Can Cognitive-Behavioral Therapy for Anxiety and Depression Be Improved with Pharmacotherapy? A Meta-Analysis

See discussions, stats, and author profiles for this publication at:
https://www.researchgate.net/publication/319908394
Can Cognitive-Behavioral Therapy for Anxiety and
Depression Be Improved with Pharmacotherapy? A
Meta-Analysis
Article
in Psychiatric Clinics of North America · September 2017
DOI: 10.1016/j.psc.2017.08.007
CITATIONS
0
READS
390
1 author:
Some of the authors of this publication are also working on these related projects:
ENIGMA View project
Transcranial magnetic stimulation, emotion regulation, and anxiety View project
David F Tolin
Hartford Hospital
265 PUBLICATIONS 13,611 CITATIONS
SEE PROFILE
All content following this page was uploaded by David F Tolin on 26 September 2017.
The user has requested enhancement of the downloaded file.

Can Cognitive-Behavioral
Therapy for Anxiety and
Depression Be Improved
with Pharmacotherapy?
A M eta-Analysis
David F. Tolin, PhD
ROOM FOR IMPROVEMENT IN COGNITIVE-BEHAVIORAL THERAPY
The efficacy of cognitive-behavioral therapy (CBT) is well established in the treatment of
anxiety and depressive disorders. Meta-analyses of controlled trials indicates a
moderate-sized superiority for CBT versus placebo (PBO) treatment,1,2 and small to
moderate-sized superiority over alternative psychological treatments such as psychody-
namic therapy.3 However, there is clearly room for improvement: across trials, as many
as half of patients receiving CBT are considered nonresponders at posttreatment.4–7
THE POTENTIAL FOR PHARMACOLOGIC AUGMENTATION
Thus, an important question is whether pharmacologic treatment, added to CBT, can
improve outcomes over CBT alone. The efficacy of medications as monotherapies for
Disclosures: Dr D.F. Tolin has received recent research funding from Palo Alto Health Sciences,
Inc. and Pfizer. He receives royalties from Guilford Press, Oxford University Press, Turner Pub-
lishing, and Springer.
The Institute of Living, Anxiety Disorders Center, 200 Retreat Avenue, Hartford, CT 06106, USA
E-mail address: [e-mail]
KEYWORDS
 Cognitive-behavioral therapy  Drug treatment  Antidepressants
 Anxiety disorders  Depression
KEY POINTS
 Antidepressant and anxiolytic medications do not consistently or markedly enhance the
effects of CBT for patients with anxiety or depressive disorders.
 Antidepressant medications may be efficacious second-line treatments for patients failing
to respond to CBT.
 Novel agents that are thought to potentiate the neurobiological mechanisms of CBT are
promising but await further study.
Psychiatr Clin N Am - (2017) -–-
http://dx.doi.org/10.1016/j.psc.2017.08.007 psych.theclinics.com
0193-953X/17/ª 2017 Elsevier Inc. All rights reserved.

anxiety and depressive disorders is reasonably well established, although significant
concerns about reporting standards (some of which could apply to CBT trials as
well as pharmacotherapy trials) have been raised. 8,9 Across studies, antidepressant
medications, including selective serotonin reuptake inhibitors (SSRIs), serotonin-
norepinephrine reuptake inhibitors (SNRIs), tricyclic antidepressants, and monoamine
oxidase inhibitors (MAOIs) show superiority over PBO for depression,10–12 although
the overall effect is small, 13 and may be clinically meaningful only in more severe
cases.14,15 Among the anxiety disorders, the antidepressant medications, as well as
anxiolytic medications including benzodiazepines and azapirones, also show a supe-
riority to PBO across studies, with medium effects overall. 16 When medications and
CBT have been compared directly, across studies the effects are roughly equivalent,
at least in the short term.2,16–21
THE NEED TO EXAMINE PLACEBO-CONTROLLED TRIALS OF
PSYCHOPHARMACOLOGIC AUGMENTATION
Previous systematic reviews have examined whether the combination of these treat-
ments is better than CBT alone. These reviews have generally shown a small advan-
tage of combined treatment over CBT alone, although there may be an increased risk
of relapse when medications are withdrawn. 22–27 Importantly, however, an examina-
tion of CBT 1 medications versus CBT alone cannot facilitate an adequate under-
standing of the incremental efficacy of medications. In the absence of a PBO
control condition, there is no way to determine the effects of the medication, either
positive or negative. The PBO response rate in anxiety disorders is substantial 28 ; in
one systematic review, pill PBO was demonstrated to increase the probability of
response in panic disorder with agoraphobia (PD/A) by 26%. 29
However, the PBO effect also may have the reverse effect in some cases: when pa-
tients with PD/A treated with alprazolam (ALP) versus PBO plus CBT rated the extent
to which they believed that their treatment gains were attributable to medication or to
their own efforts, those who attributed their gains to medication exhibited a greater
loss of gains following discontinuation than did those who had attributed their gains
to their own efforts during treatment. 30 Additionally, when patients with specific
phobia (SpP) treated with exposure plus pill PBO were told that the pill had anxiolytic
properties that would make exposure easier, they were more likely to relapse (39%)
after treatment than were patients who were told that the pill had stimulating proper-
ties that would make exposure more difficult (0%), or those who were told that the pill
had no effect on exposure (0%). 31 Thus, there may be both PBO and “nocebo” 32 ef-
fects of adding medications to CBT, beyond the specific pharmacologic effects of the
medication itself. The aim of the present review was to examine the efficacy of phar-
macologic augmentation of CBT by synthesizing studies that randomized patients to
CBT 1 medications versus CBT 1 PBO.
The present review considers 3 distinct methods of pharmacologic augmentation of
CBT. First, studies in which CBT was applied concurrently with a medication also used
as a monotherapy (eg, CBT plus an SSRI) were examined. Second, studies in which
pharmacologic treatment was administered following CBT nonresponse, rather than
concurrent administration for all patients, were examined. This strategy relates to
the practice of stepped care, 33,34 in which treatments are added sequentially only after
nonresponse to the initial trial. Third, the analysis examines studies that used novel
agents, not prescribed as monotherapies, to augment CBT. These studies were
limited to the anxiety disorders, in which there has been increasing interest in com-
pounds that do not treat the anxiety directly, but rather are thought to potentiate the
Tolin2

mechanisms by which CBT (specifically exposure therapy) exerts its effects. 35 Two
compounds that potentiate activity in the N-methyl-D -aspartic acid (NMDA) receptor
in basolateral amygdala (implicated in fear extinction 36 ) include D -cycloserine (DCS),
which enhances fear extinction in animals, 37 and Org 25,935 (ORG), a glycine uptake
inhibitor that enhances working memory in animals. 38 Three other compounds pro-
posed to augment exposure therapy increase activity in the medial prefrontal cortex
(mPFC), thought to be critical to the extinction of learned fear 39–41
: yohimbine
(YOH), an a2-adrenergic receptor antagonist that enhances fear extinction in ani-
mals42,43 ; methylene blue (MB), an autoxidizing agent that enhances fear extinction
in animals 44,45 ; and oxytocin (OXT), a neuropeptide that disrupts signals from the
amygdala to the autonomic nervous system 46,47 enhances (normalizes) resting state
functional connectivity.
METHOD
Data Selection
A search was conducted on PsycINFO (1967–January 2017) using the terms (Cogni-
tive Therapy or Cognitive Behavior Therapy or CBT or Behavior Therapy or Exposure
Therapy) and (Anxiety Disorders or Phobia or Posttraumatic Stress Disorder or Obses-
sive Compulsive Disorder or Panic Disorder or Generalized Anxiety Disorder or Social
Anxiety or Major Depressive Disorder or Major Depression or Dysthymic Disorder or
Persistent Depressive Disorder) and (Medication or Drug Therapy or Antidepressant
or Benzodiazepines or Augmentation), limited to peer-reviewed journals published
in English and listed as clinical trials, empirical studies, longitudinal studies, prospec-
tive studies, or treatment outcome studies. Recent empirical and review articles were
also searched for additional references. Criteria for inclusion were as follows:
1. Double-blind randomized controlled trials (RCTs) that included a CBT 1 medication
condition and a CBT 1 PBO condition (other arms, when included, were not
analyzed here). Studies of adults or children were accepted. Medication lead-in
or extension periods were allowed, so long as there was a simultaneous application
of CBT 1 medications.
2. Clearly defined CBT protocol (Internet-based and self-directed CBT were also
included, but eclectic psychotherapies were not).
3. Participants had a clear primary Diagnostic and Statistical Manual of Mental Disor-
ders (DSM) diagnosis of an anxiety disorder (categorized according to DSM-IV 48 ) or
unipolar depressive disorder (anxiety and depression within severe medical condi-
tions, comorbid with substance use or developmental disorders, remitted illnesses,
or treatments aimed at symptoms other than anxiety and depression were not
included; student samples were also not included unless they met DSM criteria
for an anxiety or depressive disorder).
4. Used standardized (reliable and valid) self-report or interview-based measures of
the symptoms of the disorder being treated (global measures, functional outcomes,
behavioral avoidance tests, and biomarker outcomes were not analyzed). When
more than one such measure was included, effect sizes were aggregated across
outcome measures. Responder rates were included only when they were based
on a cutoff on a standardized measure of the symptoms of the disorder.
5. Study was judged to be a clinical trial rather than a preclinical study of mechanisms
of action.
6. Provided data (eg, means and SDs, or proportions of responders when such data
were not available) or statistics (eg, change scores or F values) that could be used
to calculate an effect size. Posttreatment was defined as the last assessment
Pharmacologic Enhancement of CBT 3

collected during (or immediately following) active treatment. Follow-up was defined
as the last assessment (up to 3 years) after treatment had been discontinued (in
some cases, the last follow-up analysis was published in a separate article).
As shown in Fig. 1, 1194 articles were considered for inclusion. Of these, 1126 were
excluded (a list of excluded articles is available from the author); the most common
reasons were absence of an RCT design, or lack of either a CBT 1 medication or
CBT 1 PBO group. This process resulted in 68 studies that met inclusion criteria.
Data Analytic Strategy
Random-effects model meta-analytic strategies49 were used with Comprehensive
Meta-Analysis. 50 When possible, effect size estimates were calculated from the
mean and SD of measures at pretreatment and posttreatment (and at follow-up, where
applicable). When these data were not provided in the published article, effect sizes
were estimated from interaction F values, mean and SD change scores, or mean
change scores with t or P values within groups. Categorical outcomes (eg, response
rates) were included only when continuous data were not available.
For each comparison of CBT 1 medication versus CBT 1 PBO, the Hedges g
(weighted by inverse variance) was calculated. Hedges g is a small-sample correction
of Cohen’s d, for which values of 0.2, 0.5, and 0.8 are conventionally accepted to
represent small, medium, and large effects, respectively. 51 Calculation of g for pre-
post designs requires an estimate of the correlation (r) between the pretreatment
Fig. 1. Flowchart of study inclusion.
Tolin4

and posttreatment scores; because this was not available in published reports, r was
conservatively estimated at 0.7 according to the recommendation of Rosenthal.52 For
ease of interpretation, pooled g estimates were also converted to number needed to
treat (NNT) using the formula provided by Furukawa, 53,54 using an estimated overall
50% response rate in CBT 1 PBO. 4–7 NNT, in this case, refers to the number of pa-
tients who would have to receive CBT with medication to obtain 1 more favorable
outcome over CBT with PBO. The I2 statistic was used to assess the percentage of
variation due to true heterogeneity rather than chance and is interpreted as follows:
25% 5 little heterogeneity, 50% 5 moderate heterogeneity, and 75% 5 high hetero-
geneity. 55 Significance of heterogeneity was established using the Q statistic.
From the literature search described previously, 36 articles were included in the
concurrent treatment analysis (Table 1) that represented posttreatment and/or
follow-up analyses of double-blind, RCTs of concurrent CBT 1 medications versus
CBT 1 PBO (other arms, when used, are not reviewed here). Concurrent medications
were divided into antidepressant and anxiolytic groups. For each class, the effect of
medications was compared across diagnoses and across drug class. Only 3
double-blind RCTs of concurrent CBT 1 medications versus CBT 1 PBO for patients
failing to respond adequately to CBT monotherapy were identified and included
(Table 2). Thirty-one articles (Table 3) were included in the novel agent analysis rep-
resenting double-blind, RCTs of concurrent CBT 1 novel agents versus CBT 1 PBO
(other arms, when used, are not reviewed here).
RESULTS
Concurrent Cognitive-Behavior Therapy with Antidepressant Medication
As shown in Table 4, the pooled effect size for antidepressants was small at posttreat-
ment, with an NNT of 7.7. At follow-up (after medication discontinuation), the pooled
effect size was negative and in the negligible range. Thus, the addition of antidepres-
sants confers a small advantage over CBT monotherapy at posttreatment, although
this advantage is lost when medications are discontinued.
Antidepressant Medication by Diagnosis
There was no significant between-group heterogeneity according to diagnosis at post-
treatment (Q 5 6.32, P 5 .18) or at follow-up (Q 5 3.65, P 5 .30). Thus, the additive
effect of antidepressant medications does not appear to be significantly stronger for
any specific anxiety or depressive disorder (although a range is noted, from
g 5 0.17 for social phobia [SoP] to g 5 0.63 for posttraumatic stress disorder
[PTSD]). For anxiety disorders, the additive effect of antidepressants over PBO was
small, with an NNT of 9.1 at posttreatment, and negative and negligible at follow-
up. Similarly, for depressive disorders, the effect was small, with an NNT of 7.5 at post-
treatment, and negative and negligible at follow-up. Thus, the small benefit of additive
antidepressant medications that is eliminated on medication discontinuation is seen
across the anxiety and depressive disorders. Significant within-group heterogeneity
was found only for PD/A at posttreatment; visual inspection of Table 1 suggests
that this was driven by a single study showing a treatment-attenuating effect of the
MAOI moclobemide.
Antidepressant medication by drug class
There was significant between-group heterogeneity according to medication at post-
treatment (Q 5 32.05, P 5 .001), although no significant between-group heterogeneity
was found at follow-up (Q 5 9.20, P 5 .10). At posttreatment, tricyclics were associ-
ated with a medium effect, with an NNT of 5.2. SSRIs/SNRIs were associated with a
Pharmacologic Enhancement of CBT 5