followed, and was structured as a general medi- since considerable information was available for
cal and environmental questionnaire which ex- most of them. For instance, one man was killed
plored all important psychosocial dimensions in in prison after intermittently spending most of
considerable depth. Nearly all of the interviews his life there. His behavioral and social record
were conducted in the homes of the subjects, was available in prison records plus the results
which added to the range of possible observa- of recent psychological evaluations. A man who
tions. The short form of the MMPI was given refused the interview was a known, overt, pracafter the interview. The results of an I.Q. test ticing homosexual who had a recent felony conwere available from school or other records for viction for selling narcotics. All persons in the
nearly all subjects. If a test score was not avail- Armed Forces were known through letters from
able, the Information, Similarities, and Vocabu- their Commanding Officers or medical officers
lary subtests of the Wechsler Adult Intelligence to have been serving honorably without psychiScale (WAIS) was administered and the I.Q. atric or serious behavioral problems. One 21derived from the results. Two social class values year-old man, the least known of any of the
were assigned according to the occupational subjects, had been in Europe for the preceding
classification system of Hollingshead ( 4 ) . One 18 months in an uncertain capacity. He is
value was based on the occupation of the father known to have graduated from high school and
or surrogate father of the subjectsfirst family at to have no adverse behavioral record. In a conthe time of placement, and a second on the ference the raters agreed that it would be missubjects present occupational status or, for mar- leading to discard any cases, and that all subried females, the occupation of the husband. jects should be rated by forced choice.
The social class values move from 1 to 7 with
The MHSRS proved highly reliable as a
measure of degree of incapacity. The Intraclass
decreasing social status.
All of the investigations and interviews were Correlation Coefficient between the scores asconducted by the author in 14 states and in signed by the respective raters was 0.94, indicatCanada.
ing a high degree of accuracy. As expected,
several differences arose in the assignment of
specific diagnoses. In disputed cases a fourth
psychiatrist was asked for an opinion and differences were discussed in conference. The only
EVALUATION OF SUBJECTS
differences not easily resolved involved distincThe dossier compiled on each subject, ex- tions such as obsessive-compulsive neurosis vercluding genetic and institutional information, sus compulsive personality or mixed neurosis
was evaluated blindly and independently by two versus emotionally unstable personality. All difpsychiatrists. A third evaluation was made by ferences were within three diagnostic categothe author. Two evaluative measures were used. ries: psychoneurotic disorders, personality trait
A numerical score moving from 100 to 0 with or personality pattern disturbances. The raters
increasing psychosocial disability was assigned decided to merge these categories into one:
for each subject. The scoring was based on the neurotic personality disorder. This category
landmarks of the Menninger Mental Health- included all persons with MHSRS scores less
Sickness Rating Scale (MHSRS) (5).Where indi- than 75-the point on the scale where psychicated, the raters also assigned a psychiatric di- atric symptoms become troublesome-who reagnosis from the American Psychiatric Associa- ceived various combinations of the above three
tion nomenclature.
diagnoses. In this way, complete agreement on
Evaluations of 97 persons were done. Sev- four diagnoses was achieved: schizophrenia,
enty-two subjects were interviewed. Of the re- mental deficiency, sociopathic personality, and
maining 25 persons, six refused the interview neurotic personality disorder. One mental de(7.6 percent of those asked to participate), eight fective was also diagnosed schizophrenic and
were deceased, seven were inaccessible (active in another sociopathic. Only one diagnosis was
Armed Forces, abroad, etc.), and four were not made for all other subjects.
722
16.6 percent, a finding consistent with Kallmanns 16.4 percent. (Weinbergsshort method,
RESULTS
Table 2.
Number
Male
Age, mean
Adopted
MHSRS, mean (Total group mean = 72.8, S.D. = 18.4)
Schizophrenia (Morbid Risk = 16.6%)
Mental deficiency (I.Q.<70)
Sociopathic personality
Neurotic personality disorder
Persons +ending > 1 year in penal or psychiatric institution
Total years Institutionalized
Felons
Armed Forces, number serving
Armed Forces, number discharges, psychiatric or behavioral
Social group, first home, mean
Social group, present, mean
I.Q., mean
Years school, mean
Children, total
Divorces, total
Never married, > 30 years age
Control
Experimental
50
33
36.3
19
80.1
0
0
2
7
2
15
2
17
1
4.2
4.7
103.7
12.4
84
7
4
47
30
35.8
22
65.2
5
4
9
13
11
112
7
21
8
4.5
5.4
94.0
11.6
71
6
9
Exact
probability
0.0006
0.024
0.052
0.017
0.052
0.006
0.054
0.02 1
Heston
age of risk 15-45 years). Hoffman (6) and Oppler (7) reported rates of from 7 to 10.8 percent, of schizophrenia in children of schizophrenics. No relationship between the severity
and sub-type of the disease in the mother-child
pairs was evident.
Mental deficiency was diagnosed when a subjects I.Q. was consistently less than 70. All of
these persons were in homes for mental defectives at some time during their life and one was
continuously institutionalized. His I.Q. was 35.
The other mentally deficient subjects had 1.Q.s
between 50 and 65. No history of CNS disease
or trauma of possible causal importance was
obtained for any of these subjects. The mothers
of the mentally defective subjects were not different from the other mothers and none were
mentally defective.
Three behavioral traits were found almost exclusively within the Experimental group. These
were: (1) significant musical ability, seven persons; (2) expression of unusually strong religious feelings, six persons; and (3) problem
drinking, eight persons.
The results with respect to the effects of institutional care, social group, and type of placement will be discussed in a later paper. None of
these factors had measurable effects on the outcome.
DISCUSSION
723
724
SUMMARY
ACKNOWLEDGMENTS
Heston
Drs. Dean K. Brooks. E.I. Silk, Russel M. Guiss,
J.M. Pomerov, Superintendents of Oregon State
Hospital, Eastern Oregon State Hospital, Danmasch State Hospital, and Oregon Fairview
Home, respectively. David G. Berger, Research
Coordinator, Oregon State Board of Control;
Stewart Adams, Research Director, Los Angeles
County Probation Department; Robert Tyler,
Research Information Director, California Bureau of Corrections; Evan Iverson, State of
Washington, Department of Institutions; Anthony Hordern, Chief of Research, California
Department of Mental Hygiene; Captain
George Kanz, Oregon State Police;J.S. Gleason,
Administrator, Veterans Administration; and
Lt. General Leonard D. Heaton, Rear Admiral
E.C. Kenney, and Major General R.L. Bohannon, Chief Medical Officers of the Army, Navy
and Air Force, respectively.
Renate Whitaker, University of Oregon Medical School, and Eliot Slater and James Shields of
the Psychiatric Genetics Research Unit,
Maudsley Hospital, London, reviewed the
manuscript and made many helpful suggestions.
Finally, I wish gratefully to acknowledge the
contribution made by the subjects of this research project, most of whom freely gave of
725
References
( 1 ) Kallmann, F.J. The Genetics of Schizophrenia. New
York: J.J. Augustin, 1938.
(2) Slater, E. with J. Shields. Psychotic and neurotic
illnesses in twins. Medical Research Council Special
Report Series No. 278. London: H.M. Stationery Office, 1953.
(3) Alanen, Y.O. The mothers of schizophrenic patients. Acta Psychiat Neurol Scand Suppl. 1227, 1958.
(4) Hollingshead, A.B. and F.C. Redlich. (1958).
Social Class and Mental Illness: A Community Study. New
York: J. Wiley, 1958.
(5) Luborsky, L. Clinicians judgments of mental
health: a proposed scale. Arch Gen Psychiat (Chic.)
7:407, 1962.
(6) Hoffman, H. Studien iiber Vererbung und Entstehung geistiger Storungen. II. Die Nachkominenschaft bei
endogenen Psychosen. Berlin, Springer, 1921.
(7) Oppler, W. Zum Problem der Erbprognosebestimmung. 2 Neurol, 141:549-616, 1932.
(8) Hallgren, B. and T. Sjogren. A clinical and genetico-statistical study of schizophrenia and low grade
mental deficiency in a large Swedish rural population.
Acta Psychiat Neurol S c a d Suppl. 140. Vol. 35, 1959.
(9) Alanen, V.O., J. Rekola, A. Staven, M.
Tuovinen, K. Takala, and E. Rutanen. Mental disorders in the siblings of schizophrenic patients. Acta
Psychiat Scand Suppl. 169 39:167, 1963.
F. EXPERIMENTAL STUDIES
Goldberger at al.
727
728
Third year. The third years study at the sanitarium was continued with three wards under
observation, one in the white and two in the
colored service, as during the second year. The
modifications of and supplements to the institution diet, the hygienic conditions, the administrative routine, the mingling with other inmates
(including those with active pellagra) in the
wards and in the recreation yards were continued as during the first and second year.
The result of the third years study was exactly like that of the second year: no recurrence
and no new case among those inmates taking
the modified diet.
Result as a whole. The result of the investigation considered as a whole may be summed up
as follows: The individuals under observation,
disregarding those who were present for periods too brief to be significant, numbered 702,
Table 1. Number of specified classes of individuals observed for pellagra during specified periods,
according to orphanage of residence.
Period of observation
One year but less than
two years
At least two
years
Class
Total
MIa
Bla
ECb
Total
MI
B1
ECb
Total
MI
BIa
Pellagrins
Nonpellagrins
250
268
59
100
99
69
92
99
143
22
58
29
26
92
99
107
85
37
42
70
43
183
Goldberger et al.
and one-half months following this, approximately 40 percent of those who were affected
by the change in diet developed pellagra.
Thereupon there were added to the institution
diet, again under our direction, 4 ounces of
fresh beef, about 7 ounces of sweet milk, and
about 14 ounces of buttermilk per adult per
day; and during an observation period of 14
months immediately succeeding the adoption
of these supplements no evidence of pellagra
developed in any of the group.
DISCUSSION
729
A report of a three years study of the preventability of pellagra by means of diet, the result of
730
motion. I can think of no better or more ambitious aim for any student of epidemics than to
attempt to reduce his problems to precisely
these terms. There is ample evidence that local
movements of populations, of parasites and
hosts, of insect vectors, and especially the relation of these movements to one another, determine outbreaks of infective disease. We are far
as yet from identifying all relevant movements,
or the factors on which they depend; and of
those we know, many are hard to measure. But
there is no other hopeful method of approach.
The epidemiology of today is the child of
parasitology. Medicine is its grandmother, once
or twice removed. The universe of study which
now faces the epidemiologist is not, in its essence, composed of a number of cases of a
clinically recognizable disease, with a distribution that varies in an intriguing way in space
and time. It is composed of a variety of biological species, some acting as parasites and some
as hosts. The parasites may be viruses, or bacteria, or fungi, or protozoa, or worms. The
hosts may be men, or animals, or insects, or
plants. Some parasites, particularly certain helminths, pass through different life cycles in different hosts, so that they are dependent on at
least two host species for their continued propagation. Many parasites can propagate themselves in the tissues of two or more animal species which act as alternative hosts. Certain biting
insects act as vectors. Sometimes the parasite
passes through an essential phase of a complex
life cycle within the insects tissues. Sometimes it
is simply transferred from the blood of one host
to that of another, perhaps with multiplication
by binary fission during transit. Sometimes
non-biting insects act as mechanical vectors of
parasites, commonly from excreta to food; but
in these instances they provide only one among
many possible routes of infection. Even if we
take the simplest case in which there is no insect
vector, and in which, so far as we yet know, only
one host, say man, and one parasite are concerned, we find that our universe is not composed of sick persons and the parasite that
73 1
732
causes the disease, but that there are, among the proportions, and then to see what happens
infected hosts, all gradations from clinically typ- when we hold some factors constant and vary
ical cases, through mild and atypical infections, others. This is the method that Professor
to what are known as healthy carriers, persons Greenwood and I, together with our colleagues
who display no signs or symptoms of illness but Dr. Joyce Wilson and Dr. Bradford Hill, have
from whose tissues or excreta the causative par- been exploring for many years past (I).We have
worked with mice, and with three diseases to
asite may be isolated.
It is clear that no amassing of clinical observa- which mice are naturally prone: pasteurellosis,
tions, however careful and acute, and no cor- a bacterial disease of the respiratory tract;
relation of such observations with environmen- mouse typhoid, an intestinal bacterial disease;
tal factors, however complete the records and and a virus disease, ectromelia. All these disstatistical analyses, could have solved such prob- eases spread from mouse to mouse by direct
lems as these, or have shown us in any detail contact. The various precautions that we have
how and when we might intervene effectively. taken to control our experimental conditions
Such observations have supplied us with data have been described elsewhere.
that form an essential part of our field of study;
but when, today, we are faced with correlations
between certain diseases and certain conditions
THE EFFECT OF VARYING THE RATE OF
of climate, or housing, or occupation, we do not
ADDITION OF SUSCEPTIBLE MICE
leave it at that, we transpose our picture into
In most experiments we have started an epiterms of the effect of these environmental conditions on the hosts, or vectors, or parasites demic among a group of mice, and then added
concerned. We think, for instance, not in terms a constant number of mice each day for many
of marshes and malaria, but in terms of breed- months, or for several years. The effect of varying places for mosquito larvae, and of how these ing the rate of addition of susceptible mice to
are affected by light, shade, salinity, and a host infected herds, contact being continuous, may
of other factors. We try to discover which mis- be briefly summarized as follows.
quitoes bite which hosts and under what condiWith low rates of addition, up to one or two
tions, how far they fly, and where they hiber- mice each day, the death rate has shown wide
nate, and when. We view the problem of plague and irregular fluctuations, with occasional inin terms of rats and fleas, of typhus in terms of termissions. As the rate of addition rises, the
lice, of diphtheria in terms of carriers as well as curve of daily mortality assumes an undulating
cases, and so on. The medical or veterinary form, with 110 intermissions, and no clearly deepidemiologist becomes a biologist whether he fined waves or peaks after the few initial fluctuwill or no. The biologist can study many of the ations that always mark the earliest phase of
problems of epidemiology with no more than a epidemic spread under these conditions. The
nodding acquaintance with human or veteri- total population of the herd rises, at first steeply, then more slowly. In experiments carried on
nary medicine.
The very complexity of the natural systems for many months or years it tends towards a
with which the epidemiologist is faced often relatively constant level. Figure 1 shows the first
makes it difficult to tell which of the correla- five months experience in an epidemic of pastions that he observes are biologically signifi- teurellosis, in which six mice were added daily
cant. This difficulty is not lessened by the fact to the herd. Figure 2 shows similar periods for
that it is quite easy to invent hypotheses that two epidemics of ectromelia, in each of which
would, if they were true, fit attractively into our three mice were added daily to the herd.
puzzle. It is much less easy to determine
It may be noted that, in our limited experiwhether they are true or not; and this step has ence, there is little correlation between the averbeen omitted with a rather surprising fre- age daily death rate and the rate of addition of
susceptible mice, provided that immigration is
quency.
One way of attempting to solve some of our maintained at a steady rate. It follows that the
problems is to turn our backs on the natural effect of adding more mice each day is simply to
world, and to simplify our conditions until the increase the level of population at which equinumber of variable factors reaches manageable librium is attained.
Topley
733
.....
Figure 1. Secular death rate and population in the first five months of on epidemic of mouse
pasteurellosis; six mice added daily.
population;
.smoothed daily death rate
(all deaths);
smoothed daily death rate (specificdeaths only).
----
240
220
Zm
Q 180
5:
2120 w
g1,
80 a0-13
60 0-1 1
40 $0-09
20 0 0-07
0 3 0-05
a 0-03
0 0-01
p 300250-
18ocI/301/w/30
llI/31
Im/31
l
h
3
1
1AVl31
1dBl
1p;z/31
1IdIL41
All our evidence suggests that, with rates of scribed above by keeping the mice constituting
addition higher than have been practicable in an infected herd in separate cages, one mouse
our experiments, we should attain a steady rate to a cage. We started the experiment with 25
of mortality and a stable population, the daily mice infected with mouse typhoid and 100 nordeaths equalling the daily additions. This mal companions. Each Monday, Wednesday,
means only that such a system has no inherent and Friday the mice were assembled in a single
tendency to fluctuate, as epidemics fluctuate large cage for four hours, and on each of these
under natural conditions. There is no evidence days two normal mice were added to the herd.
that the parasite waxes or wanes in its relevant The course of events is shown in Figure 3. The
biological properties as it passes from host to herd was assembled on 19 April 1937. An epihost, or that the hosts pass through any periodic demic started, but soon died down; and after
variation in resistance. Nor does the infection about the 70th day deaths from mouse typhoid
ever die out. We must search in other directions ceased to occur. The population was rising, and
for the factors that determine the rise and fall of by the 149th day had reached 180. Since no
epidemic waves.
death from mouse typhoid had occurred for
approximately 80 days, it seemed that these
conditions, though involving close contact beTHE EFFECT OF CHANGING FROM
tween the mice for four hours on each of three
CONTINUOUS TO DISCONTINUOUS
days
a week, were not such as to ensure an
CONTACT
effective spread of infection. The conditions
In an experiment which has not yet been were therefore altered, and on the 149th day
recorded in detail we varied the procedure de- the mice were aggregated in a single large cage.
734
200
0.10 -
0.05
o+--
'
Topley
into 10 groups of 10 mice apiece, the subsequent specific mortality is much lower than in a
control herd of 100 mice maintained as a single
unit, provided that dispersal is carried out in
the initial phase of an epidemic. If, on the other
hand, dispersal is delayed until an epidemic
wave is well under way it has little effect, even
though the dispersed groups are numerous and
small. The withdrawal of mice from a herd, in
which infection is continuously spreading, to
isolation in single cages, always lowers the subsequent average mortality of these mice as compared with that of their contemporaries left in
the herd, however long they have previously
been exposed to risk (4).
GROUP-TO-GROUP INFECTION
735
736
Figan 4. Weekly attack rates (percent) for d nwphnryngeal infectiom (excluding influenza) for the
five Lent terms 1930-1934 (girls boarding schools).
8-
1930
1931
7-
1932
1933
2z
w
fw+
6-
1934
--I--
54-
321-
10
11
12
WEEKS
Figure 5. Weekly attack rates (percent) for all nasopharpgeal infectiom (excluding
influenza) for the five summer terms 1930-1934 (girlsboarding schools).
6.0
5.0
w
0 4.0
a
1930
1931 -91932
1933 -01934
5 3.0
!-
5n
-+-
2.0
1.o
6
7
8
WEEKS
9 1 0 1 1 1 2 1 3
Topley
737
Figwe 6. Weekly attack rates (percent) for all urnpharyngeal infections (excluding
influenu) for the five CMitmne terms 1930-1934 (girls boarding dmols).
8
7
6
W
3
t
E 4
:
W 3
2
1
0
WEEKS
in England and Wales was so low that no epidemic can have occurred. There are, however,
many considerations that have to be taken into
account in interpreting secular curves of this
kind; and it is possible that the decrease in any
of these diseases would have occurred apart
from the war. The cumulative evidence does,
however, suggest that the association with evacuation was not in each case fortuitous. Assuming this, it seems at least possible that school
closure in towns, and the lack of school accommodation in the evacuation areas, contributed
more to the result than the move from town to
country per se.
At the same time that we were evacuating
children from our larger towns, and decreasing
their contacts in school, we were collecting a
large part of our young adult males into huts
and billets. Young adults are, for the most part,
resistant to scarlet fever, diphtheria, measles
and whooping cough; but they are susceptible
to another disease, cerebrospinal meningitis,
which is also spread by contact infection by way
of the upper respiratory tract. Whenever we
mobilize an army we may expect an outbreak of
this disease during the first winter or spring,
when the men spend a large part of their time
in huts or billets. In spite of the low rates of
other sickness in the army, this particular expectation was fulfilled. Figure 8 shows that a
major epidemic of cerebrospinal meningitis oc-
738
Figure 7. Weekly notiffed cases of scarlet fever and diphtheria in England and Wales
(1931-1941) and weekly notified dentha for measles and whooping cough in great
towns of England and Walea (1931-1941).
1000
301
200
50
0
Great towns of England and Wales, weekly notified deaths from measles (1931-41).
100
50
0
(53 weeks)
Great towns of England and Wales, weekly notified deaths
from whooping cough (1931-41).
ltI
Topley
739
Figure 8. England and Wales, weekly notified cases of cerebro-spinal fever (1931-1941).
200
0
1931
1932
1933
(53weeks)
step in the right direction. This seems a platitude; and so it should be. But there are many
instances in which the possibilities of action
along these lines have not yet been fully exploited.
740
+
z
_--/-
##--
0
U
INFECTED
50-
0.
10
15
CAGE AGE (DAYS)
DEAD
25
20
60
10
20
30
CAGE AGE (DAYS)
40
50
Topley
the non-immunized controls; but they never attain an expectation of life approaching the normal. It should, however, be noted that the mice
in these experiments are exposed to a continuous risk of contact infection greater than would
be encountered by any human population, except under very unusual conditions. The effectiveness of antityphoid inoculation in man has
been amply proved by the experience in our
armies.
Figure 11 shows similar curves for an epidemic of the virus disease ectromelia (1).The
immunized mice in this instance were injected
with two doses of a living but attenuated virus.
The results here differ sharply from those obtained with mouse typhoid. On entry to the
cage the control mice have an expectation of life
even lower than that of the non-immunized entrants to the herd infected with mouse typhoid.
But, after a slight fall, reaching its lowest point
on the 8th day of cage age, the curve rises
sharply, and by the 30th day has attained a
value not far below the normal. Clearly, the
natural immunization that follows a non-fatal
attack of ectromelia is more effective than the
natural immunization that follows an attack of
mouse typhoid. Similarly, our vaccination has
been far more effective. The immunized mice
on entry to the cage have an expectation of life
60
d/#-~?c-..-..
50-----~---
3
5
40-
ya
3
20'---,
a
Il-
30-
10
ic
1:
----\-,.-
/*'
741
-normal
mice
immunized mice
control mice
---
742
Toblev
743
Figure 12. Course of events in thirteen epidemics, showing the number of survivors
among 100 mice at risk being plotted against time.
H2
J1 J2
HI
G2
Kl
K2
G1
L1
,
P3
P1
P2
10
20
30
DAYS
40
50
60
744
Strain
Contact
deaths
Survivors
infected
P.64
P.62
P.29
P.58
P.A. 39
731100
621100
261100
18/100
13/100
6/27
7/38
0174
1/82
0187
Just as different strains within a single parasitic species differ in their power of infecting,
and killing, a particular species of host, so different strains, or races, within a single host species differ in their resistance to attack; and the
dissociation of different infective powers that
can be demonstrated in selected strains of a
given parasite has an analogy in the dissociation
of resistance toward different parasites, or to-
105
104
103
102
10
5
5
5
5
5
1
4
5
4
4
0
5
5
4
4
1
4
4
2
2
0
3
5
Tqley
than unselected controls. It is clear, therefore,
that genetic resistance, while not specific in the
same sense as the resistance depending on antigen-antibody reactions, is limited in its range,
and is an expression of the effect of different
factors that may be inherited independently.
The racial differences in resistance depending on these factors, in man or in animals, will
affect the natural spread of epidemics by increasing their severity in some quarters of the
world, and lessening it in others. They may also,
in areas inhabited by a mixture of races, lead to
a differential incidence or death rate. It should,
however, be emphasized that it is usually a matter of the greatest difficulty, when faced with
recorded observations of this kind, to disentangle the effects of genetic factors, immunization by previous infection, and environmental
influences.
The plant pathologist and the plant geneticist
have gone further in this particular field, and
have obtained results of the greatest interest.
They have, for instance, demonstrated, in their
studies on the stem rust of wheat, that the genetic resistance of selected host strains may be
narrowly related to strain differences in the parasite (see 29). A selected strain of wheat that is
highly resistant to one strain of rust may be
highly susceptible to another; so that the possibility that has here been opened to us of eliminating susceptible hosts by selective breeding
depends on our ability to produce, by appropriate crossing, strains of wheat that are resistant to
all the strains of rust by which they are likely to
be attacked.
Apart from innate differences in resistance,
and acquired differences in specific immunity,
host resistance may be affected by a variety of
other factors, by certain dietetic deficiencies, by
exposure to severe climatic changes, and so on.
It would take us too far afield to discuss these
factors in any detail. We need only note that it is
desirable to assess their relative importance by
experiment, or by carefully controlled observations, before attempting to assign to them their
respective roles in the natural course of events.
MORE COMPLEX SYSTEMS
745
746
rest, and the disease will be reduced to negligible proportions, or even disappear.
In conclusion I should wish to express my
indebtedness to Mr. W.T. Russell, Mr. W.J. Martin and Mr. E. Lewis Faning for the help they
have given me in obtaining certain statistical
data, and in the preparation of some of the
graphs.
References
( 1 ) Greenwood, M., A.B. Hill, W.W.C. Topley, and J.
Wilson. Spec Rep Ser Med Res Coun, Lond, No. 75,
1936.
( 2 ) Topley, W.W.C.J Hyg Cam6 21:20, 1922.
( 3 ) Topley, W.W.C. and J. Wilson. J Hyg Camb 24:295,
1925.
( 4 ) Greenwood, M., A.B. Hill, W.W.C. Topley, and J.
Wilson. J Hyg Camb 39: 109, 1939.
(5) Topley, W.W.C. and G.S. Wilson. J Hyg Cam6
21:237, 1923.
(6) School Epidemics Committee, Med. Res. Coun.
Spec Rep Ser Med Res Coun Lond, No. 227, 1938.
( 7 ) Topley, W.W.C.,J. Ayrton, and E.R. Lewis. J Hyg
Camb 23:223, 1924.
(8) Greenwood, M., W.W.C. Topley, and J. Wilson. J
Hyg Camb 31:257, 403.
( 9 ) Topley, W.W.C.J Hyg Cum6 20:103, 1921.
(10) Topley, W.W.C., M. Greenwood, and J. Wi1son.J
Path B u t 34:523, 1931.
( I 1 ) Dudley, S.F. Spec Rep Ser Med Res Coun Lond, No.
75, 1923.
(12) Dudley, S.F. Spec Rep Ser Med Res Coun Lond, No.
1 1 1 , 1926.
(13) Webster, L.T.J Exp Med 37:231, 1923.
(14) Webster, L.T. J Exp Med 38:33, 45, 1923.
(15) Webster, L.T.J Exp Med 39:129, 837, 879, 1924.
(16) Webster, L.T.J Exp Med 40:109, 1924.
(17) Webster, L.T.J Exp Med 42:1, 1925.
(18) Webster, L.T. J Exp Med 43: 573, 1926.
(19) Webster, L.T.JExp Med 51:219, 1930.
(20) Webster, L.T. J Exp Med 5 2 9 0 1,909, 93 1 , 1930.
(21) Webster, L.T. and C.G. Burn.J Exp Med 44:343,
359, 1926.
(22) Webster, L.T. and C.G. Burn.] Exp Med 45:911,
1927.
(23) Webster, L.T. and C.G. Burn.J Exp Med 46:855,
871. 1927.
( 2 4 ) Topley, W.W.C., M. Greenwood, and J. Wilson.]
Path Bacteriol 34:523, 1931.
(25) Hill, A.B. Spec Rep Ser Med Res Coun Lond, No.
196, 1934.
(26) Gowen, J.W. and R.G. Schott. A m J Hyg 18:674,
1933.
(27) Webster, L . T . 1 Exp Med 57:819, 1933.
(28) Hill, A.B., J.M. Hatswell, and W.W.C. Topley. J
Hyg Camb 40:538, 1940.
(29) Craigie, J.H. Publication No. 666, Department
of Agriculture, Dominion of Canada, 1940.
(30) Hackett, L.W. Malaria in Europe. Oxford University Press, 1937.
'
747
748
Table 1. DMFa teeth per 100 children ages 6-16, based on clinical and roentgenographic examinations,
Newburghb and Kingston, N.Y., 1954-1955.
Number of children
with permanent teeth
Number of
DMF teeth
Aged
Newburgh
Kingston
Newburgh
Kingston
Newburgh
Kingston
Per cent
Difference
K-N
6-9.
10-12
13-14
16
708
52 1
263
109
913
640
44 1
119
672
1711
1579
1063
2136
447 1
5151
1962
98.4
328.1
610.1
975.2
233.7
698.6
1170.3
1648.7
- 57.9
- 53.0
- 47.9
- 40.9
DMF includes permanent teeth decayed, missing (lost subsequent to eruption), or filled.
Sodium fluoride added to Newburghs water supply beginning May 2, 1945.
Adjusted to age distribution of children examined in Kingston who had permanent teeth in the 1954-1955 examination.
Age at last birthday at time of examination.
Newburgh children of this age group exposed to fluoridated water from time of birth,
100
-4%
1I
80
60
40
20
0
K N
6-9
K N
10-12
K N
13-14
1954-1955
16
749
Table 2. Status of erupted first permanent molars in children ages 6-16, based on clinical and
roentgenographicexaminations, Newburgha and Kingston, N.Y., 1954-1955.
Percent of erupted first permanent molarsb
Caries-free
DMFd
Filled
Untreated caries
Missing
AgeC Newburgh Kingston Newburgh Kingston Newburgh Kingston Newburgh Kingston Newburgh Kingston
6-9
10-12
13-14
16
a
74.9
36.8
19.3
8.5
46.7
10.0
5.9
4.8
25.1
63.2
90.7
91.5
53.3
90.0
94.1
95.2
14.2
40.2
43.9
55.0
17.8
41.3
40.5
36.6
10.6
20.5
27.0
20.9
33.2
37.1
36.7
35.5
0.3
2.5
9.8
15.6
2.4
11.6
16.9
23.1
750
Table 3. Number and percent of children age 6-9 with caries-free deciduous cuspids, first and second
deciduous molars, based on clinical and roentgenographic examinations, Kingston and Newburgh,' N.Y.,
1954-1955.
Number of children
with all 12 teeth
present and caries-free
Number of children
examined
Ageb
Kingston
Newburgh
Kingston
6
7
8
9
216
255
277
192
184
208
213
129
24
12
5
3
Total
940
734
44
Adjusted ratec
Newburgh
Newhureh
68
58
53
13
11.1
4.7
1.8
1.6
37.0
27.9
24.9
10.1
192
4.7
26.2
4.7
25.5
~~
orosis, 26 had very mild, and six had mild fluorosis. There were no cases of moderate or
severe mottling and in no instance was there
any disfiguring discoloration. Thirty-six of the
Newburgh children examined had non-fluoride opacities. Of the 612 children examined in
Kingston, 115 had nonfluoride opacities. The
relatively infrequent occurrence of nonfluoride
enamel opacities in Newburgh compared with
Kingston tends to confirm a previous report
(13) that ingested water fluorides at the recommended concentration appear to reduce the occurrence of hypoplastic spots on the teeth.
The same groups of children examined for
enamel opacities were also examined for evidence of gingivitis. A positive score was recorded only for flagrant gingivitis, thus making
it possible to place greater emphasis on advanced disease and minimize examiner bias.
There was slightly, but significantly, more gingivitis observed among the Kingston children
than among those in Newburgh.
The final report on the pediatric findings of
the Newburgh-Kingston study pointed out that
all the scientific evidence available at the time
the study was first proposed indicated the safety
of drinking water containing about 1 ppm F at
the source. There was no reason at that time to
believe that fluorides, when added to the drinking water as part of the water treatment process,
would act in any way differently from fluorides
already present. Nevertheless, it was considered
desirable to test this remote possibility under
the carefully controlled conditions established
75 1
752
of riboflavin, pyridoxine and calcium pantothenate. Before much of the work cited, the study
by Poole et al. (12) led them to conclude, with
regard to the feeding of human infants, that all
available data on vitamin B seem to indicate
some close interaction in the infant between the
metabolism and the development of external
behavior. It, therefore, seems reasonable that
vitamin supplementation of the diets of women
with poor diets might cause improvement in the
intelligence of their children.
PLAN
Daily dietary supplement tablets were provided to 2400 women, equally divided between
a maternity clinic in Norfolk, Va. and one in
Leslie County, Ky. The study was conducted
simultaneously in both groups from October
1945 to June 1948. A double-blind technique
avoided knowledge by the subjects or investigators of the nature of the supplements. Four
tablets3 were used in a carefully organized pattern of randomization.
Tablet A (group A) = 200 mg. ascorbic acid;
Tablet B (group B) = 2 mg. thiamine, 4 mg.
riboflavin, 20 mg. niacinamide, 15 mg. iron;
Tablet C (group C) = inert placebo; Tablet D
(group D) = 2 mg. thiamine.
In Norfolk the expectant mothers were interviewed on their first registration in the free
maternity clinic. They were offered a tonic
tablet, for daily use, with the explanation that it
was expensive, but would be provided free if
they would continue in the project, and would
engage to bring their babies back for measurement at ages 3 and 4 years. The choice of the
four types of supplementation was by rotation
in order of new patient registration. No one
who handled the materials knew which type was
given to a patient. These women were seen in
the clinic twice monthly, and were visited in
753
754
TESTING OF CHILDREN
RESULTS
Intelligence testing was done by using selected portions of the Terman-Merrill Revised
Stanford-Binet Scale, employing form L for 3year-olds and for 4-year-olds. Although these
tests did not appear perfectly suited to the children in either group, the difficulties were more
evident in the rural Kentucky group where
some of the pictures were of objects totally unfamiliar to these children. In Kentucky, practical difficulties limited the testing to the 3-yearold-children.
In Norfolk, 612 women completed the program of taking supplements and 5 18 children
were available for tests at age 3; 370 children
were tested at age 4. The mean birth weights of
these children did not differ significantly for
the four groups which received the four different tablets. For the entire group the mean was
7.42 lb., the lowest recorded birth weight being
3.5 Ib. The distribution of intelligence quotients
in the 3-year-old and 4-year-old children is
shown in Tables 1 and 2.
Harrell et al.
Table 1. Frequency distributions of intelligence
quotients of 3-year-olds at Norfolk.
755
Group
IQ
57
60
3
6
9
72
5
8
81
4
5
12
5
4
12
11
11
13
16
11
16
5
12
8
90
3
6
9
102
5
8
111
4
120
6
9
132
5
8
141
4
1
1
2
5
3
1
6
5
5
11
8
12
12
6
3
2
4
1
6
1
GrouD
C
1
3
4
1
8
4
11
13
9
9
23
6
10
8
9
4
2
4
2
1
1
Total
IQ
63
66
69
72
5
8
81
4
7
90
3
6
9
102
1
4
1
4
4
8
7
9
11
16
6
14
5
10
4
2
1
1
1
1
3
1
8
6
12
27
14
29
37
34
38
63
36
40
50
25
33
20
11
13
2
3
7
2
1
150
3
6
9
162
Total 149
Mean" 100.9
~_______
8
111
4
7
120
6
9
132
5
8
141
4
7
I50
3
Total
Meana
111
103.4
(33
98.4
125
101.9
518
101.0
______
StatiStical studies of these data4 by the analysis of variance technique show really significant
differences in the IQs of the four groups. These
differences are most obvious when comparing
the three groups that received vitamins with the
group that received the placebo. The intergroup differences of the supplemented groups,
A, B and D, do not indicate significance. The
mean IQs of the three groups receiving vi4 Performed by R.E. Wheeler, Howard College, Birmingham, Alabama.
C
1
1
3
1
2
2
5
12
7
8
7
17
8
14
7
3
1
4
1
1
104
97.9
1
4
2
8
4
8
7
15
8
10
2
10
2
1
1
1
3
4
10
11
6
15
6
18
4
8
1
3
1
Total
1
1
3
5
5
9
16
19
35
22
40
24
63
29
38
12
22
5
6
2
4
5
9
4
13
9
6
2
6
1
1
3
1
4
5
2
2
1
3
1
91
101.7
98
93.6
77
97.9
370
97.7
756
Calculated F
F 99
7.78
16.6
3.02
3.84
6.71
4.24
9.66
3.82
6.69
1.oo
1.36
.5
.8
F 95
3.03
3.02
2.61
2.61
2.61
for comparison with the calculated F, the differ- birth weight data did not differ materially from
ences are seen to be within the limits of chance those in the Norfolk group, nor was there any
and thus have no independent significance. apparent intergroup difference correlated with
(F.95indicates the 5 percent level of certainty.) the four supplements when taken by the mothThe difference between the mixed vitamin B ers. No significant correlations could be estabgroup (group B) and the placebo group (group lished by analysis of variance or by scattergrams
C), referring to the 4-year-old Norfolk children, for the several features of maternal status duris 8.1 points IQ (101.7-93.6); and in the 3- ing pregnancy or at delivery.
year-old children, 5.0 points IQ (103.4-98.4).
Most psychologists and educators would regard
DISCUSSION
a difference of 8 points IQ as clinically or educationally significant; a difference of 5 points
Supplementation of the diets of the mothers
IQ might be regarded as difficult to discern, in this study began at times varying from the
but, nevertheless, significant in effect over peri- first to the ninth month of pregnancy. The
ods of time, for example, in the elementary mean duration of antepartum supplementation
school period.
was 134 days in the Norfolk group, 114 days in
Attempts to detect significant intergroup Kentucky. It is therefore apparent that any benclinical differences among the mothers were un- efits from supplementation were made available
successful. These included analyzing data on at highly variable times and usually for less than
age, parity, headaches, vomiting, hypertension, half the pregnancy. Attempts to apply statistical
constipation, hemorrhages, urinary specific methods to the significance of differences in
gravity and albumen, and incidence of venereal intelligence as correlated with the various duradiseases. The failures to establish significant dif- tions of vitamin supplementation appear unferences in these respects add to the probability justified in view of the small numbers of cases
of significance in the intelligence test com- available for any one supplement and any speparisons with and without vitamin supplemen- cial fraction of time concerned. In view of the
tation.
known effects of nutritional deficiencies in exIn Leslie County starting with 1200 women perimental animals during the first third of
during pregnancy, 8 1 1 children were available pregnancy, it is possible that more definitive
for testing at age 3 years. Table 4 gives the results might be secured were it possible to indistribution of intelligence quotients. The dif- stitute supplementation from the earliest days
ferences between the 4 groups are clearly insig- of pregnancy. Furthermore, since the developnificant. A higher mean intelligence is evident ment of the central nervous system continues
in this group than in the Norfolk group in spite during the first two years of infancy, optimum
of the fact that difficulties with the test operated results from nutritional improvement would into the disadvantage of the Kentucky group. The volve a supplementation of the diets used
Harrell et al.
7.57
72.4
5.7
8
81
4
7
90
3
6
9
102
5
8
111
4
7
120
3
6
9
132
5
S140
141
Total
Meana
a
3
2
3
6
9
13
18
11
12
22
19
26
16
15
10
7
6
3
1
4
3
3
5
6
10
24
23
23
26
17
15
23
8
8
5
5
3
1
3
1
5
8
8
12
14
16
21
12
16
21
19
20
8
4
6
3
1
1
1
2
5
12
16
8
13
21
15
20
17
8
18
8
10
10
6
I
203
105.74
213
107.62
204
107.94
Total
1
7
8
13
18
34
43
48
62
72
81
77
76
60
75
46
33
25
20
7
3
1
191
106.76
81 1
107.03
Means were computed from scores listed above using the midscore of each step as the value of all scores at that step.
The mean intelligence, determined by conventional testing methods at ages 3 and 4 years,
in children born of 612 women in Norfolk, Va.,
was significantly higher in those whose mothers
had received vitamin supplements during the
latter part of pregnancy than in those whose
mothers received an inert control tablet. Bene-
758
( 4 ) -.
The effect of early depletion of vitamin B2 upon performance in rats. IV.J Comp Psychol
20: 385-387, 1935.
(5)Balken, Eva R. and S. Maurer. Variations of
psychological measurements associated with increased vitamin B complex feeding in young children.
] Exper Psychol 17: 85-92, 1934.
(6) Beilly, J.S. and I.J. Kurland. Relationship of
maternal weight gain and weight of the newborn infant. A m J Obst Gynecol 50: 202-206, 1945.
(7) Bernhardt, K.S. Protein deficiency and learning in rats. J Comp Psychol 22: 269-272, 1936.
(8) -and R.J. Herbert. A further study of
vitamin B deficiency and learning with rats. J Comp
Psycho1 24: 263-267, 1937.
(9)ONeill, P.H. 1946. The effects on subsequent
maze learning ability of graded amounts of vitamin
B, in the diet of very young rats. Unpublished doctoral thesis. Fordham University Library, New York.
(10)-.
The effect on subsequent maze learning ability of graded amounts of vitamin B, in the diet
of very young rats. J Genet Psychol 74: 85-95, 1949.
( 1 1 ) ~.
I f mother is short on B, child may not
learn fast. Science News Letter Aug. 13, 1949.
(12) Poole, M. H., B. M. Hamil, T. B. Cooley and I.
G. Macy. Stabilizing effect of increased vitamin B,
intake on growth and nutrition of infants. Basic study.
Research Laboratory. Children's Fund of Michigan
for the Study of Growth. A m J Dw Chzld 54: 726-749,
1937.
Alvan R. Feinstein,' Harrison F. Wood,3 Jeanne A. Epstein,4 Angelo Taranta,5 Rita Simpson,G
and Esther Tursky,'
with the collaboration of Thomas Argyros, Julian Frieden, Raymond C. Haas,
Ilse Hirschfeld, Arthur J. Lewis, Carlos Manso, and Konrad Ulich
'
759
760
The third year of operation of the study continued in the manner described in the previous
report. (?). The patients received monthly examinations. Streptococcal infections were detected by monthly throat cultures, by routine
bimonthly determinations of antistreptolysin-0
titers, and by serial measurement of antistreptolysin 0, antihyaluronidase, and antistreptokinase titers in serums of several consecutive
months when clinical or bacteriologic suspicion
of an infection with Group A streptococci existed (4-6). The diagnosis of rheumatic fever
recurrences was made according to the modified Jones criteria (7).
Two procedures were used to determine how
faithfully patients maintained their oral prophylaxis. The first method consisted of a special
interview, 10 to 15 minutes in duration, with
each child. The interview included the childs
parents whenever possible and was conducted
throughout by the same physician (J.A.E.).The
questions considered the nature of supervision,
the place where the drug was kept at home, and
the daily ritual that the child followed in taking
it. Additional questions asked how many times
each month the patient failed to take the assigned medication, and whether, with the passage of time, he had become any more or less
faithful in his observance of the program. At
each previous routine monthly Clinic visit, the
examining physician had regularly questioned
the patient about the number of pills missed
that month, and these results could be compared with the answers given at the special interview. The second procedure of this special
study sought to obtain more objective data. The
RESULTS
Feinstein et al.
761
Table 1. Number of streptococcal infections in patients receiving three different prophylactic agents in
the first three years.
Prophylactic agent
Clinical
infections
Subclinical
infections
Benzathine penicilline Ga
Potassium penicillin Gb
Sulfadiazine
6
11
17
6
26
22
3d
Benzathine penicillin G
Potassium penicillin G
Sulfadiazine
6
5
5
Benzathine penicillin G
Potassium penicillin G
Sulfadiazine
12
16
22
50
Year
Totals
a
Carriers
Totals
1
7
12
38
46
3
15
9
2
4
8
11
24
22
41
31
2
5
15
23
62
68
81
22
153
Prophylactic agent
No. of
patient yr.
Benzathine penicillin G
Potassium penicillin G
Sulfadiazine
190
183
191
12
38
46
6.3
20.8
24.1
3d
Benzathine penicillin G
Potassium penicillin G
Sulfadiazine
124
117
131
11
24
22
8.9
20.5
16.8
Benzathine penicillin G
Potassium penicillin G
Sulfadbdzine
314
300
322
23
62
68
7.3
20.7
21.1
936
153
Year
Totals
Rate per
patient yr., W
a During 1st 3 yr. 2 patients on injections of penicillin had 2 streptococcal infections; of those on penicillin
tablets 9 had 2 infections, and 1 had 3; of those on sulfadiazine, 13 had 2 infections, 5 had 3, and 1 had 4.
Additional data obtained since previous report ( 3 ) caused minor modifications in figures originally
published.
762
Table 3. Evaluation of fidelity of maintenance of oral prophylaxis, with the use of the interview technic.
Fidelity of
prophylaxis
Good
Questionable
Poor
Totals
Penicillin group
Sulfadiazine group
Totals
22
9
73
19
8
84
33
9
67
26
166
55
18
69
23
8
113
100
126
100
239
100
82
penicillin and in 21 percent of those on sulfadiazine. It should be noted that these results
refer to the percentage of patients who had
streptococcal infections, and are not given in
terms of attack rates per patient years. The
striking similarity of these data suggests that the
two methods of evaluation used in this study
may be considered reliable. By comparison,
streptococcal infections occurred in 12 percent
of the patients receiving injections of penicillin.
This indicates that even when good oral prophylaxis was maintained, the patients receiving
the agents by mouth had a larger number of
streptococcal infections than those given injections. Of patients in the groups receiving the
agents by mouth those who maintained good
prophylaxis with sulfadiazine had fewer streptococcal infections than those who maintained
good prophylaxis with penicillin.
Table 6 shows the number of recurrences of
rheumatic fever that have taken place in the
first three years of this study. There has been
only one recurrence in the bicillin group, an
attack rate of 0.3 percent per patient year.
There have been 15 recurrences during the
three hundred patient years on penicillin tablets, an attack rate of 5.0 percent. There have
been six recurrences in the three hundred and
Table 4. Evaluation of fidelity of maintenance of oral prophylaxis, with the use of the pill-count technic.
~
Penicillin group
Fidelity
Good
Questionable
a. N o bottle
b. Too few pills
returned
Poor
Totals
No of patients
Sulfadiazine group
Percentage
N o of patients
Percentage
Totals
No of patients
Percentage
60
55
53
44.0
113
49.0
11
10
15
12.5
26
11.5
27
25
37
31.0
64
28.0
11
10
15
12.5
26
11.5
109
100
120
100.0
229
100.0
Feiwtein et al.
763
Table 5. Streptococcal infections in the three-year period according to fidelity of oral prophylaxis as
evaluated by two technics.
Technic of
evaluation
Fidelity of
prophylaxis
Physicians
interview
I[
Pill count
Good
Not good
Good
Not good
Prophylactic
agent
No. of
patients
No. of
patients
who had
one or more streptococcal infections
Percentage of
patients who
had one or more
streptococcal
infections
Potassium penicillin Ga
Sulfadiazine
82
84
24
17
29
20
Potassium penicillin G*
Sulfadiazine
31
42
16
21
52
50
Benzathine penicillin Gb
116
14
12
Potassium penicillin Ga
Sulfadiazine
60
53
16
11
27
21
Potassium penicillin
Sulfadiazine
49
67
22
25
45
37
~a
By mouth.
By intramuscular injection
Table 6. Recurrences of rheumatic fever in the first three years of the study.
Prophylactic agent
Benzathine penicillin G
Potassium penicillin G
Sulfadiazine
Totals
Yr.
Recurrences
Rate per
patient yr., %
315
300
322
1
15
6
0.3
5.0
1.9
937
22
Patient
Average
twenty-two patient years on sulfadiazine, an attack rate of I .9 percent. The overall total attack
rate for the nine hundred and thirty-seven patient years of observation has been 2.3 percent
per patient year.
DISCUSSION
2.3
764
SUMMARY
Feinstein et al.
765
766
Macdonald et al.
~
~~
767
~~
768
100
W
Bw 10
07
It-:
bq
2 = 0.2
u)
2
;10
z
z
w
U
W
a
(r
a
W
12
18
MONTHS
MONTHS
24
130
36\
100
W
c
g
~~
vt
10
6
tZ
W
0
[
Lu
I
12
18
24
30
MONTHS
36
42
48
18
24
30
MONTHS
36
42
48
5;i'
100
W
v)
;10
<
z
I-
a
W
12
I54
WHO BOIY
Macdonald et al.
of parasitemia. Known changes in the prevalence of parasitemia can then be matched
against these specimens or templetsand when
a reasonable fit is obtained an appropriate value
of the reproduction rate can be assigned to the
natural happenings. This method has been extensively used on the basis of field data, including the three named parameters, and measurements of two successive parasite rates at several
months interval during the same season in the
same group of people. It has been found to be
readily workable and through it a complete reconstitution can be made of natural conditions
for later use in testing the probable effect of
preventive measures.
These four parameters are required because
each of them enters into the expression which
represents the dynamic curve of parasitemia.
The reproduction rate is itself a composite expression including all the directly controlling
factors, and also the mosquito numbers and the
period of extrinsic development of the parasite.
These need not be estimated separately; they
can be included in the general representation of
a reproduction rate which is much more accurately estimated by the indirect means described above than by any attempt at direct
measurement of mosquito numbers and all the
other factors involved.
A precise model requires a precise input of
data which may be very difficult to achieve. In
consequence, a large number of programs have
been run to determine the degree of error in
the final product produced by various misrepresentations of the input data. The manbiting habit and the probability of mosquito
survival always occur in the expressions jointly
as the stability index.4 Variations in this index
below the value of 1.0 have a quite negligible
effect on the total picture and there is little
effect when it is 2.0 or less. From this it can be
taken as a working rule that if the man-biting
habit is known to be under 0.1 or if the probability of mosquito survival is known to be under
0.75 the index can be given an assumed value of
1.0 and analysis continued without great consequent error, provided this assumed value is
*Stabilityindex = a/(-log, p), representing the mean number of bites on man taken by an average mosquito during its
entire lifetime, a n d determining t h e stability o f epidemiological conditions (5). For definitions of a and p, see
Annex.
769
770
Macdonald et al.
771
Figure 2. Expected effect of an 80% effective mass treatment every 2 month on malaria subject to
reproduction rates ( 2 ) shown and a stabuity index of 1.0."
loor
z = 0.2
1
L
- 0.5
z = 1.0
O = :
1
MONTHS
MONTHS
MONTHS
1
Z=
MONTHS
1.5
I0
WHO 80455
a
Deterministic version.
772
90
80
70
60
50
5 40
g
w
2
a 30
8
P
20
10
Kankiya
input data
12
18
24
MONTHS
WHO 110158
773
Macdonald et al.
F%an 4.
Expected effect of 60%,70% and 80% effective mass treatment given once a month. and once
cveq 2 monthrb PI an adJuvant to lmagicidal attack to communities living under Kankiya holoendemic
condltions.
100
100
100
80% EFFECTIVE
70% EFFECTIVE
w
v)
2 10
8
m
5z
MONTHS
100
2
t
10
12
MONTHS
loo
60% EFFECTIVE
MONTHS
1oc
L
70% EFFECTIVE
MONTHS
80% EFFECTIVE
10
3z
s"
'Top row.
bBottom row.
and village headmen. It was started as an experimental scheme in an area of some 300
square miles (777 km2) containing 52 000 people in which a comprehensive geographical reconnaissance was completed together with a
house-to-house census of all occupants. The intention was to administer a curative dose of a
combination of chloroquine a n d
pyrimethamine to the entire population every
two months for a total of seven treatments and
to spray DDT three times a year at a dosage of
the technical product of 2 gim2. This program
was started in November 1966 and a coverage
of 87.2 percent, 84.4 percent, 77.7 percent and
82.8 percent achieved in the first four rounds of
774
Figure 5. Expected effect of an 80% effective mass treatment applied as an adjuvant to imagicidal attack on a
population of lo00 living under Kankiya holoendemlc
conditions.
1000
of the parasite rate and of its subsequent increase with the predicted patterns justify the
system of design.
Study of Outbreaks of Malaria
80% EFFECTIVE
100
10
MONTHS
775
Macdonald et al.
Figure 6. Daily incidence of new cases of malaria arising from a relapsed case occurring on 23 May in an observed
epidemic near Damascus and in a simulated epidemic with parameters as shown in the text.
OBSERVED EPIDEMIC
24
'
28
13
17
21
JULY
25
29
'2
12
16
AUGUST
20
24
28'
SIMULATED EPIDEMIC
1< 45 - C O G -
0 3 3 2 -
JUNE
'
12
JUNE
20
28
'
11
5
JULY
AUGUST
wno 80459
ACKNOWLEDGMENT
References
( 1 ) Macdonald, G. Trop Dw Bull 47:907, 1950.
(2) Macdonald, G. Trop Dis Bull 47:915, 1950.
(3) Macdonald, G. Trop Dis Bull 49:569, 1952.
(4) Macdonald, G. Trop Dis Bull 49:813, 1952.
(5) Macdonald, G. Trop Dis Bull 50:871, 1953.
(6) Macdonald, G. The EpidernioloQ and Control of Malaria. London, Oxford University Press, 1957.
(7) Macdonald, G. and C . Gockel. Bull WHO 31:365,
1964.
(8) Butts, D.C.A.J Nut Malar SOG9:44, 1950.
(9) Villarejos, V.M. A m J T r q Med 31:703, 1951.
(10) Hoekenga, M.T.JAMA 149:1369, 1952.
(11) Covell, G., G.R. Coatney,J.W. Field, and Jaswant
Singh. Chemotherapy of malaria. Geneva, World
Health Organization: Monograph Series, No. 27,
1955.
(12) Clyde, D.F. Br Med J 2:1238, 1958.
(13) Clyde, D.F. Am J Trop Med Hyg 10:1, 1961.
(14) Clyde, D.F. Malaria in Tanzania. London, Oxford
University Press, 1967.
(15) Pringle, G. and S. Avery-Jones. Bull W H O
34:269, 1966.
(16) Foll, C.V., C.P. Pant, and P.E. Lietaert. Bull W H O
32:531, 1965.
(17) Foll, C.V. and C.P. Pant. Bull W H O 34:395, 1966.
776
Macdonald et al.
4 - log, PI,
(8)
777
778
.-
Macdonald et al.
Figure 8. Flow diagram for the Monte Carlo system.
Does
the random
number exceed the
inoculation
rate?
negatives still
unchallenged?
Proceed to next step
in the program
i
Repeat or another
negative
I
I
WHO 80461
779
In an eight-year controlled clinical trial of a diet high in polyunsaturated vegetable oils and low in saturated fat and cholesterol in preventing complications of
atherosclerosis, 846 men were assigned randomly to a conventional diet or to one
similar in all respects except for a substitution of vegetable oils for saturated fat.
Fatal atherosclerotic events were more common in the control group (70 v. 48; P <
0.05). However, total mortality was similar in the two groups: 178 controls vs. 174
experimentals, demonstrating an excess of non-atherosclerotic deaths in the experimental group. This was accounted for by a greater incidence of fatal carcinomas in
the experimental group. Thirty one of 174 deaths in the experimental group were
due to cancer, as opposed to 17 of 178 deaths in the control group (P = 0.06).
INTRODUCTION
780
78 1
Control
Experimental
Total caloriedday
Protein (g/day)
2496
96.3
40.1
53.5
653
2496
97.4
38.9
102.4
365
Cholesterol (mgiday)
Polyunsaturates
(% total fatty acid)
10
39.5
RESULTS
Table 2. Numbers and sites of fatal and non-fatal carcinomas and other neoplasms diagnosed during the
diet phase and the two-year period after termination of the diet.
Diet phase
Site
Control
Other
Respiratory:
Genitourinary:
Prostate
Other
Other carcinoma
6
6
13
10
0
Post-diet phase
Experimental
10
1
5
12
12
18
16
1
6
6
16
11
5
Control
1
6
3
2
0
2
1
2
0
Experimental
~-
0
3
1
3
0
35
57
12
Skin carcinomas
Other malignanciesa
Fatal benign tumor
21
3
0
10
3
0
4
1
0
59
70
17
11
Total
0
3
1
1
2
a These include lymphosarcoma, reticulum-cell sarcoma, rhabdomyosarcoma, angiosarcorna, lymphocytic leukemia, and
astrocytoma.
782
clinical events ( I ) . However, total mortality during the diet phase was not significantly different--1781422 controls compared with 1741424
in the experimental group.
During the diet phase (see Figure 1) there
were 31 carcinoma deaths in the experimental
group and 17 in the control group (x* = 3.668,
P < 0.06). The carcinoma deaths are plotted
from the time of randomization to the time of
death. In the post-diet phase the excess continued for a year (3 experimentals, 0 controls),
but in the second year the controls exceeded the
experimentals (4 experimentals, 10 controls).
Cancers, both fatal and non-fatal, counted from
the time of randomization to the time of diagnosis are summarized in Table 2. There was a
higher incidence of the more commonly occurring visceral carcinomas in the experimental
~~
Control
Experimental
group
group
(17 carcinoma (31 carcinoma
deaths)
deaths)
230
w
25 I
u)
Myocardial infarction,
definite
Cardiac decompensation
History of angina pectoris
Cerebral infarction,
definite
Age (Y.)
Serum-cholesterol
(mg/100 ml)
35
I
0
2
2
5
5
5
6
3
68.4
6
65.5
224.5
22 1.2
Table 4. Deaths from carcinoma in diet phase adjusted for cigarette smoking.
Control
Experimental
Carcinoma
observed
Carcinoma
adjusted
No. of men
Carcinoma
observed
Carcinoma
adjusted
57
13
57
129
62
18
86
6
3
1
3
1.74
0
1.66
7.02
2.61
1.03
3.24
42
7
38
173
46
19
99
1
0
4
19
4
0
3
1.15
0
5.00
16.53
4.68
0
2.79
422
17
17.30
424
31
30.15
Cigarette smoking
No. of men
Unknown
>2 packslday
1-2 packslday
%-1 packlday
<Yz pacWday
Occasional
None
Deaths are adjusted by multiplying the number of men affected in a given cigarette-smoking stratum of the control group by
(C
E)/PC, in which C = number of controls in the stratum and E = number of experimental subjects. A corresponding
calculation is made for the experimental group (3).
Adherence (%)
Control group
0-10
10-20
20-30
30-40
40-50
50-60
60-70
70-80
80-90
90- 100
Experimental group
2
1
1
0
3
3
0
2
4
1
10
2
3
17
31
3
3
4
2
1
3
* Adherence, calculated from attendance records, is expressed as a percentage of the maximum number of meals
which could have been taken in the study dining-hall.
x2 = 10.26; P > 0.3.
Control group
0-10
10-20
20-30
30-40
40-50
50-60
60-70
70-80
SO-9Q
90- 100
Experimental group
82
47
31
21
42
40
32
42
50
35
120
46
42
30
23
33
32
37
31
30
422
424
x2 = 21.78; P 4 0.01
DISCUSSION
783
784
785
786
787
The design of the LRC-CPPT has been described in detail (6).Briefly, the LRC-CPPT was
a double-blind, placebo-controlled clinical trial
that tested the efficacy of lowering cholesterol
levels for primary prevention of CHD. Twelve
participating Lipid Research Clinics (LRCs) recruited 3806 middle-aged men with primary
hypercholesterolemia (type I1 hyperlipoproteinemia) free of, but at high risk for, CHD
because of elevated LDL-C levels. The men
were randomized into two groups that were similar in baseline characteristics. The treatment
group received the bile acid sequestrant cholestyramine resin, and the control group received a placebo; both groups followed a moderate cholesterol-lowering diet. To ensure comparability of all data across the 12 clinics over a
ten-year period, a common protocol documenting all procedures in detail was strictly adhered
to by clinical personnel, who were trained and
certified in standardized procedures (7). All aspects of the conduct of the study were carefully
monitored by the Central Patient Registry and
Coordinating Center and by the Program Office. The progress of the trial and the possibility
of serious side effects were reviewed twice a year
by a Safety and Data Monitoring Board. Any
protocol violations that were identified were
brought to the attention of this board; none
were regarded by them to put the trial into
jeopardy.
Selection of Participants
788
Postrandomization Visits
Participants attended clinics every two
months, at which time the study medication was
dispensed, dietary and drug counseling was
given, and end points and possible drug side
effects, as well as possible confounding variables
such as blood pressure and weight, were evaluated. Intervention by LRC-CPPT staff was restricted to prescription of the study medication
and the diet. At annual and/or semiannual visits, resting and graded exercise ECGs, 24-hour
dietary recalls, and complete physical examinations and medical histories were obtained. All
participants initially entered were followed up
to the completion of the trial irrespective of
their levels of adherence and the frequency of
their visits.
Lipid Measurements
Lipid levels were determined with high precision and accuracy. Comparability of the measurements of the 12 LRC laboratories was
ensured by a rigorous quality control program
especially designed for the LRC Program and
maintained by the Lipid Standardization Laboratory. The lipid levels at the second screening
(prediet) visit were used as the baseline to calculate the changes in levels of total cholesterol and
LDL-C and triglyceride observed at subsequent
visits. Since the measurement of HDL-cholesterol (HDL-C) levels at the second screening
visit was not performed according to protocol at
several clinics, the levels at the first screening
visit were used as the baseline to calculate
change in HDL-C levels.
End Points
The primary end point for evaluating the
treatment was the combination of definite CHD
death and/or definite nonfatal myocardial in-
789
farction. Appendix A gives the detailed definitions of these events as well as the definition of
suspect CHD death and suspect nonfatal myocardial infarction. Other end points included
all-cause mortality, the development of an ischemic ECG response to exercise (positive exercise test result), angina pectoris as determined
by Rose Questionnaire, atherothrombotic brain
infarction, arterial peripheral vascular disease
(intermittent claudication as determined by
Rose Questionnaire), and transient cerebral ischemic attack. Detailed definitions of these
nonprimary end points have been published
elsewhere (6).
The classification of cause of death was based
on the examination of death certificates, hospital records, and interviews with physicians, witnesses of the death, and next of kin. The diagnosis of nonfatal myocardial infarction was
based on ECGs, blood enzyme levels, and history of chest pain at the time of the clinical
event. A physician at the clinic at which the
potential end point occurred classified the end
point. In addition, each potential end point was
classified independently by two members of a
blinded verification panel. If the three reviewers agreed, the diagnosis was accepted. If
there was disagreement, the case was submitted
for definitive classification to the LRC-CPPT
Cardiovascular Endpoints Committee (6). Classification of deaths not caused by CHD was also
performed by a blinded panel.
An intraoperative event was classified on the
basis of ECG changes occurring during coronary bypass surgery or other cardiac surgery or
during the recovery period extending from the
time of surgery until discharge from the
hospital.
Statistical Methods
The hypothesis of the LRC-CPPT was that
lowering cholesterol (or LDL-C) levels would
reduce the incidence of end points, and, hence,
a one-sided test was used for the main hypothesis. The statistic reported is a stratified (using
the eight baseline risk strata) log rank (MantelHaenszel) statistic (12). This statistic compares
the life-table survival (or failure) curves in the
two groups rather than the proportion of failures. In view of the necessity for periodic review, the data were analyzed many times, and
conventional methods of computing statistical
significance no longer applied. Several statis-
790
FOIIOW-UP
All men were followed up for a minimum of
seven and up to ten years. The average period
of follow-up was 7.4 years. Between May 15 and
Aug 27, 1983, contact was made with all of the
men who were still living, including any who
discontinued visits during the course of the
trial. Thus, the vital status is known for all men
originally entered into the study. In addition,
every man or a close relative was questioned
before and at the end of the study regarding previous hospitalizations for CHD or other
reasons.
Adherence to Treatment
During the first year, the mean daily packet
count for participants attending clinic was 4.2 in
the cholestyramine and 4.9 in the placebo
group, falling to 3.8 and 4.6, respectively, by the
seventh year. Adherence to the diet as determined by a 24-hour dietary recall conducted at
six-month intervals showed no important differences between the two treatment groups (Table
1). A rise of 2 kg in body weight occurred in
each group during the seven years of the study.
Maintenance of Blind
N o cases of medical emergency required the
unblinding of participants or staff and no one
asked to be told his treatment assignment.
79 1
Total calories
Cholesterol, mg
Total fat, g
Saturated fat, g
P/Sa ratio
a
Cholestyramine resin
Postentry
Re-entry
Postentry
Prediet
On-diet
1st year
7th year
Prediet
On-diet
1st year
7th year
2264
309
95
33
0.48
2023
248
79
24
0.73
2056
255
83
26
0.69
2060
264
87
28
0.67
2278
308
97
34
0.47
2027
243
80
24
0.72
2058
251
82
26
0.67
2086
288
89
29
0.66
Total cholesterol,
mg/dL
LDLa cholesterol,
mg/dL
HDLa cholesterol,
mg/dL
HDL cholesterol/total
cholesterol
Triglycerides, mg/dL
a
Cholestyramine Resin
Postentry
Pre-entry
Postentry
Prediet
On-diet
1st year
7th year
Prediet
On-diet
1st year
7th year
291.8
279.2
275.4
277.3
291.5
280.4
238.6
257.1
216.2
204.5
198.8
197.6
215.6
205.3
159.4
174.9
45.1
44.4
44.5
45.5
45.0
44.4
45.6
46.6
0.16
158.4
0.16
153.2
0.16
162.0
0.17
173.5
0.16
159.8
0.16
156.3
0.20
172.2
0.19
182.9
*
*
792
End point
Cholestyramine
resin (N=1906)
%
Reduction
in risk
90% Confidence
interval for
% reduction
in risk
z Score
No.
No.
187b
38
9.8
2.0
155b
30
8.1
1.6
19
24
. . . . . . . . .
158
8.3
130
6.8
19
. . . . . . . . .
256b
44
13.5
2.3
222b
32
11.6
1.7
15
30
. . . . . . . . .
225
71
11.8
3.7
195
68
10.2
3.6
15
7
-23
+3
+1
+32
+27
1.92~
1.80
. . . . . . . . .
+30
0.42
Percent reduction in risk is defined as (I-RR) X 10096,where RR is the incidence rate ratio of an event in the cholestyramine
group compared with the placebo, Percent reduction in risk and z score are adjusted for follow-up time and stratification.
A subject experiencing a myocardial infarction and CHD death is counted once in this category. Hence, this line is not the
sum of the following two lines.
C The .05-level, one-sided critical value of the z score adjusted for multiple looks at the data is 1.87.
Placebo
(N = 1900)
End point
Coronary disease
Positive exercise test
Angina (Rose Questionnaire)
Coronary bypass surgery
Congestive heart failure
Intraoperative myocardial
infarction
Resuscitated coronary collapse
Cerebrovascular disease
Definite or suspect transient cerebral
ischemic attack
Definite or suspect atherothrombotic
brain infarction
Peripheral vascular disease
Intermittent claudication (Rose
Questionnaire)
a
No.
No.
345
287
112
11
19.8b
15.lb
5.9
0.8
260
235
93
8
Reduction
in risk
14.gb
12.4b
4.9
0.4
25<
20c
2 1c
28
7
5
0.4
0.3
5
3
0.3
0.2
29
40
22
1.2
18
0.9
18
14
0.7
17
0.9
-- 2 1
84
4.4b
72
3.gb
Counts all events for each individual, including events occurring after a nonfatal myocardial infarction.
Percent of those without condition at baseline.
Percent reduction in risk is adjusted for stratification.
15
differences were observed in the estimated relative risk among the strata, there was insufficient
statistical evidence to claim that the treatment
was more beneficial in one stratum than in another. The cholestyramine-treated group at
seven clinics had at least 18 percent fewer primary end points than placebo-treated men. At
four clinics there was essentially no treatment
difference; only one clinic showed an excess of
events in the drug group. The statistical hypothesis of homogeneity of effect among clinics
also was not rejected; thus, the benefit of cholestyramine resin treatment cannot be attributed to effects in only a small number of
clinics.
This stratified analysis provided an estimate
of treatment benefit adjusted for baseline strata
of what were considered to be the most important CHD risk factors when the study began.
Adjustment for a more extensive list of baseline
Placebo
..*.a
Resin
793
5
6
YEARS OF FOLLOW-UP
1
9
794
characteristics, including LDL-C, HDL-C, triglyceride, age, cigarette smoking, and systolic
blood pressure, each considered as a continuous
variable, as well as exercise test outcome, was
performed by Cox proportional hazards analysis. The adjusted estimates of treatment effect
(20.0 percent risk reduction) and z score (2.05)
were slightly greater than those obtained in the
stratified analysis. There was no significant interaction of the treatment effect with any of the
seven baseline characteristics. Thus, the proportional hazards and stratified analyses both indicate that it is highly unlikely that the treatment
benefit could have arisen from inequality of the
two treatment groups with respect to CHD risk
at baseline or from a particular subgroup of
LRC-CPPT participants.
Other Cardiovascular End Points
The frequency of other cardiovascular end
points in the two treatment groups is reported
in Table 4. Each of the CHD categories having a
large number of events showed a reduction in
incidence similar to the 19 percent reduction in
the primary e n d point. T h u s , the cholestyramine group showed reductions of 20 percent (P<.OI) in the incidence of the development of angina ascertained by the Rose Questionnaire, 25 percent (P<.OOl) in the development of a new positive exercise test result, and
21 percent (P=.O6) in incidence of coronary
bypass surgery. The two cerebrovascular disease
categories did not provide a consistent or significant pattern of benefit, but the numbers were
small. For peripheral vascular disease there was
a 15 percent (P>.l) reduction in new intermittent claudication in the cholestyramine group.
None of the other differences in Table 4 were
statistically significant, possibly because of the
small numbers.
~~
All-Cause Mortality
Although the incidence of definite and of
definite and suspect CHD death was reduced by
24 percent and 30 percent, respectively, in the
cholestyramine group, that of all-cause mortality was reduced by only 7 percent (Table 3),
reflecting an increase in deaths not caused by
CHD. Table 5 , patterned after a similar table
reported in the World Health Organization
Clofibrate Trial (I@, breaks down the all-cause
mortality into major categories. None of the
differences are statistically significant. More details are provided in Appendix B. The only
noteworthy difference (P = .08) was 11 deaths
from accidents and violence in the cholestyramine group, compared with four in the placebo
group. Of these, five in the cholestyramine and
two in the placebo group were homicides or
suicides, and six in the cholestyramine and two
in the placebo group were accidents, mainly
automobile. Each of the other major categories,
including malignant neoplasms, differed only
by one or two cases.
The possibility that a CHD event could have
been the underlying cause of a violent or accidental death was examined. All of these deaths
had been evaluated by the Cardiovascular Endpoints Committee without knowledge of treatment group, and none had met the study criteria of a CHD death. Furthermore, none had
any clinical evidence suggestive of myocardial
ischemia. Subsequent to the conclusion of the
study, all of these deaths were carefully scrutinized for the possibility of a CHD event. Seven
were due to homicide or suicide, and in none of
these was there any reason to doubt the diagnosis. Autopsy information was available for
seven of the eight accidental deaths; seven of
these deaths were due to automobile or motorcycle accidents. None showed evidence of new
Placebo
Cholestyramine resin
44
3
15
32
5
16
27
4
11
36
20
71
20
68
Possible Confounders
The results described previously show that
the cholestyramine-treated group had a reduced rate of CHD. If during the course of the
LRC-CPPT there were changes in CHD risk
factors other than total cholesterol or LDL-C
levels that were not the same in the two groups,
this could pose an alternative explanation of the
observed treatment benefit. Table 6 gives the
pre-entry, first-year, and seventh-year mean values for selected variables that include the major
known risk factors for CHD. For all of these
major risk factors, the change from baseline was
similar in the two groups, and, thus, they do not
explain the treatment benefit. In addition, very
795
Pre-entry
1st year
7th year
121
80
2.6
81
120
79
2.6
81
35
122
78
2.7
83
26
25
27
51
Cholestyramine resin
37
25
30
61
24
. . .a
58
&-entry
121
80
2.6
80
38
26
31
64
1st year
7th year
120
78
2.6
80
36
25
122
78
2.7
82
27
26
28
53
. . .a
57
Abdominal pain
Belching or bloating
Constipation
Diarrhea
Gas
Heartburn
Nausea
Vomiting
At least 1 gastrointestinal
side effect
Cholestyramine resin
Pre-entry
1st year
7th year
5
10
3
6
22
10
4
2
11
16
10
11
26
10
8
7
6
8
12
7
4
34
43
Pre-entry
1st year
7th year
10
4
5
22
10
3
2
15
27
39
10
32
27
16
6
7
9
8
4
12
12
3
2
26
34
68
29
796
stones and gallbladder disease. A few more hospitalized participants in the cholestyramine
group had, as their main diagnosis, gallstones
(16 v 11) and more cholestyramine-treated participants had an operation involving the gallbladder (36 v 25), but the differences were not
significant. Gallstones, not necessarily as the
main diagnosis, were reported in the cholestyramine group slightly more frequently
than in the placebo group (31 u 30), as were
other gallbladder and biliary tract diseases (28 v
23). No death attributable to gallbladder disease
was recorded.
Appendix E indicates that the numbers of
incident and fatal cases of malignant neoplasms
were similar in the two groups: 57 incident cases
in the placebo group, of which 15 were fatal;
and 57 incident cases in the cholestyramine
group, of which 16 were fatal. The cholestyramine group had a few more malignant neoplasms in some categories (e.g., buccal cavity
and pharynx) and less in others (e.g., respiratory system) than the placebo group, but the
numbers were small. When the various categories of GI tract cancers (buccal cavity-pharynx,
esophagus, stomach, colon, rectum, and pancreas) were considered together, there were 11
incident cases and one fatal case in the placebo
group and 21 incident and eight fatal cases in
the cholestyramine group. The total number of
incident colon cancers was identical.
COMMENT
The LRC-CPPT demonstrated that treatment with cholestyramine resin reduced the incidence of CHD. This result is in agreement
Cholestyramine resin
Pre-entry
1st year
7th year
Pre-entry
1st year
7th year
71
150
355
30
6.3
6205
70
146
355
31
6.1
6178
71
149
324
35
6.3
6043
71
149
357
30
6.2
6327
82
111
371
34
5.8
6443
74
132
334
36
6.1
6299
797
with those of previous clinical trials of choles- estrogen, low-dose estrogen, and d-thyroxine)
had to be discontinued prematurely because of
terol lowering, which have shown a general
trend of efficacy for selected CHD end points. evidence of toxicity (28, 30). The nicotinic acid
However, the earlier trials have not been re- group, in which a 9.9 percent fall in cholesterol
garded as conclusive because of such factors as levels occurred, showed a 27 percent lower incidence of nonfatal myocardial infarction but litinadequate sample size, absence of a doubletle difference in fatal CHD (31).The clofibrate
blind, failure to achieve identical treatment
groups, inadequate cholesterol lowering, o r group, in which a 6.5 percent reduction in cholesterol levels occurred, had a 9 percent lower
questionable statistical procedures (20,21 ).
Several major primary prevention trials of incidence of fatal and nonfatal CHD, but statistical significance was not attained (31).Two tridiet have reported encouraging, although not
always significant, reductions in CHD inci- als of clofibrate, the Newcastle Study (32) and
dence. They include the New York Anti-Coro- the Scottish Society of Physicians Study (33),
nary Club Study (22),the Los Angeles Veterans had previously reported a suggestion of benefit,
especially in subjects with pre-existing angina,
Administration Study (23), and the Finnish
Mental Hospital Study (24).The interpretation but the post hoc use of subgroups and discordof the results of these studies, as well as second- ance in placebo group events has led to questioning of the conclusions from these two studary prevention studies using diet, is clouded by
the ascertainment bias that may result from a ies (34).
The results of these various studies of lipidnonblinded design. Because of this and other
shortcomings, these trials have also been re- lowering drugs for the prevention of CHD indigarded as inconclusive (21).Primary prevention cate that even though some evidence of reducof CHD by diet has been evaluated during con- tion of CHD has attended their use, noteworthy
current reduction of other CHD risk factors. A and sometimes serious toxicity has occurred for
47 percent lower CHD incidence was observed each drug.
in the hypercholesterolemic participants in the
Oslo Study who were treated with a cholesterollowering diet and counseled to reduce their cigSide Effects and Clinical Chemistry Analyses
arette smoking (25). The investigators attribuThe use of cholestyramine resin resulted in
ted most of the lower CHD incidence to the
cholesterol reduction. The Multiple Risk Factor several GI side effects, although these were also
Intervention Trial (MRFIT) achieved too small common in the placebo group. They were most
an overall difference (2 percent) between the evident in the initial stages of the study and
cholesterol levels of its two treatment groups to could usually be handled by symptom-specific
treatment, but sometimes they were the basis for
assess the effect of cholesterol lowering (26).
One major primary prevention trial of a cessation of, or reduction in, the drug dose.
lipid-lowering d r u g has been reported: the These side effects, which have been previously
WHO Clofibrate Study obtained a 9 percent fall noted for cholestyramine resin, reflect the
in serum cholesterol levels and a significant 20 properties of a drug that is not metabolized in,
percent reduction in the overall incidence of or absorbed from, the GI tract. The monitoring
major ischemic heart disease events, similar in of hospitalizations showed that the two treatmagnitude to the LRC-CPPT findings (18). ment groups were similar for almost all of the
However, unlike the LRC-CPPT, this decline large number of primary diagnoses and procewas confined to nonfatal myocardial infarction, dures. Of special interest is the absence of a
whereas the incidence of fatal heart attack was significant increase in gallstones or cholecystecsimilar in both treatment and control groups. tomy. This contrasts with clofibrate, which, unOf concern in this study was the increased inci- like cholestyramine resin, is known to alter the
dence in all-cause mortality in the clofibrate lithogenicity of bile and has been associated in
group, which became more significant during a the WHO and CDP trials with an increased
four-year pasttrial follow-up (18, 27).
incidence of gallbladder disease (18, 31). The
The Coronary Drug Project (CDP) was a ma- results of a systematic radiological study for
jor secondary prevention trial of several lipid- gallstones in participants at two LRCs before
lowering drugs. Three of its groups (high-dose and after the LRC-CPPT will be published.
798
-
Malignant Neoplasms
The total incidence of fatal and nonfatal malignant neoplasms was similar in both treatment
groups. When the many different categories are
examined, various GI tract cancers were somewhat more prevalent in the cholestyramine
group. Other cancers (e.g., lung and prostate)
were more frequent in the placebo group. The
small numbers and the multiple categories prevent conclusions being drawn. However, in view
of the fact that cholestyramine resin is confined
to the GI tract and not absorbed, and [in view]
of animal experiments in which cholestyramine
resin has been found to be a promoter of colon
cancer when a cancer-inducing agent was also
fed orally (35), further follow-up of the LRCCPPT participants is planned for cause-specific
mortality and cancer morbidity.
CHD End Points
The LRC-CPPT shows that treatment with
cholestyramine resin results in a significantly
lower incidence of CHD as measured by the
primary end point of the study. The benefit of
treatment was not concentrated in any one subgroup or in a few clinics, but was widespread.
Inspection of the life-table curves shows that
benefit became apparent two years after initial
treatment. This benefit was reflected in both
categories of primary end points. The findings
were not essentially altered when the men classi-
All-Cause Mortality
There was only a 7 percent reduction of allcause mortality in the cholestyramine group,
reflecting a larger number of violent and accidental deaths. Several other primary prevention trials have reported higher noncardiovascular mortality in their active treatment groups,
resulting from a variety of medical causes (36).
Excess mortality in the LRC-CPPT cholestyramine group was confined to violent and
accidental deaths. Since no plausible connection
could be established between cholestyramine
treatment and violent or accidental death, it is
difficult to conclude that this could be anything
but a chance occurrence.
799
Goal
Experience
Sample size
Duration of follow-up, yr.
Lost to follow-up
Reduction of plasma total cholesterol levels in
placebo group
Nonadherers at yr. 7
Reduction of plasma total cholesterol levels in
men adhering" to cholestyramine resin
treatment
7-yr. incidence of primary end point in
placebo group
Reduction in primary end point
3550
7
0
3806b
7-10
0
4%
35%
4.8%
27%
28%
13.9%d
8.7%
36%
8.6%
19%
800
Caution should be exercised before extrapolating the CPPT findings to cholesterol-lowering drugs other than bile acid sequestrants. It
has been shown that bile acid sequestration
leads to a substantial reduction in plasma total
and LDL-C levels by increasing the removal of
LDL from the blood through increased activity
of specific cell-surface LDL receptors (38).This
mode of action is conceptually attractive inasmuch as it represents the enhancement of a
physiological mechanism for the control of LDL
levels. The mode of action, cholesterol-lowering
potency, and possible toxicity of other cholesterol-lowering drugs must be taken into account
before their use is advocated for the prevention
of CHD.
The LRC-CPPT was not designed to assess
directly whether cholesterol lowering by diet
prevents CHD. Nevertheless, its findings, taken
in conjunction with the large volume of evidence relating diet, plasma cholesterol levels,
and CHD, support the view that cholesterol
lowering by diet also would be beneficial. The
findings of the LRC-CPPT take on additional
significance if it is acknowledged that it is unlikely that a conclusive study of dietary-induced
cholesterol lowering for the prevention of CHD
can be designed or implemented.
80 1
802
( I ) Levy, R.I. Review: Declining mortality in coronary heart disease. Arteriosclerosis 1:3 12-325, 1981.
(2) Gordon, T., W.P. Castelli, M.C. Hjortland, et
al. The prediction of coronary heart disease by highdensity and other lipoproteins: An historical perspective. In Rifkind, B.M., and R.I. Levy (eds.), Hyperlipidemia-Diagnosis and Therapy. New York, Grune &
Stratton Inc., 1977, pp. 71-78.
(3) Stamler, J. Population studies. In Levy, R.I.,
B.M. Rifkind, B.H. Dennis, et al. (eds.), Nutrition,
Lipids, and Coronary Heart Disease. New York, Raven
Press, 1979, pp. 25-88.
(4) Arteriosclerosis: A Report by the National Heart
and Lung Institute Task Force on Arteriosclerosis, Dept. of
Health, Education, and Welfare publication (NIH)
72-137. Washington, D.C., National Institutes of
Health, 1971, vol. 1 .
(5) Levy, R.I., D.S. Fredrickson, N.J. Stone, et al.
Cholestyraniine in type 11 hyperlipoproteinemia: A
double-blind trial. Ann Intern Med 79:51-58, 1973.
( 6 ) The Lipid Research Clinics Program. The
Coronary Primary Prevention Trial: Design and implementation. J Chronic Dis 32:609-631, 1979.
( 7 ) Protocol for the Lipid Research Clinics Type IZ Coronary Primary Prevention Trial. Chapel Hill, N.C., University of North Carolina Department of Biostatistics,
1980.
(8) The Lipid Research Clinics Program, Participant recruitment to the Coronary Primary Prevention
Tria1.J Chronic Dis 36:451-465, 1983.
( 9 ) The Lipid Research Clinics Program. Recruitment for clinical trials: The Lipid Research Clinics
Coronary Primary Prevention Trial experience. Circulation 66(suppl. 4): 1-78, 1982.
(10) The Lipid Research Clinics Program. Pre-entry characteristics of participants in the Lipid Research Clinics Coronary Primary Prevention Trial. J
Chronic Dis 36:467-479, 1983.
( 1 1 ) Dennis, B., N. Ernst, M. Hjortland, et al. The
NHLBI nutrition data system. J A m Diet Assoc
77 :64 1-647, 1980.
(12) Kalbfleisch, J.D., and R.L. Prentice. The Statistical Analysis of Failure Time Data. New York, John
Wiley & Sons, 1980.
(13) OBrien, P.C., and T.R. Fleming. A multiple
testing procedure for clinical trials. Biometrzcs 35:
549-556, 1979.
(14) Majundar, H., and P.K. Sen. Nonparametric
testing for simple linear regression under progressive
censoring with staggering entry and random withdrawal. Communication in Statistics-Theory and Methods
A7:349-371, 1978.
(15) Breslow, N., and C. Haug. Sequential comparison of exponential survival curves. j Am Stat Assoc
67:691-697, 1972.
(16) Tsiatis, A.A. The asymptomatic joint distribu-
803
clofibrate to lower serum cholesterol: Mortality followup report. Lancet 2:379-385, 1980.
(28) Coronary Drug Project Research Group. The
Coronary Drug Project: Initial findings leading to
modification of its r e s e a r c h protocol. J A M A
214~1303-1313,1970.
(29) Coronary Drug Project Research Group. The
Coronary Drug Project: Findings leading to discontinuation of the 2.5 mg/day estrogen group. JAMA
226~652-657,1973.
(30) Coronary Drug Project Research Group. The
Coronary Drug Project: Findings leading to further
modifications of its protocol with respect to dextrothyroxine. JAMA 220:996-1008, 1972.
(31) Coronary Drug Project Research Group. The
Coronary Drug Project: Clofibrate and niacin in coronary heart disease. JAMA 231:360-381, 1975.
(32) Group of Physicians of the Newcastle Upon
Tyne Region: Trial of clofibrate in the treatment of
ischaemic heart disease: Five-year study. Br Med J
4:767-775, 1971.
(33) Research Committee of the Scottish Society of
Physicians. Ischaemic heart disease: A secondary prevention trial using clofibrate. Br Med J 4:775-784,
1971.
(34) Friedewald, W.T., and M. Halperin. Clofibrate
in ischemic heart disease. Ann Intern Med 76:821-823,
1972.
(35) Asano, T., M. Pollard, and D.C. Madsen. Effects of cholestyramine onl,2-dimethylhydrazine-induced enteric carcinoma in germfree rats. Proc Soc
Exp Bzol Med 150:780-785, 1975.
(36) Oliver, M.F. Serum cholesterol: The knave of
hearts and the joker. Lancet 2:1090-1095, 1981.
(37) Kannel, W.B., W.P. Castelli, T. Gordon, et al.
Serum cholesterol, lipoproteins and the risk of coronary heart disease: The Framingham Study. Ann Intern Med 74:l-12, 1971.
(38) Goldstein, J.L., T. Kita, and M.S. Brown. Defective lipoprotein receptors and atherosclerosis:
Lessons from an animal counterpart of familial hypercholesterolemia. N EnglJ Med 309:288-296, 1983.
(39) Blackburn, H., A. Keys, E. Simonson, et al.
The electrocardiogram in population study. Circulation 21:1160-1175, 1960.
804
I. Definite atherosclerotic coronary heart disease death-either or both of the following categories:
A. Death certificate with consistent underlying or immediate cause plus either of the following:
1. Preterminal hospitalization with definite or suspect myocardial infarction (see below).
2. Previous definite angina or suspect or definite myocardial infarction when no cause other than
atherosclerotic coronary heart disease could be ascribed as the cause of death.
B. Sudden and unexpected death (requires all three characteristics):
1. Deaths occurring within one hour after the onset of severe symptoms or having last been seen
without them.
2. No known nonatherosclerotic acute or chronic process or event that could have been potentially
lethal.
3. An unexpected death occurs only in a person who is not confined to his home, hospital, or other
institution because of illness within 24 hours before death.
11. Criteria for definite nonfatal myocardial infarction-any one or more of the following categories using the
stated definitions:
A. Diagnostic ECG at the time of the event.
B. Ischemic cardiac pain and diagnostic enzymes.
C. Ischemic cardiac pain and equivocal enzymes and equivocal ECG.
D. A routine Lipid Research Clinics EGG is diagnostic for myocardial infarction while the previous one
was not.
111. Suspect atherosclerotic coronary heart disease death-one or both of the following categories:
A. Death certificate with consistent underlying or immediate cause but neither adequate preterminal
documentation of the event nor previous atherosclerotic coronary heart disease diagnosis.
B. Rapid and unexpected death (requires all three characteristics):
1. Death occurring between one and 24 hours after the onset of severe symptoms or having last been
seen without them.
2. No known nonatherosclerotic acute or chronic process or event that could have been potentially
lethal.
3. An unexpected d e a t h occurs only in a person who is not confined to his home, hospital, or other
institution because of illness with 24 hours before death.
IV. Suspect myocardial infarction-any one or more of the following categories using the stated definitions:
A. Ischemic cardiac pain.
B. Diagnostic enzymes.
C. Equivocal ECG and equivocal enzymes.
D. Equivocal ECG alone, provided that it is not based on ST or T-wave changes only.
Glossary
I. Ischemic cardiac pain-severe substernal pain having a deep or visceral quality and lasting for half an
hour or more.
11. ECG (classified by Minnesota Code) (39)
A. Diagnostic-either of the following must be present:
1. Unequivocal Q or QS pattern (code 1-1).
2. Q or QS pattern (codes 1-2-1 to 1-2-7), plus any T-wave item (codes 5-1 to 5-3).
Only the first criterion applies in the presence of ventricular conduction defects.
B. Equivocal-any of the following must be present:
1. Q or QS pattern (codes 1-2-1 to 1-2-7).
2. ST junction and segment depression (codes 4-1 to 4-3).
3. T-wave item (codes 5- 1 to 5-2).
4. Left bundle-branch block (code 7-1).
111. Enzymes
A. Diagnostic enzymes-all of the following conditions:
1. Creatine kinase, SCOT, or lactic dehydrogenase values determined coexistent with the event.
2. The upper limit of normal for the local laboratory is recorded.
3. The determined value for one or more enzymes is at least twice the upper limit of the local
laboratory but does not exceed 15 times that value.
B. Equivocal enzymes-all of the following conditions:
1. Creatine kinase, SCOT, or lactic dehydrogenase values determined coexistent with the event.
2. The upper limit of normal for the local laboratory is recorded.
3. The determined value for one or more enzymes is elevated but does not fulfill criteria for diagnostic
enzymes.
805
Placebo
15
5
27
0
1
3
1
1
2
0
2
Cholestyramine
resin
16
4
11
36
2
1
2
2
1
1
6
1
4
* One placebo participant died while undergoing cardiac catheterization. Two cholestyramine resin participants died of complications ensuing from mitral valve replacement
and from carotid endarterectomy.
Listed by site in Appendix E.
Three deaths (two in the placebo group) caused by pneumonia, one placebo death caused
by staphylococcalsepticemia, and one cholestyramine resin death resulting from an undetermined infectious cause.
Cholestyramine resin
Re-entry
1st year
7th year
Pre-entry
1st year
7th year
4.3
0.04
0.52
4.8
103
1.03
2.9
98
46
114
3.1
4.5
140
4.1
7.2
228
4.2
0.04
0.52
4.8
104
1.02
3.0
96
45
113
3.0
4.5
141
4.0
7.2
234
4.2
0.04
0.61
4.7
103
0.98
3.0
101
45
4.3
0.04
0.52
4.9
103
1.03
2.9
98
46
113
3.1
4.5
140
4.1
7.2
229
4.2
0.05
0.54
4.8
105
1.01
3.0
94
45
114
3.0
4.5
140
4.1
7.2
236
4.2
0.04
0.62
4.6
103
0.98
3.0
103
3.0
4.4
141
4.3
7.3
267
100
45
103
3.0
4.4
141
4.3
7.3
270
806
Cholestyramine
resin
Placebo
13
11
11
0
5
20
5
3
4
100
12
15
1
6
30
12
0
11
97
5
2
9
9
5
0
3
2
4
1
10
5
5
1
6
3
16
11
19
0
27
22
3
29
23
16
Cholestyramine
resin (N = 1906)
(N = 1900)
Primary site
~
All cases
Deathsa
All cases
Deaths
0
1
2
0
0
1
0
0
0
0
8
1
1
0
0
2
0
~~
Buccal cavity-pharynx
Esophagus
Stomach
Colon
Rectum
Pancreas
Larynx
Lung
Leiomyosarcoma
Melanoma
Other skin
Prostate
Urinary bladder
Kidney
Brain
Thyroid
Thymus
Lymphatic tissue
Hematopoietic tissue
Unknown
Total
10
1
5
5
11
3
1
1
1
0
0
1
0
1
3
1
0
0
1
0
57
15
0
6
4
3
1
6
0
0
3
7
7
2
3
0
1
4
2
1
57b
3
0
3
0
0
0
1
0
3
0
0
1
0
0
16
Four men with malignant neoplasms (two in each treatment group) died of nonneoplastic causes. They are counted among
the incident cases but not among the deaths in this table.
One cholestyramine group participant, who survived to the end of the study, had both a prostate carcinoma and a
lymphoma; he is counted only once in the total.
PART IV
DISCUSSIONS
TERRIS:
I agree, but I also think it is terribly important to differentiate between the evaluation of health services and etiological
studies. The whole history of epidemiology has been the
history of etiological studies. That has been the main emphasis. Now we are moving into an era when people want to
take a really good look at what they are doing. We have
come to the point where we now use observational studies
and experiments to conduct etiological studies and to evaluate health services. In Latin America, for example, they
want to use epidemiology to evaluate available health services, including medical care services.
LLOPIS:
TERRIS:
NAJERA:
TERRIS:
810
is traditional health service research: its sociological, political, or economic. The field of health service research has
been taken over by the medical care people, and, as a result,
we have all sorts of studies of resources, physicians, number of beds, financial issues. Thats all that anyone studies
now-costs. But the whole emphasis is wrong. What we
want to have as the keystone of health service research are
epidemiologic studies of outcomes. We should be very bold
here and say just that.
NAJERA:
TERRIS:
Discussiom
811
812
Health service research has become more and more important in the organization of health services. But without
epidemiology, health service research is just administration. It concentrates on better administration, better management techniques.
TERRIS:
BUCK: One thing I would like to see is the use of health statistics to
TERRIS:
Discussions
813
BUCK: We should stress the importance of evaluating health services in terms of specific health outcomes. When a final
outcome is too difficult to observe, we can at least examine
an intermediate outcome.
TERRIS:
TERRIS:
NAJERA:
There was also a very nice study done in the United States
which showed how medical school students lose their social
conscience. According to the study, in the first year of
medical school up to 80 percent of the students had very
strong social interests. This is why many of them had
chosen to study medicine in the first place. But by the time
they graduated, the percentage had dropped to 20.
TERRIS:
8 14
TERRIS:
They certainly are not addressed in medical schools, particularly not by clinicians. Ask clinicians what is hypertension:
Its above 140/90, they say. There is no concept that,
compared to 100/70, anything above 120/80 is really hypertension.
BUCK:
TERRIS:
~ i s c ~ s z o815
~
816
TERRIS:
Thats true. I taught first- and second year medical students and it is already too late. By then they are learning
microbiology, anatomy-big subjects. Why should they
bother with this junk, they think. That is their attitude.
And isnt it part of the problem of the future of epidemiology and public health also a prestige problem? If
you go to a school of public health you dont have the
prestige that you have in a medical school?
NAJERA:
BUCK:
Thats the problem, isnt it? Low prestige means less money.
We are so materialistic.
TERRIS:
LLOPIS:
BUCK
LLOPIS:
Discussions
8 17
~~~
BUCK
TERRIS:
My guess is that in Latin America today there is sociopolitical unrest among many epidemiologists without
enough of an epidemiological structure to back it up. That
is why you read papers full of sociological rethoric that I
have tended to deride as being talk. But it isnt just talk. I
think we are seeing the beginning of a movement. The
whole emphasis now, interestingly enough, is on doing
actual research. I think that in all these countries where
people talk so much about social epidemiology, what is
really happening is that they are not really clear on which
way to go. Someone has to assume the job of providing
them with an adequate epidemiological knowledge base.
NAJERA:
8 18
TERRIS:
BUCK:
TERRIS:
Discussions
8 19
And yet these are all examples of changes that do not reach
the whole population; they benefit only the upper class. It
will be very difficult for this type of health education and
this type of prevention to reach everyone if the structure of
the population doesnt change. Yours is not the only solution. Sure, we need prevention, we need primary prevention, its the most important thing-but we also need something else.
TERRIS:
NAJERA:
TERRIS:
NAJERA:
TERRIS:
NAJERA:
820
TERRIS:
NAJERA:
TERRIS:
NAJERA:
TERRIS:
Discwsions
821
TERRIS:
NAJERA:
Here you have something that has been successful with one
group and has failed with others. You say we need to spend
more money; that it is essential to make every effort to
reach the less educated. But the question is how? These
people have other problems that depend on other factors.
Even if they stopped smoking, what then of the factories,
the whole economy which depends on tobacco? Up to now,
the impact on smoking cessation has been minimal, so
much so that the tobacco industry doesnt mind.
TERRIS:
Not in the United States. Thats not true. Tobacco companies are in trouble.
NAJERA:
TERRIS:
You know what my proposal is? That it would be worthwhile to put a lot of money-and the United States is
wealthy enough to do it-into subsidizing farmers to get
out of growing tobacco so that they could grow other crops.
NAJERA:
TERRIS:
NAJERA:
822
LLOPIS:
NAJERA:
TERRIS:
NAJERA:
TERRIS:
NAJERA:
BUCK:
Discussions
823
almost any study you do of any kind of morbidity or mortality shows that the inverse social class gradient continues.
There are important environmental causes for this gradient, and theyre not all a matter of lifestyle. Im all for
changing lifestyles in a healthy direction. But in doing that
you cannot neglect the other causes: housing, education,
and occupation; the conditions under which people live,
learn, and work today. We may no longer have the satanic
mills of the Industrial Revolution, but we still have lots of
jobs where theres no creativity, where work is boring, and
where there is a fear of unemployment if you object to
anything. Its been shown that some of the lifestyle problems are generated by environmental problems, particularly occupational ones. That may be one of the reasons
why less educated people are not as likely to stop smoking
or to do other things that we would like them to do. Studies
by the Social Research Institute in Michigan have found
that people on piece work, people on shift work-especially
the kind that goes against the bodys natural rhythms-are
much more often smokers. So I think we have to go beyond
lifestyle. We have to consider how people live and this
comes down to what you were saying about involving the
people themselves. People have to be encouraged to associate some of their environmental problems with their
health. T h e trade-union movement has been rather slow to
do this. Their interests are only recently beginning to veer
in this direction.
NAJERA:
824
TERRIS:
NAJERA:
But it is not real community involvement. Community involvement means that the community makes decisions and
sets priorities for an overall development in which health is
one important area. It is easy to talk about this but very
difficult to achieve it.
BUCK:
TERRIS:
You both imply that the major causes of death and disease
are related to occupation, I dont buy that at all.
BUCK:
TERRIS:
BUCK:
TERRIS:
Yes, I can. Its cigarette smoking, saturated fats, hypertension, lack of exercise.
BUCK
Discussions
825
TERRIS:
It could be genetic.
NAJERA:
LLOPIS:
NAJERA:
T h e role of epidemiology is to understand, in a comprehensive way, what is happening with health. This is my
main general criticism of the lifestyles approach. It is a
partial approach because it only looks at, say, coronary
heart disease, without looking at the web of causation.
TERRIS:
826
NAJERA:
But that approach will only take you so far. Look at us,
supposedly highly educated people; we cannot move further because special interests will not allow the shutting
down of tobacco factories, tobacco growing fields, dairy
farms and industries, and so on. Besides, the majority of
the population cannot choose lifestyles. The word style
implies the possibility of choice, and choice is not an option
for more than 80 percent of the population of the world.
TERRIS:
BUCK:
TERRIS:
We dont have good hypotheses on the causes of hypertension. We really dont. All of the social class studies have
been very unproductive. There are very minor differences
by social class. The big difference is between whites and
blacks, and there are no good hypotheses for why that is so.
You say that its occupational, but you dont have any basis
for saying that at all. You have nothing that shows that.
BUCK: The reason we dont have any very solid basis is that the
NAJERA:
You were asking why I used the term comprehensive. Comprehensive epidemiological studies take into account every
imaginable factor. For instance, consider the health services. We take for granted that our health services are good
and we dont look at them as possibly iatrogenic, as factors
that may cause disease. Then we start studying, lets say,
coronary heart disease, without taking into consideration
what the health services are doing with coronary heart
disease. We also have to review many of our diagnoses.
What is hypertension? Is it a risk factor? Is it a disease?
What does hypertension really mean? We have to review
Discussions
827
and revise these diagnoses. We cannot have a partial approach; we cannot study hypertension without studying
coronary heart disease, or studying stroke, or even diabetes. We must consider the interrelation of those factors
so we can understand the web; again the web.
TERRIS:
BUCK:
Actually though, the right time to move is when the epidemiological knowledge is solid enough that it deserves to
have an effect on public opinion. In the meantime we
should be moving on the research front to areas of less
certainty but greater potential.
TERRIS:
NAJERA:
TERRIS:
828
NAJERA:
Yes, I am, but we are talking of the role of epidemiologywhere we should go, how we should study the problems
again and again. The application is public health. And it
will be real public health if it results from epidemiology, if it
is arrived at through the epidemiologic method.
TERRIS:
Geoffrey Rose
In teaching epidemiology to medical students, I have often encouraged them to consider a question which I first heard enunciated
by Roy Acheson: Why did this patient get this
disease at this time? It is an excellent starting
point, because students and doctors feel a natural concern for the problems of the individual.
Indeed, the central ethos of medicine is seen as
an acceptance of responsibility for sick individuals.
It is an integral part of good doctoring to ask
not only, What is the diagnosis, and what is the
treatment? but also, Why did this happen,
and could it have been prevented? Such thinking shapes the approach to nearly all clinical
and laboratory research into the causes and
mechanisms of illness. Hypertension research,
for example, is almost wholly preoccupied with
the characteristics which distinguish individuals
at the hypertensive and normotensive ends of
the blood pressure distribution. Research into
diabetes looks for genetic, nutritional, and metabolic reasons to explain why some people get
diabetes and others do not. The constant aim in
such work is to answer Achesons question,
Why did this patient get this disease at this
time?
The same concern has continued to shape the
thinking of all of us who came to epidemiology
from a background in clinical practice. The
whole basis of the case control method is to
discover how sick and healthy individuals differ.
Equally, the basis of many cohort studies is the
search for risk factors, which identify certain
individuals as being more susceptible to disease;
and from this we proceed to test whether these
risk factors are also causes, capable of explain-
829
830
140-
!+
2
a
130-
a 120-
z3
a
=n 100-
w
!!
p
a
=2
90-
80-
110-
.
. *.t.. ..
.j..
,0
. . I..
. I
. I e e .
1
I
(I)
70I
0.1
I
0.5
I
1.0
.
. c.
1.5
0..
*.
I
2.0
. .
ee
2.5
3.0
..
.
I
3.5
4.0
been futile. Everyone is exposed, and other factors operate to determine the varying risk.
Epidemiology is often defined in terms of
study of the determinants of the distribution of
the disease, but we should not forget that the
more widespread is a particular cause, the less it
explains the distribution of cases. The hardest
cause to identify is the one that is universally
present, for then it has no influence on the
distribution of disease.
THE DETERMINANTS OF POPULATION
INCIDENCE RATE
Rose
831
6
.
...
/\
\\
1.
60
80
I
100
120
____
140
Kenyan nomads
London civil servants
1
160
I
180
200
minants of prevalence and incidence rates, we weight. The determinants of incidence are not
need to study characteristics of populations, not necessarily the same as the causes of cases.
characteristics of individuals.
A more extreme example is provided by the HOW DO THE CAUSES OF CASES RELATE TO
population distributions of serum cholesterol
THE CAUSES OF INCIDENCE?
levels (4) in East Finland, where coronary heart
This is largely a matter of whether exposure
disease is very common, and Japan, where the
incidence rate is low; the two distributions varies similarly within a population and bebarely overlap. Each country has men with rela- tween populations (or over a period of time
tive hypercholesterolemia (although their defi- within the same population). Softness of water
nitions of the range of normal would no supply may be a determinant of cardiovascular
doubt disagree), and one could research into mortality, but it is unlikely to be identifiable as a
the genetic and other causes of these unusual risk factor for individuals, because exposure
individuals, but if we want to discover why tends to be locally uniform. Dietary fat is, I
Finland has such a high incidence of coronary believe, the main determinant of a populations
heart disease, we need to look for those charac- incidence rate for coronary heart disease, but it
teristics of the national diet which have so ele- quite fails to identify high-risk individuals.
vated the whole cholesterol distribution. Within
In the case of cigarettes and lung cancer, it so
populations it has proved almost impossible to happened that the study populations contained
demonstrate any relation between an indi- about equal numbers of smokers and nonviduals diet and his serum cholesterol level, and smokers, and in such a situation case control
the same applies to the relation of individual and cohort studies were able to identify what
diet to blood pressure and to overweight. But at was also the main determinant of population
the level of populations it is a different story; it differences and time trends.
has proved easy to show strong associations beThere is a broad tendency for genetic factors
tween population mean values for saturated fat to dominate individual susceptibility, but to exintake versus serum cholesterol level and coro- plain rather little of population differences in
nary heart disease incidence, sodium intake uer- incidence. Genetic heterogeneity, it seems, is
s u s blood pressure, or energy intake versus over- mostly much greater within than between popu-
832
3. Physician motivation
4. Cost-effective use of resources
5.
Its first advantage is that it leads to intervention that is appropriate to the individual. A
smoker who has a cough or who is found to
have impaired ventilatory function has a special
reason for stopping smoking. The doctor will
see it as making sense to advise salt restriction in
a hypertensive. In such instances the intervention makes sense because that individual already has a problem which that particular
measure may possibly ameliorate. If we consider screening a population to discover those
with high serum cholesterol levels and advising
them on dietary change, then that intervention
is appropriate to those people in particular;
they have a diet-related metabolic problem.
The high-risk strategy produces interventions that are appropriate to the particular individuals advised to take them. Consequently, it
has the advantage of enhanced subject motivation. In our randomized controlled trial of
Rose
833
834
The potential for this approach is limitedsometimes more than we could have expectedboth for the individual and for the population.
There are two reasons for this. The first is that
our power to predict future disease is usually
very weak. Most individuals with risk factors will
remain well, at least for some years; contrariwise, unexpected illness may happen to
someone who has just received an all clear
report from a screening examination. One of
the limitations of the relative risk statistic is that
it gives no idea of the absolute level of danger.
Thus, the Framingham Study has impressed us
all with its powerful discrimination between
high- and low-risk groups, but when we see
(Figure 3) (5) the degree of overlap in serum
cholesterol level between future cases and those
who remained healthy, it is not surprising that
an individuals future is so often misassessed.
Often the best predictor of future major disease is the presence of existing minor disease. A
low ventilatory function today is the best predictor of its future rate of decline. A high blood
pressure today is the best predictor of its future
rate of rise. Early coronary heart disease is better than all the conventional risk factors as a
40
30
$20
2
n
10
0
140 150
I
I
I
I
I
I
I
I
I
1
180 220 260 300 340 380 420 460 490 50
SERUM CHOLESTEROL
Rose
835
Table 3. Five-year incidence of myocardial infarction [MI] in the U.K. Heart Disease Prevention Project.
% of men
Entry characteristic
% of MI
MI incidence
rate %
32
41
12
100
7
11
cases
15
16
100
22
< 30
30-34
35-39
40-44
245
All ages
Risk of Downs
syndrome per
1000 births
0.7
1.3
3.7
13.1
34.6
1.5
numerous, they generate half the cases. Highrisk individuals aged 40 and above generate
only 13 percent of the cases.
The lesson from this example is that a large
number of people at a small risk may give rise to more
cases of disease than the small number who are at a
high risk. This situation seems to be common,
and it limits the utility of the high-risk approach to prevention.
A further disadvantage of the high-risk
strategy is that it is behaviorally inappropriate.
Eating, smoking, exercise, and all our other lifestyle characteristics are constrained by social
norms. If we try to eat differently from our
friends, it will not only be inconvenient, but we
risk being regarded as cranks or hypochondriacs. If a mans work environment encourages
heavy drinking, then advice that he is damaging
his liver is unlikely to have any effect. No one
who has attempted any sort of health education
effort in individuals needs to be told that it is
difficult for such people to step out of line with
their peers. This is what the high-riskpreventive strategy requires them to do.
Total births in
age group (as %
of all ages)
78
16
5
0.95
0.05
100
% of total Downs
syndrome occurring
in age group
51
20
16
11
2
100
836
In mass prevention each individual has usually only a small expectation of benefit, and this
small benefit can easily be outweighed by a
small risk (8). This happened in the World
Health Organization clofibrate trial (9), where a
cholesterol-lowering drug seems to have killed
more than it saved, even though the fatal complication rate was only about 1/1000/year. Such
low-order risks, which can be vitally important
to the balance sheet of mass preventive plans,
may be hard or impossible to detect. This makes
it important to distinguish two approaches. The
first is the restoration of biological normality by
the removal of an abnormal exposure (e.g.,
stopping smoking, controlling air pollution,
moderating some of our recently acquired dietary deviations); here, there can be some presumption of safety. This is not true for the other
kind of preventive approach, which leaves intact
the underlying causes of incidence and seeks
instead to interpose some new, supposedly protective intervention (e.g., immunization, drugs,
jogging). Here, the onus is on the activists to
produce adequate evidence of safety.
CONCLUSIONS
Rose
(3) Reid, D. D., G. Z. Brett, P. J. S. Hamilton, et a!.
Cardiorespiratory disease and diabetes among middle-aged male civil servants. Lancet 1:469-473, 1974.
(4) Keys, A. Coronary Heart Disease in Seven Countries. American Heart Association Monograph Number 29. American Heart Association, New York, 1970.
(5) Kannel, W. B., M. J. Garcia, P. M. McNamara, et
al. Serum lipid precursors of coronary heart disease.
Human Pathol 2:129-151, 1971.
(6) Heller, R. F., S . Chinn, H. D. Tunstall Pedoe, et
al. How well can we predict coronary heart disease?
837
Findings in the United Kingdom Heart Disease Prevention Project. Brit MedJ 288:1409-1411, 1984.
(7) Alberman, E. and C. Berry. Prenatal diagnosis
and the specialist in community medicine. Community
Med 1:89-96, 1979.
(8) Rose, G. Strategy of prevention: lessons from
cardiovascular disease. Brit Med J 282: 1847-1851,
1981.
(9) Committee of Principal Investigators. A co-operative trial in the primary prevention of ischemic
heart disease. Br Heart J 40:1069-1118, 1978.
I A . PREVENTIVE SERVICES I
EVALUATION OF 1954 FIELD TRIALS OF POLIOMYELITIS
VACCINE
T. Francis, Jt,J.A. Napier, R. B. Voight, F. M. Hemphill, H. A. Wenner, R. F. Korns,
M. Boisen, E. Tolchinsky, and E. L. Diamond
PLAN OF STUDY
of
1. Observed Control Study. The plan of procedure announced by the National Foundation
for Infantile Paralysis and its Advisory Committee was to administer vaccine to children in the
second grade of school; the corresponding first
and third graders would not be inoculated but
would be kept under observation for the occurrence of poliomyelitis in comparison with the
inoculated second graders. This has been designated the Observed Control study. The plan
poses difficulties to objectivity in that knowledge of the vaccination status of a patient is
readily determinable and the introduction of
even unintentional bias can result; its adequacy
would depend upon a high incidence and severity of disease, a high degree of effectiveness of
vaccine, and the care with which the data were
collected. A number of states, however, had already made their decision to participate on this
basis. When finally arranged, the procedure
was followed in 127 areas of 33 states with total
population in the first, second, and third grades
of 1 080 680. The areas represent complete
counties with exception of Kansas City and the
town units of Connecticut.
2. Placebo Control Study. Since the problem
was to measure the degree of effectiveness, if
any, of an untried product, it was important to
have data which could provide an accurate
gauge of the effect, free of possible bias in diagnosis and reporting. There was introduced,
therefore, a second plan corresponding in pattern to that usually employed in scientific investigations. Children of the first, second, and
third grades would be combined. One half
would receive vaccine; the other matching half,
serving as strict controls, would receive a solution of similar appearance which should have
no influence on immunity to poliomyelitis. Each
838
Francis et al.
It is obviously essential in a study of a phenomenon arising in a limited, selected population that the members of that population be
clearly recorded, and their status with respect to
the study be established. They constitute the
denominator against which all effects are to be
measured; hence, any child in the group who
839
Number
Percent
749 236
100.0
455 474
60.8
200 745
201 229
26.8
26.9
8484
8577
1.1
1.1
36 439
4.9
280 868
37.5
12 894
1.7
Complete series of
injections:
Vaccine
Placebo
Incomplete injections:
Vaccine
Placebo
Absent at first clinic or
withdrew
840
Second
bleedings
No.
Third
bleedings
No.
100.0
50.1
49.9
11 870
100.0
39.2
60.8
20 931
a
a
a
a
Francis et al.
a reinforcing effect in persons who already possessed some antibody. The specimens from control subjects would also provide a base line for
measuring any concurrent changes which
might take place in the population through natural exposures, and permit accurate conclusions as to the antigenic activity of the vaccine.
It was also urged that blood specimens be obtained from the same persons again at the end
of the poliomyelitis season of 1954; this was
carried out in November. The laboratories
would be uninformed as to whether the specimens were from a vaccinated or control child
and the tests would largely be done on an unselected basis. . . .
5. Reactions. Because of the natural concern
with reactions which might occur with a new
product in large-scale use, the Manual of Procedures emphasized the need to record any untoward reactions which appeared to be related
or coincidental to the administration of vaccine.
A further notice from the Evaluation Center
asked for notification by phone of any circumstance of significant severity which might be
related to the inoculations.
During the course of the vaccination of children in Pittsburgh by Dr. Salk, a concurrent
study was carried out in March and April of
1954 by the Evaluation Center concerning
cause of absenteeism in vaccinated children as
compared with their classmates who had not
received vaccine. The bulk of these data had
been collected by Dr. Salk before the Center
entered the study. They were analyzed and, in
addition, a system was developed for recording
absences from school and verifying the cause of
absenteeism through the school health service.
All grades
- Total
Requests
Second grade - Requests
Complete vaccinations
Incomplete vaccinations
Absent or withdrew
First and third grades - Requests
A11 grades - Participation not requested
Participation not recorded
841
Number
1 080 680
567 210
245 895
221 998
9904
13 993
321 315
332 870
180 600
Percent of
second grade
69.2
62.4
2.8
3.9
Percent of
total population
100.0
52.5
22.8
20.5
0.9
1.3
29.7
30.8
16.7
842
the same children would again receive their inoculations from the specific vials with the same
code numbers as before. This required careful
attention to procedure and to identification of
each child. Provision had to be made for consultation with the Evaluation Center for replacements or additions when shortage developed, without revealing the code or the nature
of material in question.
The classroom groups of the first, second,
and third grades constituted the basic clinic unit
and remained so through the three inoculations. Consequently those receiving vaccine or
the corresponding placebo were segments of
the same group and the controls were clearly
designated. In general, only one lot of vaccine
or placebo was used in a placebo study area or
in a school. Because of supply problems this was
not completely uniform, and in 14 placebo
areas more than one lot was employed. Even
then, the uniformity in the 206 schools of those
areas was maintained and the controls for each
lot of vaccine were clearly designated on the
22 997 records according to corresponding
codes.
In 346 schools, with small classes, the numbers were insufficient to complete the use of a
10 cc vial-the required procedure; thus, 1529
children received only placebo. This was balanced by 1761 children in 387 other small
schools who received only vaccine. But it is
important to emphasize that the equal distribution of vaccine and placebo held at every level.
Moreover, according to our records, out of a
total of 1 237 446 injections in 419 035 persons
inoculated in placebo areas, only 748 persons
received a mixed series of vaccine and placebo.
Obserued Areas
In observed study areas where only those second grade children whose parents requested
participation were vaccinated, the problem of
establishing the control population was more
complex. After careful consideration of various
alternatives, it was decided that the total first
and third grade study population compared to
the vaccinated second grade population would
be the most critical measure that could be applied to measure the efficacy of the vaccine.
This was also the original premise on which
various state health offices agreed to participate
in the study. In observed areas there was a
Francis et al.
Stuff Supemision
During the vaccination period, April 26 to
June 15, members of the Evaluation Center
staff visited 32 of the 44 participating states to
inspect or correct procedure and to ensure
proper recording and understanding. It was
necessary to prepare additional memoranda
specifically restating the requisite information
and the need for uniformity and completeness
of the data to be supplied in the records. The
problems of vaccine replacement and the completing of inoculations in migrant children were
also handled. In the field more than 150 000
persons assisted in the total program, including
clinicians, epidemiologists, health officers,
physical therapists, public health nurses, virologists, school teachers and administrators, and
local volunteer recorders and clerks.
843
Instructions for the reporting and investigation of cases of poliomyelitis occurring in the
study population were reinforced with specific
memoranda, and forms for use in this phase of
the field trial were drawn up and distributed to
the trial areas.
Identification and Verification of Reported
Cases in the Study Population
The second phase of the study was that of
determining the occurrence of poliomyelitis
among members of the study population, of
identifying the patient, and of establishing diagnosis by integration of clinical, epidemiologic, and laboratory data which could
contribute reliable, objective information. It was
apparent that with the large number of different professional personnel involved in the widespread study, variability could be expected. By
seeking uniformity of understanding and performance it was hoped that the variations could
be reduced, and that 211 different patterns
could be avoided even though the Evaluation
Center was completely dependent upon the
ability, willingness, and collaboration displayed
at the local level where the cases of poliomyelitis
would occur.
To reduce the need for discrimination locally,
to encourage completeness of reporting and
uniformity of procedure, a single plan was instituted for the investigation of all cases in the
entire study population; that is, cases arising in
any member of the first, second, and third
grades during the spring of 1954, whether inoculated or not, whether participation had been
requested or not. Every case, paralytic, nonparalytic, suspect or doubtful was to be reported to the Center and subjected to the same
degree of study, recognizing at the same time
diagnostic problems of the early acute phase
and variation in medical interpretation. All
cases in families containing a member of the
study population were also to be reported but
subjected to limited investigation except as special studies.
Reporting of Cases
Because there may reasonably be delay between the onset of a case of poliomyelitis and its
diagnosis as poliomyelitis, it was urged that the
local agencies try to reduce any further delay in
844
delays in notification and investigation inevitably occurred. Not more than a half dozen cases
were disclosed to have escaped the established
procedures for identification and these were essentially migrant children.
Investigation of Case
Upon report of a case of poliomyelitis, the
health department needed to arrange promptly
for the successive steps in investigation. Since a
high percentage of patients was admitted to the
hospitals (88.9 percent in placebo, 85.9 percent
in observed areas) not always in the same
county, various means of carrying out the requirements were devised. The responsibility,
however, remained with the local director of the
study. In a number of instances the state or
district authorities assumed all responsibility.
Great assistance was provided in a number of
areas by the Epidemiologic Intelligence officers
of the USPHS assigned through Dr. Alexander
Langmuir to state health departments and laboratories participating in the study.
1 . Clinical-Epidemiological Report. As soon
after onset-or report-as possible, a summary
of the clinical history and examination, including spinal fluid, of the case was to be made on
the proper form (FT-6) together with the diagnosis at that time. On the same form the family
history with respect to concurrent illness or previous poliomyelitis was to be recorded. Administration of gamma globulin to family members
was also to be recorded. This information constituted the first official record received at the
Evaluation Center and contained the tentative
diagnosis. It provided also a basis on which to
proceed with subsequent studies or to eliminate
the patient from consideration if obviously his
illness was due to another cause. Much of this
was done by staff physicians but public health
nurses also contributed a large part. No case
was accepted as part of the study unless this
record was received; thus, for every case disclosed by any means an FT-6 record was obtained.
Francis et al.
collection. Copies of the forms were to be sent
to the Center-if not received, it was assumed
the specimens had not been collected and the
health officer was notified of the deficiency.
After three to four weeks a second specimen
of blood was to be obtained and, together with
the first sample, antibody titrations for diagnosis of poliomyelitis were made. The stool was
to be studied for the presence of poliomyelitis
or other virus. In some areas studies of familial
associates were also carried out by the laboratory and the local health authorities.
The complete laboratory report was made
directly to the Center on Form FT-10, although
preliminary reports were sometimes made by
letter.
At first, the same serological procedure was
employed for the testing of sera of patients in
the study areas as was used for study of pre- and
post-vaccination sera. Later it was agreed that it
should be more sensitive and titrations were
made with twofold dilutions and four tubes per
dilution to reach the endpoint. In some instances this was preceded by a preliminary
screening. In addition, Dr. Wenner and Dr.
Frisch provided to all laboratories supplies of
standard human sera with titers in the range
ordinarily encountered in man. Additional spot
checks for variability in laboratory results were
made at different times by a single pair of sera
sent to all laboratories.
Virus isolations, from stool specimens primarily, were attempted in monkey kidney cells
or HeLa cells and generally a negative specimen was retested at least once. The information gradually accumulated to indicate that
HeLa cells might be less sensitive and again
exchange of specimens among a series of laboratories was made to test this probabilitywhich proved to be correct. Subsequently, when
HeLa cells were used, passages were to be made
three times before considering the test negative.
In some instances specimens were also tested by
the inoculation of monkeys. In a number of
instances serologic tests were carried out against
mumps, choriomeningitis, and encephalitis viruses as well.
In addition, the laboratories frequently
played a large part in collection of specimens, of
maintaining the attention of health and hospital
personnel in the continued need for proper
specimens and prompt delivery. There is little
doubt that the present experience will have
845
846
~~
were taken to correct them. Physiciansinterpretation and signature as required on the FT-7
and FT-8 forms were frequently missing.
Therefore, a form letter was developed which
provided a convenient means for promptly requesting the missing information. Each participating laboratory was sent a list of all cases
on record in their area and all reported specimen collections. Thereafter, a postcard reporting this information to the laboratory on all
newly reported cases became standard operating procedure so that the laboratories would
have a complete record of cases from whom
specimens should be anticipated.
A system of logging each case by area was also
put into effect at the Center. The log served
several purposes. It provided a complete and
4 . Fatal Cases. It was requested that any fa- compact record wherein receipt of the various
tality in the study population be reported by forms required by the follow-up could be
telephone to the Evaluation Center and that posted. At intervals during the latter part of the
every effort be made to obtain a complete au- year a summary of the incomplete records in
topsy, central nervous system tissue, and other the log was mailed to each state health officer
specimens for laboratory study. Early in May, a for any cases in his areas.
In an effort to point up the importance of
special message was sent to all study areas asking that any fatal cases in inoculated children u p getting qualified review and interpretation of
to four weeks after last injection be given special reported cases on the FT-7 and FT-8 forms a
attention. The Evaluation Center and the re- general letter was mailed to all state health ofgional laboratory were to be notified immedi- ficers on August 27, re-emphasizing the need
ately, and a complete post-mortem examination for a specialists comments and interpretation
should be sought and made by a well-qualified on the back of the physical therapists forms and
pathologist with specimens provided to the lab- asking that the professional status of the physioratory. Clinical and epidemiological reviews cian be indicated below his signature. In this
were to be made. The Center.offered to assist in letter the states were again reminded to furnish
obtaining the desired personnel if necessary the Center the names and qualifications of the
and to meet expenses incurred in fulfilling the specialists they had selected for this purpose if
they had not already done so.
requirements.
Late in September it became evident that on
5 . Steps to Maintain Completeness of Znuestiga- receipt of case records attention to all phases of
tion. The substantial amount of correspondence the investigation should be re-emphasized to
required very early in the trials concerning the field and that the waiting period for each of
proper field follow-up and collection of spec- the forms should be reduced. A form (VEC-31)
imens pointed up the fact that clarification of was developed for this purpose as a Special Atthe needs was necessary beyond the printed in- tention note sent on each case, showing the
structions. Meetings were held in May and June dates on which each report was due in the office
in study areas with state and local health offi- of the Evaluation Center. The request for specials, state and local NFIP representatives, and cialists comments, signature, and status was
interested hospital personnel in 32 of the 44 stressed by a quote from the VEC letter on this
states in the trials. At these meetings staff mem- subject clipped to the VEC-31 forms.
After the middle of November lists again
bers of the Evaluation Center reviewed the requirements for follow-up and the local plans for were prepared showing missing data and sent to
state health officers requesting immediate acthis phase of the operation.
By July the pattern of the deficiencies in case tion. Telephone calls were made to a number of
studies had become quite apparent, and steps the areas.
residual injury can be more accurately measured. The specialists interpretation of the entire character and course of the case would be
expected at this time as well. This form (FT-8)
transmitted to the Center by way of the State
Health Department, was the last of the field
investigations which provided the basis for clinical classification of the patient.
The cross-responsibilities and communication difficulties often resulted in delays in receipt of the reports, even when completed expeditiously by the physical therapist. T h e
consultants diagnosis and comments were often
difficult to obtain and the records of muscle
evaluation not infrequently were simply not forwarded promptly.
Francis et aE.
The physical therapists examination was recorded on two forms (FT-7 and FT-8) which
listed muscles or muscle groups on the basis of
their anatomical mass, rather than by functional importance. The examiner was to enter a
standardized estimate of the degree of impairment for each unit, grading from normal
847
848
the data from other clinical reports and laboratory data would be integrated to furnish a final
diagnosis. In 37 instances (9 percent) in the
placebo areas the 10 to 20 day examination of
the musculature was not done, and in 11 (3
percent) the second was not obtained; in the
observed areas 81 (14 percent) first examinations and 19 (3 percent) of the second were not
done.
Score
0
1-19
20-89
90-199
200 +
Serology not
done or not
collected
Areas
Cases
No.
No.
Placebo
Observed
428
585
56
101
13.1
17.3
80
92
18.7
15.7
No lab
reports
No.
14
3.3
7.5
44
Francis et al.
Reported
isolations
Total
Virus type
I11
11
___ ___
No. %
No. % No. %
No.
obtained 14
ob-
14 days or
Total
Total
0-4
5-9
10-14
15-29
30 or more
358
76
147
76
44
13
4x rise;
homologous
type only
No rise or
irregular
changes
4x rise to
more than
one type
% 4x rise;
homologous
type only
145
33
71
28
11
191
38
65
45
33
10
20
5
11
3
0
40.7
43.4
48.3
36.8
25.0
15.4
849
850
myelitis might occur in a patient who had earlier acquired antibody against a single type of
virus, but it can be accepted that if the acute
illness now observed was poliomyelitis the type
of virus is, by these criteria, serologically indicated.
The following criteria for interpretation of
serologic results were drawn up:
Table 6. Serologic rises in cases with virus
isolation according to level of first serum
specimen..
Level of
first serum
16 or
32 or
64 or
128 or
Total
46
less
less
80
108
less
more 150
4x rise;
homologous
type only
No rise
4x rise
40
62
79
51
6
18
29
99
87
77.5
73.1
34
Francis et al.
The data from many investigations have indicated that infection with poliomyelitis virus occurs at an earlier age in lower socioeconomic
groups than in the higher. Certain factors related to these characteristics might be involved
in the decision as to whether parents requested
85 1
852
3. Mothers of participants were more likely accuracy of data on participation status of first
to spend two or more evenings a week in outside and third grade children varied greatly from
activities than were mothers of non-partici- area to area.
pants.
4. Mothers of participants were more likely
SUMMARY OF ESTIMATES OF
to have completed high school than mothers of
EFFECTIVENESS OF VACCINE
non-participants.
5. A much smaller percentage of participants
In an effort to give summary expression to
had family incomes under $4500. Participation
rate increased steadily with increasing income. the differences observed between the incidence
6. Finally, the interviewersrating of the qual- of poliomyelitis in vaccinated and control subity of the respondents neighborhood and con- jects through the different stages of analysis
dition of his house was highly correlated with which have been presented the accompanying
participation status. Participants lived in better table was compiled. It is arranged in four stages
neighborhoods, and their homes were better of analysis which sought progressively to elimikept.
nate certain cases which might be considered
less conclusively established as poliomyeliAnother refinement in the analysis dealt with t1s. . . .
As a first step, the data comprising the total
the extent of participation according to the
same characteristics but subclassified as to cases reported, total cases of poliomyelitis, total
whether the child was a member of a high, non-paralytic, and total paralytic in both
medium or low participation school. These placebo and observed control areas were subfindings have interest in connection with the jected to this examination. Through these
question of motivation but add little clarifica- stages of purification there is a progressive intion of the basic question. The results together crease in the percentage of effectiveness diswith an appreciation of the possible existence of played. There was no significant difference in
other unknown factors bearing on poliomyelitis the non-paralytic cases and any estimate of posincidence, which might influence to a different sible effect on their rate of occurrence was unreextent the participants and non-participants, liable. When they and the cases considered not
led certain consultants of the Evaluation Center to be poliomyelitis are removed so that paralytic
to suggest that in observed control areas the cases only remain, an estimate of 72 percent
unvaccinated members of the second grade be effectiveness was obtained in the placebo areas
combined with the vaccinated population to and 62 percent in the observed areas.
Because the data indicate that the cases classiform the test group for comparison with the
combined first and third grade populations as fied as bulbo-spinal paralytic are somewhat
controls. However, it was deemed preferable by more definite clinically and yield a higher perthe staff of the Evaluation Center, before analy- centage of virus recoveries, the second stage was
ses were begun, to employ as the control popu- to separate the spinal paralytic cases from the
lation in observed areas that which was origi- bulbo-spinal. The number of pure bulbar cases
nally announced to the states at the time they was too small to work with properly. In placebo
enrolled in Lhe study, namely, a comparison of areas effectiveness calculated against spinal parthe polioniyelitis experience in vaccinated sec- alytic cases was 60 percent with a lower limit of
ond grade children with the combined experi- 39 percent; against the bulbo-spinal patients it
ence in the total first and third grades. It is was 94 percent with a lower limit of 81 perdifficult to combine incidence in vaccinated and cent-an extremely successful effect. In obunvaccinated populations and have a clear view served areas the calculated effectiveness was
about the same against the spinal cases but only
of the effect in the vaccinated alone.
Other alternative procedures were consid- 50 percent and a lower limit of 19, in the bulboered and discarded, such as the use as control spinal group. These variations emphasize the
that portion of the first and third grades which influence of small numbers in addition to any
signified willingness to participate, through differences in severity of risk among placebo
submission of a parental consent form. This and observed study populations.
would seem to be the ideal group for comBecause the group classified as paralytic poparison with the vaccinates. Unfortunately, the liomyelitis, especially in the mild grades, may
Fruncis et al.
853
Vaccd
Controls
Significance
level
82
162
<0.001
49
36
76
439
Total poliomyelitis
All paralytic
All non-paralytic
57
33
24
142
115
27
<0.001
<0.001
NS
60
72
49
61
56
38
18
28
70
36
<0.001
<0.001
60
94
39
31
45
23
<0.001
<0.001
82
91
15
70
<0.001
Type I
Type I1
Type 111
13
0
2
39
6
25
<0.001
<0.05
<0.001
Diagnostic
classification
11. Paralytic-Spinal
Bulbo-Spinal
111. Laboratory confirmed
Spinal
Bulbo-Spinal
No. of cases
Est.
Lower
limit
No. of cases
Vaccd
Controls
Significance
level
Percentage
effectiveness
Est.
Lower
limit
<0.001
44
32
391
330
61
<0.001
<0.001
NS
54
62
32
51
-
20
15
199
100
<0.001
<0.001
66
50
53
19
65
68
7
9
127
71
<0.001
<0.01
83
60
64
23
80
65
20
210
<0.001
69
56
68
100
92
41
33
14
114
<0.001
<0.01
<0.001
62
80
78
33
33
47
12
34
62
The difference in distribution of non-paralytic cases is not significant at any stage of analysis.
Because of small numbers the bulbar and fatal cases are omitted.
All fatal cases occurred in controls.
contain cases which are not related to poliomyelitis virus, the next tests were conducted with
cases which had been demonstrated by laboratory studies to have undergone infection with
poliomyelitis virus. They represented a higher
degree of confidence in diagnosis. The cases
were again divided into spinal and bulbo-spinal
groups. The effectiveness of vaccine measured
against the incidence of spinal-paralytic cases
was the same in placebo and observed areas, 82
and 83 percent respectively, and the corresponding lower limits of estimated effect were
65 and 64, respectively. Enforcement of laboratory criteria apparently eliminated a substantial
number of cases which were less influenced by
vaccination and in reality, may contain cases
which are not poliomyelitis. As in the preceding
analysis, the effectiveness measured by bulbospinal cases in placebo areas was 91 percent but
that in observed areas was but 60 percent and
the significance of difference between vaccinated and control groups was sharply reduced.
The next step was to seek a measurement of
the effectiveness of vaccine in all cases of poliomyelitis from whom a specifically identified
854
Figure 1. Trend of paralytic cases among vaccinated children and their control groups
June 19, 1954 to December 31, 1954.
OBSERVED AREAS
PLACEBO AREAS
120
(I)
ul
(I)
=
l
0
#'
VACCINE
UNSAFE
VACCINE
UNSAFE
EFFECTIVE
/,
0
40
80
120
NUMBER OF PLACEBO CASES
*$-,;
t\e\d
,<\a\
100
200
300
400
NUMBER OF CONTROL CASES
855
856
source of information regarding the communicable disease problems of the nation. The
surveillance program has developed out of
this need and has been shaped by the various
problems and emergencies that have been
encountered.
Summaries of experience with four diseases-malaria, poliomyelitis, influenza and
hepatitis-will be illustrative.
THE MALARIA STORY
The first major national program that the
Communicable Disease Center undertook was
the Malaria Eradication Program, which was
based on use of DDT indoor residual spraying
in each of the 13 traditionally malarious states.
Long experience and extensive surveys during
the mid-1930s had established malaria as a
deeply rooteq endemic problem in rural sections of the Atlantic and Gulf Coastal Plains and
of the Mississippi Delta area. Furthermore,
many thousands of veterans of World War I1
returning from Africa, the Mediterranean, and
the Pacific theaters were introducing peculiarly
persistent, relapsing strains of Plasmodium viva.
Many of these veterans while on furlough, or
after separation from the service, were returning to areas of long known endemicity. The fear
of a resurgence or cyclical return of malaria was
real to many experienced malariologists. The
availability of DDT presented a challenging opportunity. A major control program was undertaken in collaboration with the states.
Actually, this extensive DDT spraying program was in operation well before effective epidemiologic services had been organized. A full,
current evaluation of the extent of the problem
had not been made. When epidemiologic studies were initiated in 1947 quite surprising findings were revealed. Simple appraisals of the
morbidity and mortality reports indicated that
these traditionall measures were grossly erroneous. The highest reported incidence occurred in Mississippi, South Carolina, and
Texas, but these three states followed the practice of requesting weekly reports of malaria by
numbers of cases seen rather than by name of
individual patient. Such a system encourages
exaggeration. Efforts to substantiate the diagnoses by laboratory study consistently failed.
Beginning in 1947, Mississippi changed its reporting requirements. The incidence dropped
857
Langmuir
from 17 764 cases to only 914 in the first year,
and appraisals among even these cases revealed
that only a very few could be verified.
It became imperative to adopt new criteria to
evaluate the presence of malaria. Laboratory
findings were sufficient to confirm the existence of a bona fide case, but the occurrence
of two or more cases in an epidemiologic relation to each other was the ultimate basis for
determining endemic presence. These new
criteria soon revealed that malaria had disappeared as an endemic disease from the South,
probably before the DDT program had gotten
underway (3).
The major events of this extraordinary malaria story are revealed in Figure 1 . The high,
sustained level of 100 000 cases a year during
the 1930s is generally believed to represent
gross underreporting. Extensive WPA surveys
at that time confirmed parasitemic rates or
spleen indexes of 50 percent or higher, indicating extensive, largely unreported endemic infection.
As just mentioned, however, at some time between 1935 and 1945 malaria mysteriously disappeared, much more rapidly than either the
reported morbidity or the mortality indicated.
The slight rise in the morbidity curve in 1945
reflects the influx of infected veterans, but the
steep decline in 1947 and subsequently is artificial and reflects the elimination of erroneous reports. The major decline clearly occurred
earlier.
The sharp peak in the curve in 1951 and
1952 reflects the Korean War and the occurrence of several thousand cases among veterans
who returned before the primaquine treatment
for all returning servicemen was instituted.
In the past five years the incidence of reported malaria in the whole country has remained below 100 cases a year. Now, the states
carry out a detailed epidemiologic investigation
on each case and report to the Communicable
Disease Center as part of an established and
continuing Malaria Surveillance Program. Almost all the cases can clearly be accounted for as
Figure 1. Reported malaria morbidity and mortality in the United States, 1932-1961. (Source: Cases and Deaths,
National Office of Vital Statisties.)
lo
1932
1934
1936
19%
1940
1942
1944
1946
1948
YEARS
1950
1952
succeeding years starting with 1958 the total reported deaths were as follows: 1958 - 0.1959
1954
1956
1958
1960
- 1,1960 - 0,1961 - 1.
1962
858
importations. They occur throughout the country without relation to previously endemic
areas. Only an occasional confirmed case arises
that is explainable in no other way than being of
indigenous origin.
The malaria story in this country should be a
sobering one to epidemiologists and health administrators. Here, a disease deeply entrenched
in large areas of the South disappeared within a
single decade, largely, if not wholly, through
natural processes rather than through planned
public health measures. Even the fact of its disappearance was not appreciated until after a
major control program had been in operation
for several years. This experience was a major
factor emphasizing the necessity of a more current and comprehensive system of surveillance.
POLIOMYELITIS
Langmuir
addition, a weekly summary and analysis of the
findings was prepared and distributed. These
reports and analyses provided regular materials
for news releases issued in the Office of the
Surgeon General.
On April 27 the decision to request withdrawal of the vaccine was made on the basis of
reports of six vaccine-associated cases. On the
evening of May 1, four days later, when the first
surveillance report was issued, 22 cases, associated with one brand of vaccine, had been recorded, and, in addition, three cases had been
associated with a different brand.
On May 7 the count had risen to 42 cases
associated with one brand and seven cases associated with other vaccines. These cases were
sharply concentrated in California and Idaho,
where most of the particular vaccine had been
used in clinics for schoolchildren in the first and
second grades. However, a small amount of the
vaccine had been distributed commercially
throughout a large part of the country, and
cases associated with the vaccine had been reported from 11 additional states.
The common-source character of the problem was becoming abundantly clear. More than
80 percent of the cases could now be related to
one brand of vaccine, which comprised less
than 10 percent of the total vaccine distributed.
The small number of cases reported in association with other brands of vaccines occurred at a
frequency well within normal expectancy for
the time ofthe year. These did not reveal localization of first paralysis to the site of inoculation, and the intervals from inoculation to onset were scattered and did not peak in a normal incubation-period curve. A relation of
these vaccines to the cases, therefore, was not
indicated.
Thus, each day as more information accumulated, the problem came into sharper focus. It
became increasingly clear that a particular
brand of vaccine was involved rather than the
vaccines of all manufacturers. Although a thorough review of all production and safety procedures was clearly demanded and, in fact, was
already in progress at the time, the other vaccine manufacturers were encouraged to stay in
production. If the surveillance program had
not been in existence and the epidemiologic
pattern had not been so distinct, many and perhaps all vaccine manufacturers would probably
have ceased production altogether.
859
860
10
INCUBATION PERIOD IN 32
CYNOMOLGUS MACAQUES (BODIAN)
s 6
z 4
2
0
10
15
20
25
this period
15
(I)
20
25
APRIL
30
10
VACCINATED CASES
(61 PARALYTIC CASES)
$ 8
Q
0
lL6
5 4
m
3
=
15
20
APRIL
25
30
10
15
MAY
20
10
25
30
FAMILY-CONTACT CASES
(80 PARALYTIC CASES)
$ 8
F0 6
rr4
w
rn
2 2
25
APRIL
30
10
15
MAY
20
25
30
10
15
JUNE
20
25
30
Langmuir
was not known precisely, but the vaccine first
became available on April 14 and was withdrawn on April 27. About 400 000 doses were
distributed during this interval. For simplicity, a
horizontal distribution of use of this vaccine was
assumed.
By the end of the first week of May the shape
of the epidemic curve of the 42 cases so far
reported justified the conclusion that a total of
less than 100 vaccine-associated cases would occur. By the end of the second week of May,
when nine family-contact cases had been recorded, a rough prediction that the total would
reach or somewhat exceed 100 was made. Actually, the final totals were, for vaccine-associated
cases, 79, with 6 1 paralytic, and for family-contact cases, 105, with 80 paralytic (5).
Much attention has been devoted to the poliomyelitis events of the spring of 1955 for two
reasons: they have been reported from the epidemiologic point of view only to a limited extent; and the experience during the emergency
and the relations that developed of necessity
between the Communicable Disease Center and
the states and many other collaborating laboratories and agencies throughout the country
were decisive in the formation of the Communicable Disease Center Surveillance Program.
The surveillance of poliomyelitis has continued. Regular reports are issued weekly during the poliomyelitis season and at intervals of
861
862
Langmuar
ing of the Epidemiology and Laboratory
branches of the Communicable Disease Center.
Thereafter, the report came out twice weekly
throughout the summer months, and then
weekly when the epidemic was under way. The
report presented extensive summaries of descriptive epidemiologic data as the information
came in. Laboratory findings were added as
they became available. Summaries of the international spread of the infection were a major
feature.
Maintaining surveillance over a disease such
as epidemic influenza involving up to tens of
millions of cases on a national basis presented
special statistical problems. The reporting and
counting of individual cases, a practice still followed in some areas, is obviously cumbersome,
slow, and of limited value. Several more effective indexes were used. These included the
following:
Simple narrative reports of epidemics and
outbreaks.
Newspaper reports of the recognized prevalence of influenza in a city or a county as measured by closing of schools or increased absenteeism in industry.
Systematic reports of absenteeism among the
employees of a large, nationwide public-service
industry.
Current analyses of excess mortality from influenza and pneumonia reported weekly by 108
cities.
Current analyses of acute-respiratory-illness
rates in the National Health Survey.
The full account of the 1957 pandemic of
Asian influenza has been published (9-1 I ) .
The first outbreaks were recognized early in
June among the crews of naval ships at Newport, Rhode Island, and independently at several bases in California. Spread was most apparent in California when sharp outbreaks
occurred in shore-based naval personnel and
among groups of students attending youthguidance camps and special conferences where
large numbers of persons were congregated in
crowded places.
On June 26 a group of 1688 delegates from
43 states and a foreign country convened at a
church conference in Grinnell, Iowa. The college facilities used normally housed about 800.
students. A group of 100 delegates from California had chartered a railroad coach for the
863
864
5
W
g
m
700
600
500
. .
I-.,
.
.
400
*. ..:
.. ,_
I
300
<
PREDICTED
200
100
'
1957
1958
1959
enza, but, during the winter of 1961-62, influenza B was widely epidemic; in fact, the severity
of this epidemic as measured by excess mortality was the highest of any influenza B epidemic since 1936.
Although influenza-pneumonia mortality is
the most sensitive index for describing influenza epidemics, these figures do not truly reflect the full toll of the disease. Influenza is
often fatal to the aged and the chronically ill,
although a wide variety of diagnoses may appear on the death certificates. Excess total mortality is, without doubt, a more accurate measure of the costs in terms of death attributable to
influenza. Taking the six-month period from
October to March, 1957-58, and the threemonth period from January to March, 1960,
together, excess total mortality amounts to
86 000 deaths. Approximately two thirds of
these patients were sixty-five years of age or
older, and 25 percent were from forty-five to
sixty-four years of age. Only a third of the cases
were classified in the death records as influenza
pneumonia; half were charged to cardiovascular-renal disease, and 15 percent to all other
causes.
This impressive fact of 86 000 excess deaths is
the basis for the recommendation of the Surgeon General's Advisory Committee on Influenza Control to urge the annual immunization
of the aged and the chronically ill. Since it is
recognized that many recent studies have dem-
1960
1961
1962
1963
onstrated the protective value of present influenza vaccines, it is most reasonable to assume
that the immunization of those at greatest risk
of dying would provide a firm measure of protection. It is not visualized that mass immunization campaigns are warranted, but rather annual immunization of the aged and chronically
ill should become an accepted routine of good
medical and geriatric practice.
HEPATITIS
Langmuir
spective analysis of morbidity and mortality statistics helped to characterize the problem (13),
but no systematic surveillance was undertaken
until 1960. At first, quarterly reports summarizing current trends and reporting significant
news were issued; more recently a monthly
schedule of reporting has been attempted.
Events of considerable consequence have occurred in profusion during the past two years.
The year 1961 was a record year, with 72 000
cases being reported. The seriousness of serum
hepatitis resulting from faulty sterilization techniques was vividly shown in an outbreak involving over 40 cases and 15 deaths in one physician's practice. The importance of raw shellfish,
both oysters and clams, as an occasional factor
in the spread of infectious hepatitis was established for the first time in this country. These
and other events have come to recognition at
least in part as a result-and in several cases as a
direct result-of the surveillance program, at
both the state and national levels.
Figure 5 shows the incidence of reported
cases of hepatitis from 1954 to the present.
Note the seasonal swing, with highest frequency
occurring during winter months and lowest frequency during the summer. A seven-year periodicity is also evident; peaks occurred in 1954
and 1961. Such a rhythmic pattern would have
been fascinating to William Farr and all subsequent epidemic theorists, because relatively simple natural laws must govern such a regular
phenomenon if only they could be divined.
A comparison of morbidity and mortality
trends reveals one epidemiologic conclusion
7ow
'
'
'
1959
'
1960
'
-I
1963
865
A. Arithmetic Scale
10
1952 '53
'54
'55
'56
'57
'58
DEATH RATE
'59 1960
0.8
DEATH RATE
o-2
'1960
866
THE FUTURE
Langmuir
sulting from faulty sterile technics and other
now obscure causes.
CONCLUSIONS
867
References
( I ) Farr, W. Cholera epidemic, 1848-1849. In Vital Statistics: A memorial volume of selections from the
reports and writings of: Edited by N.A. Humphrey.
London, E. Sanford, 1885, pp. 333, 334.
(2) Andrews, J.M. United States Public Health
Service, Communicable Disease Center. Pub Healih
Rep 61:1203, 1210. 1946.
(3) Andrews, J.M., G.E. Quinby, and A.D. Langmuir. Malaria eradication in United States. Am J
Pub Health 40:1405-1411, 1950.
(4) United States Department of Health, Education, and Welfare, Public Health Service. A n Evalualion of the Efficacy of Gamma Globulin in the Prophylaris of
Paralytic Poliomyelatis (IS Used in the United States, 1953:
Report of the National Advisory Committee fir Evaluation
of Gamma Globulin in Prophylaxis of Poliomyelitis. Washington, D.C., Government Printing Office, 1954.
(Public Health Monograph No. 20.)
(5) Langmuir, A.D., N. Nathanson, and W.J. Hall.
Surveillance of poliomyelitis in United States in 1955.
Am J Pub Health 46:75-88. 1956.
( 6 ) Frost, W.H. Epidemiology of influenza. In Papers of Wade Hampton Frost, M.D.: A contribution to
epidemiological method. Edited by K. F. Maxcy. New
York: Commonwealth Fund, 1941, pp. 321-339.
(7) Collins, S.D. Review and study of illness and
medical care with special reference to long-time
trends. Pub Health Monogr No. 48, 1957, pp. 1-86.
(8) Farr, W. Cholera epidemic, 1848-1849. In Vztal Statistics: A memorial volume of selectiom from the
reports and writings of: Edited by N.A. Humphreys.
London, E. Sanford, 1885, page 330.
(9) Burney, L.E., et al. Asian variant: influenza:
type A. Pub Health Rep 73:99-178, 1958.
(10) Trotter, Y., Jr., et al. Asian influenza in United
States, 1957-1958. A m J Hyg 70:34-50, 1959.
(11) International Conference on Asian Influenza.
University of California School of Medicine and Institute of Allergy and Infectious Diseases, National
Institutes of Health, U. S. Public Health Service. Am
Rev Resp Dis 83:l-219, 1961.
(12) Dunn, EL., D.E. Carey, A. Cohen, and J.D.
Martin. Epidemiologic studies of Asian influenza in
Louisiana parish. A m ] Hyg 70:351-371, 1959.
(13) Sherman, I.L., and H.F. Eichenwald. Viral
hepatitis: Descriptive epidemiology based on morbidity a n d mortality statistics. A n n I n t Med
44:1049-1069, 1956.
( 2 4 ) Dougherty, W.J. and R. Altman. Viral hepatitis
in New Jersey, 1960-1961. Am J Med 32:704-736,
1962.
(15) Henderson, D.A. Relationship of infectious
hepatitis to consumption of raw clams from Raritan
Bay. Paper presented at Conference in the Matter of
Pollution of the Interstate Waters of Raritan Bay
(NewJersey-New York) and Its Tributaries, New York
City, August 22 and 23, 1961.
A change of strategy was instituted in the West and Central Africa Smallpox
Eradication Program in the fall of 1968. High priority was given to eliminating
smallpox foci rather than limiting activities to mass vaccination. The progress of
the program since September 1968 shows that the change in emphasis resulted in
active surveillance activities which discovered more smallpox cases than the official reporting system. All known smallpox outbreaks were investigated in 1969.
Vaccination totals did not suffer from the change in emphasis. The attack on
smallpox foci successfully prevented the expected seasonal increase in 1969 and
resulted in an interruption of smallpox transmission throughout West and Central Africa.
INTRODUCTION
868
Foege et al.
months was 1:5. Thus, the most efficient use of
epidemic control teams could be made during
the fall, the period of lowest transmission.
A decision was made to undertake epidemiologic control activities as a priority measure even if mass vaccination efforts were reduced. In early September 1968, epidemiologic
control activities were coordinated in eight nations (Nigeria, Togo, Dahomey [now Benin],
Mali, Upper Volta, Niger, Guinea, and Sierra
Leone), all of the countries in the 20-country
region which were infected with smallpox. A
neologic term, eradication escalation, was
used to identify these activities and is used in
this report because to persons involved in the
West and Central Africa Smallpox Eradication
Program it became synonymous with a set of
methods.
869
house searches were performed to locate patients and immunize unvaccinated persons.
Outbreak Investigation
Vigorous attempts were made to locate all
sources of smallpox infection. Reported cases
were promptly investigated, the full extent of
each outbreak was delineated, and the target
area for epidemic control efforts was defined.
This area included villages, markets, and compounds which were contiguous with the infected village, or which received frequent visitors from the infected village. T h e events
which produced the outbreak were established
and, insofar as it was possible, the chain of infection was traced to determine the source of
the original case. Laboratory specimens were
collected to verify the diagnosis of smallpox.
METHODS
Four principal methods were employed during the eradication escalation effort: active surveillance, outbreak investigation, outbreak control, and rapid communication of disease
intelligence.
Surveillance
Participating countries employed a variety of
techniques to locate smallpox cases. The usual
disease-reporting system was utilized as a
method for determining where control teams
should work. Since this approach requires waiting for reports of smallpox, it constitutes a passive surveillance system. In addition, active surveillance techniques were used, here defined as
attempts to find smallpox cases not reported
through the usual system. Newspapers, radio,
and letters were used to alert the public and
solicit information about smallpox cases. Widely
dispersed informal surveillance systems were
developed through the cooperation of other
health services (malaria, leprosy), other government personnel (teachers, rural mail carriers,
agricultural project personnel), local authority
figures (village and area chiefs), and volunteer
agencies, such as missions. Several countries instituted special active surveillance projects in
areas suspected to have smallpox. Visits by project personnel were made to each village leader
in the suspicious area or exhaustive house-to-
Outbreak Control
Various outbreak control techniques were
used. The objective was to vaccinate a geographically or sociologically contiguous area
around each patient. The extent of this area was
determined by the outbreak investigation. The
approach is intermediate between selective vaccination of close personal contacts (practiced to
control imported smallpox in many non-endemic countries) and the indiscriminant mass
vaccination campaigns which must be performed in the absence of adequate outbreak
investigation.
Communications
Unofficial weekly telegraphic reports were
sent to the Smallpox Eradication Program
headquarters at the Centers for Disease Control
in Atlanta and to its Regional Office in Lagos,
Nigeria, to facilitate rapid communication of
smallpox information. The cables reported the
weekly number of smallpox cases, their geographic location, and the status of control activities. These data were circulated in an informal weekly report and distributed to the
involved countries, enabling health officials to
take rapid preventive measures or to intensify
active surveillance activities when outbreaks
threatened their borders.
870
Smallpox Incidence
Figare I. Reported smallpox by month, 1960-1967 average, 1968 and 1969, in West and Central Africa.
2oool
1000
VJ
a 1400
0
U
0
a
w
fz
/*\\
;
1960-1967
2oo
60
40
,-
200
0 1I
0
JAN
Source: WHO
I
\
\
\
I
,/
//
\\
\
\
\
\
'\\-.,
1960-1967 Average
\
\.
- ','_-r-r
.4
4
'
1
MAR
1
1
JUNE
1968
SEPT
DEC
1
MAR
JUNE
1969
1
SEPT
DEC
Foege et al.
871
Figure 2. Percentage of population not vaccinated in the Smallpox Eradication Program area compared with the ratio (To) of reported smallpox cases to the expected
smallpox cases, where expected cases are based on the 1960-1967 monthly averages.
Eradication
escalation
activities started
1
Percent of
not vaccinated
30
I-
20
Smallpox cases
54-
321
J F M A M J J A S O N D J F M
1968
1969
Time period
Jan-Jun 1968
(monthly average)
Jul-Sept 1968
(monthly average)
October 1968
November 1968
December 1968
.January 1969
Cases reported
Cases disthrough officovered
cia1 reporting through active
system
surveillance
3865
(644)
612
(204)
88
69
60
54
195
(33)
216
(72)
177
146
80
76
%
detected
through active
Total cases surveillance
4060
(677)
828
(276)
265
215
140
130
4.8
26.1
66.8
67.9
57.2
58.4
aDahomey,Ghana, Guinea, Mali, Niger, Northern States Nigeria, Sierra Leone, Togo, and
Upper Volta.
tensive response to each outbreak. A smallpox and 1969 marked improvement occurred in the
outbreak is defined as one or more cases occur- response to reported outbreaks. Table 2 indiring in an area previously free of smallpox, or cates a steady improvement in the percentage of
cases occurring in smallpox-infected areas reported outbreaks which were investigated. In
which are epidemiologically related. In 1968 1969 all outbreaks were investigated.
872
Time reported
No. of
reported
outbreaks
No.
investigated
109
(18)
91
(30)
71
86
(14)
77
(26)
69
97.2
(24)
28
(23)
28
100.0
January-June 1968b
(Monthly average)
July-September 1968
(Monthly average)
October-December
1968
(Monthly average)
January 1969
investigated
78.7
84.6
Dahomey,Ghana, Guinea, Mali, Niger, Northern States Nigeria, Sierra Leone, Togo, and
Upper Volta.
bIncomplete for Northern States Nigeria
Because of the emphasis placed on the eradication escalation activities and the diversion
of vaccinators and supervisors from the mass
vaccination campaign, it was feared initially
that ongoing vaccination activities would be
crippled. Table 3 demonstrates that the converse was true. Investigation and control activities diverted only a small proportion of project personnel to outbreak control activities. The
participating countries frequently acquired
short-term personnel to augment the program
during the initial critical months.
DISCUSSION
Dixon, while dismissing the notion that smallpox eradication can be achieved by vaccinating
80 percent of the population within four or five
years, points out that if more study were given
to the foci of smallpox, it might be possible to
eradicate the disease from an area by vaccinating a far smaller proportion of the total population (6).
Epidemiologic control methods were advocated over a century ago in England, when it
was clear that mass vaccination methods were
not eliminating smallpox. In 1896, the Royal
Commission pointed out the need for a complete system of notification to diminish the
prevalence of smallpox (7). Despite repeated
references in the literature, the practical concept of surveillance and epidemiologic control
has developed slowly and has thus far had
greatest application in non-endemic countries
attempting to detect and eliminate importations.
The use of these techniques in smallpox endemic countries is a logical extension if endemic
countries are viewed as non-endemic areas with
Total
%
increase
4 546 424
6 092 684
34
2 137 504
6 683 928
2 287 581
8 380 265
7
25
aMinus Sierra Leone. Sierra Leone did not begin its vaccination campaign until 1968
Foece et al.
islands of endemicity, each equivalent to an
importation. Removal of the islands is reasonable at any time, but particularly crucial at the
seasonal trough period when the fewest islands
exist.
Selective epidemiological control (8) has
been successful in West and Central Africa. Intelligent use of vaccination based on knowledge
of where the disease is and when, where, and to
whom the disease is likely to spread is more
economical in time, vaccine, and personnel
than blind mass vaccination. Mass vaccination
campaigns should continue in endemic areas,
but we consider the use of surveillance, investigation, and selective epidemiologic control
techniques to be of equal and, under certain
circumstances, of even greater importance than
systematic mass vaccination activities.
873
References
(1) Top, F.H., Sr. Communicable and Infectiow Diseases. 6th edition. St. Louis, C.V. Mosby co., 1968, pp.
165- 173.
(2) Felsenfeld, 0. The Epzdemiology of Tropical Diseases. Springfield, Ill. Charles c . Thomas, 1966, pp.
336-343.
(?) WHO Expert Committee on Smallpox. First Report. WHO Technical Report Series 283:20, 1964.
(4) National Institute of Communicable Diseases of
India. Evaluation of the National Smallpox Eradication
Program in Karnal District, Haryana, New Delhi, 1968.
(5) Henderson, D.S. Suroeillance-The Key to Smallpox Eradication. Presented at the WHO Inter-Regional
Course on Methods of Epidemiological Surveillance.
Prague, August 1968.
(6) Dixon, C.W. Smallpox. London, J. & A. Churchill Ltd., 1962, p. 359.
(7) Royal Commission on Vaccination: A Report.
London, 1896.
(8) Dick, G. Smallpox. A reconsideration of public
health policies. Progr Med Virol 8:l-29, 1966.
This paper provides additional data based on periodic breast cancer screening with mammore extensive experience than previously re- mography and clinical examination result in a
ported regarding the impact of periodic breast reduction in mortality from breast cancer in the
cancer screening with mammography and clini- female population? (3,4). Two systematic rancal examination on mortality from breast cancer dom samples, each consisting of 31 000 women
( I , 2). The new observations, covering a follow- aged 40 to 64 years with at least one years
up period of five years, reinforce the first pre- membership in HIP were selected. Study group
liminary reports that women in such a screening women were offered a screening examination
program have substantially lower mortality and three additional examinations at annual infrom breast cancer than a similarly constituted tervals. All screening examinations have been
control group. During the five years of follow- completed.
Women in the control group followed their
up, there were 40 deaths due to breast cancer
among the 31 000 women aged 40-64 years usual practices in receiving medical care. No
invited for screening, almost two-thirds of special effort was made to encourage them to
whom accepted, as compared with 63 among have general physical examinations. On the
the 3 I 000 women in the control group. Reduc- other hand, they were not discouraged from
tion in mortality is concentrated entirely among having such examinations, which are part of
women over 50 years of age; under age 50, their benefits in HIP.
Each screening examination consisted of a
mortality from breast cancer does not differ
between the study and control groups. Clinical clinical (usually conducted by a surgeon) and Xexamination of the breast and mammography ray examination and an interview with the pacontribute independently to the early detection tient to obtain relevant demographic informaof breast cancer under screening conditions, tion and a health history. The clinician conbut detection by mammography has been of ducted his examination with no knowledge of
special importance in reducing mortality from the roentgenographic findings and recorded
his observations and recommendations for folbreast cancer, in the short-run, at least.
low-up medical care on study forms designed
for this purpose.
METHODOLOGY
Cephalocaudad and lateral X-ray views of
In December 1963, the Health Insurance each breast were taken by specially trained techPlan of Greater New York (HIP), a prepaid, nicians using a modified Egan technique (5,6).
group-practice plan, started a long-term ran- Mammograms were separated from the clinical
domized trial directed at the question, Does report for independent readings by two of the
radiologists on the staff. Final responsibility for
resolving differences rested with the chief radiSource: Seventh National Cancer Conference Proceedings, Sep- ologist on the study team. Later, clinical infortember 27-29, American Cancer Society, 1972.
mation derived from the screening examination
1 T h i s study was s u p p o r t e d i n p a r t by contracts
PH43-63-69 and NIH 69-88 from the National Institutes of was reviewed in conjunction with the radiologic
Health.
* Director, Department of Research and Statistics, Health findings by the chief clinician on the project
staff, and a recommendation was made for rouInsurance Plan of Greater New York.
Associate Clinical Professor of Preventive Medicine, New tine examination one year later, early recall beYork Medical College; Director of Radiology, LaGuardia
cause of suspicious findings, biopsy, or aspiraHospital, New York.
Associate Director of Surgery, Chief of Breast Service, tion.
Beth Israel Medical Center; Associate Clinical Professor of
Surgical and pathological findings in breast
Surgery, Mt. Sinai School of Medicine, New York.
Director of Operations, H I P Mammography Study, cancer cases are obtained from hospital charts.
Health Insurance Plan of Greater New York.
The projects coordinating pathologist reviews
874
Shapiro et al.
slides and conducts special studies of tissue
blocks when they are available. Each case of
microscopically confirmed carcinoma of the
breast is investigated to establish whether a mastectomy was performed prior to the womans
entry into the study (if so, she is excluded from
the study), the type of surgery performed, the
histologic type, nodal involvement, and size of
lesion.
Deaths are identified through intensive follow-up of all confirmed breast cancer cases, and
by matching death records on file in various
health departments against the total file of records for study women (including those who refused screening examinations) and for control
women in order to locate deaths attributed to
breast cancer. As a final check in this process,
several months after the fifth anniversary of the
womens entry date, an attempt is made for each
study and control group woman, even if she is
no longer enrolled in HIP, to determine her
survival status and any history of breast surgery.
The techniques being used include a mail survey, medical and enrollment record review, and
a check against the records of death from all
causes for unresolved cases.
A critical requirement in the various followup procedures is to identify breast cancer cases
and deaths with similar degrees of success in the
study and control groups. The evidence is reassuring. Three-fourths of the women in both
groups are still enrolled in HIP at the end of
five years, and the enrollment and medical record systems should provide comparable information for study and control women. The mail
survey designed to locate missing breast cancer
cases appears to be almost equally effective in
the study and control groups; the response
rates are 84 percent and 82 percent for the two
groups, respectively.
875
Characteristic
Study gmupb
Not
Control
Total Examined examined group
Total
100.0
100.0
100.0
100.0
Age, yr.
40-44
45-49
50-54
55-59
60-69
24.2
23.7
22.5
18.4
11.2
25.3
24.1
22.4
17.8
10.4
22.3
22.9
22.7
19.3
12.8
24.5
23.6
21.9
18.7
11.3
Religion
Protestant
Catholic
Jewish
29.1
38.1
32.8
28.0
36.3
35.7
31.1
41.4
27.5
29.2
37.9
32.9
Education
Elementary
school
High school
College
22.6
46.5
30.9
19.5
46.8
33.7
28.3
45.9
25.8
22.1
45.0
32.9
Marital status
Never married
Ever married
8.7
91.3
7.5
92.5
10.9
89.1
9.3
90.7
Prior pregnancies
Never pregnant 20.3
61.9
1-3
4 or more
17.8
19.4
61.5
19.1
21.9
62.7
15.4
23.0
58.6
18.4
29.1
70.9
33.4
66.6
21.2
78.8
25.9
74.1
89.1
10.9
93.0
7.0
88.2
11.8
SCREENING PARTICIPATION
876
Completed high
school
College
852
1003
10
10
27
25
63
66
711
1684
828
14
11
13
25
27
32
61
62
55
12
12
25
28
63
60
16
11
16
29
28
32
55
62
52
Population
Breast cancers
Rate
Per
No.
1000
Study-screened
Detection due to initial exam.b
Detection due to annual reexam.
Detection not due to screening
223
55
77
91
2.25
2.72
1.51
0.92
Study-refused screening
Control
73
284
1.37
1.86
877
Shabzro et al.
Modality
Total
Total
Mammography only
Clinical only
Clinical and
mammography
132
44
59
Total
Mammography only
Clinical only
Clinical and
mammography
100.0
33.3
44.7
29
22.0
40-49
60 or50-59 older
Number
31
65
6
27
19
26
36
11
14
6
12
Percentb
100.0 100.0
19.4
41.5
61.3
40.0
19.4
18.5
11
100.0
30.6
38.9
30.6
a initial evidence for biopsy recommendation made independently by the two modalities.
Percentages in this and subsequent tables may not add to
100.0 due to rounding.
CONTROL
STUDYTOTAL
SCREENED
REFUSED SCREENING
-----..--------__----
SCREENED
DETECTED
ON SCREENING
MAMMOGRAPHY ONLY
77%
CLINICAL ONLY
76Yo
BOTH
DETECTION NOT DUE
TO SCREENING
878
Table 5. Percent distribution of breast cancers by histologic evidence of axillary node involvement or
other evidence of metastases.
Axillary node metastasis (histology)
Population
Positive
Unknowna
Extent of
disease
unknown
Other
evidence of
metastasisb
Number
Negative
296
57.1
34.5
4.4
4.1
Study, screened
Detection due to screening
Mammography only
Clinical only
Clinical & mammography
223
132
44
59
29
62.8
70.5
77.3
76.3
48.3
32.3
22.7
15.9
18.6
41.4
1.3
1.5
3.4
3.6
5.3
6.8
5.1
6.9
91
73
284
51.6
39.7
45.8
46.2
41.1
42.3
1.1
13.7
7.0
1.1
5.5
4.9
Control
breast cancer cases between the study and control groups (Table 6). By far the most frequent
surgical procedure in each of the subgroups was
radical mastectomy. As would be expected from
the data on stage of disease at diagnosis in Table
5, the proportion with radical mastectomy is
somewhat higher among the breast cancer cases
Table 6. Percent distribution of breast cancers by
type of surgery and use of local radiation therapy.
Study
~
Treatment
modality
Number
Type of surgery
Control
284
Radicala
73.9
Modified radicalb
Otherc
14.1
12.0
Total
296
Not
Screened screened
223
73
83.4
87.0
72.6
9.8
6.8
9.4
3.6
11.0
16.4
36.Ie
63.2
33.2
66.8
45.2
52.1
49.7e
49.7
Shapiro et al.
in the screened study group (87 percent) than
among cases diagnosed in the not screened
study group (73 percent) or the control group
(74 percent). An appreciably lower proportion
of the study group women with breast cancer
had local radiation therapy than was observed
in the control group. However, most of the difference is due to the relatively large proportion
of cases with no evidence of axillary node involvement in the study group (see adjusted figures in footnote e, Table 6).
MORTALITY FROM BREAST CANCER
Death Rates
Flgure 2. Deaths due to breaet cancer: 5-year follow-upafter entry into study.
'1
563
4.1
4-
1
2.6
Studya
Control
Study=
879
Control
880
No. of person-years
Control
Study
Control
Studyb
Total
152 742
151 660
63
40
8
12
18
25
11
6
8
15
61
30
30
30
781
567
327
067
61
30
30
29
337
350
114
859
70
60
I-
50
/63
w 40
Control
a:
W
0
I-
30
v)
3
U
20
m
u.
0
U
W
.+"/
Shapiro et al.
Table 8. Breast cancer deaths by age at death, five
years of follow-up.
Age at death
Total
40-49 years
(40-44)
(45-49)
50-59 years
(50-54)
(55-59)
60-69 years
a Includes women screened a n d women who refused
screening.
881
882
45
;
:<
?
CONTROL
GROUP
35
30
25
:u '5
0
'
10
GROUP
42%
40
5 20
I
STUDY
35%
128%
26%
rn
m
m
m
m
m
rn
TOTALa
STUDY
REFUSED
SCREENING
Total
Screeneda
Detection
Due To
Not Due To
Screeninga Screening
SCREENED
Table 9. Five-year case fatality rates among confirmed breast cancer cases, rates per 100.
Control compared with study groups
Population
Control
Study (screened and refused screening)
Screened, total
Detection due to screening
Detect. not due to screening
Refused screening
Five-year case
fatality
rate
Difference
Standard
error
of difference
42.1
27.9
25.8
16.9
38.7
14.1
16.3
25.2
3.4
6.6
6.3
5.2
8.0
34.5
7.6
7.5
.03
.O1
<.Ol
.67
.31
Note: Rates for study group cases allow for one year average lead time in cancer detection due to screening.
Probability that difference as large or larger is due to chance factors.
refused screening combined) than in the control group, both among women with no histologic evidence of axillary node involvement and
among the other cases.
Consistent with the observations based on
mortality due to breast cancer discussed in the
earlier section, case fatality rates do not suggest
any improvement related to the screening pro-
883
Shapira et al.
Figure 5. Cumulativecase fatality rates among women with breast cancer by interval
of follow-up after diagnosis.
g50
$40-
xu
$30oui
b5
U2 0 -
cd ci
m a
$510-
OIL
a w 8-
a,
a
2. 6 4
7.8
I
alncludes women screened and those who refused screening. Allowance made for 1 year lead time
Case fatality
rateb
16.9 ( 2 3.9)
2.3 ( * 2.2)
22.0 ( 2 6.4)
29.1 ( * 10.7)
a Initial evidence for biopsy recommendation made independently by the two modalities.
Rates for all modalities obtained from life-table values
covering a six-year follow-up period, which is equated with
five-year period for cases not detected through screening.
Numbers in parentheses refer to standard error due to sampling variability.
gram among women with breast cancer diagnosed at ages 40-49 years (Table 12). Considerably lower case fatality rates are observed at
both ages 50-59 and 60 and over.
General Mortality
Mortality due to causes other than breast cancer is being determined through the follow-up
Population
dif-
Evidence of
metastasisb
26.1
17.4
60.8
45.2
8.7
15.6
7.9
10.6
0.27
0.14
a Rates for study group cases allow for one year average
lead time in cancer detection due to screening.
b Evidence of node involvement (histology) or other evidence of advanced diseases.
884
50-59
60-69
31.8
39.4
48.8
24.2
41.5
21.0
Difference
Population
-7.6
24.6
20.5
Standard error of
difference
13.9
8.5
13.0
0.59
<0.01
0.11
Rates for study group cases allow for one year average
lead time in cancer detection due to screening.
Table 13. Number and rate of deaths from all causes, excluding breast cancer, by age at death; five-year
follow-up after entry.
Study
Age at death
Control
Total
Screened
Refused
screening
Number
All ages
40-49
50-59
60-69
680
98
294
288
670
106
296
268
54.3
23.6
52.5
103.8
53.9
25.3
53.0
100.4
338
53
147
138
332
53
149
130
All ages
40-49
50-59
60-69
41.5
18.9
40.2
81.7
77.4
38.4
77.2
132.6
Note: Data refer to mortality over a five-year period among women whose entry dates ended Dec. 31, 1965, representing 82
percent of total population in the study and control groups.
Shupiro et ul.
885
80
77
70
60
CJY
54
a
w
>-
50
54
42
0
0
0
40
a
W
z
W
30
20
10
C0NTR0L
Total
Screened
Refused
Screening
STUDY
886
Control
Total
680
217
79
62
76
312
151
670
20 1
75
48
78
299
170
Screened
Refused
screening
Number
Genitourinary system
Other
Circulatory system
Other
338
119
41
29
49
136
83
332
82
34
19
29
163
87
Malignant neoplasms
Digestive system
Genitourinary system
Other
Circulatory system
Other
54.3
17.3
6.3
4.9
6.1
24.9
12.1
53.9
16.2
6.0
3.9
6.3
24.0
13.7
41.5
14.6
5.0
3.6
6.0
16.7
10.2
77.4
19.1
7.9
4.4
6.8
38.0
20.3
Note: Data refer to mortality over a five-year period among women whose entry dates ended Dec. 31, 1965, representing 82
percent of total population in the study and control groups.
Shapzro et al.
ACKNOWLEDGMENTS
The authors appreciate the help and cooperation of the clinicians and radiologists of the 23
medical groups associated with the project and
the pathologists in the hospitals involved.
Important contributions were made by Raymond Fink, Ph.D., in the design and conduct of
the study and Ruth Roeser, M.S., and Mary
Tang, M.S., in the analytical phase. Acknowledgment is also made for the fine contributions
of Dr. Stanley Gross who was the coordinating
pathologist for the study and Drs. Maurice M.
Pomeranz, Mortimer J. Lacher, and Filomen
Lopez who aided in the review and analysis of
the medical and radiologic reports of screening
observations.
References
( I ) Shapiro, S., P. Strax, and L. Venet. Periodic
breast cancer screening in reducing mortality from
breast cancer. JAMA 215:1777-1785, 1971.
( 2 ) Venet, L., P. Strax, W. Venet, and S. Shapiro.
Adequacies and inadequacies of breast examinations
by p h y s i c i a n s i n m a s s s c r e e n i n g . C a n c e r
28: 1546-1551, 1971.
887
task force recommended that the Ontario Cancer Treatment and Research Foundation consider funding such a study in the Toronto area,
the findings of which are reported here.
METHOD
888
889
RESULTS
890
Table 1. Frequency of screening by Pap smear among cases and controls during the five years before the
year of diagnosis of the index case in relation to age, income, and education.
Frequency of Pap-smear screening in:
All subjects
Age
20-34
35-44
45-59
60 +
Totala
Income ($ p.a.)
<6000
6-9999
10-14 999
15 OOO+
Total=
Highest grade
achieved
<9
9-1 1
12+
Total=
~~~
Relative
risk
Cases
Controls
67/212(32%)
591/1060(56%)
2.7b
7/16 (44%)
18/36 (50%)
311102 (30%)
11/58 (19%)
67/212(32%)
66/107 (62%)
124/185 (67%)
266/460 (58%)
134/304 (44%)
590/1056(56%)
2.1
2.0
3.1b
3.4c
2.8b
9/45 (20%)
9/41 (22%)
12/37 (32%)
16/35 (46%)
46/158(29%)
811184
651124
128/217
1601226
434/751
(44%)
(52%)
(59%)
(71 %)
(58%)
3.1d
3.9'3
3.0d
2.9'3
3.2b
22/78 (28%)
19/64 (30%)
19/43 (44%)
601185 (32%)
109/251
200/339
214/328
5231918
(43%)
(59%)
(65%)
(57%)
2.0e
3.4b
2.4e
2.5b
Total figures vary because of missing values. Total relative risk is standardized to the distribution of the characteristic in the
combined population of cases and controls.
b p<O.OOOl
c p<O.OOl
p<O.Ol
pC0.05
a
highest levels of income and education. However, within each level the relative risk of cervical cancer in those who failed to have a Pap
smear remained consistently high (2.0-3.9),
and most risks were statistically significant.
Standardized overall estimates were all highly
significant.
Other Known Risk Factors
As might be expected, there were also significant differences between cases and controls regarding some of the other known risk factors
for carcinoma of the cervix. Thus, the proportion of divorced and separated women among
the cases was 16 percent, significantly higher
than the 6 percent among controls (p<O.OOOl).
The mean age at marriage was 22.1 years in
cases, compared with 23.1 in controls (p<0.05),
and in those women for whom the information
was available (167 cases, 801 controls) the mean
age at first intercourse was 20.0 years in cases vs
21.5 in controls (p<O.OOl). However, the overall
(standardized) relative risk of cervical cancer
associated with failing to have a Pap smear remained virtually unchanged when standardized
More cases than controls were regularly employed at the time of the interview (38 percent
vs 29 percent), and this could account for some
of the difference in Pap-smear frequency since
working women might have less opportunity to
visit a doctor. In fact, however, regular employment was associated with a higher proportion of
screening Pap smears, both in the cases (36
percent vs 28 percent) and the controls (62 percent vs 52 percent). Furthermore, the relative
risk associated with failure to have a Pap smear
was identical (2.9, p<O.OOI) whether the analysis was restricted to cases and controls who were
regularly employed or to those who were not.
Access to Medical Care
89 1
892
References
( I ) Apostolides, A. and M. Henderson. Evaluation of cancer screening programs. CA 39: 1779-1787,
1977.
(2) Christopherson, W.M., J.E. Parker, W.M. Mendez, and F.E. Lundin, Jr. Cervical cancer death rates
and mass cytology screening. CA 26:808-811, 1970.
(3) Boyes, D.A., G. Knowelden, and A.J. Philips.
Lappreciation des mesures de contr6le du cancer.
Bull Cancer 1:83-88, 1973.
(4) Timonen, S., U. Nieminen, and T. Kauraniemi. Cervical cancer. Lancet i:401-402, 1974.
(5)Mac Gregor, E.J. and S. Teper. Mortality from
893
(9) Fleiss, J.L. Statistical methodsfor rates and proportiom. New York, John Wiley and Sons, 1973, pages
109- 129.
( 1 0 ) Mantel, N. and W. Haenzel. Statistical aspects
of the analysis of data from retrospective studies of
disease. J Nail Cancer lnst 22:719-748, 1959.
(11) Mantel, N. Chi-square tests with one degree of
freedom: extension of the Mantel-Haenzel procedure. Am Stat AssocJ 58:690-700, 1963.
(12) Spriggs, A. and M.M. Buddington. Protection
by cervical smears. Lancet i:143, 1976.
I B . MEDICAL CARE I
MEASURING THE QUALITY OF MEDICAL CARE THROUGH
VITAL STATISTICS BASED ON HOSPITAL SERVICE AREAS:
1. COMPARATIVE STUDY OF APPENDECTOMY RATES1
Paul A. Lembckez
The changing character of medical and public health problems in recent years has intensified the need for new methods of measuring
the quality of medical care. The term quality of
medical care may mean different things to different people. It is usual to express the quality
of medical care in terms of qualifications of
personnel, adequacy of equipment, and the
technical excellence of medical services performed, and it is assumed that such factors are
correlated positively with results favorable to
the patient. Granted that a well trained professional staff working in a well equipped and organized institution will deliver a better quality
of medical care generally than a poorly trained
staff working with inadequate resources, it is
true nevertheless that such measurements are at
best only relative and indirect, and are not adequate scientific methods.
The old chestnut, The operation was a success but the patient died, is a homely illustration of the fact that attention is sometimes directed to the wrong criteria of success of
medical treatment. The best measure of quality
is not how well or how frequently a medical
service is given, but how closely the result approaches the fundamental objectives of prolonging life, relieving distress, restoring function, and preventing disability. For example,
unnecessary surgical operations, no matter how
well done, do not contribute to such objectives,
and any considerable number of unnecessary
operations for a given disease or condition
would indicate a poor quality of care for that
condition.
Measurements of quality should be expressed
in terms that are uniform and objective, and
that permit meaningful comparisons between
Source: American Journal of Public Health 42276-286, 1952.
Presented before the Medical Care Section of the American Public Health Association at the Seventy-ninth Annual
Meeting in San Francisco, Calif., November 1, 1951.
* Associate Professor of Public Health Administration,
School of Hygiene and Public Health, Johns Hopkins University, Baltimore, Md.
communities, institutions, groups and time periods, and with general standards. Unfortunately, there are as yet few simple and easy
methods to meet the need for such type of
measurement, and much thought and experimentation will be required to develop them.
The present comparative study of appendectomy rates in a number of hospital service areas
is directed to only one small segment of the
whole problem, and there is no thought that the
quality of medical care generally can be characterized by the single index of appendectomy
rate. This study is but one type of experimental
approach to the general problem of methodology, and is the first of a series that have
been undertaken in the hope of developing improved methods of measuring the quality of
care of certain hospitalized illnesses through
vital statistics based on hospital service areas.
THE HOSPITAL SERVICE AREA:
DEFINITION AND USEFULNESS
895
Lembcke
The situation was rather different in the region under study. For example, in one county,
four different cities or villages had general hospitals. These hospitals varied considerably from
one another in respect to size, type of medical
staff, and occupation, etc., of the population
served. In such circumstances, county-wide statistics based on hospital data would submerge
and hide differences that otherwise might appear if the statistics were based on the population of geographic areas principally served by
the hospital or hospitals in the locality.
Similar problems are encountered when a
hospitals services extend over city, county, or
state boundaries.
Studies of the place of residence of all patients, except newborn infants, who were hospi-
Table 1. Appendectomies and appendicitis deaths and population in 23 hospital service areas in and
around Rochester, N. Y.
Primary appendectomies, 1948
Rate per
1000
Hospital
service
Population
In area
hospitals
Other
hospitals
Total
Crude
Age-sex
adjusted
Pa
1948
Average,
1944-1948
area
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
(9)
A
B
C
D
E
8862
12 108
13 277
20 620
8227
3
30
33
42
20
19
7
4
13
4
22
37
37
55
24
2.5
3.1
2.8
2.7
2.9
2.9
2.9
2.9
3.0
3.0
.50
.43
.45
.60
.75
8.3
4.9
-
6.6
3.0
2.9
-
457 965
1402
47
1449
3.2
3.2
1.8
2.8
18 906
18 234
15 249
4352
19 561
18 081
7663
11 019
7498
60
31
50
12
62
66
23
33
29
4
21
2
2
12
7
8
10
4
64
52
52
14
74
73
31
43
33
3.4
2.9
3.4
3.2
3.8
4.0
4.0
3.9
4.4
3.5
3.5
3.5
3.6
3.8
4.0
4.1
4.1
4.2
.54
.49
.50
.69
.17
.09
.25
.16
.18
5.5
6.6
1.1
5.5
2.6
4.1
4.4
11 133
19 877
24 766
3566
73 718
30 916
17 120
34 619
21
77
103
14
400
189
99
229
26
10
16
4
8
4.2
4.4
4.8
5.0
5.5
6.3
6.4
6.9
4.5
4.9
5.3
5.5
5.7
6.2
6.9
7.1
.04
4.0
10
9
47
87
119
18
408
194
109
238
857 332
3028
252
3280
3.8
4.0
K
L
M
N
0
P
S
T
U
V
W
Total
a
c
Deaths per
100 000
.07
5.5
9.1
-
2.9
5.4
10.1
3.2
4.9
4.5
4.7
5.2
2.2
3.5
2.7
6.5
-
3.6
5.3
Probability of difference between age-sex-adjusted rate and standard rate of 3.2 being due to chance variation.
Less than 0.001.
Less than 0.0001,
896
CALCULATION OF APPENDECTOMY
INCIDENCE
Appendectomy is an important operation, too. An estimated 600 000 primary appendectomies are done each year
in the United States, costing perhaps $100 to $150 million for
medical and hospital care, and resulting in as much as $50
million loss of income to wage earners. Six hundred thousand
appendectomies would require 3 million or more hospital
days, tying up more than 10 000 hospital beds having a capital replacement value of $150 to $200 million. This amount
of care would require the equivalent of full-time service by
about 2500 doctors, 6000 nurses, and 10 000 other hospital
employees.
Lembcke
the operating room journal was consulted originally, and in all cases it was referred to as a
check on the index of operations, where such
existed. Data recorded on separate cards for
each patient included: name of hospital; patients residence, age or year of birth, and sex;
whether primary or secondary appendectomy;
whether recovered or died. A very interested
and helpful attitude was shown by all hospitals
both in and outside of the region.
Primary appendectomies in 1948 numbered
3280 among the 857 332 residents of the 23
hospital service areas, a crude rate of 3.83 per
1000 population per year. In addition, 1372
secondary appendectomies were recorded in
the hospitals of the region, a crude rate of 1.60.
The crude primary and secondary rates together were 5.43 per 1000 population. This
figure is roughly one-third of the average birth
rate of the past 20 years, suggesting that if the
same rates were to hold, the average expectancy
of having the appendix removed at some time
during life is about one in three.
Table 1 shows for each hospital service area
its population and the respective numbers of
primary appendectomies done in and outside
of the hospital service area. Because the hospital service areas differed somewhat in age and
sex composition, rates were adjusted, as shown
in column 6, to those that would be expected if
the age and sex distribution of each hospital
service area were the same as that of New York
State at the Census of 1940.
The age-sex-adjusted primary appendectomy rates range from 2.9 to 7.1 per 1000 population. As a standard for comparison, there
has been chosen the rate of 3.2 per 1000 among
the 457 965 persons in the Rochester hospital
service area. This area was chosen because it has
a large population, lending stability to rates,
and because it contains a medical school and
teaching hospital, and several other hospitals
generally believed to be well staffed and
equipped. Because of the small population
groups in many hospital service areas, minor
differences from the standard may be due to
chance variation. The probability that the observed difference was due to chance variation is
shown in column 7 of the table. Assuming that
probability of 0.05 or less is needed for statistical significance, seven hospital service areas
show rates above the standard by amounts
that are statistically significant.
897
SECONDARY APPENDECTOMIES
898
Such an association is not apparent, however, area. Similar findings were obtained in several
in columns 6 and 8 of Table 1. In fact, the 19 other areas and hospitals.
A curious difference between hospital service
resident appendicitis deaths reported to the
state department of health in 1948 are so few in areas with high and those with low appendecnumber that the difference between the com- tomy rates is found when urban and rural rates
bined rate of 2.78 appendicitis deaths per are computed for the six hospital service areas
100 000 population for the eight hospital serv- containing cities. In the Rochester hospital service areas with the highest appendectomy rates ice area, the urban and rural appendectomy
and the rate of 1.75 for the standard is not rates were 2.71 and 2.86 per 1000, respectively,
statistically significant. However, if one consid- and average annual appendicitis mortality rates
ers the average appendicitis mortality over the for the period 1944-1948 were 2.73 per
five year period 1944-1948, a statistically signif- 100 000 for the urban population and 2.80 for
icant difference is found between the stan- the rural. The results in Hospital Service Area J
dard of 2.75 for that period and the appen- were similar, although somewhat higher. In the
dicitis death rate of 5.10 for the eight highest four hospital service areas having high total
areas.
rates, the urban appendectomy rate was 8.11
Of course, it is not known whether the hospi- per 1000 and the rural rate was 3.44. The corretal service areas with the highest appendectomy sponding mortality rates were 5.26 for the urrates in 1948 occupied the same relative posi- ban and 3.94 for the rural area. This difference
tion in the four preceding years. It does seem in urban-rural rates is not due to the different
reasonable to infer, however, that within the up- age distribution of the various population
per and lower limits of the appendectomy rates groups. A reasonable inference is that in areas
observed in this study, a low incidence of sur- served by hospitals where operations are not
gical operations for appendicitis was not associ- controlled strictly, the simple fact of greater accessibility to hospital and medical care is a deated with high appendicitis death rates.
ciding factor.
Few references to appendectomy incidence in
the
general population were found in the literCOMPARISON OF APPENDECTOMY RATES
ature. Those that were found tended to correIn a later section of this paper the possibility spond with the overall rates in this study.
In a questionnaire survey of 2968 college stuis suggested that in the future the surgical treatment of appendicitis may be largely supplanted dents averaging 19.5 years of age, Stiles and
by antibiotic therapy alone, and that appendec- Mulsow (3)elicited a history of previous appentomy may become as obsolete as mastoidectomy. dectomy in 9.6 percent of males and 10.9 perIn the meantime, it is necessary to answer the cent of females. Comparable cumulative perquestion, In the present state of our knowl- centages were developed from the data in this
edge, what is a reasonable annual rate for pri- study, using 1948 age-specific rates and New
mary appendectomy? The difficulties in an- York State life table data (4). At the 1948 rates,
swering this question may be eased somewhat 12.2 percent of males and 17.2 percent of
by the results of further investigation in some of females in all 23 hospital service areas would
the hospital service areas studied. For example, have had primary appendectomy by age 20. In
the figures shown in Table 1 prompted a study the standard area, however, comparable figby the medical staff of the hospital serving the ures were much lower-7.7 percent for males
area designated as U in the table. They found and 8.4 for females.
The Navy reported for 1949 a total of 3831
that among 16 physicians with surgical privileges, one-half of the primary appendectomies appendectomies in a mean strength of 535 410
were done by only two, neither of whom re- persons of median age 23.2, a rate of 7.2 per
ceived referred cases; and that clearcut indica- year (5).In a 10 year period at the University of
tions for surgical operation were lacking in a Minnesota, Fowler and Boehner (6) found an
large proportion of patients operated on by the annual incidence of 4 per 1000 among 15 000
two physicians. There can be no reasonable students averaging about age 20. Such a rate
doubt that many unnecessary operations have was no more than one-half the actual incidence,
been done on residents of this hospital service they thought, because many students had ap-
Lembcke
Data now available permit calculation of the crude primary appendectomy rate among the 18 000 persons residing
in the Cooperstown, N. Y., hospital service area served primarily by the Mary Imogene Bassett Hospital. The annual
average resident rate over the five year period 1946-1950 was
1.9 per 1000 general population. Of the 171 primary appendectomies on which this rate is based, 8 1 percent were done in
the Mary Imogene Bassett Hospital, and 19 percent in hospitals outside of the Cooperstown hospital service area.
899
900
Figure 1. Appendicitis death rate per 100 OOO population, New York State, 19001950, and Metropolitan Lie Insurance industrial policyholders, 1911-1950.
18
16
14
12
W
10
$ 8
I
t-
$ 6
u,
k
0
n
z
w
a
a
_ _ _ _ _ _ _ M E T R O P O L I T A NLIFE INSURANCE
INDUSTRIAL POLICYHOLDERS
1900
10
20
30
40
I
50
Lembcke
901
when surgical facilities were lacking, appendicitis was often treated successfully with penicillin or other antibiotics only (14, 15).Recent
studies such as those of Yaeger, Ingram, and
Holbrook (16) on aureomycin in experimental
peritonitis in dogs; Yaeger, Lynn, and Barnes
(171, Crile (It?), and J o s h and Drake (19)on the
use of aureomycin and penicillin in peritonitis
of appendiceal origin in children and adults,
strongly support the importance of the antibiotics, even though most of the human cases were
subjected to surgery routinely, without regard
to presence or absence of abscess, after antibiotic therapy was completed.
The question of whether surgery is needed in
addition to antibiotic therapy is still open. As
was true of the early advocates of conservative
management, Crile and Fulton (15)seem to feel
that surgery may be dangerous where the peritoneum is involved. They point out that in the
majority of cases, acute appendicitis is selflimited and subsides spontaneously and completely in 48 hours. They consider that the majority of so-called ruptured appendices are
periappendiceal abscesses that are ruptured by
the surgeon during the operation. Also, they
believe that if the patient is first seen when the
infection has extended into the free peritoneal
cavity, immediate operation is unlikely to improve chances of recovery. These observations
and inferences suggest that surgical treatment
of appendicitis should be required only rarely
because, according to Crile (18)only 5 percent
of cases go on to the stage of appendiceal peritonitis, and if treated with antibiotics only onehalf of the peritonitis cases will require surgical
treatment-usually for drainage of residual abscess. It seems possible also that considerably
less than 5 percent of cases of acute appendicitis
would progress to peritonitis if they were
treated promptly with antibiotics. Surgical treatment for the relief of symptoms due to fecaliths,
stenosis, or external constricting agents should
be required very rarely, according to Jennings,
Burger, and Jacobi (20), who found only 14
such cases in a total of 1680.
The findings of this study, and the publications cited, are interpreted as strongly suggesting that the antibiotics are the essential factor in
successful treatment of appendicitis with peritonitis, and uncomplicated appendiceal infections as well. In view of the possibility of a great
reduction in the amount of surgical treatment
902
1. Resident vital statistics derived from hospital records and based on the total population of
a hospital service area represent one useful type
of measurement of the quality of medical care.
Clearly demarcated, mutually exclusive areas
served principally by one or more hospitals can
be defined by residence studies of hospital patients, and detailed population data for such
hospital service areas can be assembled from
that supplied by the U. S. Census for small
subdivisions.
2. The outstanding values of resident vital statistics based on hospital service areas are: (1) rates are
comparable between areas and with a standard;
(2) responsibility for results is associated closely with
the hospital medical staff and governing board;
(3) the likelihood of corrective action is enhanced by
the objectivity of the rates.
3. This method is best applied to diseases and
operations assodated almost exclusively with hospitals. Limitations on the effectiveness of the results
are encountered when a hospital service area is
served in large part by several hospitals.
4. Application of this method to a study of
primary appendectomies in 23 hospital service
areas would seem to indicate that considerably
more operations of this type are done than are
necessary in the light of our present knowledge-excesses of from 25 to 100 percent in
various areas.
5. The findings of this study, together with
observations by others reported in the recent
medical literature, are interpreted as showing
that the antibiotics are the essential factor in the
currently successful treatment of acute appendicitis, with or without peritonitis, and that the
value of surgery is questionable in most cases.
In view of these findings, a controlled clinical
study of the relative efficacy of surgery with or
without antibiotic therapy, and of antibiotic
therapy alone, is recommended.
References
(1) Sinai, N. and D.E. Paton. Hospitalization of the
People of Two Counties. Bureau of Public Health Economics Research Series No. 6, Ann Arbor, Mich.,
University of Michigan, School of Public Health,
1949.
(2) The Regional Hospital Plan-The First Years Experience. Council of Rochester Regional Hospitals,
1946.
( 3 ) Stiles, K.A. and F.W. Muslow. Incidence of
appendicitis from a survey of college students. Am J
Digest Dir 13:39-40, 1946.
(4) State Regional Life Tables, 1939-41. National Office of Vital Statistics, Public Health Service, Washington, D.C., Federal Security Agency, 1948.
(5) Statistics of Diseases and Injuries in the US.
Navy. Bureau of Medicine and Surgery, Dept. of the
Navy, 1949.
(6) Fowler, L.H. and J.J. Boehner. Appendicitis in
college students. Staff Meeting Bull Hosps University of
Minnesota 11:344-361, 1940.
(7) Green, H.W. and R.M. Watkins. Appendicitis
in Cleveland. Surg Gynecol Obstet 83:613-624, 1946.
(8) Stat Bull Metropol L f e Insur Co Vol. 32, 1951.
(9) Fitz, R.H. Perforating inflammation of the vermiform a p p e n d i x . T r a n s Assoc A m P h y s i c i a n s
1:107-144, 1886.
(10) Homans, J. A Textbook of Surgery Baltimore,
Md., Thomas, 1932.
(11) Willis, A.M. Mortality in appendicitis. Surg
Gynecol Obstet 42:318-322, 1926.
( 1 2 ) Reynolds, J.J. Conservative management of
appendiceal peritonitis. New Orleans M t3 S J
87:32-39, 1934.
(13) Adams, J. and P.M. Bancroft. The conservative
management of appendiceal peritonitis in children. J
Pediat 12:298-312, 1938.
(14) Berkely, W.L. and H.C. Watkins. Chemotherapy in the management of acute appendicitis. U S .
N a v M Bull 42:l-6, 1945.
(15) Crile, G., Jr. and J.R. Fulton. Appendicitis with
emphasis on use of penicillin. U.S. Navy M Bull
451466-473, 1945.
(16) Yeager, G.H., C.H. I n g r a m , a n d N.A.
Holbrook, Jr., Comparison of effectiveness of newer
antibiotics in experimental peritonitis. A n n Surg
129:797, 1949.
(17) Yeager, G.H., W.D. Lynn, and T.G. Barnes.
Treatment of peritonitis of appendiceal origin with
aureomycin. South Surgeon 16:1192, 1950.
(18) Crile, G., Jr. Peritonitis of appendiceal origin
treated with massive doses of penicillin. Surg Gynecol
Obstet 83:150-162, 1946.
(19) J o s h , B.S. and M.E. Drake. Aureomycin in
the treatment of ruptured appendices in children.
Pediatrics 7:684-690, 1951.
(20) Jennings, J.E., H.H. Burger, and M. Jacobi.
Acute appendicitis: A clinical and pathological study
of 1680 cases. Arch Surg 44:896, 1942.
In 1957 the Social Medicine Research Unit of the Medical Research Council first
showed that there were differences in case-fatality of certain conditions between
teaching and non-teaching hospitals. That report was the beginning of a series and
went back to 1957. Now the latest available figures have been studied and the story
has been brought up to 1959. It is clear that the differences persist.
It has previously been shown that there was a
higher case-fatality in appendicitis, hyperplasia
of prostate, and other conditions in non-teaching hospitals of the National Health Service
than in teaching hospitals (I, 2). These studies
were based on data collected in 1951-1955 by
the General Register Office in the national Hospital Inpatient Enquiry which they made jointly
with the Ministry of Health. Collection of data
has continued, and analysis of the years
1956-1959 is now presented.
The hospitals participating in the Enquiry
report an effectively random 10 percent sample
of their discharges. The proportion of beds
included in the Enquiry has increased from 79
percent of the total teaching hospital beds in
England and Wales and 67 percent of the nonteaching in 1956, to 95 percent and 97 percent,
respectively, in 1959.
RESULTS
Table 1 shows the conditions where the authors feel that comparison is most valid. Numbers are substantial and the diagnostic label is
unequivocal. The Appendix Table includes all
the data available to us, except for diabetes
which is dealt with separately. Immediate admissions and other admissions are, in general, distinguished in the data, the latter referring largely to cases from the waiting list and
transfers from other hospitals. The proportion
of older patients is greater in the non-teaching
hospitals, and direct standardization for age was
used to adjust for this. Standardization for sex
was also done where applicable.
Source: Medical Care 1:71-76, 1963.
From the Social Medicine Research Unit, Medical Research Council, London Hospital, E.l.
903
904
Condition
Non-teaching
hospitals
Standardized
Case
fatality
Case
fatality
Deaths
Deaths
378
23
4060
29 (P<.OOl)
Peptic ulcerb
(a) With perforation
(b) Without perforation
All immediate admissions
33
61
94
8.1
5.0
5.8
316
594
910
Appendicitis
(a) With peritonitis
All immediate admissions
10
14
2.8
0.44
121
186
4.3
0.60
23
6.1
312
9.7 (P<.Ol)
16
2.6
207
3.6
Hyperplasia of prostate
(a) With acute retention
1957-1959
(b) Without mention of acute
retention 1957-1959
All immediate admissions,
1956-1959
20
16
10
10
5.3
6.4
270
14
9.5
262
12
46
9.4
636
13 (P<.05)
68
2.7
602
3.4 (P<.05)
16
1.3
106
2.3 (PC.05)
10
1.1
81
Hyperplasia of prostate
21
3.5
190
1.5
6.0 ( R . 0 1 )
a The standardized case fatality was derived by direct standardization on the age and sex distribution of the teaching
Hospital admissions.
Where differences are statistically significant the probability level is shown.
Admissions with hematemesis or melena are not separated in the data supplied.
Hyperplasia of Prostate
Analysis of case-fatality in this condition is for
many reasons particularly worthwhile. Table 2
has given details by length of stay. In Table 4
the case-fatality for different age groups in both
Two main lines of explanation may be proposed. Firstly, it is possible that the difference in
case-fatality represents superior treatment in
the teaching hospitals. Secondly, the non-teach-
Lipworth et al.
Table 2. Case-fatality of hyperplasia of pmstate in teaching and non-teaching hospitals
by duration of stay 1956-1957.
ing hospital patients may be socially or otherwise at an initial disadvantage. It has already
been shown that, proportionate to their number
of beds, teaching hospitals had more consultants and other staff. Hospital staffing is at present under review as a result of the Platt Reports
and some improvement may result from this.
We know of no recent published evidence of
social differences between the patients of the
two types of hospital. If these were substantial
they could well be a factor; thus it is well known
that chronic bronchitis is more prevalent in the
lower social classes. This aspect needs more
study. In the meantime there is little evidence
Teaching
Non-teaching
Duration of stay
hospitals,
hospitals,
in days
case-fatality, % case-fatality,b %
~~
0-3
4-7
8-2 1
22-28
29 +
All durations
0.2
1.3
1.3
1.3
2.1
6.2
905
1.0
1.4
4.2
1.2
2.5
10.2
I.C.D.
No.
Condition
260.1260.5
260
Teaching hospitals
CaseCases Deaths fatality, %
Cases
Non-teaching hospitals
CaseStandardized
Deaths fatality, % case-fatality, %
132
4.5
734
130
18
14 (P<.O01)
274
19
6.9
1215
155
13
13 (P<.Ol)
406
1000
1406
435
25
42
67
3
6.2
1949
6777
8726
2244
285
298
583
37
15
4.4
6.7
1.6
14 (P<.OOl)
3.7
5.9
1.6
4.2
4.8
.69
International Statistical Classification of Diseases, Injuries and Causes of Death, World Health Organization (1957) as
modified by the General Register Office.
Table 4. Case-fatality of hyperplasia of prostate in teaching and non-teaching hospitals showing different
age groups, 1956-1959.
Non-teaching hospitals
Rate %
Cases
Deaths
Rate %
Non-teaching
hospital
standardized
13%
Teaching hospitals
Cases
Deaths
45 years
45-54
55-64
65-74
75 +
Total
Immediate Admissions
1
16
103
2
224
21
143
23
487
46
1.9
9.4
16
9.4
26
122
906
1834
1721
4609
2
1
41
200
392
636
7.7
0.8
4.5
11
23
14
45 years
45-54
55-64
65-74
75 +
Total
Other Admissions
6
45
206
233
112
602
1.5
2.6
11
3.5
35
121
76 1
1111
682
2710
1
13
68
108
190
0.8
1.7
6.1
16
7.0
As in Table 1
3
6
12
21
6.0%
906
9 8-
7 6 -
4
Non-teaching
55-
I-
z
w
5n
E d -
3 -
21
1 -
1953-55
Deaths
Case-fatalitya
52
9.9%
201-500
beds
70
13%
Over 500
beds
15
9.7%
1956-57
1958-59
Lipworth et al.
907
Appendix. Case-fatality in teaching and non-teaching hospitals for all data examined.
Teaching hospitals
Condition
I.C.D.
No.
Cases
Non-teaching hospitals
CaseDeath fatality, %
Cases
CaseStandardized
fatality, % Case-fatality, %
Deaths
Appendicitis and
peritonitis
Acute appendicitis with
peritonitis
Peritonitis (cause not stated)
Appendicitis without
mention of peritonitis
All immediate
appendicitis admissions
1568
369
100
1668
9
378
468.1
161
540.0
541.0
542.0
540.0
54 1.O
542.0
11 716
3927
34
30
9
23
925
12 641
133
4060
14
32
12
29 (P<.OOlj
1217
547
628
36
35
22
4
6.4
3.5
11
4626
5405
184
340
242
12
7.3
4.5
6.5
6.6
4.1
5.4
1211
61
5.0
10 215
594
5.8
5.3 (N.S.)
86
315
6
16
17
0
19
5.4,
0
738
1606
30
120
192
4
16
12
13
17
8.5
8.6
407
33
8.1
2374
316
13
10
633
94 3
42
1618
51
39
4
94
8.1
4.1
9.5
5.8
5364
701 1
214
12 589
460
434
16
910
8.6
6.2
7.5
7.2
362
48
10
16
2.8
33
2435
486
12 1
198
5.0
41
2821
0.14
26 791
65
0.24
0.22
550-552
3183
14
0.44
29 226
186
0.64
0.60 (N.S.)
561
560
560-56 1
375
515
890
23
9
32
6.1
1.8
3.6
2899
4211
7110
312
81
393
11
1.9
5.5
9.7
1.6
4.7 (N.S.)
585
613
16
2.6
5389
207
3.8
3.6 (N.S.)
610
158
16
10
1754
270
40
2.5
480
45
540.1
541.1
542.1
540.1
541.1
542.1
540
54 1
542
540-542
550.1
576
550.0
55 1
552
24
.082
(
6
)
(N.S.)
8.0
5.3
6.4
6.4 (N.S.)
4.3
35
584
786.1
15
9.4
14
7.8
908
Teaching hospitals
CaseCases Death fatality, % Cases
I.C.D.
No.
Hyperplasia of prostate
without acute retention
of urine 1957-59
All immediate admissions
of hyperplasia of
prostate 1957-59
All immediate admissions
1956-59
Retention of urine
Skull fractures and head
injuries
(Except fracture of face
bones)
Non-teaching hospitals
CaseStandardized
Deaths fatality, % Case-fatality, %
610
210
20
9.5
2107
262
12
12
610
368
36
9.8
3861
532
14
13
610
786.1
800
801
487
78
46
3
9.4
3.8
4609
804
636
73
14
9.1
13 (PC.05)
7.3 (N.S.)
2508
68
2.7
18 288
602
3.3
380
36
9.5
1556
550
35
22
108
0 (-)
803
804
850-856
3.4 (P<.05)
468.1
25 (PC.001)
-gastric
With
-duodenal
operation-gastrojejunal
-All ulcers
Without
-gastric
operation-duodenal
-gastrojejunal
All ulcers
All non-gastric
immediate-duodenal
-gastrojejunal
All ulcers
540
54 1
542
540-542
540
54 1
542
540-542
540
54 1
542
540-542
369
773
46
1188
273
424
14
711
642
1197
60
1899
9
6
1
16
2
2
0
4
11
8
1
20
2.4
0.78
2.2
1.3
0.73
0.47
0
0.56
1.7
0.67
1.7
1.1
1531
2747
97
4375
1283
1760
50
3093
2814
4507
147
7468
50
55
1
106
34
20
0
54
84
75
1
160
3.3
2.0
1.0
2.4
2.7
1.1
0
1.7
3.0
1.7
0.68
2.1
3.1
1.9
1.2
2.3 (Pi.05)
2.1
0.83
0
1.3 (Pi.05)
2.6
1.5
0.72
1.9 (P<.Ol)
Appendicitis
550-552
810
0.25
9625
16
0.17
0.19 (N.S.)
All hernia of
abdominal cavity
560
56 1
3872
10
0.26
22 621
74
Peritonitis
576
12
17
83
25
584
585
924
10
1.1
4953
81
1.6
1.5 (N.S.)
Hyperplasia of Prostate
610
602
21
3.5
2710
190
7.0
6.0 (P<.Ol)
Retention of urine
786.1
17
127
4.7
3.2 (P<.05)
813
46
5.7
4.4 (Pi.001)
0.33
30
0.29 (N.S.)
22 (N.S.)
800
801
803
804
850-856
260
~~
14
36
~
Admissions with hernatemesis o r melena are not separated in the data supplied.
Breakdown in sub-categories shown is only available for 1957-59.
I.C.D. No. refers to the international statistical classification of diseases (seventh revision, 1955) adopted by the World Health
Organization.
Tests of statistical significance were applied to all totals and the results are as shown.
The Standardized case fatality was derived by direct standardization on the age and sex distribution of the teaching hospital
admissions.
Terminology used in the breakdownof ischemic heart disease (immediate admissions) and under other sub-headings is that
used in the General Register Office.
a
Comparisons of the self-care status of treated and untreated patients before and
after a one-year period of treatment suggest that the rehabilitation potential of a
disabled public assistance population tested by the provision of maximum medical
rehabilitation programs is very low.
BACKGROUND AND PURPOSE OF STUDY
'
909
9 10
Bed capacity
Welfare clients
(no.)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
58
135
134
60
83
353
272
67
50
328
180
248
180
240
59
41
103
120
30
78
327
226
19
31
255
109
157
156
142
50
Total
2447
1844
York City Inter-
91 1
Total
Diagnostic category
Number
Percentb
35
18
15
15
8
4
11
35
17
15
14
8
4
11
35
18
14
14
8
4
10
35
12
14
12
8
2
11
140
65
58
55
32
14
43
34
16
14
13
8
3
11
106
104
103
94
407
100
Total
a
A = Control group; B = Nursing home treatment group; C = Rehabilitation center treatment group; D = Control group.
All percentages are approximate.
912
Male
Total
Age
(years)
No.
Percent
No.
Percent
No.
Percent
Under 69
70-79
80
62
81
93
26
34
40
76
65
26
46
39
15
138
146
119
34
36
30
Total
236
100
167
100
403a
100
(59%)
(41%)
a Discrepancy between total number in this table and in succeeding tables is due to lack of information on small numbers of
patients.
RESULTS
Table 4. Number and group assignment of study population according to nursing home.
Nursing
home
Number of
patients
screened
Number of
eligible
patients
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
36
105
123
29
80
356
255
19
38
279
130
171
167
162
57
2007
Total
Assigned group
A
10
39
26
6
16
80
50
4
14
52
16
32
31
25
6
3
12
11
0
6
29
15
0
5
18
7
3
15
7
3
3
25
18
3
6
16
5
4
12
8
3
7
26
17
1
3
18
4
407
106
32
31
25
6
104
103
94
913
Shown in Table 5 are the medical and rehabilitation prognoses offered by the screening physiatrists. In only 25 percent of the cases were
good rehabilitation prognoses given and only
2 1 percent of the population had medical prognoses which implied a possible return to full
activity. Only 43 patients, approximately 10 percent of the sample, were regarded as having
reasonable medical and rehabilitation prognoses for return to full activity.
The limited numbers of disabled persons
available for rehabilitation treatment in this
population and the still smaller proportion with
good medical and rehabilitation prognoses support earlier survey and study reports by physiatrists of a low rehabilitation potential for physically impaired nursing home populations (4,9).
Pretreatment Independence. Despite these early
physiatric findings in the nursing home, the
results of the independent pretreatment selfcare tests, as shown in the figure, indicated that
rather high proportions of this population were
independent or nearly independent in four out
of five self-care criteria areas. Except for dressing-an activity which was discouraged by both
the social conditions of living and the imperatives of patient management in the nursing
home-this population's measured levels allowed for little possibility of showing improvement over time. For most of the population,
only maintenance or worsening in self-care levels was possible.
An interesting by-product of this test finding
was the marked contrast between the high levels
of self-care skills obtained at the test site away
from the nursing home and the low levels of
self-care skills reported by our clinical teams in
their day-to-day observations of this patient
group in the nursing home. One interpretation
of this discrepancy would suggest that the major
rehabilitation problem was the lack of employment of physical self-care skills already present
in this population in the nursing home rather
than a need to restore physical function to perform self-care tasks. Questions may also be
raised concerning differences in function under different conditions as well as differences
between test results and the ability to sustain
activity in daily life.
Comparisons. Comparisons of the differences
between initial and post-treatment self-care statuses indicate that neither of the two rehabilita-
"
LOCOMOTION
TRANSFER SKILLS
60
40
20
0
4
TOlLETlNG
4
FEEDING
60
40
20
c
-Deoendence
.
1
4
DRESSING
4
___+
Independence
914
Fair
Percent
No.
Poor
Percent
No.
Totals
Percent
No.
Percent
No.
~~
Total
a
43
58
1
42
57
1
31
188
3
14
85
1
12
61
18
13
67
20
102
100
(25%)
222
100
(53%)
91
100
(22%)
86
307
22
21
74
5
415a
100
Table 6. Percent distribution of differences between initial and final self-care scores according to
assigned group (excluding losses).
Self-care areas
Locomotion
Group
Difference"
Transfer skills
Dressing
Feeding
Toileting
A (control)
B (nursing home)
C, (hospital transfer)b
Ce (hospital refusa1)c
D (control)
20
23
13
18
20
50
48
55
63
39
30
29
32
19
41
12
13
10
11
13
69
72
73
56
74
19
15
17
33
13
14
18
17
18
33
60
59
50
67
55
25
23
34
15
11
11
12
7
11
8
77
65
72
78
79
12
23
21
11
13
18
26
10
12
32
49
51
67
44
51
33
23
23
44
17
All groups
20
49
31
10
70
20
21
58
21
10
74
16
22
52
26
showed no greater potential for gain or maintenance in status when medical rehabilitation care
was supplied than when it was not.
Attempts were also made to look for group
differences by comparing patients rather than
attributes or traits. This was done by arithmetically combining, for each patient, the scores
of the pretreatment evaluations and then comparing this sum with the sum of the posttreatment scores. The results are shown in Table 7.
Here, too, no consistent pattern favorable to the
therapeutic programs was obtained. Part of this
result may be due to the previously noted high
levels of initial self-care status and the outstanding population characteristic of failure to show
significant change (more than
or - 1 score
unit) in measured self-care levels. In Table 8,
the distribution of obtained score differences
for the entire population is shown. The percentage of persons who receive identical scores ini-
Total
all
groups
Change in
statusa
Improved +
Same 0
Worsened -
20
13
31
16
15
29
16
11
22
25
20
16
77
59
98
64
60
49
61
234
Total
9 15
Table 8. Distribution of differences between initial and final test scores for all groups (excluding losses)
according to self-care area.
Difference
in
scores
Locomotion
Transfer skills
No.
Percent
No.
3
12
35
1.2
4.9
11.0
14.3
4
8
14
17
120
48.9
+1
28
12
7
1
11.4
4.9
2.9
0.4
-4
-3
-2
-1
+2
+3
+4
27
Percent
Toileting
Feeding
Dressing
No.
Percent
No.
Percent
No.
Percent
1.7
3.4
5.7
7.0
0
2
12
49
.9
4.9
20.2
3
2
4
30
1.2
0.8
1.6
12.3
0
6
9
37
0
2.5
3.7
15.0
169
69.8
125
51.7
180
74.1
143
58.1
21
4
4
1
8.7
1.7
1.7
0.4
45
8
1
0
18.5
3.3
0.4
0
18
5
1
0
7.4
2.1
0.4
0
32
12
7
0
13.0
4.5
2.9
0
enced patient hospitalizations and mortality. Attempts were made to insure that each patient
The evidence does not allow for an interpreta- so referred received an adequate therapeutic
tion that either numbers of hospitalizations or trial before being referred back to the nursing
patient deaths were favorably influenced by the home. Still, there were frequent questions from
therapeutic programs. These findings would the operational staffs at these hospital centers as
appear to confirm the early pretreatment esti- to the validity of referral of these patients to
mates made by the screening physiatrists of the their facilities.
The impressions of our therapeutic teams
restricted rehabilitation and medical prognoses
who treated patients in the nursing home also
of this group.
pointed up the apathy, lack of motivation, and
Reuctaom to Treatment. The clinical response of
reluctance of patients to participate in the forthis population to the two therapeutic programs
mal physical exercise and conditioning prodeveloped in this study provides additional evigrams. Many patients questioned the relevancy
dence bearing on the question of its rehabilitaof these exercises to their immediate needs,
tion potential. The most general conclusion
though they enjoyed the social interaction the
which can be drawn is that the goals of more
therapeutic sessions entailed. There was eviindependent self-care and ambulation were not
dence that few patients carried out the exercise
considered by this population as being relevant
programs when our staff was not present unless
to their immediate concerns and problems.
the nursing home staff itself provided the stimThis was illustrated most dramatically by the
ulation and the initiative. The populations imresistance shown by those patients (group C)
mediate concerns were with their everyday comdesignated to be transferred from the nursing
plaints and worries. T h e quality of food,
home to hospital rehabilitation centers. Less
interpersonal difficulties with other patients,
than half of this group agreed to participate in and bodily aches and pains were the problems
the program, and they did so only after inten- they were anxious to have our staff deal with.
sive efforts on the part of the project casework
staff. The reasons given for such lack of cooperation related to patients concerns for their
SUMMARY
future well-being, negative and hostile attitudes
A medical rehabilitation program was offered
toward hospitals, fear of losing their present
nursing home situation, and disillusionment to a physically impaired population (1) with a
background of chronic disease and disability
with rehabilitative care received in the past.
Pretreatment orientation and planning ses- and a long history of economic and social desions were held with the administrative staffs at pendency; (2) drawn from the lower sothese rehabilitation centers. Nevertheless, most cioeconomic strata in our society, with few
patients were not regarded by the therapeutic meaningful family ties and attachments; and
personnel as good candidates for rehabilitation. (3) with many previous hospitalizations and
916
There is much interest in the role of the gen- of a year in the practice, with disease known to
eral practitioner in early detection of chronic the general practitioner on the left, and the
disease and in its prevention. In this paper 1 undetected cases on the right. The annual
have used epidemiological methods to show the number of new cases and the age-groups of all
nature and size of some of the problems in patients, with data on precursors and associated
England and Wales, by adjusting the relevant signs for some diseases, are given.
Official sources have been used to compile
data to a hypothetical average general practhe vital statistics given in Table 3 and for the
tice.
The practice has been given a list of 2250 (the estimate of certain other events in the practice.
nearest round number to the average list of
general practices in England and Wales in
COMMENT ON TABLES
1960-I), and the age and sex distribution of
England and Wales in 1960 (2) (Table 1). I have
estimated the number of patients in this prac- Subclinical Disease
tice who would be known to have certain disCommunity surveys have shown that epilepsy
eases according to morbidity statistics (2-16). (I#), psychoneurotic illness (8),
chronic bronchitis (3),and rheumatoid arthritis (IO),are more
prevalent than is suggested by the morbidity
Table 1. Population distribution in the average
general practice.
statistics of 106 general practices in England in
1955-1956 (11). N o doubt there are undiscovered cases of these diseases in most gen65+
Total
15-44
45-64
0-14
eral practices. Some of these potential patients
may be people living with a disability which has
1085
276
104
Male
264
441
been
recognized, and for which they have not
164
1165
304
446
Female
251
sought treatment in the year of the inquiry. But
268
2250
887
580
this can hardly be true of diabetics and the
Total
515
tuberculous, where the figures support the dictum that for every known case there is another
By the methods described by Morris (17), I have undiscovered.
tried to complete the clinical picture by esDetection of diabetes has been made easier by
timating the numbers of people with un- the use of glucose-oxidase paper strips, a simple
detected or potential disease who might be and reliable method of urine-testing. Commufound on search. These numbers are mostly nity studies (4, 15) show that middle-aged and
based on surveys of whole communities for a elderly patients are most likely to have unrecogparticular condition, and I have assumed that nized diabetes, and the general practitioner has
the hypothetical average general practice will many opportunities to test their urine. Those
contain the same proportion of people with with symptoms can be helped, and some comthese conditions. Table 2 shows the experience plications perhaps reduced.
The figures for tuberculosis (2, 6, 18) prove
that this disease is still a clinical and public
Source: The Lancet July 6, 1963, pages 28-31.
health problem. The highest rates of infection
1 Work done while holding a grant from the Postgraduate
Medical Foundation, University of Sydney, at the Medical are among elderly men and delinquent and psyResearch Council Social Medicine Research Unit, T h e chopathic members of the community (6),some
London Hospital, London, E. 1.
* School of Public Health and Tropical Medicine, Univer- of whom are likely to be found in the average
general practice.
sity of Sydney, Australia.
917
918
Table 2. Experience of one year in the average general practice; both sexes, all ages, unless specified.
Total present in the practice, including undetected
and potential disease
,,
,,
,,
,,
,,
,,
,,
,,
,,
,,
,,
,,
,,
,,
,,
,,
,,
,,
,,
,,
lung
breast (3 in 4
years)
stomach (3 in 4
years)
prostate and
rectum ( I
in 2 years)
cervix (1 in 4-5
years)
6-7
1
0-12
7
1
0-75
0-74
0-52
0-22
1
1
1
12
7
5
14
12
20
Staphylococcal diseases ( I I )
Overt skin infection
Glaucoma (11)
Aged 45+
110
8
24
Bronchitis (11)
Males aged 45-64
Females aged 45-64
24
19
11
7-8
27
62
12-14
2-3
3-4
11-12
0.5
2-3
10
23
23
114
37
35
29
15
14
Significant bacteriuria ( I I )
Females aged 15
40
Nasal carriers of Staph. aureus (13)
500- 1500
>,
,, (penicillin-rtsistant)
100- 300
Early chronic glaucoma (5)
Aged 45+
17
Casual diastolic blood-pressure 100 mm.
Hg. and over (12)
Males aged 45
30
Females aged 45+
131
Symptoms and signs of bronchitis (3)
Males aged 45-64
47
Females aged 45-64
24
Defihite and probable rheumatoid arthritis (10)
Aged 15+
25
Epilepsy (14)
13-14
Conspicuous psychiatric morbidity (7)
Males 15+
58
Females 15+
102
I,
I,
Last
919
2
1
10
5
4
1
1
39
17
26
Ophthalmic
General medical
Mental Hospital:
All admissions
3-4
3-4
5
208
96
1-2
9-10
14
2-3
1
51
46
32
31
31
First admissions
2-3
New outpatients
8
Diagnostic Services:
Referred to hospital pathological
laboratory
59
Referred to hospital X-ray department
99
Referred to hospital mass radiography by practitioner
10
Annual number examined by mass
177
radiography
Domiciliary visits by consultants
16
Doniicaliary Services:
Health visitor
600 visits
1134 visits to 45 paHome nurse
tients
Home help
16 patients
Maternal and Chald Welfare:
Patients attending antenatal clinic
17
Patients attending postnatal clinic
2
Domiciliary confinements:
Attended by midwife
14
Doctor booked
12
Attended by doctor
2
First attendance infant welfare
clinic
31
Seen by school medical service
107
64 1
277
364
100
17
17
25
5
ably) staphylococcal skin infection in the average general practice (boils, carbuncles, impetigo, styes, and so on), there may be between 500
and 1500 nasal carriers of Staph. aureus-perhaps a fifth of them penicillin-resistant (13).
Anemia
If recent community surveys (9) record the
true prevalence of anemia, the average general
practitioner deals with only about one patient in
seven of those who are anemic. In the course of
a year many of the 200 with anemia in the
practice will consult their doctor for incidental
disease, if not because they have symptoms due
to the anemia itself; and many are probably
working below full efficiency. A small minority,
especially of the elderly, have grave disease of
which anemia is an early sign (21). Clinical ex-
920
Last
Table 4. Coronary-artery disease in men
aged 45-64.
Vkible:
1 Death (2).
5 Cases (11).
Submerged:
1 1 E.C.G. evidence of left ventricle hypertrophy
921
OTHER EVENTS
Social Pathology
The figures for illegitimate births and for divorce are given in Table 3. Others can be derived from various sources. Each year one adult
criminal will be sent to prison, and five or six
children under the age o f 17 will be charged
(38).
15 Casual diastolic blood-pressure 100 mm. Hg with offenses (27). About 100 people in the
or over (12).
practice will receive National Assistance (29).
24 Serum-cholesterol 300 mg. per 100 ml., or
Twenty-five
to fifty people over retiring age will
over (39).
live alone (30) and about 40 children under the
28 Healed infarcts (40).
52 Smoking more than 20 cigarettes per day (25). age of 15 come from broken homes (31).There
55 Obese, more than 10% over ideal weight (41). will probably be between five and ten problem
140 Moderate to severe atheroma of coronary arfamilies (32); and there will be four chronic
teries (42).
alcoholics with mental and physical complica? ? ? Insufficient exercise.
Worried by responsibility.
tions, and about another dozen who are adOther emotional stress.
dicted to alcohol (33). Each year ten abortions
276 At risk.
will escape the notice of the medical profession,
compared with the three or four who receive
proper medical care (34). Many of these numbers
are only crude estimates, and regional and
Although mass serum-cholesterol estimations
social
variations could cause wide deviations
are hardly justifiable until we have a cheap mifrom
the
average.
cromethod, ordinary clinical examination could
detect more men with high blood-pressure, Uncommon Events
some of whom might need treatment. Cheap,
In the same way figures can be adjusted to
and easily portable, transistorized electrocarshow how seldom some conditions will turn up
diographs could be used by general practiin the average general practice. Diseases which
tioners to screen vulnerable groups in their
were common a generation ago are now rare;
practice-for
example, men who are overthe average general practitioner might wait
weight, who lack opportunity or incentive to
eight years to see a case of rheumatic fever in a
take exercise, or who smoke too much. These
child under the age of 15; 60 years to see a case
should benefit from the good counsels of their
of typhoid or paratyphoid fever, and as long as
family doctor. The question marks in the table
400 years to see a case of diphtheria (18). He
indicate suitable subjects for research in general
will probably see a patient with schizophrenia
practice; no one has any clear idea about the
once in two years ( 3 3 , a patient with leukemia
leisure activities of middle-aged men, nor about
or other malignant disease of the lymphatic and
the relation, if any, between these and health. IS
hemopoietic system once in four years, and a
the vicarious stress of watching competitive
patient with a cerebral tumor once in eight to
sports related to the incidence of stress disease, ten years (2).
such as hypertension?
The diversity of activities in the average genPatients who commit suicide are at the tip of
eral practice, and the range of tasks expected of
another iceberg. In the average general practice
the average family doctor have been shown.
there will be one suicide (2) in four years, and,
Changing conditions of practice in recent years
when it was still an indictable offense, one atshould not allow his clinical skills to atrophy.
tempted suicide came before the courts (27) in
the same period. But each year at least two more
COMMUNICATIONS
people will have made suicidal attempts (28). A
still larger number have depressive illness seIf the general practitioner is to draw together
vere enough to make them wish to end their the various parts of the National Health Servlives. These people do not always get appropri- ice, much will depend on the efficiency of his
ate medical treatment, many with milder de- lines of communication. The number of conpression are even less likely to do so.
tacts with the hospital services (Table 3) is prob-
922
A model has been used to show a years experience in an average general practice, particularly of the chronic diseases.
A considerable amount of undetected disease, some of which is serious and some controllable, might be found fairly easily without
adding greatly to the burden of the days work.
Ordinary clinical skills and new diagnostic
aids may be used to detect cases of actual and
potential disease in general practice.
The detection and control of diabetes, some
forms of hypertension and their sequelae,
glaucoma, anemia, some kinds of cancer, and
coronary-artery disease have been discussed.
This investigation was suggested by Prof. J. N.
Morris. To him and other colleagues at the
M.R.C. Social Medicine Research Unit, I am
grateful for much helpful criticism.
References
Using the World Health Organization definition of health as a starting point, we have
sought indexes of positive physical, emotional,
and social health for use as outcomemeasurements. For the purposes of this trial, these outcome measurements had to be objective, positive in orientation, and capable of application to
several hundred patients by nonclinical interviewers. Satisfactory measures of physical function that had been developed elsewhere (2-4)
were incorporated into a household survey.
However, we were unable to find satisfactory
positive measures of emotional and social function that were reasonably objective and could be
employed and scored by nonclinicians. As a result, our research group had to develop and
validate, in an independent investigation, the
emotional and social function measurements
used in this study.
METHODS
The basic design of the Burlington Randomized Trial is described in detail elsewhere
(1).In summary, 1598 families receiving clinical
services from two family physicians in a middleclass suburb were randomly allocated, in a ratio
of 2: 1, to a conventional group (designated RC),
923
924
in which they continued to receive their primary clinical services from a family physician
working with a conventional nurse, or to a
nurse practitioner group (designated RNP). Patients in the RNP group received their firstcontact, primary clinical services from one of
two nurses who had successfully completed an
educational program that stressed clinicaljudgment in the evaluation and management of
conditions arising in primary care (5). Accordingly, the nurse practitioner either totally managed each patients office visit by providing reassurance o r specific therapy, o r requested
consultation from the associated physician.
Outcome Measures
Four outcome measures were applied to
members of the RC and RNP groups.
Mortality. A surveillance system identified
deaths of RC and RNP patients during the oneyear experimental period. Decedents were categorized by age, sex, cause of death, and group
assignment, and crude mortality rates were generated. On two separate occasions, the clinical
records for each decedent were assembled,
purged of any notation that would indicate the
experimental group to which they had been
randomized, and submitted to the President of
the Ontario College of Physicians and Surgeons. Members of this professional body,
which serves a licensing and disciplinary function for physicians in the province, reviewed
each case to determine whether, in their opinion, the death could have been prevented.
Physical Function. Specific outcome measurements were applied to the same patients
(drawn by random sampling from each of the
families in the study and designated the interview cohort) both before and at the end of the
experimental period, to permit paired comparisons in which patients could serve as their
own controls. The measurement of physical
function determined the patients mobility, vision, hearing, and ability to execute activities of
daily living. The three indexes of physical function were ( 1 ) the proportion of patients with
unimpaired mobility, vision, and hearing on the
day of the interview; (2) the proportion of patients able to execute their usual daily activities
during the 14 days before the interview; and (3)
the proportion of patients free from an illness
~~
Sackett et al.
925
RNP families left the practice because of dissatisfaction. By the final two months of the experiment, the proportion of RNP patient visits managed entirely by the nurse practitioners had
stabilized at 67 percent.
Statistical Analyses
Table 1 summarizes the distributions of family size, sex, age, and annual household income
for the RC and RNP cohorts just before the
one-year experimental period. The groups are
highly similar, and none of the observed differences approach statistical significance. The initial similarity of the RC and RNP groups is
further supported in Table 2, which summarizes the physical function of members of the
RC and RNP groups just before the one-year
experimental period. Large and identical portions of patients in the RC and RNP groups had
unimpaired mobility, vision, and hearing on the
day of the interview. Similarly large and comparable proportions of patients in each group
had been able to carry out their usual daily
activities throughout the 14 days before this interview. A review of the beta levels for the
Table 1. Comparison of the RC and RNP
interview cohorts at the start of the trial..
RC
614
RNP
340
2.8
42
58
2.7
43
57
5
5
8
5
33
31
6
4
5
7
8
29
35
8
4
4
15
13
28
15
16
4
13
12
24
14
23
a RC = patients receiving conventional care; RNP = patients receiving care from nurse practitioners.
926
when the clinical records of decedents were reviewed by appointees of the Ontario College of
Physicians and Surgeons, no deaths of RNP
patients were judged to have been preventable.
pb
RNP
%
Unimpaired mobility, vision, and hearing
Unimpaired in usual daily
activities
Free from bed disability
86
86
0.03
87
86
89
83
0.09
0.22
a RC = patients receiving conventional care; RNP = patients receiving care from nurse practitioners.
Indicates probability that we have failed to detect a real
difference of 2 5 % in physical function between RC and RNP
patients.
Mortality
It was anticipated during the design of the
trial that the number of deaths during the experimental period would be small. As shown in
Table 3, there were only 18 deaths in the RC
group and 4 deaths among RNP patients. The
mean age at death was similar for decedents in
the RC and RNP groups, and the difference in
crude death rates for the two groups was not
statistically significant. On the two occasions
RNP
group
pb
~~
RNP
groupa
no.
By cause at death
8
2
Cancer
4
1
Myocardial infarctionb
Other cardiovascular dis4
ease
2
1
Other
By age at death
10 to 29 years
2
3
1
30 to 49 years
7
2
50 to 69 years
6
1
70 years and over
59.3 years 57.0 years
Mean age at death
4
18
Total deaths
2.7
Death rate per thousand 6.0
a RC = patients receiving conventional care; RNP
tients receiving care from nurse practitioners.
Includes sudden death.
pa-
88
86
0.10
90
87
90
86
0.02
0.05
a RC = patients receiving conventional care; RNP = patients receiving care from nurse practitioners.
Indicates probability that we have failed to detect a real
deterioration of physical function among RNP patients 2
5%.
Sackett et al.
927
>
z
W
E
U
.... ..
..
2ot.: :
I
.....
LL
I-
z
W
0
LT
.....
n...
?NO
.... ...*.....
O
0.0
0.1
.....
W
I
0.2
0.3 0.4 0.5 0.6 0.7 0.8
EMOTIONAL FUNCTION INDEX VALUE
I-.
0.9
1.0
tioner nor the family physician have any clinically significant impact on health outcomes and
that this trial has merely compared equally ineffective, neutral alternatives for the delivery of
primary care. We have deliberately excluded a
no treatment control group for two reasons.
First, we concluded with our collaborators that
it would be unethical to withhold clinical services from a control group of patients in this
investigation, just as it has been judged unethical to withhold treatment from control
groups in randomized clinical trials of surgical
and chemotherapeutic approaches to cancer
(8). Our trial is analogous to the trial in which
therapy with a new pharmacologic agent is compared with current standard therapy. Second,
primary care practices of this magnitude, studied over this duration of time, generate a volume of clinical conditions (both statistically and
clinically significant in number) whose outcomes are profoundly affected by the skill of
detection and the appropriateness of management. This is substantiated, for example, by the
numbers of patients identified as requiring the
diagnosis and treatment of occlusive cardiovascular and infectious disorders, both in this and
in other investigations of primary care (1,9,10).
The second potential pitfall, volunteer bias,
was avoided by incorporating random allocation into the experimental design, and the comparability of the RC and RNP groups at the
start of the trial, as shown in Tables 1 and 2,
attests to the success of this procedure. Further-
928
....
70 60 -
....
>
0
.:
b .
0 .
3 40-
z
I-
z
a
RNP
30-
10
0.0
0.1
0.2
0.4
0.5 0.6 0.7 0.8
SOCIAL FUNCTION INDEX VALUE
0.3
0.9
1.0
RNP
groupb
no.
_ _ _ _ _ _ _ _ ~
39/71 55 18/41 44
58/67 87 27/34 79
55/73 75 39/51 76
32/64 50 19/42 45
42/51 82 23/30 77
51/69 74 28/40 70
Sackett et al.
929
~~
SUMMARY
The relationship between age-specific mortality rates and some indices of health
facilities and some environmental and dietary factors has been studied in 18 developed countries. The indices of health care are not negatively associated with mortality, and there is a marked positive association between the prevalence of doctors
and mortality in the younger age groups. No explanation of this doctor anomaly has
so far been found. Gross national product per head is the principal variable which
shows a consistently strong negative association with mortality.
Health services in developed countries, both
state and private, are based upon many common assumptions about what constitutes adequate health care. Doctors and their paramedical colleagues in different developed
countries receive similar educations to more or
less the same standards, and their approaches
to clinical and preventive medicine are unlikely
to differ in fundamental principles. However,
these countries show marked differences in
their mortality rates and in health costs per
head. In this paper, we seek to discover some
factors to explain these differences in mortality.
MATERIALS AND METHODS
930
that is, we included variables such as prevalence of doctors, and availability of hospital beds, which many people consider to be
self-evidently related to mortality.
2. A factor had to show a large product-moment correlation with at least one mortality rate, or it had to show a consistent
pattern of association with several mortality rates.
Cochrune et ul.
93 1
Australia
Austria
Belgium
Canada
Denmark
England and Wales
Finland
France
German Federal
Republic
10.2
6.7
39.5
3.9
12.6
10.2
4.7
Median
13.7
35.4
52.3
23.6
27.3
106.0
5.2
Maximum
18.5
56.0
97.7
68.8
50.8
399.7
7.1
2440
3810
3.7
2805
83.9
124.3
75.8
7.2
3195
90.5
148.3
120.0
17.5
3410
108.2
173.8
138.5
1.6
1949
10.0
77.2
4236
16.3
324.2
6652
49.4
40.5
80.7
94.8
630
932
Mortality rates
Minimum
Median
Maximum
8.5
16.5
11.0
5.3
3.2
6.8
8.0
16.9
43.6
107.9
21.5
22.9
18.2
8.5
4.1
8.8
10.9
22.9
56.2
150.0
54.5
31.7
29.6
10.2
4.7
13.0
15.9
32.2
72.8
183.2
4.
5.
6.
7.
Table 4. Correlation coefficients between the death rates and the input variab1es.a
Mortality rates
Age groups (years)
Maternal Perinatal
Doctors
Nurses
Beds
Pediatricians
Obstetricians
Midwives
% GNPon health
Cigarettes
Alcohol
Calories
Protein
Fat
Sugar
Population density
GNP per caput
Education index
Intervention index
a
0.45
-0.39
0.04
0.40
0.04
-0.10
-0.12
0.17
0.68
0.41
0.43
0.10
-0.61
0.17
-0.29
-0.13
-0.15
0.60
-0.53
-0.32
0.47
0.18
-0.15
0.01
0.22
0.52
0.59
0.37
0.29
-0.57
0.24
-0.48
-0.22
0.15
Infant
0.67
-0.50
-0.10
0.51
0.18
-0.14
-0.10
0.22
0.61
0.58
0.33
0.23
-0.56
0.21
-0.46
-0.20
-0.02
1-4
5-14
15-24
25-34
35-44
45-54
55-64
0.37
-0.28
0.07
0.23
-0.17
-0.29
-0.23
0.11
0.33
0.58
0.44
0.32
-0.40
0.07
-0.41
0.28
-0.13
0.42
0.37
0.18
0.31
0.29
-0.33
0.27
0.31
0.32
0.41
0.20
0.46
-0.31
-0.03
0.18
-0.43
0.12
0.32
0.12
0.37
0.35
0.48
-0.57
0.39
0.36
0.26
0.31
0.47
0.43
-0.17
-0.30
0.25
-0.79
0.44
0.23
0.06
0.06
0.37
0.54
-0.28
0.30
0.35
0.27
0.30
0.50
0.37
-0.20
-0.35
0.17
-0.61
0.48
0.04
0.19
-0.02
0.15
0.36
0.00
0.00
0.32
0.09
0.31
0.50
0.21
-0.05
-0.45
-0.13
-0.47
0.30
-0.27
0.27
-0.14
-0.11
0.09
0.26
0.23
0.46
-0.18
0.38
0.49
0.10
0.26
-0.30
-0.36
-0.27
0.26
-0.20
0.11
-0.14
-0.12
0.04
0.28
0.36
0.49
-0.14
0.52
0.43
0.16
0.21
-0.10
-0.53
-0.21
0.07
~~
Cochrane et al.
933
934
Table 5. Regression analysis of mortality rates on the seven variables with greatest explanatory power.
Input variables
Mortality rate
Maternal
Perinatal
Infant
Age groups (years)
1- 4
5-14
15-24
25-34
35-44
45-54
55-64
Population Intervention
Sugar
density
index
consumption
Doctors GNP Cigarettes Alcohol
1
*
172
3
1
0
-4
-3
-3
-1
- 15
- 11a
- 16a
25
-8 a
1
0
1
-5
-7
- 9=
1
5
2
5
8a
1o a
4
7
7
% Total sums of
squares explained
by 7 variables
18
0
-3
0
-2
2
-2
0
- 29
-8 a
-4
72
90
97
1
-1
0
0
-1
-3
-3
1
-2
- 7.
-7
-6
-2
- 16
-1oa
-9
-4
-3
-5
-6
-8
-11
-8
-3
-3
55
42
79
65
57
55
62
5a
-9
-4
-1
The figures in the first seven columns are the percentage changes in the death rates following a one-standard-deviation
increase in the input variables, the other input variables remaining fixed.
a t value for inclusion of variable in regression exceeds 2. Note, however, that even when not formally significantthe values
given are best estimates.
pared have an identical set of independent (input) variables. The final column of Table 5
displays the percentage of the total (corrected)
sums of squares of a given death rate explained
by the set of seven variables.
The findings for infant mortality are particularly interesting. The seven variables explain 97
percent of the variance in infant mortality rates.
In fact GNP per head alone explains 21 percent
of the variance, while the prevalence of doctors
alone explains 45 percent of the variance. Doctors and GNP per head together explain 82
percent of the variance. Doctors and GNP per
head are not themselves highly correlated
(r = 0.2) in these developed countries.
Other points of interest in Table 5 are the
positive association between alcohol consumption and maternal and infant mortality; the positive association between cigarette consumption
and mortality, which is strongest with infant,
perinatal, and maternal mortality, and is also
strong in the two oldest age groups; and the
negative association between population density
and mortality in young adults. Gross national
product is negatively associated with mortality
in all age groups except 5 to 34 years, where it
becomes negligible. Sugar consumption is negatively associated with mortality in all age
groups. The intervention index becomes prominent in the age groups 15 to 24 and 25 to 34
years, although, as noted earlier, caution is required in interpreting it in the latter age group.
Caution is also advisable in interpreting the re-
Cochrane et al.
~
~~
935
Secondly, the correlation and regression coefficients are still positive, but much weaker, if doctors are replaced by obstetricians and pediatricians, both of whom are likely to influence
infant mortality more strongly than other doctors.
In general, however, we must admit defeat
and leave it to others to extricate doctors from
their unhappy position.
It is also difficult to explain the roles of population density and the intervention index (proportion of health service spending coming from
government funds), both of which are strongly
negatively associated with mortality in young
adults. An interpretation of the effect of the
intervention index is that the more nationalized
the health services, the more effective is delivery of health care for potentially lethal illness.
This may well be so, but in referring to our
analysis, great caution is necessary before drawing this inference. Why should the intervention
index be most strongly associated with deaths in
young adults? A tentative explanation might be
that deaths in young adults are primarily due to
accidents, particularly road accidents, and nonfatal outcome following an accident may be dependent upon an efficient accident service. Private medicine would not be interested in
funding an accident service which by its nature
must often provide treatment before questions
of fees are broached, and therefore this may be
highly dependent upon public financing. If
state financing of health services in developed
countries is truly effective, however, then the
intervention index should be strongly associated with other causes of death, particularly
perinatal and infant mortality; and this we have
not found. We originally thought that the population density effect could be explained by a
negative association with road accidents but this
is not so, and, in the light of other findings, we
can hardly argue that nearness to medical help
is important.
The results for gross national product per
caput accord with expectations. Its association
with mortality is strongest in the youngest and
oldest age groups, and its negative sign is consistent with the idea that increasing overall
wealth reduces mortality. In the intermediate
age groups, its negligible association with mortality is consistent with the view that older children and young adults, having survived thus
far, are little affected, with respect to mortality,
936
by those social and environmental factors which ers on health services. As to the overall value of
correlate with wealth. This view is, of course, the results, we consider them to be interesting
tenable only for societies in which overall wealth and provocative, and perhaps capable of generexceeds subsistence level, as is true of all our 18 ating worthwhile new hypotheses which may be
tested in appropriate studies.
developed countries.
I t is not surprising that cigarette consumpWe thank Robert Maxwell and many internation should be associated with mortality rates in tional organizations for help with this study.
Reprints from A.L. Cochrane, Rhoose Farm
the older age groups. The strong association
with infant and perinatal mortality is not easy to House, Rhoose, South Glamorgan.
explain, although it is perhaps now generally
recognized that smoking in pregnancy has a
deleterious effect on the fetus but we cannot References
claim t o have shown this.
( 1 ) Stewart, C.T. Allocation of resources to health.
Our finding that sugar consumption is not J Hum Resour 6(1):103-122, 1971.
positively associated with mortality is inconsis(2) Hinds, M. W. Letter. N Engl J Med 291:741,
tent with the belief that unrefined sugar is gen- 1974.
(3) Richardson, J. The Dependency Hypothesiserally harmful, and associated with coronary
That More Doctors Will Result in Lower Quality
heart disease in particular (6). We would not, Health. Research paper No. 113, School of Economhowever, wish to place too much weight upon ics and Financial Studies. Sydney, Macquarie Univerour own findings. In any case, severe doubts sity, 1976.
( 4 ) Illich, I. Medical Nemsk-the Exprofmution o j
about the harmful role of sugar have been
Health. London, Calder and Boyars, 1975.
raised by other workers in studies designed to
(5) West, R.R. and C.R. Lowe. Regional variations
test this issue (7, 8).
in need for and provision and use of child health
We believe that one overall conclusion may be services in England and Wales. Br Med J 293434446,
drawn from this study. I t is that health service 1976.
(6) Yudkin, J. Diet and coronary thrombosis. Lancet
factors are relatively unimportant in explaining
2:155-162, 1957.
the differences in mortality among our 18 de(7) Bennett, A.E., R. Doll, and R. W. Howell. Sugar
veloped countries. There is nothing new in this. consumption a n d cigarette smoking. Lancet
The case has been argued particularly well by 1:1011-1014, 1970.
(8) Medical Research Council. Working Party on
Fuchs (9).As a corollary to this, it could also be
the relationship between dietary sugar intake and arargued that there is probably a considerable terial disease. Lancet 2:1265-1271, 1970.
element of inefficiency in the way some devel(9) Fuchs, V.R. Who shatt live? New York, Basic
oped countries spend so much more than oth- Books, 1974.
APPENDIX
Sources of Data and Indices
Mortality data and World Health OrganizaAlcohol (liters per
population density tion. World Health Statistics
head per year)
data
Annual. Vol. I . Vital Statistics
and Causes of Death for 1970.
Geneva, WHO, 1973.
Cochrane et al.
% GNP spent on
health
Percentage of health
expenditure covered by public expenditure
(intervention
index)
937
Education index.
UNESCO, Paris (through
Percentage of the the kindness of Dr. S.
cohort continuing Fauchette).
education after
age 18 (both
sexes)
Dietary data
ABSTRACT
Fisher et al.
whether radical mastectomy and total mastectomy followed by radiation produced an equivalent outcome. Previous studies of results (lifetable estimates) at three years (5) and at five
years (6) have failed to demonstrate a significant
difference in outcome among the three treatments in patients with clinically negative nodes
and between the two treatments in patients with
clinically positive nodes. This report presents
the 10-year findings of our trial.
METHODS
939
Statistical Analysis
The end points considered for the overall
treatment comparisons were disease-free survival, distant-disease-free survival, and overall
survival. Times to those end points were calculated from the date of mastectomy. Recurrences
of tumor in the chest wall and operative scar
were classified as local treatment failures. Recurrences in the internal mammary, supraclavicular, and subclavicular nodes in all patients and in the ipsilateral axillary nodes of
patients treated by radical mastectomy or total
mastectomy and regional irradiation were con-
940
sidered to be regional treatment failures. Patients with clinically negative nodes who were
treated by total mastectomy and subsequently
had positive nodes requiring an axillary dissection were not deemed to have had a treatment
failure at that time unless the nodes could not
be completely removed. The decision not to
consider delayed nodal involvement a treatment
failure was stipulated in the protocol before the
start of the study. Such an event was not included in the determination of disease-free survival. Distant disease was evaluated in two ways:
when it occurred as a first treatment failure and
when it occurred as any distant treatment failure-i.e., the first or subsequent local or regional failure. Patients were considered to be
free of disease at a point in time if they were
alive and had no local, regional, or distant evidence of breast cancer and no other primary
tumor.
The major goal of this study was to determine
whether treatments used as alternatives to radical mastectomy increased the risk of tumor recurrence or death. Actuarial life-table estimates
and associated standard errors (8)were calculated for each treatment group and each end
point. Comparisons of the survival distributions
were made within each nodal group by means
of the summary chi-square (log-rank) test (9,
10). A two-sided P value G0.05 was considered
to be significant. Standard errors for specific
points selected from the life tables are provided
in the text.
RESULTS
Disease-Free Survival
There were no significant differences (P =
0.2, Figure 1A) in disease-free survival over the
entire period of follow-up among the groups of
patients with clinically negative nodes treated by
radical mastectomy, total mastectomy plus radiation, or total mastectomy alone. At 10 years
47 2 2.6, 48 -t 2.7, and 42 f2.6 percent of each
group, respectively, were alive and free of disease. When disease-free survival was examined
in terms of events occurring during the first
and second five-year periods of follow-up, it was
observed that any differences among groups
occurred within the first five years after surgery
(P = 0.08, Figure 1B). At the end of the fifth
year the disease-free survival was 60 % 2.5 percent for the radical-mastectomy group, 65 2.5
percent for the patients undergoing total mastectomy plus radiation, and 56-I 2.5 percent for
those treated by total mastectomy alone. An
additional 15 percent of patients in all three
groups had a treatment failure between the 5th
and 10th year. There were no differences in the
probability of failure among the three groups
during the second five years (P = 0.8, Figure
1C). In each group, approximately 75 percent
of patients who were free of disease at the end
of 5 years remained free of disease at the end of
the 10th year. The differences noted during the
initial five years were due to the higher incidence of local or regional disease occurring as
the first evidence of disease in the total-mastectomy group, not because of an increase in distant disease occurring as a first treatment failure (Figure 2). The group treated by total
mastectomy and radiation had a lower incidence of local and regional recurrence than did
the other two groups.
There was no significant difference in disease-free survival between the two groups with
clinically positive nodes, either overall (P = 0.2,
Figure 1A) or during the first (P = 0.2, Figure
1B)or the second (P = 0.9, Figure 1C) five-year
period. Only 4 5 & 2 . 9 percent of patients
treated by radical mastectomy and 40 f2.9 percent of those undergoing total mastectomy plus
irradiation were free of disease at the end of the
5th year, and by the 10th year only 29 2 2.7 and
25 f2.6 percent, respectively, remained free of
disease. Nearly two thirds of the patients who
were free of disease at the end of the fifth year
remained so during the next five years. There
was little difference between the two groups
with respect to the occurrence of local, regional,
or first distant disease (Figure 3).
The total incidence of events-i.e., first treatment failures, second cancers, and unrelated
deaths-did not differ significantly among the
three groups of patients with clinically negative
nodes or between the two groups with clinically
positive nodes (Table 1). The proportion of first
distant treatment failures, as well as the distribution of distant disease according to site, was
similar for the three negative-node groups and
for the two positive-node groups. There was
little difference in the sites of first distant disease between the negative-node and positivenode groups. Of 532 reported distant failures,
one half (51.5 percent) were in two organ systems, 29.3 percent in the skeletal system, and
Fisher et al.
941
Figure 1. Survival free of disease through 10 years (A), during the flrst 5 years (B), and during the
m n d 5 years for patients free of disease at the end of the 5th year (C). Patients were treated hy
radical mastectomy (solid cirele), total mastectomy plus radiation (X), or total mastectomy alone
(open circle). There were no significant differences among the three groups of patients with
clinically negative nodes (solid line) or between the two groups with clinically positive nodes
(broken line).
I-
0
K
W
40
cn
w
a:
W
L
L
W
cn
= 0.2
= 0.2
I
>
20
-AI
.L
4u
-BI
YEARS0
= 0.8
;\
= 0.2
362
Z--8 294
1
I
-CI
P = 0.8
10
218
95
228
92
206
86
130
40
117
40
Whereas no major differences in the occurrence of regional disease were observed among
the negative-node groups, there were differences between the two positive-node groups.
Patients treated with radical mastectomy had a
lower incidence of ipsilateral axillary-node recurrence ( I .O percent) than did patients undergoing total mastectomy and irradiation (1 1.9
percent). On the other hand, supraclavicularnode recurrences were higher in the radicalmastectomy group (5.8 vs. 0 percent).
Distant-Disease-Free Survival
942
Part
Figure 2. Local or regional and distant treatment failures as the first evidence of disease in
patients with clinically negative nodes who were treated by radical mastectomy (solid circle), total
mastectomy and radiation (X), or total mastectomy alone (open circle). There were no significant
differences in distant disease occurring aa a first treatment failure among the three groups,
whereas local and regional disease was best controlled in the group reaiving radiation.
20
LOCAL REGIONAL
-"
I
..
..,.
"
,.
3
4
5
6
7
YEARS POST MASTECTOMY
-X
Figure 3. Local or regional and distant treatment failures as the first evidence of disease in
patients with clinically positive nodes who were treated by radical mastectomy (solid circle) or total
mastectomy and radiation (X). There was no significant difference in distant or local and
regional disease between the two groups.
60
2o
0
20
DISTANT
L
-
LOCALREGIONAL
P = 0.3
4
5
6
7
YEARS POST MASTECTOMY
10
943
Fisher et al.
Table 1. Distribution of first treatment failure, second cancers, and death from unrelated causes,
according to treatment.*
Patients with
Clinically Negative Nodes
Event
Total
R M (N
Patients with
Clinically Positive Nodes
294)
197 (54.4)
183 (52.0)
214 (58.6)
205 (70.2)
217 (73.8)
16 (4.4)
11 (3.0)
3 (0.8)
2 (0.6)
4 (1.1)
3 (0.9)
1 (0.3)
0 (0.0)
28 (7.7)
19 (5.2)
6 (1.6)
3 (0.8)
21 (7.2)
17 (5.8)
1 (0.3)
3 (1.0)
5 (1.7)
2 (0.7)
1 (0.3)
2 (0.7)
9 (2.5)
5 (1.4)
4 (1.1)
0 (0.0)
0 (0.0)
0 (0.0)
12 (3.4)
11 (3.1)
1 (0.3)
0 (0.0)
0 (0.0)
0 (0.0)
15 (4.1)
4 (1.1)
11 (3.0)
0 (0.0)
0 (0.0)
0 (0.0)
22 (7.5)
3 (1.0)
17 (5.8)
0 (0.0)
0 (0.0)
2 (0.7)
35 (11.9)
35 (11.9)
0 (0.0)
0 (0.0)
0 (0.0)
0 (0.0)
96 (26.5)
11 (3.0)
2 (0.6)
30 (8.3)
22 (6.1)
8 (2.2)
4 (1.1)
3 (0.8)
2 (0.6)
0 (0.0)
0 (0.0)
14 (3.9)
109 (31.0)
18 (5.1)
2 (0.6)
28 (8.0)
24 (6.8)
8 (2.3)
8 (2.3)
1 (0.3)
0 (0.0)
0 (0.0)
0 (0.0)
20 (5.7)
107 (29.3)
19 (5.2)
l(0.3)
31 (8.5)
22 (6.0)
7 (1.9)
7 (1.9)
1 (0.3)
1 (0.3)
0 (0.0)
0 (0.0)
18 (4.9)
104 (35.6)
13 (4.4)
2 (0.7)
33 (11.3)
23 (7.9)
5 (1.7)
11 (3.8)
1 (0.3)
0 (0.0)
0 (0.0)
0 (0.0)
16 (5.5)
116 (39.4)
12 (4.1)
2 (0.7)
34 (11.6)
27 (9.2)
6 (2.0)
11 (3.7)
1 (0.3)
3 (1.0)
1 (0.3)
0 (0.0)
19 (6.5)
Combinations
Local and regional
Local and distant
Regional and distant
Widespread
Unknown
12 (3.3)
3 (0.8)
5 (1.4)
0 (0.0)
3 (0.8)
1 (0.3)
9 (2.6)
0 (0.0)
2 (0.6)
1 (0.3)
2 (0.6)
4 (1.1)
13 (3.4)
3 (0.8)
3 (0.8)
2 (0.5)
3 (0.8)
2 (0.5)
22 (7.5)
1 (0.3)
9 (3.1)
7 (2.4)
2 (0.7)
3 (1.0)
17 (5.7)
1 (0.3)
6 (2.0)
6 (2.0)
2 (0.7)
2 (0.7)
16 (4.4)
0 (0.0)
16 (4.4)
15 (4.3)
2 (0.6)
13 (3.7)
11 (3.0)
1 (0.3)
10 (2.7)
7 (2.4)
0 (0.0)
7 (2.4)
11 (3.7)
2 (0.7)
9 (3.0)
Local
Chest wall
Scar
Both
Regional
Axilla
Supraclavicular
Internal mammary
Subclavicular
>1
Distant
Opposite breastb
Integumentary
Skeletal
Respiratory
Hemic and lymphatic
Digestive
Genitourinary
Nervous
Endocrine
Cardiac and sense organs
>I
48 (13.2)
34 (9.6)
40 (11.0)
29 (9.9)
33 (11.2)
Alive, event-free
165 (45.6)
169 (48.0)
151 (41.4)
87 (29.8)
77 (26.2)
Values are numbers of treatment failures, with percentages of patients shown in parentheses. R M denotes radical
mastectomy, T M R total mastectomy plus radiation, and TM total mastectomy.
Includes second primary tumors.
944
60
40
P = 0.4
80
60
P = 0.4
40
YEARS0
RISK
--
10
W 362
225
97
352
234
95
365
221
93
292
140
43
H 294
134
47
X-4
Overall Survival
Fisher et al.
945
Figure 5. Survival through 10 years (A), during the first 5 years (B), and during the second 5 years
for patients alive at the end of the 5th year (C). Patients were treated by radical mastectomy (solid
circle), total mastectomy and radiation (X), or tow mastectomy alone (opencircle). There were no
significant differences among the tbree groups of patients with clinically negative nodes (solid line)
or between the two groups with positive nodes (brokenline).
I-
w
w
100
80
60
P
>
5
cc
-A-
40
(I)
P
I
= 0.7
I
---I
>
k
4
m
$
n
100
80
60
YEARS
40 I
362
352
o--O 365
*-* 292
H
#AT
x-x
x-4
= 0.3
294
= 0.4
-CI
270
265
270
180
10
119
106
106
57
60
169
tions occurred within 24 months after total mastectomy; only three (4.6 percent) occurred during the second five years of follow-up. The
hazard rate over time indicates that the highest
risk of axillary-node involvement was concentrated in the first 24 months after operation.
Over the rest of the 10-year period, a continuous but low risk remained evident.
We first reported results from this prospective randomized clinical trial comparing radical
mastectomy with alternative treatments for primary breast cancer in 1977, when the average
follow-up time was 36 months (26 to 62 months)
(5). The findings at 10 years (average time in
the study, 126 months) confirm and extend the
earlier results. They continue to indicate no
significant difference in disease-free, distantdisease-free, or overall survival among patients
without clinical evidence of axillary-node involvement who were treated by three distinctly
different treatment regimens: radical mastec-
DISCUSSION
946
Figare 6. Relation of treatment to survival according to tumor location. Patients were treated by radical mastectomy (mud circle), total mastectomy and radiation (X), or total mastectomy alone (open circle). The outcome for
patients with clinically negative or positive nodes and lateral tumors or medial and central tumors was not affected
by the treatment.
IZ
60-
4030
w
J-
5>-
p=O6
I
P = 0.6
I
100
RM
a TMR
+
a
202
X 195
122
112
40
41
85
95
55
53
17
18
at
tomy, total (simple)mastectomy with local and nodes as well. In patients treated by total masregional radiation, or total mastectomy with tectomy alone, positive nodes might have been
subsequent removal of axillary nodes if they expected to serve as a source of further tumor
became clinically positive. Among patients with dissemination, resulting in an increase in disclinical evidence of nodal involvement treated tant treatment failure and a higher mortality if
either by radical mastectomy or total mastec- temporal considerations are seminal in the mettomy plus local and regional radiation, there astatic process. Since such events did not occur,
continues to be no significant difference be- there is support for our concept that regional
tween the two treatment groups.
lymph nodes are indicators rather than inIt has previously been reported that 40 per- stigators of distant disease. Moreover, just as
cent of the patients in this trial who werejudged unremoved nodes were not deleterious, our
to have clinically negative nodes and who un- findings indicate that their removal did not adderwent a radical mastectomy had histologic versely affect prognosis, refuting the view that it
evidence of positive nodes (11). Consequently, may be beneficial not to remove regional nodes
since this was a randomized trial, it can be justi- in patients with clinically perceptible cancers
fiably assumed that approximately 40 percent (12).
of the patients with negative nodes in the other
Particularly interesting was the observation
two treatment groups had tumor-positive lymph that even though positive nodes were not re-
Fisher et al.
947
ease had the same estimated probability of remaining disease-free (free of disease overall or
free of distant disease) and of surviving during
the second five years regardless of the treatment. The findings were the same for the positive-node groups. The rate of treatment failure,
overall or distant, was lower in the second fiveyear period than in the first regardless of nodal
status or treatment. There was no difference,
however, in the death rate between the first and
second five-year periods. Patients with positive
nodes who were free of distant disease at the
end of five years had about the same probability
of remaining free of distant disease during the
next five years as did the negative-node group.
Thus, the 5-year results were highly predictive
of the outcome at 10 years, and conclusions
based on the early results remain valid. We do
not mean to imply that patients surviving for
five years are not at risk for subsequent failure.
The risk of failure in the next five years is not,
however, related to the initial local and regional
treatment. It will be important to discern
whether the same conclusion remains valid during the third five-year period.
The current findings confirm the validity of
our previous results and our conclusion that the
location of tumor in the breast does not influence the prognosis and does not provide justification for varying local and regional therapy
(13). Our results fail to support the view that
radiation of internal mammary nodes in patients with inner-quadrant lesions improves survival (14,15).N o survival benefit was noted in
our patients with clinically negative or positive
nodes who were treated by total mastectomy
and regional radiation that included internal
mammary nodes.
Almost all available information on patterns
of metastases in patients with breast cancer has
been derived from autopsy studies (16-20). Relatively few data from breast-cancer clinical trials
document the incidence and organ site of the
first treatment failure in patients who received
no systemic therapy after surgery (21-23). Our
current findings indicate that variations of primary treatment within a group of patients with
the same clinical nodal status failed to alter the
distribution of first treatment failures that were
distant in location. Moreover, the distribution
overall and in specific organ sites was remarkably uniform regardless of the therapy and the
nodal status. As expected, radiation therapy al-
948
References
(1) Fisher, B. The surgical dilemma in the primary therapy of invasive breast cancer: A critical appraisal. C u w Probl Surg 1:53, 1970.
(2) McWhirter, R. The value of simple mastectomy and radiotherapy in the treatment of cancer of
the breast. Br J Radio1 21:599-610, 1948.
(3) Handley, R.S. The technic and results of conservative radical mastectomy (Pateys operation). Prog
Clan Cancer 1:462-470, 1965.
(4) Fisher, B. Laboratory and clinical research in
breast cancer-a personal adventure. Cancer Res
40~3863-3874.
(5) Fisher, B., E. Montague, C. Redmond, et al.
Comparison of radical mastectomy with alternative
treatments for primary breast cancer. A first report of
Fisher et al.
cancer: Results of the NSAP clinical trial. Ann Surg
172171 1-732, 1970.
(23) Valagussa, P., G. Bonadonna, and U.Veronesi.
Patterns of relapse and survival following radical mastectomy: Analysis of 7 16 consecutive patients. Cancer
41:1170-1178, 1978.
(24) Fisher, B. Breast-cancer management: Alter-
949
APPENDIX. List of participants in protocol no. 4 of the national surgical adjuvant breast project.
Responsible
pathologist
Institution
Responsible surgeon
Responsible
radiation oncologist
Herbert Volk
C. William Kaiser
Claude H. Organ
Bernard Gardner
William Donegan
Richard M. Hirata
James McManus
C.W. Konvolinka
John R. Benfield
Benjamin F. Rush, Jr.
Claude R. Hitchcock
Richard G. Margolese
Isidore Cohn, Jr.
Donald Cooper
Walter Lawrence, Jr.
John F. Weiksnar
Richard H. Evans
Richard G . Rosen
Gerson J. Lesnick
Thomas Nealon, Jr.
Willis Maier
Kent Westbrook
Marshall J. Orloff
Tapas K. Das Gupta
Richard L. Lawton
T. James Guzik
Francis E. Rosato
Bernard Fisher
W. Bradford Patterson
Anatolio B. Cruz, Jr.
Edward B. Rowe
Roger S. Foster
Harvey Butcher
Alexander J. Walt
N.A. Ghossein
John Molnar
Merrill Feldman
Leonard Berman
J.R. Zastera
Wade Bardawil
Joseph Bohorquez
Yale Rosen
James Thomson
Carlos M. Perez-Mesa
Terry Powell
William Starke
George Schwarz
William Finkelstein
David Beiler
C. James Favino
Frank M. Hirose
Richard Small
Michael Lyons
John Mallams
John I. Coe
Manoucher Azad
Claude LaChance
J.J. Hazel
Ronald A. Welsh
Joseph V. Schlosser
Janet Parker
Gerald Justh
Saul Kay
E. Richard King
Eugene Schwartz
William Henkin
Lionel Cohen
Miriam L. Christ
Flora Mincer
Norwin Becker
John Boland
Mamoru Kaneko
George Schwarz
William E. Delaney 111
Marie LoPonte
Paul Putong
Eleanor Deed
H.K. Leathers
Sidney Saltzstein
Carl Von Essen
Jose Manaligod
Edwin J. Liebner
Howard B. Latourette Frederick W. Stamler
Quintus Crews
Frances Wachter
Lawrence Davis
Horatio T. Enterline
John Parsons
Robert Totten
Robert Cooper, Jr.
Eileen Paterson
Peter Zanca
George Bannayan
Richard Marshall
Marvin H. Olson
G. Stephen Brown
David Duffel1
Carlos A. Perez
Walter Bauer
Alan Scheer
Barbara Rosenberg
ABSTRACT
This study was a randomized trial to determine whether anastomosis of the superficial
temporal artery to the middle cerebral artery
decreased the rate of stroke and stroke-related
death among patients with symptomatic disease
of the internal carotid and middle cerebral arteries.
METHODS
950
95 1
RESULTS
Entry of Patients
952
cluded 115 eligible patients who refused to enter the trial and 52 patients whose clinicians
insisted that they undergo bypass surgery; for
11 other patients, no reason was given. The
mean age of the patients who were eligible but
not entered was 58 years-similar to that of the
trial patients. Of the 1377 eligible patients, 714
(52 percent) were randomly assigned to medical
and 663 (48 percent) to surgical therapy. Randomization created balanced treatment groups
with respect to important prognostic characteristics and underlying vascular lesions (Table
1). Although 74 percent had some abnormalities on neurologic examination at entry, 93
percent had either minimal or no functional
impairment.
No. of patients
Age (mean yr)
Sex (%)
Male
Female
Randomization diagnosis (%)
Transient ischemic attack
Minor stroke
Other medical problems (%)
Hypertension
Diabetes
Angina pectoris
Prior myocardial infarction
Intermittent claudication
Medications at entry (%)
Platelet antiaggregants
Antihypertensive agents
Blood pressure at entry
(mean mm Hg)
Systolic
Diastolic
Most distal angiographic lesion (%)a
Middle cerebral LrteEy
Stenosis
Occlusion
Internal carotid artery
Stenosis (above C-2)
Occlusion, no symptomsb
Occlusion, recurrent symptoms
Medical
Surgical
7 14
56
663
56
82
18
81
19
34
66
33
67
48
18
8
9
11
52
17
10
11
13
57
31
54
33
144
85
145
85
13.0
11.1
14.4
12.1
16.7
38.7
20.6
15.4
37.0
2 1.1
A final assessment was obtained for all patients between December 1984 and May 1985.
The final adjudication of all end points was
completed by June 6, 1985.
Follow-up Performance
No patient was lost to follow-up, none were
withdrawn, and the average duration of followup among surviving patients was 55.8 months
(range, 28 to 90). A total of 21 428 individual
follow-up assessments were completed from a
potential total of 24 160. The completeness of
follow-up was similar in the smaller (<25 patients) and the larger centers: 86 and 92 percent, respectively. Less than 10 patients each
were randomized from 19 centers. Their
pooled contribution was only 101 patients
(7 percent of the total).
Nine medical patients (1.3 percent) crossed
over and underwent EC-IC bypass on the same
side as the lesion for which they had been randomized. Another six medical patients (0.8 percent) underwent EC-IC bypass on the opposite
side. Of the 663 patients randomly assigned to
the surgical group, 652 (98 percent) underwent
surgery, which was performed an average of
nine days after randomization.
Repeat angiograms were obtained in 92 percent of the patients at a median time of 32 days
after surgery, and on the final review 96 percent of these studies revealed patent anastomoses. The graft patency rates were 95 percent in
the smaller and 96 percent in the larger centers,
and they were high in all three regions: 94 percent in North America, 97 percent in Europe,
and 98 percent in Asia. Fourteen percent of the
original stenotic lesions of the middle cerebral
artery had progressed to occlusion as seen on
the postoperative angiograms. Three patients
(0.5 percent of those in whom there were postoperative studies) had strokes within a day after
postoperative angiography that were considered complications of the radiologic procedure.
Two of the three recovered without a serious
permanent deficit. Removing these three patients from the analysis did not change the results.
Medical regimens likely to affect the risk of
stroke were equally applied to both groups: Aspirin was used for an average of 75 percent of
the follow-up period among medical patients, as
compared with 74 percent among surgical pa-
domization. In the medical cohort of 714 patients, a total of 24 patients (3.4 percent) had
some kind of cerebral or retinal ischemic event,
9 (1.3 percent) had major strokes, and 1 died of
myocardial infarction. Comparing the rate of
major perioperative strokes in the surgical
group (4.5 percent) with the spontaneousstroke rate in the medical group (1.3 percent)
showed an excess of 3.2 percent in fatal or nonfatal strokes in the surgical group.
Events
The primary study question was: Does anastomosis of the superficial temporal artery to
the middle cerebral artery, despite perioperative stroke and death, reduce the rate of
subsequent events of stroke and stroke-related
death in the patients studied? The answer is
no. Fatal and nonfatal strokes occurred both
more frequently and earlier in patients randomly assigned to surgery (Figure 1). We were
able to test and reject, with a statistical power
greater than 99 percent, our original hypothesis
of a surgical benefit consisting of a one-third
reduction in fatal and nonfatal stroke. In fact,
the Mantel-Haenszel chi-square analysis gen-
Figure 1. Results of the primary analysis (all strokes, both fatal and nonfatal),
showing the failure of bypass between the superficial temporal artery and the middle cerebral artery to reduce stroke In the surgical (663 patients) as compared with
the medical cohort (714 patients) after an average follow-up of 55.8 months. The
analysis uses Kaplan-Meier cumulative-failure curves.
(I)
0.4-
Lu
MEDICAL
SURGICAL
__
______
(N =
714)
(N = 663)
0.3-
12
18
24
30
953
36
STUDY MONTH
42
48
54
60
954
955
Figure 2. Results of secondary analyses, showing the failure of bypass in the total surgical cohort, as compared with
the total medical cohort, to reduce the Occurrence of major stroke and stroke death (A), all strokes and all deaths
(B), all ischemic strokes ipsilateral to the side of symptoms for which randomizationwas carried out (C), and major
ischemic strokes ipsilateral to the side of such symptoms (D).
A
04
oc3
W =
W
0.3
MEDICALSURGICAL,,,
(N
= 714)
B
MEDICAL-
(N = 714)
(N = 663)
12 18 24
STUDY MONTH
30 36 42 48 54 60
STUDY MONTH
D
0
0.4
MEDICAL
0.3
SURGICAL
(N
0.4
MEDICAL-
(N = 714)
714)
--- (N = 663)
0.2
L2izzLL
_____.-..
*--
0.1
STUDY MONTH
Treatment group
Impairment
Medical
Surgical
No. of patients
714
663
percent
None
Major
Death
Cerebrovascular
Myocardial infarction
Sudden death
Other cardiovascular
Other
4.8
4.5
2.7
2.1
5.6
18 24
30 36 42
48
54
60
STUDY MONTH
Minor
12
56
19
5
20
4.7
3.8
1.5
3.2
3.8
57
19
7
17
956
No. of patients
Medical
Surgical
714
663
percent
61
23
5
11
None
Minor
Major
Death
61
23
6
10
Medical
No. of patients
Better
Same
Worse
Dead
42
14
6
5
17
3.5
Surgical
37
14
6
5
12
3.8
Patients6
No.
Observed
Surgical Group
Expected
No.
Observed
Expected
MantelHaenszel
chi-square
number .f patients
All patients
714
205
218.3
663
205
191.7
1.72
438
133
148.0
418
148
133.0
3.23
276
147
72
51
69.9
61.7
245
140
57
64
59.1
53.3
0.13
4.04
72
59
26
14
27.1
20.5
77
50
29
22
27.9
15.5
0.10
4.74
43
79
87
17
18
17.4
16.9
31
80
14
16
13.6
17.1
0.02
0.15
27
31.5
109
41
36.5
1.32
350
364
98
107
112.1
105.9
337
326
113
92
98.9
93.1
3.81
0.02
Geographical region
North America
Europe
Asia
352
247
115
115
60
30
126.8
64.9
26.8
327
230
106
120
63
22
108.2
58.1
25.2
2.37
0.77
0.78
a Values listed under the heading Observed indicate the observed number of patients in each treatment group who had a
stroke. Those listed under Expected indicate the number of patients in each treatment group who would be expected to have
a stroke if surgery had no effect, taking into account differences in sample size and duration of follow-up.
ICA denotes internal carotid artery, MCA middle cerebral artery, and TIA transient ischemic attack.
N o symptoms were experienced between angiographic demonstration of the occlusion and randomization.
Symptoms were experienced between angiographic demonstration of the occlusion and randomization.
Severe stenosis is stenosis of 70 percent or more of the luminal diameter.
DISCUSSION
Figure 3. Analysis of the effect of surgerJi according to center size, for all fatal and
nonfatal strokes, showing that the average difference between the observed and
expected number of strokes was similar in small and large centers. Each participating center Is represented by a dot on the scattergram.
(I)
I-
z
W
>
SURGERY WORSE
n
W
Yn
2
IX
W
>
n
W
(I)
m
0
SURGERY BETTER
-8
-lo
957
16
25
36
CENTER SIZE
49
64
81
958
vent subsequent important ischemic events provided the rationale for this trial. I t was
recognized that the cerebral and retinal ischemic events experienced by the patients could be
either of hemodynamic origin, resulting from
poor circulatory perfusion, or caused by arteryto-artery emboli, and that in many of them it
would not be possible to tell which mechanism
was responsible (4,5). However, advocates of the
operation postulated that it might benefit patients with ischemia due to either of these
causes by improving cerebral perfusion, providing additional collateral circulation, or in patients with ischemia of embolic origin, providing a sufficient increase in perfusion to
promote early passage and disintegration of the
embolus.
The results indicate with strong statistical
power that fatal and nonfatal strokes were not
prevented by anastomosis of the superficial
temporal to the middle cerebral artery. A surgical benefit as small as 3 percent (clinically
insignificant) can be rejected at the 0.05 confidence level. This negative result held for all
patients and for individual subgroups, whether
the patients had transient ischemic attacks or
minor strokes, whether their lesions were occlusive or stenotic, and whether they involved
the internal carotid artery, the middle cerebral
artery, or both.
We believe that our conclusion that EC-IC
surgery fails to prevent stroke or even reduce
transient ischemic attacks is both statistically
powerful and clinically credible because our
study involved (1) a large number of patients
with a long period of follow-up, (2) a uniformity
of disease process in the population studied, (3)
a randomization method that produced balanced treatment groups, (4) a complete and
accurate record of all entry and event data, (5)
external blinded adjudication of eligibility
and event data, (6) a uniformity of ancillary
treatment for the total patient population, and
(7) the achievement of effective anastomoses
with acceptably low morbidity and mortality.
As in the previous studies of internal mammary ligation (6) and the gastric freeze (7), this
negative result from a randomized trial is at
variance with previously published reports on
case series. Direct comparisons are impossible,
because no other published series study has incorporated a concomitantly randomized control
group. Studies have been reported that have
959
960
961
~~~~
S.J. Peerless; principal neuroradiological investigator, A.J. Fox; senior staff, B. Valberg, J. Peacock. Methods Center (McMaster University),
Hamilton, Ont.: principal epidemiologic investigator, D.L. Sackett; chief epidemiologist, R.B.
Haynes; chief statistician, D.W. Taylor; senior
staff, C.Collis, J. Mukherjee, P. Flanagan. Steering Committee: University of Western Ontario,
H.J.M. Barnett (chairman), S.J. Peerless, A.J.
Fox, B. Valberg, V.C. Hachinski; McMaster
University, D.L. Sackett, D.W. Taylor, R.B.
Haynes, J. Mukherjee. Monitoring Committee: M.
Goldstein (chairman 1977- 1982; National Institute of Neurological and communicative Disorders and Stroke)*, M.D. Walker (chairman
1982-present, National Institute of Neurological and Communicative Disorders and
Stroke)*, J.B. Benedict (National Institute of
Neurological and Communicative Disorders
and Stroke)*, W. Weiss (National Institute of
Neurological and Communicative Disorders
and Stroke), J.R. Marler (National Institute of
Neurological and Communicative Disorders
and Stroke), J.P. Whisnant (Mayo Clinic). H.G.
Schwartz (Washington University), A. Heyman
(Duke University-Medical Center). W.H.
Feindel (Montreal Neurological Institute).
Participating Centers (in order of number of
eligible patients entered): University of Western
Ontario, London, Ont.: H.J.M. Barnett*, C.W.
McCormick, V.C. Hachinski, S.J. Peerless*, G.G.
Ferguson; neuroradiologists, J. Allcock, A.J.
Fox*; additional contributors, K. Meguro, R.
Cote, D. Moulin, P.C. Gates, S. Lauzier. University of Toronto, Toronto, Ont.: R. Wilson, G.
Sawa, H. Schultz, M.C. Chiu. University of Tenn e s s e e , M e m p h i s , T e n n . : A. H e c k , J.
Robertson*, B. Gerald; National Institute of
Neurosurgery, Budapest, Hungary: L. Ronai,
E. Pasztor* (sponsoring neurosurgeon), J. Vajda, M. Horvath, I. Nyary, G. Deak. University
of Essen, Essen, Federal Republic of Germany;
A. Buch, H.M. Mehdorn, C. Nahser. Universita
di Firenze, Florence, Italy; L. Amaducci*, D.
Inzitari, S. Briani (deceased), R. Gagliardi, A.
Nori. Neurochirurgia und Neurologica University Klinik Giessen, Federal Republic of Germany: 0. Busse, E. Grote, C. Hornig, R. Schonmayr. Kyoto University Medical School, Kyoto,
Japan: M. Kameyama, I. Akiguchi, H. Shio, H.
H a n d a ( s p o n s o r i n g n e u r o s u r g e o n ) , Y.
Yonekawa*. Hopital Pellegrin, Bordeaux,
France: J.M. Orgogozo, J.J. Pere, J.P. Castel,J.M.
Caille. New York University, New York: W.K.
Has*, E.S. Flamm". University of Pecs, Pecs,
Hungary: M. Bodosi, G. Gacs, F.T. Merei. Westeinde Ziekenhuis Den Haag, Holland: J.Th.T.
Tans, C.A.F. Tulleken*, P. Hoogland. Univer-
962
References
( I ) Yasargil, M.G., ed. Mzcrosurgery Applied to
Neuro-Surgery. Stuttgart, Georg Thieme, 1969, pages
105- 115.
(2) ECIIC Bypass Study Group. The International
Cooperative Study of Extracranialhtracranial Arterial Anastomosis (EC/IC Bypass Study): methodology and entry characteristics. Stroke 16:397-406,
1985.
( 3 ) Sackett, D.L. and M. Gent. Controversy in
counting and attributing events in clinical trials. N
Engl J Med 301:1410-1412, 1979.
( 4 ) Barnett, H.J.M. Pathogenesis of transient ischemic attacks. In Cerebrovascular Diseases (P. Scheinberg, ed.) New York, Raven Press, 1976, pages 1-21.
(5) Barnett, H.J.M. Progress toward stroke prevention. Neurology (NY) 30:1212-1225, 1980.
(6) Barsamian, E.M. The rise and fall of internal
mammary ligation in the treatment of angina pectoris
and the lessons learned. In Costs, Risks, and Benefzts of
Surgery U.P. Bunker, B.A. Barnes, and F. Moseler,
eds.). New York, Oxford University Press, 1977,pages
212-220.
(7) Miao, L.L. Gastric freezing: an example of the
evaluation of medical therapy by randomized clinical
trials. In Costs, Risks, and Benefits of Surgery U.P.
Bunker, B.A. Barnes, and F. Moseler, eds.). New
York, Oxford University Press, 1977, pages 198-21 1 .
(8) Norrving, B., B. Nilsson, and J. Risberg. rCBF
in patients with carotid occlusion: Resting and hypercapnic flow related to collateral pattern. Stroke
13:155-162, 1982.
(9) Tsuda, Y., K. Kimura, Y. Iwata, et al. Improvement of cerebral blood flow and or C 0 2 reactivity
after superficial temporal artery-middle cerebral artery bypass in patients with transient ischemic attacks
and watershed-zone infarctions. Surg Neurol
22~595-604, 1984.
(10) Yonas, H., D. Cur, B.C. Good, et al. Stable
xenon CT blood flow mapping for evaluation of patients with extracranial-intracranialbypass surgery.J
Neurosurg 62:324-333, 1985.
(11) Yonekura, M., G. Austin, and W. Hayward.
Long-term evaluation of cerebral blood flow, transient ischemic attacks, and stroke after STA-MCA
anastomosis. Surg Neurol 18:123-130, 1982.
(12) Laurent, J.P., P.M. Lawner, and M. OConnor.
Reversal of intracerebral steal by STA-MCA anastomosis. J Neurosurg 57:629-632, 1982.
(I?) Powers, W.J., W.R.W. Martin, P. Herscovitch,
M.F. Raichle, and R.L. Grubb, Jr. Extracranial-intracranial bypass surgery: Hemodynamic and metabolic
effects. Neurology (NY) 34:1168-1174, 1984.
(14) Baron, J.C., M.G. Bousser, A. Rey, A. Guillard,
D. Comar, and P. Castaigne. Reversal of focal misery
perfusion syndrome by extra-intracranial arterial bypass in hemodynamic cerebral ischemia: A case study
with 150 positron emission tomography. Stroke
12:452-459, 1981.
(15) Grubb, R.L., Jr., R.A. Ratcheson, M.E. Raichle, A.B. Fliefoth, and M.H. Gado. Regional cerebral
blood flow and oxygen utilization in superficial tem-
963
964
perficial temporal to middle cerebral artery anastomosis: Clinical outcome in patients with ischemia
of infarction in internal carotid artery distribution.
Arch Neurol36: 1-4, 1979.
(46) Rhodes, R.S., R.F. Spetzler, and R.A. Roski.
Improved neurologic function after cerebrovascular
PART V
DISCUSSIONS
TERRIS:
BUCK:
TERRIS:
967
968
Yes, that is a problem with any test. You also have a lot of
venereal disease testing in groups that are not at risk.
BUCK:
TERRIS:
NAJERA:
I like the outline. When we talk about the role of epidemiology regarding high-risk groups, I think we should
go beyond the known high-risk groups and try to find new
ones. I think we should insist that high-risk groups be
defined by their mortality and morbidity. And we should
encourage general studies. The type of vital statistics or
population data that we now have do not always allow us to
study the population according to occupation o r social
class. We should insist on more precise demographic data.
At present it is hard to determine groups, since categories
Discussions
969
TERRIS:
970
NAJERA:
LLOPIS:
NAJERA:
BUCK:
Disczcssions
971
TERRIS:
BUCK:
TERRIS:
NAJERA:
BUCK:
NAJERA:
The reason in the United States is that the dominant ideology insists that there are no social classes in the United
States. Didnt you know that this is supposed to be the
country without social classes?
Well, you are giving a philosophical reason.
They dont want to study social classes.
What is the reason in Canada?
We dont want to study them, either. Its the same reason.
T h e countries without social classes!
BUCK:
NAJERA:
I think that you have raised a very, very, important pointthe definition of disease. Maybe its not so necessary for
acute diseases, but it is for chronic diseases. In order to
help the study of disease, we need to think a lot about
972
TERRIS:
BUCK:
Youve really put your finger on it. Cassel, for example, did
not confine himself to illness manifested in emotional disturbances. T h e studies he did of rapidly urbanized Ap-
Discussions
973
BUCK:
TERRIS:
BUCK:
TERRIS:
BUCK
Some believe you find it everywhere, but that its not diagnosed the same in all classes.
NAJERA:
974
BUCK:
Discussions
975
TERRIS:
In the United States and England we have a well-established tradition of epidemiological research, and PAHO
has sent people from many Latin American countries to get
some of the best training in the world at elite institutions in
these two countries. When these people returned home,
however, not very much happened in their countries in
terms of research. For some reason they got involved in
teaching, or whatever. Somehow we should also make it our
task to indicate that the job of an epidemiologist is to stop
talking a n d d o some work, research work. If epidemiologists dont do some decent research studies, theyre
not fulfilling their jobs. And we have to emphasize this,
because this seems to be a real problem in Latin America. I
think it probably is so in most of the developing world that
doesnt have a tradition of research. This is where it must
be developed.
BUCK:
976
TERRIS:
BUCK:
TERRIS:
BUCK:
NAJERA:
Discussions
977
The other thing that happens is that in a cardiology institute, for example, they will have one epidemiologist, very
well trained in the London School of Hygiene. Then, in a
neurology institution or in a peripheral vascular disease
institute you will also find one lone epidemiologist. In each
institute there will be an epidemiologist surrounded by 100
clinicians. But if you have one epidemiologist surrounded
by hundreds of clinicians and laboratory people, he is
dead, he wont do anything. What they should d o is keep an
epidemiologist in each institute, but also let them be part of
a collective, of a center where there are two, three, or more
individuals not affiliated with institutes, people who are the
theoreticians in the group. That way, they can meet regularly with the epidemiologists from all of the institutes, and
they can talk to each other. What people really need is to
talk to each other, to discuss problems. There should be
mechanisms developed for epidemiologists from different
centers to get together and discuss what theyre doing.
NAJERA:
TERRIS:
LLOPIS:
978
NAJERA:
LLOPIS:
I have been involved in many training programs for epidemiologists, especially in surveillance, and on the whole
they all have been highly disappointing. I n my opinion, we
have to change our whole approach. I think we shouldnt
start with training programs, but rather with research programs. I dont believe that any effort we make toward
training people will be successful if they have no place to
work o r to develop their skills and interests when they come
back. People should be trained in cooperative research
programs so that we could have both things at the same
time: a place where they return to work that may provide
the critical mass and this center of excellence that we were
Discussions
979
TERRIS:
980
N A;TERA:
Discussions
981
BUCK: I agree with you entirely about the elective course. But the
TERRIS:
982
BUCK:
Ive always thought I was a big failure, too. But the other
day, for reasons I wont bore you with, I sat down and tried
to figure out the number of people I knew that I had
influenced for sure. It was a very small number, but if
multiplied by itself it might be enough. Its like using the
net reproduction rate which measures how many daughters
will be born to a cohort of newborn girls. If the average is
one, you achieve replacement. So we could figure out how
many epidemiologists we have to produce to get enough.
First, of course, we have to decide how many would be
enough. We obviously dont have to turn 50 percent of
every medical class into epidemiologists. It might be very
dangerous if we did.
NAJERA:
BUCK:
NAJERA:
Discussions
983
NAJERA:
Perhaps this will be a way of evolving into the new profession I talked about. I n Spain you still have to be a physician,
but now we have more women than men in the field of
prevention and epidemiology. There has been a shift in
that.
BUCK:
I think there is a potential problem here. Unless you provide the non-medical people with much more than methodological courses in epidemiology and statistics, especially
if their background is very general, these courses will not
really prepare them for creative epidemiological research,
nor for administrative positions in public health or health
care administration. Milton Roemer had the right idea
when he said that students should be given a rich mixture
of human biology, economics, political science, administrative theory, statistics, and epidemiology. This is not the
same thing as the streamed medical curriculum, because it
offers much more than you can offer today in a medical
school.
NAJERA:
TERRIS:
I would like to emphasize a point I made when we discussed why the London School of Hygiene was so important to the movement of transition from the old to the new
epidemiology. That is that the key factor in the whole
process was the close collaboration of epidemiologists with
statisticians. You see, there is not going to be good research
984
if there isnt a team of medical or non-medical epidemiologists-and in Latin America its going to be mostly
medical, lets not delude ourselves on this point-and statisticians working very closely together. Medical epidemiologists are not sufficiently sure of themselves on
methodology; they need the statisticians. There are dangers in working with the statisticians, they can cause difficulties, but we need them. I think this is crucial. The critical
mass must include both epidemiologists and statisticians.
Of that Im convinced.
LLOPIS:
to it must
be active rather than passive. We have to present a cogent
and logically impeccable alternative.
TERRIS:
The Rockefeller Foundation people are selling this program, with real money to back it up, all over Asia, Africa,
and Latin America. They are going to divert promising
people into doing drug trials. Both the Rockefeller Foundation program and the Robert Wood Johnson Foundations clinical scholars program avoid public health
schools like the plague. I think it is an absurdity. Here we
have the Third World with all its terrible problems of
famine, malnutrition, infant diarrhea, malaria, and all the
other infectious and noninfectious diseases, and all this
money is being spent to teach clinicians how to do clinical
trials. These foundations operate under a false banner.
They are misusing the term epidemiology. Why? Because
of the great prestige of epidemiology in the world today,
because of the fact that the schools of public health are the
outstanding centers of teaching and research in epidemiology. This is threatening. They want the medical
schools to continue to be dominant; they want the clinicians
to keep their political power; they want to make sure that
health services dont infringe on the narrow professional
interests of the clinicians.
BUCK:
Discussioizr
985
BUCK:
TERRIS:
LLOPIS:
On the other hand, we have made the point that the schools
of public health need updating.
TERRIS:
That should be a major role for epidemiologists to concentrate on in the future: get the schools of public health into
the new era.
LLOPIS:
TERRIS:
LLOPIS:
Dr. E. D. Acheson
Chief Medical Officer
Department of Health and Social Security
Alexander Fleming House
London, England
986
Appendix
987
988
Appendix
Appendix
989
NOTES
NOTES