Article

Perceived Stress and Cognitive Vulnerability Mediate the Effects of Personality Disorder Comorbidity on Treatment Outcome in Major Depressive Disorder: A Path Analysis Study

Harvard University, Cambridge, Massachusetts, United States
Journal of Nervous & Mental Disease (Impact Factor: 1.69). 10/2007; 195(9):729-37. DOI: 10.1097/NMD.0b013e318142cbd5
Source: PubMed
ABSTRACT
Although personality disorder (PD) comorbidity has been associated with poor treatment outcome in major depressive disorder (MDD), little is known about mechanisms mediating this link. Converging evidence suggests that maladaptive cognitive patterns, particularly in interaction with stressors, might lead to poor treatment outcome in MDD subjects with PD pathology. The goal of this study was to test the role of PD comorbidity, cognitive vulnerability, and perceived stress in treatment outcome in MDD. Three hundred eighty-four MDD outpatients were enrolled in an 8-week open-label treatment of fluoxetine. Structural equation modeling and path analyses revealed that the effect of PD vulnerability on treatment outcome was fully mediated by increased pretreatment cognitive vulnerability and depression severity, which led to increased stress perception after treatment and poorer antidepressant response. Depressogenic cognitions might be continuously activated by chronic distress in MDD subjects reporting axis II pathology, leading to stress exacerbation and eventually poorer treatment outcome.

Full-text

Available from: Lee Baer
ORIGINAL ARTICLE
Perceived Stress and Cognitive Vulnerability Mediate the
Effects of Personality Disorder Comorbidity on Treatment
Outcome in Major Depressive Disorder
A Path Analysis Study
Michele Candrian, MA,* Amy Farabaugh, PhD,* Diego A. Pizzagalli, PhD,† Lee Baer, PhD,*
and Maurizio Fava, MD*
Abstract: Although personality disorder (PD) comorbidity has been
associated with poor treatment outcome in major depressive disorder
(MDD), little is known about mechanisms mediating this link.
Converging evidence suggests that maladaptive cognitive patterns,
particularly in interaction with stressors, might lead to poor treat-
ment outcome in MDD subjects with PD pathology. The goal of this
study was to test the role of PD comorbidity, cognitive vulnerability,
and perceived stress in treatment outcome in MDD. Three hundred
eighty-four MDD outpatients were enrolled in an 8-week open-label
treatment of fluoxetine. Structural equation modeling and path
analyses revealed that the effect of PD vulnerability on treatment
outcome was fully mediated by increased pretreatment cognitive
vulnerability and depression severity, which led to increased stress
perception after treatment and poorer antidepressant response. De-
pressogenic cognitions might be continuously activated by chronic
distress in MDD subjects reporting axis II pathology, leading to
stress exacerbation and eventually poorer treatment outcome.
Key Words: Personality disorder, cognitive vulnerability,
perceived stress, path analyses.
(J Nerv Ment Dis 2007;195: 729 –737)
S
tudies indicate that 20% to 50% of inpatients and 50% to
85% of outpatients with a current major depressive dis-
order (MDD) meet criteria for one or more personality
disorders (PDs) (Yen et al., 2006). These high rates of PD
comorbidity underscore the need to better understand the link
between these disorders and evaluate the potential implica-
tions of PD comorbidity on MDD.
Although inconsistencies among studies abound, the
presence of PD comorbidity is generally hypothesized to have
adverse effects on the course and treatment of MDD. In line
with this hypothesis, comorbid PD in MDD has been asso-
ciated with: longer time to achieve treatment response (Pilko-
nis and Frank, 1988); higher rates of relapse (Hart et al.,
2001; Ilardi et al., 1997); shorter time to recurrence (Cyr-
anowski et al., 2004); chronicity (Riso et al., 1996); and
poorer response to antidepressant treatment (Peselow et al.,
1992; Sato et al., 1993).
Several studies, however, did not find a link between
comorbid PD and poor treatment response (Fava et al.,
1994b, 1997, 2002; Mulder et al., 2003), and recent reviews
have challenged the view that comorbid PD negatively im-
pact treatment outcome in depression (Kool et al., 2005). A
recent meta-analysis involving the highest numbers of studies
(n 34) and patients (1663 MDD subjects with comorbid PD
and 1860 MDD subjects without comorbid PD) found, how-
ever, that comorbid PD was associated with a double risk of
poor outcome irrespective of treatment modality (drugs, psy-
chotherapy, or combined treatment) (Newton-Howes et al.,
2006).
Although many, albeit not all, studies have shown a
link between PD comorbidity and poor treatment outcome in
MDD, it is important to stress that the causal mechanisms or
mediating variables underlying this association remain
largely unknown. A greater understanding of mechanisms
underlying the relationship between personality pathology
*Depression Clinical and Research Program, Department of Psychiatry,
Massachusetts General Hospital, Harvard Medical School, Boston, Mas-
sachusetts; and †Department of Psychology, Harvard University, Cam-
bridge, Massachusetts.
Supported by NIMH grant R01 MH48483-05. DAP was supported by NIMH
(R01 MH68376) and NCCAM (R21 AT002974) grants.
Dr. Fava has received research support from Abbott Laboratories, Lichtwer
Pharma GmbH, and Lorex Pharmaceuticals as well as honoraria from
EPIX Pharmaceuticals, Bayer AG, Compellis, Janssen Pharmaceutica,
Knoll Pharmaceutical Company, Lundbeck, Dov Pharmaceuticals, Bio-
vail Pharmaceuticals, Inc., BrainCells, Inc., Cypress Pharmaceuticals, Fabre-
Kramer Pharmaceuticals, Inc., Grunenthal GmBH, MedAvante, Inc., Sepra-
cor, and Somerset Pharmaceuticals. In addition, Dr. Fava has received
both research support and honoraria from Aspect Medical Systems,
Astra-Zeneca, Bristol-Myers Squibb Company, Cephalon, Eli Lilly &
Company, Forest Pharmaceuticals Inc., GlaxoSmithkline, J & J Pharma-
ceuticals, Novartis, Organon Inc., Pharmavite, Pfizer Inc, Roche, Sanofi/
Synthelabo, Solvay Pharmaceuticals, Inc., and Wyeth-Ayerst Laborato-
ries. Dr. Pizzagalli has received research support from GlaxoSmithkline.
Send reprint requests to Michele Candrian, MS, Department of Psychiatry,
Massachusetts General Hospital, 15 Parkman St., WACC 812, Boston,
MA 02114. E-mail: mcandrian@partners.org.
Copyright © 2007 by Lippincott Williams & Wilkins
ISSN: 0022-3018/07/19509-0729
DOI: 10.1097/NMD.0b013e318142cbd5
The Journal of Nervous and Mental Disease Volume 195, Number 9, September 2007 729
Page 1
and the course of MDD might not only help in reconciling
inconsistent findings in the literature but could also inform
the development of more efficacious treatment approaches.
Diathesis-stress theories of depression might provide a
powerful framework for identifying mediating variables un-
derlying links between comorbid PD and poor treatment
outcome in depression. In general, diathesis-stress theories
postulate that specific factors predispose individuals to de-
velop depression when confronted with negative life stress
(Gotlib and Hammen, 2002). Among various diatheses, the
role of cognitive vulnerabilities in the etiology and course of
depression has received substantial empirical scrutiny. Ac-
cording to cognitive theories of depression, an individual’s
interpretation of negative events increases his or her vulner-
ability to developing and maintaining depression after these
events occur (Abramson et al., 1989; Beck, 1967). Beck’s
cognitive theory of depression, in particular, proposes that
dysfunctional attitudes—rigid and extreme beliefs about the
self, the future, and the world that often entail themes of
deriving one’s worth from being perfect or needing approval
from others—are activated in response to specific stressors,
leading to an increased likelihood to develop depression
(Beck et al., 1979).
A convergence of several lines of evidence raises the
possibility that maladaptive cognitive patterns, in particular
in interaction with stressors, might lead to poor treatment
outcome in MDD subjects with PD comorbidity. First, irre-
spective of depressive status, PDs have been associated with
elevated dysfunctional attitudes (e.g.,, Ilardi and Craighead,
1999; O’Leary et al., 1991), which in turn have been shown
to negatively impact the course and treatment of depression
(e.g.,, Alloy et al., 2006; Dunkley et al., 2006; Riso et al.,
2003; Thase et al., 1992). Of primary relevance to the present
study, elevated dysfunctional attitudes at baseline predicted
poor response to both psychological (e.g.,, Scott and Har-
rington, 1996) and pharmacological (e.g.,, Fava et al., 1994a;
Zuroff et al., 1999) treatments. Among clinically depressed
subjects, those with PD comorbidity have been found to
report significantly higher dysfunctional attitude scores than
depressed subjects without PD comorbidity (Marton et al.,
1989). Thus, dysfunctional attitudes and depressogenic cog-
nitive patterns might be important mediating variables influ-
encing treatment outcome in MDD subjects with PD comor-
bidity.
Second, PDs predispose individuals to the experience
of negative life events (American Psychiatric Association,
1994) and are characterized by increased stress reactivity. In
a community sample, for example, Daley et al. (1998) found
that PD symptoms predicted interpersonal chronic stress and
self-generated episodic stress over 2 years, which in turn
increased the risk for depression. These findings are impor-
tant, particularly since environmental factors, including life
stressors, have been found to potentiate the effects of cogni-
tive dysfunctions. Accordingly, in both clinical (e.g., Lewin-
sohn et al., 2001) and nonclinical (Flett et al., 1995) samples,
dysfunctional attitudes have been found to interact with
stressful life events to prospectively predict depressive symp-
toms or onset of depression. Of interest, recent studies sug-
gest that dysfunctional attitudes (1) fully mediated the rela-
tion between depressive symptoms and stressors (Church et
al., 2005); and (2) influenced both actual and perceived daily
stress, which in turn predicted depressive symptoms (Dunk-
ley et al., 2003). Overall, these findings suggest that individ-
uals endorsing depressogenic cognitive styles are more likely
to make negative inferences in response to negative life
events, in turn increasing their vulnerability to depression
(Abramson et al., 1989; Beck, 1967). Moreover, stress per-
ception seems to be an important mediator explaining the
relationship between dysfunctional attitudes and depressive
symptoms (Dunkley et al., 2003).
THE PRESENT STUDY
On the basis of the literature reviewed above, we
hypothesized that 3 factors—PD vulnerability, cognitive vul-
nerability, and stress exacerbation—would influence treat-
ment outcome in MDD. Specifically, we expected that (1)
certain personality traits would be linked to increased cogni-
tive vulnerability, (2) cognitive vulnerability would lead to
increased stress appraisal after the treatment, and (3) in-
creased stress appraisal (as well as increased depression
severity) would lead to poor treatment outcome. These hy-
potheses were incorporated within a model postulating that
maladaptive cognitive patterns leading to increased stress
exacerbation mediated the effects of PD comorbidity on
treatment outcome (Figure 1). Structural equation modeling
and path analyses were used to test these hypotheses and the
possible causal relations among PD vulnerability, cognitive
vulnerability, perceived stress, and treatment outcome.
METHODS
Participants
The current study presents new findings from a larger
study that evaluated the efficacy of an 8-week open-label
treatment of fluoxetine 20 mg/d for MDD. The parent study
was conducted at the Depression Clinical and Research Pro-
gram at Massachusetts General Hospital (Farabaugh et al.,
2002, 2006; Fava et al., 2002) and included 384 outpatients
between the ages of 18 and 65. All enrolled subjects met
criteria for MDD as assessed with the Structured Clinical
Interview for DSM-III-R, Patient Edition (SCID-P; Spitzer et
al., 1989) and had a score of 16 on the 17-item Hamilton
Rating Scale for Depression (HRSD; Hamilton, 1960) at
baseline. The following conditions led to exclusion from the
study: pregnancy, breast-feeding, use of birth control, suicide
risk, history of neurological illness, serious unstable medical
illness, organic mental disorders, substance abuse during the
last year, schizophrenia, delusional disorder, bipolar disorder,
severe antisocial PD, and mood-congruent or incongruent
psychotic features. Subjects were also excluded if they re-
ported: (1) a history of multiple adverse drug reactions, (2)
nonresponse to or intolerance of fluoxetine (60 80 mg/d), (3)
failure to respond to at least one adequate antidepressant
treatment during their current major depressive episode, (4)
current use of other psychotropic drugs, and (5) hypothyroid-
Candrian et al. The Journal of Nervous and Mental Disease Volume 195, Number 9, September 2007
© 2007 Lippincott Williams & Wilkins730
Page 2
ism. Throughout the acute treatment, subjects were seen
biweekly for safety and efficacy assessments.
The study protocol and procedures were approved by
the Massachusetts General Hospital Institutional Review
Board; participants provided written informed consent before
entering the study.
Clinical Assessments and Questionnaires
The main goal of the present study was to evaluate
possible causal relations among PD vulnerability, cognitive
vulnerability, perceived stress, and treatment outcome in
MDD. To this end, both before and immediately after the
8-week treatment, subjects were administered the self-rated
Perceived Stress Scale (PSS; Cohen et al., 1983), the Dys-
functional Attitude Scale (DAS; Weissman and Beck, 1978),
and the Cognitions Questionnaire (CQ; Fennell and Camp-
bell, 1984) to assess individual differences in stress appraisal,
dysfunctional attitudes, and depressive cognitive style, re-
spectively. To assess the presence of any PD, the SCID-II
(including its screening questionnaire) (First et al., 1997) was
administered at both time points. All clinical assessments
(SCID-P, SCID-II, and HRSD) were carried out by clinicians
fully trained in their administration.
The PSS has been widely used in the literature to assess
the degree to which participants appraise their daily life as
unpredictable, uncontrollable, and overwhelming (e.g., In the
last week, how often have you felt that you were unable to
control the important things in your life?”). Previous research
has shown that this scale better predicts stress-related psycho-
logical symptoms, physical symptoms, and health service utili-
zation than commonly used life event scales (Cohen et al.,
1983). This self-rated scale includes 14 items scored on a 5-point
scale and possesses satisfactory internal and short-term reliabil-
ity (coefficient alpha reliability: 0.84; 2-day test-retest reliabil-
ity: 0.85; Cohen et al., 1983).
The DAS was developed to assess dysfunctional and
rigid cognitions, which have been linked to the onset and
maintenance of depression in Beck’s cognitive theory of
depression (Beck, 1967). Specifically, this 40-item self-rated
questionnaire assesses maladaptive attitudes, including per-
fectionistic standards of performance (e.g., “If I fail at my
work, then I am a failure as a person”), sensitivity to social
criticism and need for approval (e.g., “If others dislike you,
you cannot be happy”), expectations of control (e.g., “I
should always have complete control over my feelings”), and
rigid ideas about the world. Each item is rated on a 7-point
Likert scale ranging from totally agree to totally disagree. A
total score and 2 scale scores (Perfectionism and Need for
Social Approval) can be computed; in the present study, the
total score was used. Higher scores indicate greater endorse-
ments of dysfunctional beliefs. DAS scores, either alone or in
conjunction with stressors, have been found to predict de-
pressive symptoms (Hankin et al., 2004; Ilardi and Craig-
FIGURE 1. Initial, fully mediated model postulating indirect effects of personality vulnerability and cognitive vulnerability on
treatment outcome in MDD. Ovals depict latent (unmeasured) variables (personality vulnerability and cognitive vulnerability),
whereas rectangles symbolize measured variables. Straight arrows depict paths (presumed influences). The letters “d” denote
“disturbances” (i.e., residual errors). For disturbances, a coefficient equal to 1 was selected so that the residual errors had the
same scale of measurement as the respective measured variables (Keith, 2006). HRSD: Hamilton Rating Scale for Depression
(Hamilton, 1960); DAS: Dysfunctional Attitude Scale (Weissman and Beck, 1978); CQ: Cognitions Questionnaire (Fennell and
Campbell, 1984); PSS: Perceived Stress Scale (Cohen et al., 1983). The subscript “pre” and “post” denote pretreatment and
posttreatment scores, respectively.
The Journal of Nervous and Mental Disease Volume 195, Number 9, September 2007 Mediators of Treatment Outcome in MDD
© 2007 Lippincott Williams & Wilkins 731
Page 3
head, 1999), highlighting the validity of this scale. Satisfac-
tory internal consistency (Cronbach’s alpha 0.89) and test-
retest reliability over an 8-week period (r 0.84) have been
reported (Weissman and Beck, 1978). In the present sample,
the test-retest reliability during the 8-week treatment period
was satisfactory (r 0.70, p 0.0001, n 142).
The CQ was developed to provide an overall measure
of depressive cognitive style. This self-report measure has
been derived from the revised learned helplessness model
(Abramson et al., 1978), which conceptualizes depression as
a response to negative events perceived as uncontrollable and
attributed to stable and internal causes. Specifically, the CQ
assesses 5 dimensions of negative thinking in relation to
different types of hypothetical events and their consequences.
The 5 dimensions probed are: emotional impact (e.g., aver-
siveness), attribution of causality, generalization across time,
generalization across situations, and perceived uncontrolla-
bility. A total score providing an overall measure of depres-
sive distortions was used. Previous studies have shown that
the total CQ score possesses satisfactory internal reliability
and validity (Fennell and Campbell, 1984; MacLeod and
Williams, 1990; Mitchell and Campbell, 1988). In the present
study, the CQ scale had satisfactory test-retest reliability (r
0.66, p 0.0001, n 115).
Statistics
To avoid the possibility that PD diagnoses may be con-
founded by the patient’s depressed state (Fava et al., 1994b,
2002; Zimmerman, 1994), the statistical analyses considered
only MDD subjects who either (1) met DSM-III-R criteria for
cluster A (n 42), cluster B (n 40), or cluster C (n 120)
at both the pre- and posttreatment assessments; or (2) did not
meet any PD criteria at either assessment (n 93).
For the statistical analyses, 3 data analytic strategies were
used. First, zero-order correlations between measures of depres-
sion (HRSD), cognitive vulnerability (DAS, CQ), and perceived
stress (PSS) were computed to evaluate relations among the
variables under investigation. Second, structural equation mod-
eling and path analysis were used to assess the fit between: (1)
models hypothesizing specific causal relations between PD vul-
nerability, cognitive vulnerability, stress perception, and treat-
ment outcome; and (2) the observed set of correlations between
the variables in the models. Note that the goal of the path
analyses was not to test all possible models but instead to test
models derived from previous theories and empirical findings.
Third, to test the specificity of findings emerging from the
second step, path analyses were separately performed for cluster
A, cluster B, and cluster C PDs. For path analyses, PD vulner-
ability was entered as a dichotomous variable (see Keith, 2006,
for detail concerning the use of dichotomous variables in path
analyses).
Figure 1 shows an initial model postulating a specific
causal flow from a latent exogenous variable (PD vulnerabil-
ity) through 2 sets of intervening variables (first set: pretreat-
ment HRSD and cognitive vulnerability; second set: post-
treatment PSS) to an outcome variable (posttreatment
HRSD). Thus, this model postulates indirect effects of PD
vulnerability on treatment outcome in MDD. The effects are
mediated by cognitive vulnerability and depression severity
before treatment leading to increased stress perception after
the treatment, in turn modulating treatment outcome. Both
PD vulnerability and cognitive vulnerability were defined as
latent variables.
For path analyses, we used AMOS (Arbukle, 2003;
version 5.0), which uses maximum-likelihood estimation to
test the fit of a hypothesized model to the observed variance-
covariance matrix. In line with the recommendation of Hoyle
and Panter (1995), various measures of fit were used to
evaluate various models. First, chi square was used to assess
the statistical fit of the model; nonsignificant chi square
means that the model and the actual data are consistent with
one another. Next, we considered the ratio of the chi square
value to the df in the model (absolute fit); ratios between 1
and 2 reflect better-fitting models (Carmines and McIver,
1981). To assess incremental fit, the Comparative Fit Index
(CFI) was used as a goodness-of-fit index. Goodness-of-fit
index provides an estimate of the total covariance accounted
for by the model, and CFI values over 0.95 represent a good
fit of the model to the data (Bentler, 1990). Finally, to assess
parsimony-adjusted fit, we used the root mean square error of
approximation (RMSEA); values lower than 0.05 are inter-
preted as suggesting a close fit of the model (Browne and
Cudeck, 1993).
Although the initial model postulated a full mediation of
PD on treatment outcome, an alternate model was evaluated by
adding direct paths from the exogenous latent variable (PD
vulnerability) and the mediating variable (cognitive vulnerabil-
ity) to the outcome variables (posttreatment HRSD). Note that
the initial model was nested in the alternate model (i.e., it can be
derived from the other by deleting paths). Accordingly, the
difference between the respective chi square values was com-
puted to assess whether the initial and revised models fit the data
differently. The Akaike Information Criterion (AIC) was used to
evaluate competing models. Following prior recommendations
(Keith, 2006), the model with the lower AIC value was favored.
RESULTS
Zero-Order Correlations
Before conducting a path analysis, zero-order correla-
tions were computed to determine whether the variables
under investigation were related to each other. As shown in
Table 1, most of the correlations were significant, justifying
the use of path analysis.
Initial, Fully Mediated Model
Figure 1 illustrates the initial model postulating indirect
effects of PD and cognitive vulnerability on treatment out-
come in MDD and the resulting path coefficients. As shown
in the figure, all standardized coefficients were significant and
large (i.e., above 0.25; Keith, 2006). The path between
pretreatment HRSD and posttreatment PSS score was also
significant but in the moderate range. All fit indices indicated
a good fit of the model to the data,
2
11.06, p 0.85
(df 17; n 231),
2
/df 0.65. The RMSEA was smaller
than 0.001, with a 90% confidence interval of 0.000 to 0.034,
and the CFI was 1.0. As shown in Figure 1, PD vulnerability
was significantly and positively correlated with pretreatment
Candrian et al. The Journal of Nervous and Mental Disease Volume 195, Number 9, September 2007
© 2007 Lippincott Williams & Wilkins732
Page 4
cognitive dysfunctions and pretreatment HRSD scores, which
in turn were both significantly and positively correlated with
increased stress perception after treatment. Elevated stress
perception was positively correlated with depression severity
after the 8-week treatment. Sobel’s tests (Sobel, 1982) con-
firmed that the indirect path between PD vulnerability and
post-treatment stress perception (Z 4.43, p 0.00001;
mediating variable: cognitive vulnerability), and the indirect
path between cognitive vulnerability and treatment outcome
(Z 2.75, p 0.007; mediating variable: posttreatment
stress perception) were both significant. Thus, the effect of
PD vulnerability on treatment outcome was fully mediated by
increased cognitive vulnerability and depression severity,
leading to increased stress exacerbation after treatment. Table
2 summarizes the effect coefficients for the initial model.
Revised Model
The initial model does not include direct paths between
(1) PD vulnerability and posttreatment PSS, (2) PD vulnera-
bility and posttreatment HRSD, and (3) Cognitive vulnera-
bility and posttreatment HRSD. The initial, fully mediated,
and overidentified model was compared with a revised, just-
identified model including these 3 additional paths. The
revised model fit the data equally well (
2
9.30, p 0.81,
df 14, n 231,
2
/df 0.66; RMSEA 0.001, 90%
confidence interval: 0.000 0.041; CFI 1.00). A test of the
difference between the 2 competing models indicated that the
initial model did not fit the data significantly less well than
the just-identified revised model (
2
1.76, df 3, p
0.62). Following established procedures (Keith, 2006), the
initial model was favored because (1) was more parsimonious
(df 17) than the revised model (df 14) and (2) had an
equivalent fit. Evaluation of the critical ratio (t coefficient/
SE
coefficient
) for the additional direct paths leads to a similar
conclusion in favor of the initial model. In fact, the critical
ratio for the path between PD vulnerability and posttreatment
PSS; (t 1.94, p 0.23); PD vulnerability and posttreatment
HRSD (t ⫽⫺0.36, p 0.72); and cognitive vulnerability and
posttreatment HRSD (t 0.02, p 0.98) indicated that these
paths were not significant. Finally, the AIC was lower for the
initial model, again consistent with the notion that the model
without direct path should be favored.
Cluster-Specific Model
To assess whether the initial model was specific to a
given DSM-based PD cluster, a path analysis of the initial
model was performed for cluster A, cluster B, and cluster C
separately. For each cluster, the model provided a good fit of
the data (cluster A:
2
2.18, p 0.90, df 6,
2
/df 0.36;
RMSEA 0.001; CFI 1.0; cluster B:
2
7.46, p 0.28,
df 6,
2
/df 1.24; RMSEA 0.043; CFI 0.99; cluster
C:
2
6.91, p 0.33, df 6,
2
/df 1.15; RMSEA
0.027; CFI 0.996). For each cluster, the direct path be-
tween PD and posttreatment PSS was not significant (Table
3). Interestingly, only for cluster A, a fully mediated model
was observed (Figure 2). For both cluster B and C, the path
coefficient between cognitive vulnerability and posttreatment
PSS was not significant (Table 3).
DISCUSSION
In recent years, inconsistent findings have emerged
around the question of whether PD comorbidity might have
adverse effects on the course and treatment of MDD (Kool et
al., 2005; Mulder, 2006; Newton-Howes et al., 2006). Al-
though several studies have shown a link between PD comor-
bidity and poor treatment outcome in MDD (Newton-Howes
et al., 2006), little is known about causal mechanisms or
mediating variables underlying this link. The main goal of the
present study was to evaluate the effects of potential medi-
ating variables on treatment outcome after an 8-week open-
label treatment with fluoxetine in a clinical sample charac-
TABLE 1. Zero-Order Correlations Among the Variables Under Investigation
Pretreatment HRSD Pretreatment DAS Pretreatment CQ Posttreatment PSS Posttreatment HRSD
Pretreatment HRSD 1.000 0.214** (n 160) 0.180* (n 139) 0.260*** (n 172) 0.407*** (n 231)
Pretreatment DAS 1.000 0.589*** (n 132) 0.135 (n 138) 0.154 (n 160)
Pretreatment CQ 1.000 0.258** (n 118) 0.161 (n 139)
Posttreatment PSS 1.000 0.655*** (n 172)
Posttreatment HRSD 1.000
Mean 19.52 147.86 28.19 25.87 9.41
SD 3.32 35.71 11.10 9.37 6.40
N 231 160 139 172 231
HRSD indicates Hamilton Rating Scale for Depression (Hamilton, 1960); DAS, Dysfunctional Attitude Scale (Weissman and Beck, 1978); CQ, Cognitions Questionnaire (Fennell
and Campbell, 1984); PSS, Perceived Stress Scale (Cohen et al., 1983).
*p 0.05, **p 0.01, ***p 0.001.
TABLE 2. Effect Coefficients for the Initial Model
Postulating Indirect Effects of Personality and Cognitive
Vulnerability on Treatment Outcome (Posttreatment HRSD)
Variable Direct Indirect Total
Personality vulnerability 0.219 0.219
Cognitive vulnerability 0.146 0.146
Pretreatment HRSD 0.267 0.107 0.374
Posttreatment PSS 0.576 0.000 0.576
Note: Direct, indirect, and total effects were calculated after standardizing all
variables. Personality vulnerability and cognitive vulnerability were entered as latent
variables. For personality vulnerability and cognitive vulnerability, no direct effects on
treatment outcome were postulated.
HRSD indicates Hamilton Rating Scale for Depression (Hamilton, 1960); PSS,
Perceived Stress Scale (Cohen et al., 1983).
The Journal of Nervous and Mental Disease Volume 195, Number 9, September 2007 Mediators of Treatment Outcome in MDD
© 2007 Lippincott Williams & Wilkins 733
Page 5
terized by substantial PD comorbidity. On the basis of
previous findings, we hypothesized that maladaptive cogni-
tive patterns and increased stress appraisal might mediate the
effects of PD on treatment outcome. These hypotheses were
confirmed. Specifically, path analyses revealed that PD co-
morbidity significantly and positively correlated with cogni-
tive vulnerability (dysfunctional attitudes and depressogenic
cognitive patterns), which in turn was positively correlated
with stress appraisal after the treatment; increased stress
perception was in turn significantly and positively correlated
TABLE 3. Standardized Regression Weights Emerging From the Cluster-Specific
Path Analyses
Cluster A Cluster B Cluster C
Personality vulnerability 3 Cognitive vulnerability 0.622*** 0.703*** 0.566***
Personality vulnerability 3 Pretreatment HRSD 0.207* 0.309*** 0.251***
Pretreatment HRSD 3 Posttreatment PSS 0.236** 0.211* 0.191**
Cognitive vulnerability 3 Posttreatment PSS 0.429* 0.307 0.189
Personality vulnerability 3 Posttreatment PSS 0.016 0.018 0.076
Pretreatment HRSD 3 Posttreatment HRSD 0.244*** 0.276*** 0.257***
Posttreatment PSS 3 Posttreatment HRSD 0.510*** 0.545*** 0.580***
HRSD indicates Hamilton Rating Scale for Depression (Hamilton, 1960); DAS, Dysfunctional Attitude Scale
(Weissman and Beck, 1978); CQ, Cognitions Questionnaire (Fennell and Campbell, 1984); PSS, Perceived Stress
Scale (Cohen et al., 1983).
*p 0.05, **p 0.01, ***p 0.001.
FIGURE 2. Fully mediated model investigating the effects of cluster A PD comorbidity, cognitive vulnerability, baseline depres-
sion severity, and perceived stress on treatment outcome. See Figure 1 for more details.
Candrian et al. The Journal of Nervous and Mental Disease Volume 195, Number 9, September 2007
© 2007 Lippincott Williams & Wilkins734
Page 6
with depression severity after treatment. Notably, a fully
mediated model was compared with a partially mediated
model that included direct paths between (1) PD and treat-
ment outcome, (2) PD and stress perception, and (3) cognitive
vulnerability and treatment outcome. The partially mediated
model was not a significantly better fit to the data than the
fully mediated model, and the additional 3 paths, including
the one between PD and treatment outcome, were not signif-
icant. Moreover, Sobel’s tests confirmed that the indirect path
between PD vulnerability and posttreatment stress perception
and the one between cognitive vulnerability and treatment
outcome were significant. Together with the presence of
nonsignificant direct paths, findings from the Sobel’s tests
indicate that the relation between PD and treatment outcome
can be considered fully mediated (Dunkley et al., 2006).
Interestingly, although each DSM-based PD cluster was as-
sociated with elevated cognitive vulnerability, the path coef-
ficient between pretreatment cognitive vulnerability and post-
treatment perceived stress was significant only for cluster A,
indicating that the fully mediated model provided an excel-
lent statistical fit for MDD subjects reporting enduring cluster
A pathology (paranoid, schizoid, and schizotypal PD).
The present findings implicate cognitive vulnerability
and perceived stress in the mediation of treatment outcome
for MDD subjects presenting with enduring personality pa-
thology. These results are consistent with and extend a large
body of previous work. First, PDs are characterized by deeply
ingrained and inflexible patterns of relating, perceiving, and
thinking (DSM-IV, American Psychiatric Association, 1994),
and previous studies have documented elevated dysfunctional
attitudes in subjects with axis II pathology (e.g., Ilardi and
Craighead, 1999; O’Leary et al., 1991). According to cogni-
tive theories of depression, and in particular Beck’s cognitive
theory, maladaptive, negatively focused cognitive schemata
involving themes of failure, personal inadequacy, and hope-
lessness about the self, the world, and the future are activated
in response to specific stressors, leading to an increased
likelihood to develop depression (Abramson et al., 1989;
Beck, 1967, Beck et al., 1979). Consistent with this hypoth-
esis, a multitude of studies have found that dysfunctional
attitudes and depressogenic cognitive patterns influence the
onset and course of depression. In prospective studies, for
example, individuals endorsing dysfunctional attitudes and
negative cognitive style experienced more episodes, more
severe episodes, and more chronic courses of depression
during a 2.5-year follow-up period compared with control
subjects (e.g., Alloy et al., 2006). Similarly, in a clinical
sample characterized by substantial PD comorbidity, DAS
(perfectionism) scores predicted depressive symptoms 3
years later (Dunkley et al., 2006). Of primary relevance to the
present study, elevated dysfunctional attitudes at baseline
predicted poor response to both psychological (Jarrett et al.,
1991; Scott and Harrington, 1996) and pharmacological
(Fava et al., 1994a,b; Zuroff et al., 1999) treatments. Finally,
elevated dysfunctional attitudes have been related with early
onset and longer duration of depression (Luty et al., 1999),
increased risk for relapse (e.g., Thase et al., 1992), and
chronic course (Riso et al., 2003). Findings emerging from
the present study are consistent with these prior reports and
indicate that the presence of elevated dysfunctional attitudes
and depressogenic cognitions before treatment predict higher
depressive symptoms after an 8-week fluoxetine treatment in
MDD subjects reporting PD comorbidity.
Interestingly, in the present study, the effect of cogni-
tive vulnerability on treatment outcome was mediated by
increased stress perception after the treatment. Accordingly,
MDD subjects with axis II pathology reporting rigid and
extreme beliefs about the self and the world before the
treatment reported higher level of stress, which in turn was
associated with higher depressive symptoms after the treat-
ment. These findings are intriguing, particularly since cogni-
tive vulnerability models have suggested that maladaptive
cognitive schemata may remain latent until primed by a
distress or negative life event (Ingram et al., 1998; Miranda
and Persons, 1988), and activation of cognitive vulnerability
during follow-up periods has been hypothesized to contribute
to relapse and recurrence of depression (Segal et al., 1992).
Because PD is characterized by chronic, clinically significant
distress (American Psychiatric Association, 1994, p. 633), it
is possible that depressogenic cognitions are continuously
primed and activated in MDD subjects reporting enduring
axis II pathology, leading to poor treatment outcome.
In the current study, presence of any DSM-based PD
cluster was associated with increased cognitive vulnerability.
Only for MDD subjects with cluster A comorbidity, however,
elevated dysfunctional attitudes and depressogenic cognition
at baseline predicted increased stress perception after the
treatment, indicating that the fully mediated model provided
an excellent fit only for cluster A. These findings are intrigu-
ing, particularly in light of prior evidence that symptoms of
cluster A (as well as cluster B, but not cluster C) pathology
predicted interpersonal chronic stress and self-generated ep-
isodic stress over 2 years, which in turn increased the risk for
depressive symptoms (Daley et al., 1998). Although stress
mediated the relationship between cluster A pathology and
later symptoms of depression in both the present and Daley et
al.’s (1998) study, it is important to emphasize that the
reasons for this specificity are not entirely clear and that
previous studies investigating the effects of DSM-based PD
pathology on treatment outcome in depression have yielded
somewhat inconsistent findings (Daley et al., 1999; Hart et
al., 2001; Ilardi et al., 1997; Peselow et al., 1992). One
possibility is that the emotional withdrawal, lack of warmth,
and odd/eccentric behavior characteristic of cluster A pathol-
ogy may lead to restricted social support, which is an impor-
tant buffer against the physiological (Heinrichs et al., 2003)
and psychological (Ystgaard et al., 1999) effects of stress.
Although the present findings await replication from future
studies, they suggest that the link between cognitive vulner-
ability and stress exacerbation might be particularly impor-
tant for MDD subjects reporting cluster A comorbidity.
The limitations of the present study deserve mention.
First, a key mediating variable (posttreatment stress ap-
praisal) and the outcome variable (posttreatment HRSD) were
measured concurrently. Accordingly, the present findings
cannot demonstrate any causal relation between increased
The Journal of Nervous and Mental Disease Volume 195, Number 9, September 2007 Mediators of Treatment Outcome in MDD
© 2007 Lippincott Williams & Wilkins 735
Page 7
stress perception and depressive symptoms after the treat-
ment. It is possible that depressive symptoms influenced
stress perception, or that bidirectional relations exist between
these variables. To assess causality, prospective designs as-
sessing mediating variables and outcome variables at differ-
ent time points will be required. Second, we did not investi-
gate single PD diagnoses or subgroups of patients differing in
clinical and sociodemographic variables that have been asso-
ciated with differential treatment response (Fava et al., 1997).
Although conceptually of great interest, subgrouping would
have produced sample sizes too small for the SEM analyses.
Moreover, the use of DSM-based clusters has received sup-
port in several factor and cluster-analytic studies (Bagby et
al., 1993). Third, although the present SEM provided an
excellent fit to the data, it is important to keep in mind that it
is always possible that other models not tested in the present
study might fit the data equally well or even better. The
present model was, however, developed based on previous
empirical findings and current etiological theories of depres-
sion, and the findings confirmed the a priori hypotheses.
Fourth, analyses were primarily based on self-report assess-
ments. Although the questionnaires used in the present study
have been widely used in the literature and possess satisfac-
tory reliability and validity, reporting biases cannot be ex-
cluded. Finally, in light of the rather extensive exclusion
criteria used in the current study, future work should evaluate
the generalizability of the present findings to community
samples, which will likely be more heterogeneous.
In spite of these limitations, the present findings indi-
cate that the relation between PD and treatment outcome was
fully mediated by intervening variables. Specifically, the
SEM analyses revealed that the presence of PD comorbidity
was associated with increased maladaptive cognitive patterns
(dysfunctional attitudes and depressogenic cognitions) lead-
ing to elevated stress appraisal after the treatment, which in
turn was associated with higher depression severity after an
8-week fluoxetine treatment. More generally, the present
findings underscore the need to address underlying cognitive
and personality vulnerability, in addition to symptoms of
depression, in treatments for depression (Hayes et al., 1996;
Zuroff et al., 1999).
REFERENCES
Abramson LY, Metalsky GI, Alloy LB (1989) Hopelessness depression: A
theory-based subtype of depression. Psychol Rev. 96:358 –372.
Abramson LY, Seligman MEP, Teasdale JD (1978) Learned helplessness in
humans: Critique and reformulation. J Abnorm Psychol. 87:49 –74.
Alloy LB, Abramson LY, Whitehouse WG, Hogan ME, Panzarella C, Rose
DT (2006) Prospective incidence of first onsets and recurrences of depres-
sion in individuals at high and low cognitive risk for depression. J Abnorm
Psychol. 115:145–156.
American Psychiatric Association (1994) Diagnostic and Statistical Manual
of Mental Disorders (4th ed). Washington (DC): American Psychiatric
Press.
Arbuckle JL (2003) Analysis of Moment Structures (AMOS) (User’s Guide
Version 5.0). Chicago (IL): SmallWaters Corporation.
Bagby RM, Joffe RT, Parker JDA (1993) Re-examination of the evidence for
the DSM-III personality disorder clusters. J Personal Disord. 7:320 –328.
Beck AT (1967) Depression: Clinical, Experimental and Theoretical As-
pects. New York: Hoeber Medical Division, Harper & Row.
Beck AT, Rush A, Shaw B, Emery G (1979) Cognitive Therapy of Depres-
sion. New York: Guilford Press.
Bentler PM (1990) Comparative fit indexes in structural models. Psychol
Bull. 107:238 –246.
Browne MW, Cudeck R (1993) Alternative ways of assessing model fit. In
KA Bollen, JS Long (Eds), Testing Structural Equation Models (pp
136 –162). Newbury Park (CA): Sage.
Carmines E, McIver J (1981) Analyzing models with unobserved variables:
Analysis of covariance structures. In G Bohrnstedt, E Borgatta (Eds), Social
Measurement: Current Issues (pp 65–115). Beverly Hills (CA): Sage.
Church NF, Brechman-Toussaint ML, Hine DW (2005) Do dysfunctional
cognitions mediate the relationship between risk factors and postnatal
depression symptomatology? J Affect Disord. 87:65–72.
Cohen S, KamarckT, Mermelstein R (1983) A global measure of perceived
stress. J Health Soc Behav. 24:385–396.
Cyranowski JM, Frank E, Winter E, Rucci P, Novick D, Pilkonis P, Fagiolini
A, Swartz HA, Houck P, Kupfer DJ (2004) Personality pathology and
outcome in recurrently depressed women over 2 years of maintenance
interpersonal psychotherapy. Psychol Med. 34:659 669.
Daley SE, Hammen C, Burge D (1999) Depression and axis II symptom-
atology in an adolescent community sample: Concurrent and longitudinal
associations. J Personal Disord. 13:47–59.
Daley SE, Hammen C, Davila J, Burge D (1998) Axis II symptomatology,
depression and life stress during the transition from adolescence to
adulthood. J Consult Clin Psychol. 66:595– 603.
Dunkley DM, Sanislow CA, Grilo CM, McGlashan TH (2006) Perfectionism
and depressive symptoms 3 years later: Negative social interactions,
avoidant coping and perceived social support as mediators. Compr Psy-
chiatry. 47:106 –115.
Dunkley DM, Zuroff DC, Blankstein KR (2003) Self-critical perfectionism
and daily affect: Dispositional and situational influences on stress and
coping. J Pers Soc Psychol. 84:234 –252.
Farabaugh A, Mischoulon D, Yeung A, Alpert J, Matthews J, Pava J,
Fava M (2002) Predictors of stable personality disorder diagnoses in
outpatients with remitted depression. J Nerv Ment Dis. 190:248 –256.
Farabaugh AH, Sonawalla SB, Fava M, Pedrelli P, Papakostas GI, Schwartz
F, Mischoulon D (2006) Differences in cognitive factors between “true
drug” versus “placebo pattern” response to fluoxetine as defined by pattern
analysis. Hum Psychopharmacol. 21:221–225.
Fava M, Bless E, Otto MW, Pava JA, Rosenbaum JF (1994a) Dysfunctional
attitudes in major depression. Changes with pharmacotherapy. J Nerv
Ment Dis. 182:45– 49.
Fava M, Bouffides E, Pava JA, McCarthy MK, Steingard RJ, Rosenbaum JF
(1994b) Personality disorder comorbidity with major depression and
response to fluoxetine treatment. Psychother Psychosom. 62:160 –167.
Fava M, Farabaugh AH, Sickinger AH, Wright E, Alpert JE, Sonawalla S,
Nierenberg AA, Worthington JJ III (2002) Personality disorders and
depression. Psychol Med. 32:1049 –1057.
Fava M, Uebelacker LA, Alpert JE, Nierenberg AA, Pava JA, Rosenbaum JF
(1997) Major depressive subtypes and treatment response. Biol Psychiatry.
42:568 –576.
Fennell MJ, Campbell EA (1984) The cognitions questionnaire: Specific
thinking errors in depression. Br J Clin Psychol. 23:81–92.
First MB, Spitzer RL, Gibbon M, Williams JBW, Janet BW (1997) Struc-
tured Clinical Interview for DSM-IV Personality Disorders, (SCID-II).
Washington (DC): American Psychiatric Press, Inc.
Flett GL, Hewitt PL, Blankstein KR (1995) Perfectionism, life events and
depressive symptoms: A test of a diathesis-stress model. Curr Psychol.
14:112–137.
Gotlib I, Hammen C (2002) Handbook of Depression. New York: Guilford
Press.
Hamilton M (1960) A rating scale for depression. J Neurol Neurosurg
Psychiatry. 23:56 62.
Hankin BL, Abramson LY, Miller N (2004) Cognitive vulnerability-
stress theories of depression: Examining affective specificity in the
prediction of depression versus anxiety in three prospective studies.
Cogn Ther Res. 28:309 –345.
Hart AB, Craighead WE, Craighead LW (2001) Predicting recurrence of
major depressive disorder in young adults: A prospective study. J Abnorm
Psychol. 110:633– 643.
Hayes AM, Castonguay LG, Goldfried MR (1996) Effectiveness of targeting
the vulnerability factors of depression in cognitive therapy. J Consult Clin
Psychol. 64:623– 627.
Candrian et al. The Journal of Nervous and Mental Disease Volume 195, Number 9, September 2007
© 2007 Lippincott Williams & Wilkins736
Page 8
Heinrichs M, Baumgartner T, Kirschbaum C, Ehlert U (2003) Social support
and oxytocin interact to suppress cortisol and subjective responses to
psychosocial stress. Biol Psychiatry. 54:1389 –1398.
Hoyle RH, Panter AT (1995) Writing about structural equation models. In
RH Hoyle (Ed), Structural Equation Modeling (pp 158–176). Thousand
Oaks (CA): Sage.
Ilardi SS, Craighead WE (1999) The relationship between personality pa-
thology and dysfunctional cognitions in previously depressed adults.
J Abnorm Psychol. 108:51–57.
Ilardi SS, Craighead WE, Evans DD (1997) Modeling relapse in unipolar
depression: The effects of dysfunctional cognitions and personality disor-
ders. J Consult Clin Psychol. 65:381–391.
Ingram RE, Miranda J, Segal ZV (1998) Cognitive Vulnerability to Depres-
sion. New York: Guilford Press.
Jarrett RB, Eaves GG, Grannemann BD, Rush AJ (1991) Clinical, cognitive
and demographic predictors of response to cognitive therapy for depres-
sion: A preliminary report. Psychiatry Res. 37:245–260.
Keith TZ (2006) Multiple Regression and Beyond. Boston: Pearson Educa-
tion, Inc.
Kool S, Schoevers R, de Maat S, Van R, Molenaar P, Vink A, Dekker J
(2005) Efficacy of pharmacotherapy in depressed patients with and with-
out personality disorders: A systematic review and meta-analysis. J Affect
Disord. 88:269 –278.
Lewinsohn PM, Joiner TE, Rohde P (2001) Evaluation of cognitive diathe-
sis-stress models in predicting major depressive disorder in adolescents.
J Abnorm Psychol. 110:203–215.
Luty SE, Joyce PR, Mulder RT, Sullivan PF, McKenzie JM (1999) The
relationship of dysfunctional attitudes to personality in depressed patients.
J Affect Disord. 54:75– 80.
MacLeod AK, Williams JM (1990) Overgeneralization: An important but
non-homogeneous construct. Br J Clin Psychol. 29:443– 444.
Marton P, Korenblum M, Kutcher S, Stein B, Kennedy B, Pakes J (1989)
Personality dysfunction in depressed adolescents. Can J Psychiatry. 34:
810 813.
Miranda J, Persons JB (1988) Dysfunctional attitudes are mood-state depen-
dent. J Abnorm Psychol. 97:76 –79.
Mitchell S, Campbell EA (1988) Cognitions associated with anxiety and
depression. Pers Indiv Diff. 9:837– 838.
Mulder R (2006) Personality disorder and outcome in depression. Br J Psy-
chiatry. 189:186 –187.
Mulder RT, Joyce PR, Luty SE (2003) The relationship of personality
disorders to treatment outcome in depressed outpatients. J Clin Psychiatry.
64:259 –264.
Newton-Howes G, Tyrer P, Johnson T (2006) Personality disorder and the
outcome of depression: Meta-analysis of published studies. Br J Psychi-
atry. 188:13–20.
O’Leary KM, Cowdry RW, Gardner DL (1991) Dysfunctional attitudes in
borderline personality disorder. J Personal Disord. 5:233–242.
Peselow ED, Fieve RR, DiFiglia C (1992) Personality traits and response to
desipramine. J Affect Disord. 24:209 –216.
Pilkonis PA, Frank E (1988) Personality pathology in recurrent depression:
Nature, prevalence and relationship to treatment response. Am J Psychi-
atry. 145:435– 441.
Riso LP, du Toit PL, Blandino JA, Penna S, Dacey S, Duin JS, Pacoe EM,
Grant MM, Ulmer CS (2003) Cognitive aspects of chronic depression.
J Abnorm Psychol. 112:72– 80.
Riso LP, Klein DN, Ferro T, Kasch KL, Pepper CM, Schwartz JE, Aronson
TA (1996) Understanding the comorbidity between early-onset dysthymia
and cluster B personality disorders: A family study. Am J Psychiatry.
153:900 –906.
Sato T, Sakado K, Sato S (1993) Is there any specific personality disorder or
personality disorder cluster that worsens the short-term treatment outcome
of major depression? Acta Psychiatr Scand. 88:342–349.
Scott J, Harrington J (1996) A preliminary study of the relationship among
personality, cognitive vulnerability, symptom profile and outcome in
major depressive disorder. J Nerv Ment Dis. 184:503–505.
Segal ZV, Shaw BF, Vella DD, Katz R (1992) Cognitive and life stress
predictors of relapse in remitted unipolar depressed patients: Test of the
congruency hypothesis. J Abnorm Psychol. 101:26 –36.
Sobel ME (1982) Asymptotic intervals for indirect effects in structural
equations models. In S Leinhart (Ed), Sociological Methodology (pp
290 –312). San Francisco: Jossey-Bass.
Spitzer RL, Williams JBW, Gibbon M, First MB (1989) Structured Clinical
Interview for DSM-III-R—Patient Edition. New York: Biometrics Re-
search Department, New York State Psychiatric Institute.
Thase ME, Simons AD, McGeary J, Cahalane JF, Hughes C, Harden T,
Friedman E (1992) Relapse after cognitive behavior therapy of depression:
Potential implications for longer courses of treatment. Am J Psychiatry.
149:1046 –1052.
Weissman AN, Beck AT (1978) Development and validation of the Dys-
functional Attitude Scale (DAS). Paper presented at the 12th Annual
Meeting of the Association for the Advancement of Behavior Therapy,
Chicago, IL.
Yen S, McDevitt-Murphy ME, Shea MT (2006) Depression and personality.
In DJ Stein, DJ Kupfer, AF Schatzberg (Eds), The American Psychiatric
Publishing Textbook of Mood Disorders (pp 673– 686). Washington (DC):
American Psychiatric Publishing, Inc.
Ystgaard M, Tambs K, Dalgard OS (1999) Life stress, social support and
psychological distress in late adolescence: A longitudinal study. Soc
Psychiatry Psychiatr Epidemiol. 34:12–19.
Zimmerman M (1994) Diagnosing personality disorders. A review of issues
and research methods. Arch Gen Psychiatry. 51:225–245.
Zuroff DC, Blatt SJ, Sanislow CA III, Bondi CM, Pilkonis PA (1999)
Vulnerability to depression: Reexamining state dependence and relative
stability. J Abnorm Psychol. 108:76 89.
The Journal of Nervous and Mental Disease Volume 195, Number 9, September 2007 Mediators of Treatment Outcome in MDD
© 2007 Lippincott Williams & Wilkins 737
Page 9
  • Source
    • "Pizzagalli, 2011) and psychosocial variables (e.g. Candrian et al. 2007) with depression treatment response , subtyping distinctions based on empirically derived symptom profiles have been disappointing because of profile instability (Hasler & Northoff, 2011; Baumeister & Parker, 2012; van Loo et al. 2012). However, an alternative approach to symptom-based subtyping, given the desire to predict treatment response and course of illness, would be to define subtypes using recursive partitioning (Strobl et al. 2009; Zhang & Singer, 2010) and related machine learning methods (van der Laan & Rose, 2011; James et al. 2013) that search for synergistic associations of baseline measures with subsequent outcomes. "
    [Show abstract] [Hide abstract] ABSTRACT: Background: Although variation in the long-term course of major depressive disorder (MDD) is not strongly predicted by existing symptom subtype distinctions, recent research suggests that prediction can be improved by using machine learning methods. However, it is not known whether these distinctions can be refined by added information about co-morbid conditions. The current report presents results on this question. Method: Data came from 8261 respondents with lifetime DSM-IV MDD in the World Health Organization (WHO) World Mental Health (WMH) Surveys. Outcomes included four retrospectively reported measures of persistence/severity of course (years in episode; years in chronic episodes; hospitalization for MDD; disability due to MDD). Machine learning methods (regression tree analysis; lasso, ridge and elastic net penalized regression) followed by k-means cluster analysis were used to augment previously detected subtypes with information about prior co-morbidity to predict these outcomes. Results: Predicted values were strongly correlated across outcomes. Cluster analysis of predicted values found three clusters with consistently high, intermediate or low values. The high-risk cluster (32.4% of cases) accounted for 56.6-72.9% of high persistence, high chronicity, hospitalization and disability. This high-risk cluster had both higher sensitivity and likelihood ratio positive (LR+; relative proportions of cases in the high-risk cluster versus other clusters having the adverse outcomes) than in a parallel analysis that excluded measures of co-morbidity as predictors. Conclusions: Although the results using the retrospective data reported here suggest that useful MDD subtyping distinctions can be made with machine learning and clustering across multiple indicators of illness persistence/severity, replication with prospective data is needed to confirm this preliminary conclusion.
    Full-text · Article · Jul 2014 · Psychological Medicine
  • Source
    • "Poorer treatment responses have been associated with older age, longer duration of current depressive episode, a history of multiple episodes, melancholic or psychotic features (McGrath et al., 2008; Kilts et al., 2009), and low compliance (Weiss et al., 1997). Converging evidence suggests that maladaptive behavioral patterns, particularly in interaction with stressors, ineffective utilization of medical treatment, and noneffective coping styles, might lead to poor treatment outcomes in MDD patients (Candrian et al., 2007; Leskelä et al., 2009). However, few reliable predictors have been found and most findings of antidepressant treatment prediction have not been replicated. "
    [Show abstract] [Hide abstract] ABSTRACT: There is growing evidence that individual differences among patients with major depressive disorder (MDD) on psychological and demographic measures may predict the therapeutic response to selective serotonin reuptake inhibitors (SSRIs). In this retrospective chart review, 108 outpatients with current major depressive episodes were treated with citalopram, paroxetine, or fluvoxamine. The Hamilton Depression Rating Scale and the Minnesota Multiphasic Personality Inventory-2 were administered before and after 8 weeks of SSRIs treatment. Clinical response was defined as a 50% or greater decrease in the 17-item Hamilton Depression Rating Scale total score (final visit minus baseline). This naturalistic short-term follow-up outcome study demonstrates that among depressive outpatients who responded to an 8-week trial, 57.4% achieved a good response to SSRIs. Statistical analysis showed that SSRI treatment may be 3.03 times more advantageous for MDD outpatients who are younger than 39 years. The patients with an elevated score of above 66T on the Social Introversion Minnesota Multiphasic Personality Inventory-2 scale are approximately 0.37 times as likely to be SSRI responders as are patients with a Social Introversion score less than 66T. Thus, it seems that in MDD outpatient age is the strongest predictor of response to SSRIs.
    Full-text · Article · Mar 2012 · International clinical psychopharmacology
  • Source
    • "The poorer treatment outcomes of some depressive subtypes is partly explained by the patients' level of negative or dysfunctional cognitions.33 Depressed patients' interpretation of negative events also may increase the likelihood of maintaining depression and of poor response to medication.34,35 In the midst of an episode of MDD, ineffective treatment trials may constitute a specific stressor that, interpreted in a negative context, could combine with dysfunctional attitudes to result in increasingly resistant depression in some patients. "
    [Show abstract] [Hide abstract] ABSTRACT: Current treatment of Major Depressive Disorder utilizes a trial-and-error sequential treatment strategy that results in delays in achieving response and remission for a majority of patients. Protracted ineffective treatment prolongs patient suffering and increases health care costs. In addition, long and unsuccessful antidepressant trials may diminish patient expectations, reinforce negative cognitions, and condition patients not to respond during subsequent antidepressant trials, thus contributing to further treatment resistance. For these reasons, it is critical to identify reliable predictors of antidepressant treatment response that can be used to shorten or eliminate lengthy and ineffective trials. Research on possible endophenotypic as well as genomic predictors has not yet yielded reliable predictors. The most reliable predictors identified thus far are symptomatic and physiologic characteristics of patients that emerge early in the course of treatment. We propose here the term "response endophenotypes" (REs) to describe this class of predictors, defined as latent measurable symptomatic or neurobiologic responses of individual patients that emerge early in the course of treatment, and which carry strong predictive power for individual patient outcomes. Use of REs constitutes a new paradigm in which medication treatment trials that are likely to be ineffective could be stopped within 1 to 2 weeks and other medication more likely to be effective could be started. Data presented here suggest that early changes in symptoms, quantitative electroencephalography, and gene expression could be used to construct effective REs. We posit that this new paradigm could lead to earlier recovery from depressive illness and ultimately produce profound health and economic benefits.
    Full-text · Article · Dec 2009 · Dialogues in clinical neuroscience
Show more