6.

Direct and Interactive Effects of Distress Tolerance and Anxiety Sensitivity

on Generalized Anxiety and Depression
Cogn Ther Res
DOI 10.1007/s10608-014-9623-y
Springer Science+Business Media New York 2014
Nicholas P. Allan Richard J. Macatee Aaron M. Norr Norman B. Schmidt, Department of
Psychology, Florida State University, USA

Abstract
Anxiety sensitivity and distress tolerance are both hypothesized risk factors for generalized
anxiety disorder (GAD) and major depressive disorder (MDD). However, it is unclear whether
these factors synergistically influence GAD and MDD and related symptoms. Using latent
variable methods, direct and interactive relations between anxiety sensitivity and distress
tolerance with worry and depressive symptoms and with GAD and MDD diagnoses were
examined in 347 outpatients. Interactive effects of anxiety sensitivity and distress tolerance were
found for worry and GAD/MDD. The interactions gener-ally suggested that anxiety sensitivity
confers a greater risk for worry and GAD/MDD at higher levels of distress tolerance, and that
distress tolerance confers a greater risk for worry and GAD/MDD at lower levels of anxiety
sensitivity. Given the interactive effects of anxiety sensitivity and distress tolerance for
GAD/MDD, interventions targeting both risk factors may prove more efficacious than targeting
each individually.

Introduction
Generalized anxiety disorder (GAD) and major depressive disorder (MDD) often co-occur
(e.g., Kessler et al. 1999). This has led investigators to speculate whether these disor-ders may
reflect the same disorder process(see Hettema2008 for a review), citing evidence from twin
studies indicating closely shared genetic susceptibility (Kendler et al. 2007; Roy et al. 1995),
pharmacotherapy response (Kuzma and Black 2004), and personality traits (i.e., neuroticism;
Watson et al. 2005). However, some studies have documented dis-order-specific environmental
(e.g., parental overprotection in GAD) and temperamental (e.g., low positive emotionality in
MDD) risk factors (Beesdo et al. 2010; Moffitt et al. 2007). Given the evidence that environmental
factors may be responsible for differential disorder expression (Roy et al. 1995), it is important
to identify common and distinct risk factors that may reflect environmentally-mediated expres-
sion of these conditions. In addition, given the data sug-gesting that risk factors such as
neuroticism only contributes a modest amount of variance to the genetic correlation between
MDD and GAD (Kendler et al. 2007), it is vital to delineate other risk factors that may be linked
to the shared genetic basis of the disorders.
Anxiety sensitivity (AS), or a fear of anxiety related sensations, is an ideal risk factor to
examine given that evidence exists supporting an environmentally-mediated learning model
(Reiss and McNally 1985; Schmidt et al. 2000) and a genetically-based predisposition model of
AS (Stein et al. 1999; Zavos et al. 2012). AS has reliably been linked to pathological worry (Norr
et al. 2013; Viana and Rabian 2008) and individuals with a GAD diagnosis (for which excessive
worry is a defining feature; Diagnostic and Statistical Manual of Mental Disorders, 5th ed.;
American Psychiatric Association 2013) typically have
elevated levels of AS compared to control subjects (Rodriguez et al. 2004; Taylor et al. 1992).
The association between AS and worry remains significant after accounting for symptoms of
depression (Keough et al. 2010; Rector et al. 2007). Further, although no studies have reported
the prospective relations between AS and worry and GAD specifically, several studies have
indicated that high AS confers risk for anxiety more generally (e.g., Allan et al. 2014a; Schmidt
et al. 2006).
Similar to the literature on AS and GAD, studies have shown that AS is related to symptoms of
depression in clinical and non-clinical populations (Cox et al. 1999, 2001), and that individuals
with an MDD diagnosis exhibit elevated levels of AS (Otto et al. 1995; Taylor et al. 1996). A
longitudinal study in an undergraduate population found that AS prospectively predicts
symptoms of depression controlling for baseline depressive symptoms (Grant et al. 2007).
Further, the relation between AS and depressive symptoms remains after accounting for
symptoms of GAD and other anxiety disorders, suggesting this association is not simply
accounted for by diagnostic comorbidity (e.g., Tull and Gratz 2008). Together, these and other
studies indicate that AS is a well-established risk factor for MDD and GAD and the associated
features of worry and depressive symptoms (Naragon-Gainey 2010; Olatunji and Wolitzky-
Taylor 2009).
Distress tolerance (DT), or the ability to tolerate aversive emotional states, is another risk
factor that appears to be environmentally-mediated given that it changes with treat-ment
(Bornovalova et al. 2012; McHugh et al. 2013) and is associated with genetic vulnerabilities
(Amstadter et al. 2012). Low DT has also been associated with worry across numerous non-
clinical samples (Starr and Davila 2012), even when controlling for co-occurring depressive
symp-toms (Huang et al. 2009; Keough et al. 2010). To our knowledge, only one study has
examined the relations between DT and worry in a clinical sample. Macatee et al. (2014) found
that individuals diagnosed with GAD, independent of co-occurring MDD, reported lower DT
relative to a healthy control group. Additionally, lower DT was found to prospectively predict
greater worry one month later controlling for baseline worry, suggesting that DT may act as a
risk factor for the development of pathological worry.
Similar to the literature on DT and worry, greater depressive symptoms have been associated
with low self-reported and behaviorally-indexed DT across multiple non-clinical samples
(Dennhardt and Murphy 2011; Anestis et al. 2012; Buckner et al. 2007; Ellis et al. 2010; Gorka
et al. 2012), though this literature is not entirely consistent (Brown et al. 2009; Cummings et al.
2013). Furthermore, low DT has been linked to greater depressive symptoms independent of co-
occurring worry (Starr and Davila 2012). Although fewer studies have been conducted in
clinical samples, Ellis et al. (2010) found that individuals diagnosed with MDD, independent of
co-morbidity, dem-onstrated lower behaviorally-indexed DT relative to a healthy control group.
In the self-report domain, data from two treatment studies in clinical samples suggests that DT
increases with treatment and low DT is associated with greater depressive symptoms at post-
treatment (Williams et al. 2013; McHugh et al. 2013).
Although AS and DT appear to be independently associated with worry and depressive
symptoms, few studies have examined these constructs simultaneously (e.g., Ke-ough et al. 2010;
Norr et al. 2013; Starr and Davila 2012) despite suggestions that the effect of a risk factor may
appear different in isolation than it appears in conjunction with other risk factors (Garber and
Hollon 1991). Recent studies have posited that risk and vulnerability factors may operate
synergistically to elevate anxiety and depression symptoms via augmenting stress created and
depleting coping resources (e.g., Kleiman and Riskind 2012; Riskind et al. 2010). In support of
this model, Kleiman and Riskind (2012) found two cognitive vulnerability factors (i.e., negative
cognitive style and looming cognitive style) acted synergistically to confer risk for more severe
anxiety and depression symptoms.
Findings for the unique effects of AS and DT when both variables are included have been
somewhat inconsistent across studies. Controlling for other vulnerability factors (i.e., discomfort
intolerance, intolerance of uncertainty) and negative affect, Norr et al. (2013) found AS and DT
to be uniquely associated with worry, but they failed to replicate this finding in a second sample,
though the associa-tion between DT and worry approached significance. Starr and Davila
(2012) found that AS and DT were both sig-nificantly associated with depressive symptoms after
con-trolling for co-occurring worry, but the possibility of an interaction between AS and DT was
never tested. To our knowledge, only one study has examined a potential AS and DT interaction
in the prediction of worry. Keough et al. (2010) found that AS and DT were both significantly
associated with worry after controlling for co-occurring depressive symptoms, though evidence
for an interaction was not found. To our knowledge, no studies have included an interaction
between AS and DT to explain the associa-tion between these constructs and depressive
symptoms, though one study using a clinical sample found a non-significant interaction in the
prediction of suicidal ideation (a symptom of depression) specifically (Capron et al. 2013). In
contrast to Starr and Davilas (2012) study, Ca-pron et al. (2013) found a significant main effect
for AS only. To summarize, extant data suggests that both AS and DT contribute uniquely, but
not synergistically, to worry, whereas evidence is less consistent for the non-redundant role of
DT in depressive symptoms There are two significant limitations to existing research on AS, DT,
and GAD and MDD diagnoses and associated worry and depressive symptoms. First, most
existing studies rely upon non-clinical samples, and no studies have examined AS and DT in
conjunction in clinical samples. This is a noteworthy limitation of the literature given the
importance of establishing the relative effects of these constructs in individuals at a high risk for
psychopathology (e.g., Norr et al. 2013). For example, elucidating common and distinct risk
processes in GAD and MDD is important for enhancing our understanding of nosology (e.g.,
Beesdo et al. 2010). Second, though some studies have utilized latent variable techniques to
examine the association between AS and GAD/MDD and related symptoms (e.g., Allan et al.
2014a), we are not aware of any studies that have assessed the relations between DT and
GAD/MDD and related symptoms at the latent variable level. Fur-thermore, no research exists
that uses latent variable modeling to examine AS and DT simultaneously. Exam-ining these
relations at the latent variable level can reduce concerns regarding the reliability and validity of
the mea-sured variables (MacCallum and Austin 2000). Further, synergistic effects are more
likely to be found in latent variable models than in regression-based approaches (i.e., the
approach used by Keough et al. 2010) due to the added power to detect interaction effects
(Jaccard and Wan 1995).
To address these limitations, we conducted analyses on a large outpatient sample to assess the
relative contribution and potential interaction of AS and DT in the prediction of
worry/depressive symptoms and GAD/MDD diagnoses. Based upon prior findings demonstrating
that both AS and DT contribute uniquely to worry (Keough et al. 2010; Norr et al. 2013) and
evidence suggesting that AS and DT are related (Mitchell et al. 2013; Bernstein et al. 2009), we
predicted that (1) AS and DT would be synergistically associated with worry and GAD diagnosis
such that the strongest relations between AS and DT would be at elevated levels of the other
predictor. Similarly, based on prior find-ings demonstrating that both AS and DT contribute to
depressive symptoms (Tull and Gratz 2008; Williams et al. 2013), we predicted that (2) AS and
DT would be synergis-tically associated with depression symptoms and MDD diagnosis, again
such that the strongest relations between AS and DT would be at elevated levels of the other
predictor.
Participants
The sample consisted of 347 adult outpatients seeking treatment at the Florida State University
(FSU) Anxiety and Behavioral Health Clinic (ABHC). The ABHC provides services to the
community. Referrals primarily present for anxiety-related issues and come from a catchment
area that includes northwest Florida and southern Georgia.

Discussion
The aims of the current study were to examine the main and interactive effects of AS and DT on
worry and depressive symptoms and GAD/MDD diagnoses. With regard to worry and GAD
diagnoses, the results were generally consistent with past findings that AS and DT account for
variance in worry and GAD (e.g., Macatee et al. 2014; Rodriguez et al. 2004). Although DT and
AS accounted for unique variance in worry and GAD diag-noses, a significant interaction effect
was also found, a finding inconsistent with prior research in non-clinical samples (Keough et al.
2010). However, we used latent variable modeling in a clinical sample whereas Keough et al.
(2010) used regression-based approaches in a college sample, two likely explanations for the
discrepant findings.
For both worry and GAD diagnoses in the current study, the relative importance of AS or DT
depended on the level of the other risk factor, such that having low DT or high AS attenuated the
contribution of the other risk factor.
With regard to depressive symptoms and MDD diag-nosis, the results were generally
consistent with past find-ings that AS and DT account for variance in levels of depression (Cox
et al. 2001; Starr and Davila 2012). An interaction between AS and DT best explained their
asso-ciations with the likelihood of presenting with an MDD diagnosis, but only AS was
associated with depressive symptoms. The lack of an association between DT and depressive
symptoms is surprising, given the general pat-tern of findings in our results (i.e., significant
interactions between DT and AS across worry, GAD, and MDD), as well as prior research by
Starr and Davila (2012), which found associations between DT and depressive symptoms in a
non-clinical sample. One possible explanation for this discrepancy is that we used the ASI-3 to
assess AS whereas Starr and Davila used the original ASI. The ASI-3 captures the lower-order
AS cognitive concerns dimension, which is most strongly related to depressive symptoms (Allan
et al. 2014a), more reliably than does the ASI (Taylor et al. 2007). Therefore, use of the ASI-3 in
the current study may have allowed us to more fully capture the variance shared between AS and
depression symptoms, variance which might have been accounted for by DT in the study by Starr
and Davila. To test this, future work is needed examining the unique variance DT and AS
cognitive concerns share with depressive symptoms.
The finding that the effects of AS on worry/GAD and MDD were attenuated at low DT and that
the effects of DT on worry/GAD and MDD were attenuated at high levels of AS contrasts
previous findings with other risk/vulnerability factors (e.g., Kleiman and Riskind 2012) for which
risk factors operated synergistically on anxiety and depression. One possible explanation for this
finding is that the overlap between AS and DT might also be the mechanism through which these
constructs relate to GAD and MDD. Both of these risk factors are considered tolerance
constructs (e.g., Leyro et al. 2010), and it has been suggested that they are lower-order facets of
a higher-order distress intolerance construct (McHugh and Otto 2012). This is supported by the
moderate correlations between AS and DT in the cur-rent study as well as prior studies (rs from
-.40 to -.47; e.g., Keough et al. 2010). Therefore, it might be that the mechanisms through which
AS and DT impact GAD and MDD are similar such that the effect of mean levels of one
construct is better accounted for by at-risk levels of the other construct.
Together, these findings indicate that DT and AS may be important, transdiagnostic risk factors
for GAD and MDD. These findings have potential implications for intervention
efforts aimed at treating GAD and MDD through amelio-ration of risk factors such as AS and
DT. Interventions targeting risk factors such as AS and DT are an important avenue in reduction
of anxiety and depressive symptoms. There are several studies that have demonstrated efficacy in
reducing DT (McHugh et al. 2013; Williams et al. 2013). There are also several studies that
have demonstrated efficacy in reducing AS (Keough and Schmidt 2012; Schmidt et al. 2007), and
at least one study has found evidence for reductions in anxiety and depressive symp-toms
through an intervention targeting AS (e.g., Schmidt et al. in press). Although the present findings
suggest that the effects of AS and DT depend on the level of the other risk factor, this does not
mean that the presence of one risk factor protects an individual from the influence of the other
risk factor. Rather, it is likely to mean that the presence of one risk factor masks the effect of the
other risk factor such that if interventions were to specifically target one of these risk factors
successfully, the other risk factor may then manifest as a risk factor. Therefore future
interventions targeting AS and DT in unison may demonstrate increased efficacy over
interventions targeting a single risk factor given that individuals with high AS might not fully
benefit from a DT intervention and individuals with low DT might not fully benefit from an AS
intervention.
There are several limitations to consider regarding the current study. All measures were
collected concurrently and therefore no claims of causality can be made. Although AS and DT
have been prospectively linked as risk factors for worry and depression (e.g., Macatee et al.
2014; Schmidt et al. 2006; Schmidt et al. in press), no research has examined the interactive
effects of AS and DT on worry and depression prospectively. Such a study would allow for
examination of an opposing model as well (i.e., examination of whether worry and depression
prospec-tively impact AS and DT). Only two of several important risk factors for worry and
depression were included in this study. Studies including other risk factors such as rumi-native
responding, frustration intolerance, and discomfort intolerance (e.g., Leyro et al. 2010; Starr
and Davila 2012) in addition to AS and DT would provide further evidence of the robustness of
these findings. In addition, because several of these constructs overlap with AS and DT, it is
possible that similar interactive effects would be found between them and AS and DT. In
addition, potential moderating effects, such as gender (e.g., Carter et al. 2001; Kessler et al.
1993) were not considered.
To conclude, this study demonstrated the relations AS and DT share with GAD and MDD are
each dependent on the relative level of the other risk factor. Strengths of this study include the
use of latent variable modeling, which is more powered to capture interaction effects due to the
reduction in measurement error (Jaccard and Wan 1995). In addition, this study explored these
relations in a clinical sample, and included self-reports of psychopathology as well as diagnoses
given through clinical interviews. Although prospective and experimental studies are needed to
determine whether this relation is causal, these findings suggest that future studies should
consider the interaction among risk factors when examining their effect on GAD and MDD. This
may be especially important in interven-tion and prevention studies.