Accepted Manuscript

Title: Cognitive behavioral and pharmacological treatments of

OCD in children: A systematic review and meta-analysis.
Eili N. Riise Gro Janne Wergeland
Author: Lars-Goran Ost
Bjarne Hansen and Gerd Kvale
PII:
DOI:
Reference:

S0887-6185(16)30220-1
http://dx.doi.org/doi:10.1016/j.janxdis.2016.08.003
ANXDIS 1869

To appear in:

Journal of Anxiety Disorders

Received date:
Revised date:
Accepted date:

24-4-2016
12-8-2016
12-8-2016

Please cite this article as: Ost,

Lars-Goran., Riise, Eili N., Wergeland, Gro Janne.,
Hansen, Bjarne., & Kvale, and Gerd., Cognitive behavioral and pharmacological
treatments of OCD in children: A systematic review and meta-analysis.Journal of
Anxiety Disorders http://dx.doi.org/10.1016/j.janxdis.2016.08.003
This is a PDF file of an unedited manuscript that has been accepted for publication.
As a service to our customers we are providing this early version of the manuscript.
The manuscript will undergo copyediting, typesetting, and review of the resulting proof
before it is published in its final form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that
apply to the journal pertain.

Cognitive behavioral and pharmacological treatments of OCD in children: A systematic

review and meta-analysis.

Lars-Gran sta,b,c, Eili N. Riisec,d, Gro Janne Wergelande, Bjarne Hansenc,d, & Gerd Kvalec,d

Department of Clinical Neuroscience, The Karolinska Institute, Stockholm, Sweden

Department of Psychology, Stockholm University, Sweden

Department of Clinical Psychology, University of Bergen, Norway

Haukeland University Hospital, OCD-team, 5021 Bergen, Norway

Department of Child and Adolescent Psychiatry, Haukeland University Hospital, Bergen,

Norway

*Corresponding author at Department of Clinical Neuroscience, Section of Psychology,

Karolinska Institutet, S-171 77 Stockholm, Sweden.
e-mail address: ost@psychology.su.se (L-G st), Phone: +46-737 121 285

HIGHLIGHTS
CBT yielded very large effect sizes compared to wait list and placebo.
SRIs yielded a moderate effect size compared to placebo.
The addition of SRIs did not potentiate the effect of CBT.
CBT had higher response and remission rate than SRI.
Methodological quality was positively related to the CBT effect size.

Abstract
Obsessive-compulsive disorder (OCD) is ranked by the World Health Organization (WHO)
among the 10 most debilitating disorders. The treatments which have been found effective are
cognitive behavior therapy (CBT) and serotonin reuptake inhibitors (SRI). This meta-analysis
includes all RCTs of CBT (25) and SRI (9) for OCD in youth using the Childrens YaleBrown Obsessive Compulsive Scale (C-YBOCS). CBT yielded significantly lower attrition
(12.7%) than SRI (23.5%) and placebo (24.7%). T h e e f f e c t s i z e s f o r
c o m p ariso n s of

C B T

p lace b o ( 0.9 3 ), an d
w e r e si g n i f ic a nt,

w i th

w a iti n g -list ( 1. 5 3 ),

S R I w i t h

w h ereas

C B T

a n d

C o m b o

n ot.

R e gar din g resp o nse rate

C o m b o

( C B T + S R I ) vs.

( 6 6 % )

place b o (0.5 1)
vs. S R I ( 0.2 2)

C B T
C B T

w e re sign ifica ntl y

S R I ( 4 9 % ), w h i c h
( 2 9 % ) a n d W L C

( 0.1 4 )

w e re

(7 0 % ) a n d
hig h er tha n

w as h i g h er tha n p lace b o

( 1 3 % ) .

A s f o r r e m i s s i o n CBT

(53%) and Combo (49%) were significantly higher than SRI (24%), placebo (15%), and WLC
(10%), which did not differ from each other. Combo was not more effective than CBT alone
irrespective of initial severity of the samples. T h e r a n d o m i z e d
c o ntr olle d trials ( R C T s ) h a ve a n u m b er of
m e t h o d o l o g ical pr o blem s

a n d

rec o m m e n d atio ns f or im p r o v i n g researc h

m e t h o d o l o g y

are disc ussed as

w ell as clinical

i m p licatio ns of the fin din gs.

Keywords: c h i l d r e n , o b s e s s i v e - c o m p u l s i v e
d is or d er, e x p os u re a n d res p o nse prev e nti o n,
c o g n iti ve thera p y,

m e ta -a n a l ysis, s y ste m a ti c

re vie w

1. Introduction
Obsessive-Compulsive disorder (OCD) is a debilitating anxiety disorder characterized
by obsessions and compulsions. Between 1 and 3% of children and adolescents are affected
(Canals, Hernndez-Martnez, Cosi, & Voltas, 2012; Zohar, 1999) and 75% of affected
children and adolescents also have comorbid conditions, with depression, other anxiety
disorders, oppositional defiant disorder, attention-deficit/hyperactivity disorder and Tourettes
disorder being the most common (Geller et al., 2000; Geller et al., 2001a). OCD causes
functional impairment and affects family, social life and academic performance (Piacentini,
Bergman, Keller, & McCracken, 2003). The disorder tends to be chronic with a substantial
proportion of patients still struggling with OCD-symptoms in adulthood (Micali et al., 2010;
Stewart et al., 2004). As many as 30-50% of adults with OCD report symptom onset in
childhood (Rasmussen, 1990).
CBT in the treatment of OCD entails either exposure and response prevention (ERP),
cognitive therapy (CT) or a combination of the two. CBT for pediatric OCD has demonstrated
to be superior to various comparison conditions and active medications in a growing body of
randomized controlled trials (RCTs) over the last fifteen years (McGuire et al., 2015). The
robust findings support CBT as an evidence-based treatment and it is recognized as the
treatment of choice for childhood OCD (AACAP, 2012; NICE, 2005). There is also empirical
support for the use of serotonin reuptake inhibitor (SRI) medication in treating pediatric OCD.
Both tricyclics (clomipramine) (March, Johnston, Jefferson, Kobak, & Greist, 1990; De
Vaugh-Geiss et al., 1992) and selective SRIs such as fluoxetine (e.g. Geller et al. 2004) and
sertraline (March et al., 1998) have been evaluated in several RCTs. Although SRIs have
produced significant symptom reductions relative to placebo, the yield only moderate effect
sizes as compared to large effect sizes for CBT (e.g. McGuire et al., 2015; Ivarsson et al.,
2015). There is also evidence indicating that clinical remission after pharmacological

treatment is not significantly different from placebo (POTS, 2004) and that only 30% of
patients respond to SRIs (Franklin et al., 2011). In line with these findings, the clinical
recommendations are that pharmacological treatment alone (without CBT) should be offered
only if CBT-therapists are unavailable (AACAP, 2012). However, SRI treatment in
combination with CBT is recommended for patients with moderate to severe symptoms or
treatment resistant OCD (AACAP, 2012) suggesting that there is an additional gain in
combining CBT and SRI compared to CBT alone. The empirical support for this
recommendation has been questioned. For instance, Storch et al. (2013) found no significant
difference in treatment effect when comparing CBT+SRI and CBT+Placebo and recent metaanalyses by Snchez-Meca, Rosa-Alczar, Iniesta-Seplveda, and Rosa-Alczar (2014) and
Ivarsson et al. (2015) found that CBT+SRI was not superior to CBT only. The current metaanalysis will include RCTs of CBT or SRI or the combination of the two to evaluate and
compare the efficacy of the different interventions. Hence, the results will contribute to
further illuminate the empirical basis of clinical recommendations in the treatment of OCD.
Previous meta-analyses have also examined treatment for pediatric OCD delivered as
CBT, SRIs or the combination of CBT and SRI. However, many of these have several
limitations. First, the earliest published meta-analyses included both RCTs and non-RCTs
(Abramowitz, Whiteside, & Deacon, 2005; Freeman et al., 2007; O'Kearney, Anstey, & Von
Sanden, 2006). Subsequent meta-analyses have only included RCTs, but Watson and Rees
(2008) only had five CBT-studies, whereas Sanchez-Meca et al. (2014) included 11, McGuire
et al. (2015) had 9, and Skarphedinson et al. (2014) had 13, due to various inclusion criteria
applied. Our meta-analysis included 25 CBT-studies, which means that the overlap with the
most recent meta-analyses varies from 36% to 52%, and it has the most complete coverage of
CBT-studies so far.

Systematic evaluations of research methodology are important in meta-analyses in

order to investigate methodology as a moderator of effect size, as there is a chance that such
issues may influence outcome. Furthermore, evaluations of methodology should also identify
common shortcomings in the literature in order to further develop the quality of research in
the field. Although recent meta-analyses by Snchez-Meca et al. (2014) and McGuire et al.
(2015) investigated overall methodology as a moderator of effect size, they did not provide a
summary or analysis of methodological quality in their included trials. We will investigate
methodology as a moderator of effect size and add to the current literature by providing an
updated status and analysis of methodological quality in both CBT and SRI studies allowing
for well-funded recommendations for improvement in future trials.
Background and treatment data may also explain differences in outcome between
studies. Meta-analyses of treatment efficacy aim to summarize the evidence of the
effectiveness of different treatment methods. One assumes that the possible superiority of
CBT over SRI is caused by the fact that CBT is a more efficacious treatment than SRI.
However, it could be that there are systematic differences between SRI- and CBT-studies in
terms of treatment data (e.g. attrition) or background data (e.g. pre-treatment severity) and
that these differences affect outcome. Thus, systematic evaluations of treatment and
background data are necessary to inform the field whether differences in effect sizes are
attributable to the actual treatment and not to other variables.
In summary, the current article aims to add important information to the literature by
providing an updated meta-analysis of the efficacy of CBT and SRIs in the treatment of
childhood OCD. Compared to previous meta-analyses the current analysis aims to give a
more systematic examination of treatment moderators both in terms of patient and treatment
characteristics and a more thorough evaluation of methodological issues. The aims of the
present paper are:

1. To evaluate and compare the efficacy of CBT and SRIs in the treatment of pediatric
OCD, both in terms of primary outcome measure and response/remission.
2. To investigate potential moderators of treatment outcome.
3. To evaluate differences in effect due to type of CBT or type of SRIs.
4. To investigate whether there are systematic differences between CBT and SRI trials in
terms of background or treatment data.
5. To provide a thorough evaluation of methodological stringency in CBT and SRI
studies.
6. To provide clinical recommendations as well as recommendations for methodological
improvements in future trials.

2. Method
2.1. Literature search
PsycINFO and PubMed were searched from the start of the data bases to March 1st 2016
with the following search words: Obsessive compulsive disorder OR OCD AND randomized
controlled trial OR RCT OR random* AND youth OR child* OR adolescent OR pediatric.
All abstracts were read and when there was an indication of a group of patients receiving
the particular cognitive-behavioral or drug treatment being compared with another condition
in a RCT the full-text was retrieved. Studies using single case designs were excluded since
there is no consensus yet regarding the calculation of effect sizes for these designs. The
reference lists in the retrieved articles were then checked against the database search and any
other articles that might fulfil the inclusion criteria were retrieved.
2.1.1. Inclusion criteria
In order to be included in the review and meta-analysis a study had to:

Be published, or in press, in an English language journal.

Randomly allocate participants to either treatment and control, or to two or more

active treatments.

Have participants who all were diagnosed with OCD according to DSM or ICD, and
had a maximum age of 18.

Use Childrens Yale-Brown Obsessive Compulsive Scale (CY-BOCS) or the Y-BOCS

clinician version pre- and post-treatment. The C-YBOCS (Scahill et al., 1997) is a
reliable clinician-administered 10 item rating scale assessing severity of obsessions
and compulsion.

2.1.2. Exclusion criteria

OCD not being the principal diagnosis.

The study is not a RCT.

The study is not testing a form of CBT or a SRI.

The study is not using CY-BOCS or Y-BOCS.

The study is a secondary analysis of a previously published RCT.

Figure 1 shows a flowchart of the inclusion of studies in the present meta-analysis, which was
conducted according to the PRISMA criteria (Liberati et al., 2009).

2.2. Categorization of potential moderators

2.2.1. Conditions
The various conditions in the RCTs were classified as CBT (various types of cognitive
behavioral treatments exclusively), SRI (various serotonin reuptake inhibitors), Combo (the
combination of CBT and SRI), Placebo (either pill or psychological, e.g. relaxation), and
WLC (waitlist control).
2.2.2. Declining participation

Patients who fulfilled diagnostic criteria and were offered participation but then
declined and those who accepted participation but did not show up for the first session were
counted as decliners.
2.2.3. Attrition
Patients who participated in at least one session but failed to complete the agreed upon
number of sessions were counted as dropouts. People who completed therapy but did not
show up for the post-treatment assessment were not included in this category.
2.2.4. Type of CBT
Cognitive behavioral treatments were categorized into three subcategories: ERP with
no cognitive therapy, CT with no ERP (but with behavioral experiments), or the combination
(ERP+CT).
2.2.5. Format and parental involvement
Format of therapy was classified as individual, group, or family treatment. Degree of
parental involvement was classified as low if parents are not present during sessions but
informed about progress of therapy, moderate if parents are present during some therapy
sessions full-time or only part-time of all sessions and high if parents are present full-time
during all therapy sessions.
2.2.6. Statistical analysis
Statistical analysis was categorized as intent-to-treat (ITT) if all randomized
participants are included in the statistical analysis and completers if dropouts were deleted.
2.2.7. Reliability of categorizations
The categorizations were done independently by the first and the second author
yielding a mean agreement of 90%. Disagreements were solved after consensus discussions.

2.3. Methodological quality

10

2.3.1. The psychotherapy outcome study methodology rating scale (POMRS)

The scale consists of 22 items covering various important aspects of the methodology
in psychotherapy outcome research (see st, 2008). Each item is rated as 0 = poor, 1 = fair,
and 2 = good, and each step has a verbal description of one or more sentences. The internal
consistency of the scale was good with a Cronbachs of .81. The inter rater reliability of the
scale (between the first and the third author), based on 25% randomly selected studies was
ICC(3, 1) = .92 for the total score. The kappa coefficients on the individual items varied
between .50-1.00, with a mean of .78, indicating a good to excellent inter-rater reliability.
2.3.2. Quality Rating System (QRS; Moncrieff et al., 2001)
The QRS, consisting of 23 items rated on a 0-2 scale, was used for the SRI-studies.
The third author rated all SRI-studies and 40% of these were independently rated by the first
author. The agreement between the raters was ICC(3, 1) = .88. The kappa coefficients on the
individual items varied between .40-1.00, with a mean of .68, indicating a good to excellent
inter-rater reliability.

2.4. Risk of bias

The first author used the Cochrane Collaboration tool for assessing risk of bias
(Higgins, Altman, & Sterne, 2011). The following domains were rated: random sequence
generation, allocation concealment, blinding of participants and personnel (drug studies only),
blinding of outcome assessment, incomplete outcome data, and selective reporting.

2.5. Meta-analysis
In the current meta-analysis the clinician-administered C-YBOCS (or Y-BOCS) was
used to calculate effect size for each study. This was the primary outcome measure in the
included studies. When a study presented intent-to-treat (ITT) data these were used, if not

11

completer data were used. The effect size (ES) was calculated as: (Mactive treatment
Mcomparison)/SDpooled, separately for post- and follow-up assessment. Within-group ES was
calculated as (Mpre Mpost)/SDpre according to recommendations by Morris and DeShon
(2002). The mean ES was computed by weighting each ES by the inverse of its variance.
Before pooling the effect sizes we screened for statistical outliers, defined as being
outside M 2SD. Two (3.9%) of the ESs were outliers, and winsorising (Lipsey & Wilson,
2001) was used by reducing outliers to the exact value of M+2SD. The software
Comprehensive Meta-Analysis, version 2.2.057 (CMA; Biostat Inc., 2006) was used for all
analyses and to correct for small sample sizes Hedgess g was calculated. A random effects
model was used since it cannot be assumed that the ESs come from the same population as
studies comparing CBT conditions with waitlist, placebo, medication and other forms of CBT,
and SRI with placebo are included in the meta-analysis.
Heterogeneity among ESs was assessed with the Q- and the I-square statistic. The
possibility of publication bias was analyzed with the trim-and-fill method of Duval and
Tweedie (2000) as well as Eggers regression intercept (Egger, Davey Smith, Schneider, &
Minder, 1997). Moderator analyses of continuous variables were carried out with metaregression and for categorical variables with sub-group analysis using the mixed effect model.

3. Results
3.1. Description of the studies
3.1.1. Background data
Background data for the 34 included studies (25 CBT, 9 SRI) are displayed in
Appendix Table A.1. The most common country of origin was USA (21), followed by Great
Britain (6), Australia (2), Brazil (2), and Germany, Holland, and Norway (1 each). The
diagnostic systems used were DSM-IV (28), DSM-III-R (4), and DSM-III (2). In total 1990

12

patients participated in the 34 studies. Proportions of patients who fulfilled inclusion criteria
but declined participation for various reasons were provided by 22 studies (65%). A
significantly higher proportion of CBT- (76%) than SRI-studies (33%) provided this
information (Fishers exact probability test, p = 0.040, two-tailed). The declining proportion
varied from 0% to 53.3%, with a mean of 7.4% (SD 9.7) and Mdn of 4.3%. There was no
significant difference between CBT- and SRI-studies in this respect.
Proportion of females was given by 33 studies and the mean was 45.9% (SD 9.3;
CBT- 48.8%, SRI-studies 43.9%. The mean age of the samples varied from 5.8 to 15.0 with a
mean of 12.5 (SD 2.1; CBT- 12.4, SRI-studies M 13.0). Only 14 studies (42%) reported age
of onset in their samples and this varied from 5.1 to 13.3 with a mean of 8.5 (SD 2.1). Percent
comorbidity was described by 28 studies (82%) and it ranged from 11% to 97% with a mean
of 54.3 (SD 2.3). There was a significantly higher (Qbetween (df 1) = 34.34, p = 0.0001)
proportion of patients with comorbidity in CBT- (68.9%) than in SRI-studies (31.3%).
Finally, the mean pre-treatment severity of OCD (on CY-BOCS) was 24.6 (CBT 24.6, SRI
24.5, Combo 25.6, and placebo 24.9).
3.1.2. Treatment data
Treatment data for the included studies are presented in Appendix Table A.2,
separately for CBT- and SRI-studies. Format was individual treatment in 14 studies, family
treatment in 9, and group treatment in two studies. The treatment period varied from 3 to 20
weeks with a mean of 12.7 (SD 3.8). Number of sessions varied from 9 to 14 with a mean of
12.5 (SD 1.6), and the total hours of therapy ranged from 10 to 22.5, with a mean of 15.4 (SD
4.2). Follow-up assessment was reported for 15 (60%) of the CBT-studies with a mean
follow-up period of 5.5 (SD 8.0) months. However, for those studies reporting follow-up
assessment the mean was 9.2 months.

13

In the 9 pharmacological treatment studies comparing active drug with placebo the
following drugs were used: clomipramine (2), fluoxetine (3), fluvoxamine (2), paroxetine (1),
and sertraline (1). The treatment period varied from 8 to 43 weeks with a mean of 14.7 (SD
11.3). However, when one outlier with 43 weeks was reduced to 10 weeks, the mean was
11.0 weeks. The proportion of patients reporting any side effects varied from 37% to 100% in
the drug conditions (mean 74.9, SD 22.1), and from 12.6% to 86.4% in the placebo conditions
(mean 55.3, SD 28.7). Finally, follow-up assessment was presented in only one study
(Neziroglu et al., 2000) with a period of 14 months.

3.2. Methodological data

3.2.1. Methodology ratings
Since three CBT-studies used waitlist as control condition item 22 (Equality of therapy hours)
could not be rated, and maximum score would be 42 for these studies. Thus, instead of using
the raw scores we calculated a percentage of the maximum score. The same was done for the
SRI-studies but on the Quality Rating System all studies could have a maximum score of 46.
The means were 56.3 (SD 9.0) for CBT and 68.8 (SD 10.4) for SRI. These means were not
statistically compared since they come from different rating scales.
3.2.2. Risk of bias
The risk of bias categorization is presented in Appendix Table A.3. It was unclear
regarding random sequence generation (56% of studies) and allocation concealment (79%)
since almost all studies just described that the participants were randomly allocated without
any further information. A low risk of bias was found in 19 studies (56%) concerning
blinding of outcome assessment, in 26 studies (76%) regarding incomplete outcome data (they
used intent-to-treat analysis), and all 34 studies when it comes to selective reporting.

14

Regarding blinding of participants there was a low risk in 4 (44%) and an unclear risk in 4
SRI-studies.
In order to score the risk of bias a low risk was given 1 point, an unclear risk 0.5, and a
high risk 0. The CBT-studies were scored on five variables and the SRI-studies on six. Thus,
the percentage of maximum score was calculated in order to compare the two groups of
studies. The mean score for CBT-studies (82.811.4) was significantly higher (t(32) = 4.43,
p<0.0001) than that of SRI-studies (64.87.1), which means that there was a lower risk of bias
in CBT-studies.
3.2.3. Specific methodological issues
3.2.3.1. Statistical power. Using a continuous variable, like CY-BOCS, and comparing two
conditions, as 85% of the studies do, the following cell sizes are needed to have 80% power
for an -level of .05: 26 if a large ES (d = 0.80), 64 if a medium (d = 0.50), and 400 if a small
ES (d = 0.20) is expected. A large majority (22, 65%) had a cell size which was <26, nine
(26%) had cell sizes of 26-63, and three (9%) had cell sizes of 64 or more. The mean cell size
at the start of the studies was 26.8 (SD 22.7).
When dividing the RCTs on primarily CBT- and SRI-studies we find that 28% of the
CBT- and 56% of the SRI-studies had cell sizes of 26, a non-significant difference (Fishers
test. two-tailed, p = 0.22).
3.2.3.2. Reliably diagnosing the participants. Of the 25 CBT-studies 22 (88%) used a
diagnostic schedule compared to 2 (22%) of the SRI-studies, a significant difference (Fishers
exact test, two-tailed, p = 0.0007). If the 5 studies directly comparing CBT and SRI are
excluded 100% of CBT-studies used interview schedules (Fishers exact test, p = 0.0001). Of
the 24 RCTs using an interview schedule only 4 (17%) reported IRR. Three of these (Barret
et al., 2004; Farrell et al., 2013; Lewin et al., 2014b) found perfect IRR (kappa-coefficient
1.0) and the forth (POTS, 2004) also had a high of 0.88.

15

3.2.3.3. Reliability of the primary outcome measure. The IRR of C-YBOCS should to be
demonstrated in each individual study by letting an independent rater blindly rerate a
proportion (e.g. 20%) of the interviews. Only 7 studies (21%) reported the IRR for their
application of CY-BOCS and the reliability coefficients varied between .81 and .98, with a
mean of .92. None of the SRI-studies reported IRR compared to seven (24%) of the CBTstudies, a non-significant difference (Fishers exact test, p = 0.149).
3.2.3.4. Independence and blindness of raters. Of the 34 studies in the current meta-analysis
21 (62%) reported having used independent assessors (IAs). In 10 studies (29%) it was
unclear whether the assessor was independent or not, and in two studies the assessors were
not independent. A total of 27 studies (79%) reported that the assessors were blind. When
independence and blindness were combined 20 (61%) of the studies had assessors who were
independent and blind, 2 (6%) had assessors who were neither, and 10 (30%) had unclear
independence/blindness status. Comparing CBT- and SRI-studies regarding proportion of
studies using independent assessors (54% and 89%, respectively) and blinded assessors (79%
and 89%, respectively) did not yield significant differences.
3.2.3.5. Adherence and competence ratings.
Adherence was assessed in 11 (44%) of the CBT-studies and competence in only 4
(16%) of the studies. Various rating scales were used for adherence and competence in these
studies and the reported means were recalculated as percentage of the scaless maximum
score. The mean for adherence was 89.6% and for competence 89.0%. The large proportion of
studies not assessing these important factors means that it is difficult to know if the patients
across the CBT-studies actually received the treatment they were supposed to get and how
competently it was delivered.
3.2.3.6. Credibility ratings. Fully 31 of the 34 studies (91%) were comparisons between two,
or more, active treatments. However, none of the studies included credibility ratings, and

16

only one (Lewin et al., 2014b) had expectancy ratings. This issue is particularly important
when it comes to studies using a placebo control condition. Thirteen of studies had a
condition with a psychological or pharmacological placebo but none assessed credibility.
3.2.3.7. Assessment of response and remission. Only 24 studies (71%) reported data on
response and 16 (47%) on remission. Three different criteria were used for treatment
response. Eleven studies used a certain percentage of reduction of the pre-treatment CYBOCS score. This varied from 25% (Asbahr et al., 2005) to 40% (Geller et al, 2001b), which
in CY-BOCS scores corresponded to 4.7 and 10.5 points, respectively. One study used the
cutoff score 15 (Skarphedinsson et al., 2015). The second most common criterion (11) was
a 1 or 2 on the Clinical Global Impression-Improvement scale. Finally, one study (Farrell et
al., 2013) used Reliable Change Index on CY-BOCS without any information of the required
change in points. Concerning remission 12 of the studies used a cutoff score on CY-BOCS
and of these the most common score was 10 (8), followed by 12 (3), and 14 (1). One
study used Clinically Significant Change on CY-BOCS (Farrell et al., 2013), and three
(Barrett et al., 2004; Bolton et al., 2008; Bolton et al., 2011) used free of OCD-diagnosis as
the remission criterion.
A comparison of CBT- and SRI-studies showed that there was no significant
difference regarding proportion of studies assessing response (68% for CBT and 78% for
SRI). However, 64% of the CBT-studies assessed remission, whereas none of the pure SRIstudies did so, a significant difference (Fishers exact test, two-tailed, p = 0.0011).

3.3. Meta-analysis
3.3.1. Attrition
A subgroup analysis of the attrition rates for the different conditions in this metaanalysis yielded a significant Qbetween (df 4) = 17.13, p = 0.002. The dropout rate was

17

significantly lower in CBT- (12.7%) than in SRI-conditions (23.5%) and placebo studies
(24.7%), but none of the other conditions differed significantly from each other. A
comparison within CBT studies yielded a significant Qbetween (df 2) = 8.52, p = 0.014. ERP
(17.8%) had a significantly higher attrition than CT (5.5%) and ERP+CT (9.9%), which did
not differ significantly.
3.3.2. Primary measure
Table 1 shows the results on CY-BOCS at post-treatment for all studies and divided
on the various types of comparisons. The overall ES was moderate (g = 0.52) but
significantly different from zero. It also had significant heterogeneity, which will be followed
up with moderator analyses (see 3.3.3.) All CBT- (g = 0.53) and all SRI-studies (g = 0.48)
had similar effect sizes, whereas combined CBT+SRI treatments (Combo) achieved a large
ES (g = 0.80). The latter was, however, based on only two studies. When dividing CBTstudies on different control groups we find that the effect size was very large for WLC- (g =
1.53) and large for placebo (g = 0.93) comparisons. The SRI vs. placebo comparison was
moderate (g = 0.51) but significant. When CBT was compared to any active treatment, SRIs,
Combo, or a different type of CBT the ESs were small and non-significant.
A sub-group analysis of type of CBT yielded a non-significant Qbetween (df 2) = 3.85, p
= 0.146. CT (k = 4) had a g = 1.04 (95% CI 0.45-1.63), ERP (k = 8) g = 0.68 (95% CI 0.181.18), and ERP+CT (k = 18) g = 0.35 (95% CI -0.04-0.73).
3.3.2.1. Placebo comparisons. A total of 14 studies included placebo control groups (SRI 9,
CBT 5). The overall ES for these studies (g = 0.68, 95% CI 0.46-0.90) was significantly
different from zero (z = 6.02, p = 0.0001) and significantly heterogeneous (Q = 36.5, I2 =
64%). A subgroup analysis using mixed effects found no significant difference (Qbetween =
2.65, p = 0.10) between SRI- (g = 0.52) and ERP-studies (g = 0.93).

18

3.3.2.2. SRI-studies. A sub-group analysis within the SRI-studies yielded a significant Qbetween
(df 4) = 9.98, p = 0.041). The different SRIs had the following ES: Clomipramine 1.12,
Fluoxetine 0.52, Paroxetine 0.52, Sertraline 0.28, and Fluvoxamine 0.05. However, there was
only one study each on Paroxetine and Sertraline. When these were deleted the difference
was still significant (Qbetween (df 2) = 6.50, p = 0.039).
3.3.2.3. Family-based CBT. There were a total of 10 family-based CBT (FCBT) comparisons
yielding a large ES (g = 0.92, 95% CI 0.32-1.52) which was significantly different from zero
(z = 3.02, p = 0.003) and heterogeneous (Q = 81.9, I2 = 89%). Subgroup analysis showed that
FCBT was significantly better than placebo (k = 4, g = 0.93) but not better than individual
CBT (k = 2, g = 0.24) in head-to-head comparisons (Peris & Piacentini, 2013; Reynolds et al.,
2013).
3.3.3. Moderator analyses
Continuous variables on which at least 75% of the studies provided information were
analyzed with the meta-regression module in the CMA program using the fixed effects
analysis (see Table 2). Mean age of the study sample was a negative moderator, i.e. lower ES
with increasing age. However, when the three studies which only had children in the age
range 3-8 were deleted from the analysis age was no longer a significant moderator. Pretreatment CY-BOCS score was also a significant negative moderator, i.e. lower ES with
increasing OCD severity. Analyzing this variable separately for CBT- and SRI-studies we
find opposite patterns; it was a negative moderator in CBT-studies but a positive moderator in
SRI-studies, i.e. higher ES with increasing OCD severity. Since the SRI-studies all have
placebo control we did a separate analysis with only placebo controlled CBT-studies and
found that pre-treatment CY-BOCS score was a positive moderator, i.e. the same result as for
SRI-studies. The methodological quality score of the CBT-studies was also a significant
positive predictor of ES. However, for SRI-studies methodological quality was a non-

19

significant negative moderator. Risk of bias was not a significant moderator, neither for all
studies, nor for CBT- and SRI-studies separately. Percent of eligible patients declining to
participate in the RCT was a negative moderator of ES. Of the remaining variables analyzed
weeks of treatment was a negative predictor, but this was entirely due to one outlier
(Neziroglu et al., 2000) with 43 weeks for the Combo treatment. When using the number of
weeks for the SRI condition in that study the moderator was no longer significant. Proportion
of females, of comorbidity, of patients on drug treatment for OCD during the CBT study,
treatment variables (for CBT), and attrition rate did not significantly moderate the ES.
However, the proportion of comorbid anxiety disorders was a positive moderator of ES.
For categorical variables sub-group analysis (mixed effects) in the CMA program was
used (see Table 3). The only variable that turned out significant was type of comparison group
with passive control (WLC) yielding significantly higher ES than active controls. However,
there was no difference between studies using ITT- or completer analysis. Within CBTstudies, treatment format, degree of parental involvement, and whether or not adherence to the
manual or therapist competence was assessed did not affect ES significantly.
3.3.4. Publication bias
The possibility of publication bias was investigated, using Duval and Tweedies trimand-fill method and Eggers regression intercept. For all studies, CBT studies, SRI studies,
and all the subsequent comparisons publication bias turned out not to be a problem. In 6 of the
9 comparisons no study was trimmed and in 3 only 1 study was trimmed. In no case was
Eggers regression intercept significant.
3.3.5. Response
The results regarding response to treatment are shown in Table 4. The subgroup
analysis yielded a significant Qbetween (df 4) = 70.40, p < 0.0001. CBT- (69.6%) and Combo(66.1%) had significantly higher response rates than SRI-conditions (48.9%), placebo

20

(29.3%), and WLC (13.0%). The response rate for SRI was significantly higher than that for
placebo and WLC, which did not differ from each other. A comparison within CBT-studies
yielded a significant Qbetween (df 1) = 6.44, p < 0.011; ERP+CT (77.8%) had a significantly
higher response rate than ERP (60.9%). A comparison within SRI-studies did not yield a
significant difference between clomipramine (56%) and SSRIs (48%).
At follow-up response data were provided by only 8 CBT-conditions, 2 Combo, and 1
SRI condition. The response rates were 79.7% for CBT, 69.5% for Combo, and 60.0% for
SRI, a non-significant difference (Qbetween (df 2) = 1.26, p = 0.53). Even if the follow-up
response rates are nominally higher than the post rates when post and follow-up response
rates were compared the sub-group analysis showed no significant differences for any of the
sub-groups.
3.3.6. Remission
Concerning remission (see Table 4) the sub-group analysis showed a significant
Qbetween (df 4) = 41.29, p < 0.0001. CBT- (52.7%) and Combo- (49.1%) had significantly
higher remission rates than SRI-conditions (24.1%), placebo (14.7%), and WLC (10.1%).
SRI, placebo, and WLC did not differ from each other on this measure. A comparison within
CBT-studies did not show a significant difference between ERP (46.8%), CT (54.1%), and
ERP+CT (55.3%).
At follow-up 11 CBT-conditions and only one Combo-condition provided remission
data. The remission rate was 66.3% for CBT and 69.2% for Combo. Sub-group analysis
comparing post and follow-up remission rates for CBT-conditions (k = 33) yielded a
significant Qbetween (df 1) = 6.38, p = 0.012), indicating that the remission rate increased from
post to follow-up assessment, approximately 9 months later.
3.3.7. Within-group effect size

21

The within-group effect sizes are shown in Table 5. For all conditions combined the
ESs was very large and significantly different from zero both at post-treatment (g = 1.35) and
follow-up (g = 2.66) assessment. The overall ES doubled from post to follow-up, which is
primarily due to the fact that placebo- and WLC-conditions are not included at follow-up
since they are treated after post-assessment. When post-treatment and follow-up ESs were
compared the difference was nonsignificant for CBT-conditions (Qbetween (df 1) = 1.05, p =
0.31), for SRI-conditions (Qbetween (df 1) = 0.06, p = 0.81), and for Combo conditions (Qbetween
(df 1) = 2.50, p = 0.11).
Sub-group analysis of the effect sizes at post-treatment yielded a significant Qbetween
(df 4) = 142.53, p < 0.0001. CBT (g = 2.43) and Combo (g = 2.54) had significantly higher
ES than SRI (g = 1.49), placebo (g = 0.84), and WLC (g = 0.22). The ES for SRI was
significantly higher than those of placebo and WLC, and placebo had a higher ES than WLC.
The sub-group analysis of follow-up ESs yielded a non-significant Qbetween (df 2) = 3.88, p =
0.144.
In order to evaluate if pre-treatment OCD-severity influenced the within-group ES
meta-regression analysis was used. For all CBT-conditions (k = 32) the point estimate of the
slope was positive and significant (z = 2.72, p = 0.0065) indicating larger ES as severity
increased. For the Combo conditions (k = 9) there was a significant negative point estimate
of the slope (z = -2.26, p = 0.024) indicating smaller ES as severity increased, whereas it was
negative but nonsignificant (z = -1.52, p = 0.13) for the SRI-conditions (k = 14).

4. Discussion
The aim of the present study was to provide an updated meta-analysis of the efficacy
of CBT and pharmacological treatment of pediatric OCD. This is the largest meta-analysis in
the field to date with 34 included RCTs, 25 of which were CBT- and 9 pure SRI-studies. The

22

results from the current study support findings from previous meta-analyses by demonstrating
that both CBT and SRI are efficacious interventions in the treatment of pediatric OCD.

4.1. Treatment efficacy

The mean effect sizes for CBT (g = 0.53), SRI (g = 0.48), and combined treatment (g = 0.80)
all reached statistical significance hereby demonstrating their efficacy in the treatment of
pediatric OCD. The mean effect size for SRIs (0.48) is consistent with the results of Watson
and Rees (2008) and McGuire et al. (2015) who found effect sizes of 0.48 and 0.50,
respectively. In order to compare our results for CBT with previous meta-analyses we have to
combine the ES for WLC- (g = 1.53) and placebo comparisons (g = 0.93), yielding a mean of
1.28. Watson and Rees (2008) got 1.45, Snchez-Meca et al. (2014) 1.74, and McGuire et al.
(2015) 1.21. The higher ES of Snchez-Meca et al. is probably due to fewer studies than the
present meta-analysis and a different formula for calculating ES.
Further, our results showed that within-group ES were very large for all conditions,
and CBT and Combo had significantly higher ES than SRIs. Thus, our results are consistent
with those of previous meta-analyses (e.g. Snchez-Meca et al., 2014) in demonstrating the
superiority of CBT over pharmacological treatment.
The present meta-analysis also investigated the efficacy of CBT and SRI in terms of
treatment response and remission. CBT (69.6%), SRI (48.9%) and Combo (66.1%) all had
significantly higher response rates than placebo (29.3%) and waitlist controls (13.0%), and
CBT and Combo had significantly higher response rates than SRI. The findings are consistent
with the results of McGuire et al. (2015). In terms of remission our results showed that about
half of all patients in CBT- (52.7%) and Combo studies (49.1%) no longer met diagnostic
criteria after treatment, whereas remission rate in SRI-studies (24.1%) was not significantly
different from that of placebo (14.7%) and waitlist controls (10.1%). There is no doubt that

23

CBT and SRIs reduce symptom severity, but it is noteworthy that the remission rate in SRIstudies was not significantly different from non-active treatment, and that 75% still met
diagnostic criteria for OCD post treatment.
For the CBT-studies we also investigated whether there was a difference in treatment
efficacy depending on the subtype of CBT applied. We found a larger ES for CT-studies than
for ERP and ERP+CT, but the difference was not significant and may be biased by a variety
of comparison conditions. When comparing family-based CBT and individual CBT we found
no significant difference, a finding consistent with that of McGuire et al. (2015). For the SRIstudies there was a significant difference in ES due to the type of drug. Clomipramine yielded
higher ES than the various SSRIs, which corroborates the finding reported by Snchez-Meca
et al. (2014).

4.2. Moderators
Some characteristics of the patient sample were found to be significant moderators of
treatment outcome; age, baseline severity, proportion declining participation, and proportion
of comorbid anxiety disorders. The significant association between age and ES was, however,
completely due to the three studies of very young patients (age 38); when excluding these
studies age no longer moderated ES. Further research on the youngest age group is necessary
to find out whether the relatively large effect sizes in these studies were caused merely by low
age or by other factors. Still, for the older children in the age range 818, the findings from
the current and previous meta-analyses (e.g. McGuire et al., 2015; Snchez-Meca et al., 2014)
indicate that age does not moderate outcome.
When analyzing all the studies together we found that pre-treatment OCD severity
moderated ES negatively, i.e. lower ES was associated with increasing severity. Separate
analyses showed that for the CBT-studies increasing severity was associated with lower ES,

24

but in the SRI-studies the effect was opposite. The use of different comparison conditions
explains this finding. Since all drug studies used placebo controls we analyzed CBT-studies
with placebo controls and for this subgroup there was a positive association between severity
and ES. McGuire et al. (2015) also included baseline OCD-severity as a potential moderator
but did not find a significant association. This may be due to the smaller number of
comparisons (10) compared to the present meta-analysis (34).
Further, when analyzing within-group ES for the CBT- conditions, we found that
higher pre-treatment severity yielded larger ES. This finding suggests that CBT is even more
effective in samples with severe OCD. The opposite effect was found for SRI and Combo
treatment; high severity at pre-treatment was associated with smaller effect sizes. However,
this finding should be interpreted with caution as the effect for SRI-studies was not significant
and for Combo studies it was based on only two studies.
Proportion of participants declining participation in the RCT was a significant
negative moderator for ES; the larger proportion of patients declining, the smaller the ES.
There are no obvious explanations for this finding and the issue is poorly investigated in
previous therapy research in general, and in the treatment of childhood OCD, in particular.
Our recent meta-analysis on CBT for adult OCD (st et al., 2015) did investigate proportion
of patients declining as a possible moderator but found no significant relationship between the
proportion of patients declining and ES.
Total proportion of comorbidity was not a significant moderator, but increasing
proportion of comorbid anxiety disorders was associated with larger ES. McGuire et al.
(2015) also found the same thing in CBT-studies with placebo as comparison. They suggest
that this finding may be explained by a more robust CBT response in OCD patients with
comorbid anxiety disorders as a result of a more fear-based psychopathology.

25

For CBT-studies higher methodological quality score was associated with higher ES.
This is important and indicates that it pays off to focus on designing a study with stringent
methodology. It runs counter to the meta-analysis on acceptance and commitment therapy by
st (2014) who found that the methodology score was a negative moderator, i.e. higher score
was associated with lower ES. For the SRI-studies in the present meta-analysis, we found that
methodological quality was not significantly associated with ES.
For CBT-studies the degree of parental involvement or treatment format (individual,
family or group) did not affect ES significantly. The results from the current study show that
CBT for pediatric OCD is effective when delivered in different formats, and that the active
involvement of parents is not a crucial factor for the treatment effects. This finding is
consistent with the findings of a meta-analysis on the effect of parental involvement in the
treatment of anxiety disorders overall by Thulin, Svirsky, Serlachius, Andersson, and st
(2014) who found no significant difference between parent-involved treatment and child-only
treatment, but a trend in favor of child-only.

4.3. Background data

There were no significant differences between CBT- and SRI-studies in terms of mean
age, proportion of females, and pre-treatment severity. There was no significant difference in
declining rate (7.4% overall) between CBT- and SRI-studies, and a significantly higher
proportion of CBT-studies (76%) than SRI-studies (33%) providing this information. A
comparison with our meta-analysis on adult OCD (st et al., 2015) shows similar proportions
of CBT-studies providing information on declining rate, 86% and 76%, respectively, but a
higher declining rate in the adult (15%) compared to child (7.5%) studies.
The proportion of patients with comorbidity was significantly higher in CBT-studies
(68.9%) than in SRI-studies (31.3%). Together with the higher proportion of SRI-studies not

26

disclosing the declining rate means that the samples in SRI-studies may be more highly
selective than CBT-samples since an unknown proportion of potential participants are not
included as they may prefer psychological treatment. Such a preference is illustrated by the
study of Lewin, McGuire, Murphy, and Storch (2014a).

4.4. Treatment data

Pharmacological treatment studies had significantly higher attrition rate than CBTstudies. This finding is consistent with previous meta-analyses which also found higher
attrition rates in the SRI-studies (McGuire et al., 2015; Snchez-Meca et al., 2014; Watson &
Rees, 2008). CBT- and SRI-studies had similar treatment periods with a mean of 12.5 and
14.7 weeks, respectively. Only one of the SRI-studies (11 %) reported follow-up data
(Neziroglu et al., 2000), whereas 60% of CBT-studies did so. The meta-analysis on adult
CBT-studies (st et al., 2015) found that 81% had follow-up assessment. The lack of followup in the pharmacological studies makes it impossible to evaluate the long term effect of SRIs
in the treatment of OCD, with or without continued medication, and there is an urgent need
for more knowledge about the distinct long-term efficacy of CBT- and SRI-treatment.

4.5. Methodological data

Snchez-Meca et al. (2014) found that pharmacological studies (5.5) had a higher
score than CBT-studies (4.9) on their 6-item methodology quality scale, but they did not
indicate if that difference was significant. The scales used in the current study evaluate the
studies more thoroughly and include important methodological issues such as the reliability of
the diagnostic procedure or the severity of the disorder. In our opinion the use of more
comprehensive scales in the current study gives a better reflection of the methodological
quality in the field.

27

Specific methodological issues were investigated and we found that CBT-studies

(88%) to a significantly higher degree than SRI-studies (22%) used a validated semistructured interview schedule to diagnose the participants. Only four (16 %) of the CBTstudies and none of the SRI-studies reported IRR. A recent meta-analysis on OCD-treatment
for adults (st et al., 2015) also found that IRR was rarely reported and it seems to be a
common methodological weakness in this research area.
The use of standardized diagnostic instruments is also important to enable the reliable
assessment of potential comorbid disorders. CBT-studies reported significantly higher
proportion of comorbidity than SRI-studies and because of the low degree of applying
diagnostic schedules in SRI-studies it is difficult to know if this was due to a real lower
comorbidity rate in the SRI-studies, and/or if it was a result of poor assessment of comorbid
disorders.
The validity of the outcome measure depends on the use of blinded and independent
assessors. If the assessor has a specific interest in the results, a belief that one treatment is
better than the other, or knowledge of which treatment the patient has been given, this might
affect the judgment and threaten the validity of the results. There were no significant
differences between SRI- and CBT-studies on the use of independent and blinded assessors;
62 % of the studies reported having used an independent assessor and 79 % reported that the
assessor was blinded. Although there is room for improvement in this respect, it is
encouraging that the majority of the included studies used blinded assessors. However, it
should be self-evident that RCTs in OCD also assess the IRR of the outcome measure; it is
not enough to refer to the original psychometric study of the CY-BOCS. None of the SRIstudies reported IRR compared to four of the CBT-studies (16%). A similar proportion (19%)
was obtained in the adult meta-analysis (st et al., 2015).

28

Another important methodological issue is the assessment of response and remission.

In total 71% of the studies reported response rate, and there were no significant difference
between CBT- and SRI-studies in this respect. The total percentage of studies reporting
remission was only 47 (all of them were CBT-studies); none of the pure SRI-studies provided
this information. There is also a major challenge regarding the inconsistency of the criteria
used to evaluate response and remission, and there is a need for consensus on this issue to
enable accurate analyses on the percentage of patients likely to respond to treatment. MataixCols et al. (2016) proposed a response definition of 35% reduction on CY-BOCS and a
remission definition of no longer meeting criteria of a structured diagnostic interview or 12
on CY-BOCS plus a CGI-Severity rating of 1 or 2.

4.6. Treatment of moderate-severe OCD

The American Academy of Child and Adolescent Psychiatry (2012) Committee on quality
issues recommends the combination of SSRI and CBT for patients with moderate to severe
OCD. Snchez-Meca et al. (2014) supported this conclusion in their meta-analysis, without
finding any significant difference in ES between CBT (d = 1.74) and Combo (d = 1.71).
However, the empirical basis for this recommendation is weak. We found the following
results on this issue: (1) Combo vs. CBT comparisons g = 0.14, (2) Response rate Combo
66%, CBT 70%, (3) Remission rate Combo 49%, CBT 53%. (4) Within-group effect size
Combo 2.54, CBT 2.43. None of these differences were significant. While proportion of
comorbidity overall was not a significant moderator, proportion of comorbid anxiety disorders
was a significant positive moderator. Furthermore, pre-treatment CY-BOCS score was a
significant positive moderator for the within-group ES in CBT-, but a significant negative
moderator in SRI-studies.

29

The meta-analysis of Ivarsson et al. (2015) on SRI for OCD in children also concluded
that We find no support in the literature we reviewed that moderate to severe OCD should be
treated with COMBO from the start as recommended in the US AACAP parameter (p. 101).
Based on their results and ours CBT should be recommended as the first-line treatment
irrespective of comorbidity and pre-treatment severity.

4.7. Limitations
A meta-analysis is never better than the included studies and we cannot conclude how
effective CBT, SRI or Combo would be for patients not included in RCTs. It might me that
OCD-patients with the highest severity and certain types of comorbidities never apply for
treatment in a RCT and, thus, the treatments are not tested for these patients. Another
limitation of the current meta-analysis is the exclusion of RCTs not using the interviewversion of CY-BOCS (or Y-BOCS). However, only four studies were excluded for this
reason and it is unlikely that they would have changed the outcome had they been included.
A third limitation is that none of the SRI-studies provided data on remission and only one had
follow-up assessment. A final limitation is that multiple comparisons were done without
correction, since this is not included in the applied software.

4.8. Clinical implications

Based on our systematic review and meta-analysis the following clinical implications can be
offered. CBT leads to better treatment effects than SRIs and should be the first-line treatment
for youth with OCD. There is no additional gain in combining CBT and SRI compared to
CBT alone since the combination does not lead to better effects for moderate-severe OCD.
However, adding CBT to SRI leads to a better effect than that of SRI alone. The format of
CBT (individual, group, family) does not seem to affect the outcome and neither does the

30

degree of parental involvement in the therapy. Finally, the treatment effects are maintained or
furthered at follow-up.

4.9. Recommendations for future research

Based on the methodological review conducted in the present meta-analysis we offer some
suggestions for improved methodological stringency to consider when planning future RCTs.

Do a proper power analysis before starting a RCT and adjust the design accordingly.
The field does not learn much from underpowered RCTs yielding non-significant
differences on the primary outcome measures.

Assessment of severity of OCD-symptoms with the CY-BOCS should be conducted

by trained, independent, and blinded assessors, and IRR should be reported based on a
random sample (20%) of all assessed participants.

Provide data on treatment response and remission rates in the conditions.

Carry out short- and long-term follow-up of the treatment effects.

Use a psychometrically evaluated interview schedule for diagnostic purposes and

report IRR.

Provide information on the number of declining patients and their reasons.

Assess medication at inclusion and report it.

Use a proper random sequence generation and describe the concealment of allocation.

Assess credibility when two, or more, active treatments are compared.

Videotape therapy sessions and assess therapists' adherence to the manual used and
their competence in carrying out the treatment.

Use three (or more) therapists with proper training and randomize patients, not only to
condition, but also to therapist in order to enable analysis of therapist effect.

31

Funding
No external funding was obtained for this meta-analysis.
Contributors
LG designed the meta-analysis, wrote the coding schema, rated the studies, meta-analyzed
the included studies, and wrote the first draft of method and results. ENR did the literature
search, coded the studies, and wrote the first draft of introduction and discussion. GJW did the
literature search, rated the SRI-studies, and co-rated the CBT-studies. BH and GK co-wrote
the first draft of introduction and discussion with ENR. All authors participated in the revision
and approved the final manuscript.
Conflict of interest
None of the authors have any conflict of interest to report.

Appendix. Supplementary data

Supplementary data to this article can be found online at

References
* Indicates studies included in the meta-analysis.
AACAP. (2012). Practice parameter for the assessment and treatment of children and
adolescents with obsessive-compulsive disorder. Journal of the American Academy of
Child & Adolescent Psychiatry, 51(1), 98-113. doi.org/10.1016/j.jaac.2011.09.019
Abramowitz, J. S., Whiteside, S. P., & Deacon, B. J. (2005). The effectiveness of treatment
for pediatric obsessive-compulsive disorder: A meta-analysis. Behavior Therapy,
36(1), 55-63. doi.org/10.1016/S0005-7894(05)80054-1
*Asbahr, F. R., Castillo, A. R., Ito, L. M., Latorre, M. R. D. d. O., Moreira, M. N., & LotufoNeto, F. (2005). Group cognitive-behavioral therapy versus sertraline for the treatment
of children and adolescents with obsessive-compulsive disorder. Journal of the
American Academy of Child & Adolescent Psychiatry, 44(11), 1128-1136.
doi.org/10.1097/01.chi.0000177324.40005.6f
*Barrett, P., Healy-Farrell, L., & March, J. S. (2004). Cognitive-behavioral family treatment
of childhood obsessive-compulsive disorder: A controlled trial. Journal of the
American Academy of Child & Adolescent Psychiatry, 43(1), 46-62.
doi.org/10.1097/00004583-200401000-00014
Biostat Inc. (2006). Comprehensive Meta-Analysis, version 2. Statistical software.
Englewood, NJ: Biostat Inc.
32

*Bolton, D., & Perrin, S. (2008). Evaluation of exposure with response-prevention for
obsessive compulsive disorder in childhood and adolescence. Journal of Behavior
Therapy and Experimental Psychiatry, 39, 11-22. doi.org/10.1016/j.jbtep.2006.11.002
*Bolton, D., Williams, T., Perrin, S., Atkinson, L., Gallop, C., Waite, P., & Salkovskis, P.
(2011). Randomized controlled trial of full and brief cognitive-behaviour therapy and
wait-list for paediatric obsessive-compulsive disorder. Journal of Child Psychology
and Psychiatry, 52, 1269-1278. doi.org/10.1111/j.1469-7610.2011.02419.x
Canals, J., Hernndez-Martnez, C., Cosi, S., & Voltas, N. (2012). The epidemiology of
obsessivecompulsive disorder in Spanish school children. Journal of Anxiety
Disorders, 26(7), 746-752. doi.org/10.1016/j.janxdis.2012.06.003
*de Haan, E., Hoogduin, K. A., Buitelaar, J. K., & Keijsers, G. P. (1998). Behavior therapy
versus clomipramine for the treatment of obsessive-compulsive disorder in children
and adolescents. Journal of the American Academy of Child & Adolescent Psychiatry,
37(10), 1022-1029. doi.org/10.1097/00004583-199810000-00011
*DeVeaugh-Geiss, J., Moroz, G., Biederman, J., Cantwell, D., Fontaine, R., Greist,
Landau, P. (1992). Clomipramine hydrochloride in childhood and adolescent
obsessive-compulsive disorder - a multicenter trial. Journal of the American Academy
of Child and Adolescent Psychiatry, 31, 45-49. doi.org/10.1097/00004583199201000-00008
Duval, S., & Tweedie, R. (2000). Trim and fill: A simple funnel-plot-based method of testing
and adjusting for publication bias in meta-analysis. Biometrics, 56(2), 455-463.
doi.org/10.1111/j.0006-341X.2000.00455.x.
Egger, M., Davey Smith, G., Schneider, M., & Minder, C. (1997). Bias in meta-analysis
detected by a simple, graphical test. British Medical Journal, 315(7109), 629-634.
doi.org/10.1136/bmj.316.7129.469.
*Farrell, L. J., Waters, A. M., Boschen, M. J., Hattingh, L., McConnell, H., Milliner, E. L.,
Storch, E. A. (2013). Difficult-to-treat pediatric obsessive-compulsive disorder:
feasibility and preliminary results of a randomized pilot trial of D-cycloserineaugmented behavior therapy. Depression and Anxiety, 30, 723-731.
doi.org/10.1002/da.22132
*Franklin, M. E., Sapyta, J., Freeman, J. B., Khanna, M., Compton, S., Almirall, D.,
March, J. S. (2011). Cognitive behavior therapy augmentation of pharmacotherapy in
pediatric obsessive-compulsive disorder: The pediatric OCD treatment study II (POTS
II) randomized controlled trial. JAMA. 306, 1224-1232.
doi.org/10.1001/jama.2011.1344.
Freeman, J. B., Choate-Summers, M. L., Moore, P. S., Garcia, A. M., Sapyta, J. J., Leonard,
H. L., & Franklin, M. E. (2007). Cognitive behavioral treatment for young children
with obsessive-compulsive disorder. Biological Psychiatry, 61(3), 337-343.
doi.org/10.1016/j.biopsych.2006.12.015
*Freeman, J. B., Garcia, A. M., Coyne, L., Ale, C., Przeworski, A. M. Y., Himle, M., . . .
Leonard, H. L. (2008). Early childhood OCD: preliminary findings from a familybased cognitive-behavioral approach. Journal of the American Academy of Child &
Adolescent Psychiatry, 47(5), 593-602. doi.org/10.1097/CHI.0b013e31816765f9
*Freeman, J., Sapyta, J., Garcia, A., Compton, S., Khanna, M., Flessner, C., . . . Franklin, M.
(2014). Family-based treatment of early childhood obsessive-compulsive disorder: the
pediatric obsessive-compulsive disorder treatment study for young children (POTS
Jr)-a randomized clinical trial.(Report). JAMA Psychiatry, 71(6), 689-698.
doi.org/10.1001/jamapsychiatry.2014.170.
Geller, D., Biederman, J., Faraone, S. V., Frazier, J., Coffey, B. J., Kim, G., & Bellordre, C.
A. (2000). Clinical correlates of obsessive compulsive disorder in children and
33

adolescents referred to specialized and non-specialized clinical settings. Depression

and Anxiety, 11(4), 163-168. doi.org/10.1002/1520-6394(2000)11:4<163::AIDDA3>3.0.CO;2-3
Geller, D., Biederman, J. J., Faraone, J. S., Agranat, J. A., Cradock, J. K., Hagermoser, J. L., .
. . Coffey, J. B. (2001a). Developmental aspects of obsessive compulsive disorder:
findings in children, adolescents, and adults. The Journal of Nervous and Mental
Disease, 189(7), 471-477. doi.org/10.1097/00005053-200107000-00009
*Geller, D. A., Hoog, S. L., Heiligenstein, J. H., Ricardi, R. K., Tamura, R., Kluszynski, S., &
Jacobson, J. G. (2001b). Fluoxetine treatment for obsessive-compulsive disorder in
children and adolescents: A placebo-controlled clinical trial. Journal of the American
Academy of Child & Adolescent Psychiatry, 40, 773-779. doi.org/10.1097/00004583200107000-00011
*Geller, D. A., Wagner, K. D., Emslie, G., Murphy, T., Carpenter, D. J., Wetherhold, E.,
Gardiner, C. (2004). Paroxetine treatment in children and adolescents with obsessivecompulsive disorder: A randomized, multicenter, double-blind, placebo-controlled
trial. Journal of the American Academy of Child & Adolescent Psychiatry, 43, 13871396. doi.org/10.1097/01.chi.0000138356.29099.f1
*Gorenstein, G., Gorenstein, C., de Oliveira, M. C., Asbahr, F. R., & Shavitt R. G. (2015).
Child-focused treatment of pediatric OCD affects parental behavior and family
environment, Psychiatry Research, 229, 161-166.
doi.org/10.1016/j.psychres.2015.07.050
Higgins, J.P.T., Altman, D.G., & Sterne, J.A.C. (2011). Chapter 8: Assessing risk of bias in
included studies. In J.P.T. Higgins, & S. Green (Eds.), Cochrane handbook for
systematic reviews of interventions version 5.1.0 [updated March 2011]. The
Cochrane Collaboration.
Ivarsson, T. Skarphedinsson, G., Kornr, H., Axelsdottir, B., Biedil, S., Heyman, I. . . .
March, J. (2015). The place of and evidence for serotonin reuptake inhibitors (SRIs)
for obsessive compulsive disorder (OCD) in children and adolescents: Views based on
a systematic review and meta-analysis. PsychiatryResearch. 227, 93-103.
doi.org/10.1016/j.psychres.2015.01.015
Lewin, A. B., McGuire, J. F., Murphy, T. K., & Storch, E. A. (2014a). Editorial perspective:
The importance of considering parent's preferences when planning treatment for their
children the case of childhood obsessivecompulsive disorder. Journal of Child
Psychology and Psychiatry, 55(12), 1314-1316. doi.org/10.1111/jcpp.12344
*Lewin, A. B., Park, J. M., Jones, A. M., Crawford, E. A., DeNadai, A. S., Menzel, J., . . .
Storch, E. A. (2014b). Family-based exposure and response prevention therapy for
preschool-aged children with obsessive-compulsive disorder: A pilot randomized
controlled trial. Behaviour Research and Therapy, 56, 30-38.
doi.org/10.1016/j.brat.2014.02.001
Liberati, A., Altman, D.G., Tetzlaff, J., Mulrow, C., Gotsche, P.C., Ioannidis, J.P.A.,
Moher, D. (2009). The PRISMA statement for reporting systematic reviews and metaanalyses of studies that evaluate health care interventions: Explanation and
elaboration. PLoS Medicine, 6, e1000100. doi.org/10.7326/0003-4819-151-4200908180-00136
*Liebowitz, M. R., Turner, S. M., Piacentini, J., Beidel, D. C., Clarvit, S. R., Davies,
Simpson, H. B. (2002). Fluoxetine in children and adolescents with OCD: a placebocontrolled trial. Journal of the American Academy of Child and Adolescent Psychiatry,
41, 1431-1438. doi.org/10.1097/00004583-200212000-00014
Lipsey, M.W., & Wilson, D.B. (2001). Practical meta-analysis. Thousand Oaks, CA: Sage
Publications.
34

*March, J. S., Johnston, H., Jefferson, J. W., Kobak, K. A., & Greist, J. H. (1990). Do subtle
neurological impairments predict treatment resistance to clomipramine in children and
adolescents with obsessive-compulsive disorder? Journal of Child and Adolescent
Psychopharmacology, 1, 133-140. doi.org/10.1089/cap.1990.1.133
*March, J. S., Biederman, J., Wolkow, R., Safferman, A., Mardekian, J., Cook, E. H.,
Wagner, K. D. (1998). Sertraline in children and adolescents with obsessivecompulsive disorder: A multicenter randomized controlled trial. JAMA, 280, 17521756. doi.org/10.1001/jama.280.20.1752
Mataix-Cols, D., Fernandez de la Cruz, L., Nordsleten, A. E., Lenhard, F., Isomura, K., &
Simpson, H. B. (2016). Towards an international expert consensus for defining
treatment response, remission, recovery, and relapse in obsessive-compulsive disorder:
A Delphi survey. World Psychiatry, in press.
*Mataix-Cols, D., Turner, C., Monzani, B., Isomura, K., Murphy, C., Krebs, G., & Heyman,
I. (2014). Cognitive-behavioural therapy with post-session D-cycloserine
augmentation for paediatric obsessive-compulsive disorder: pilot randomised
controlled trial. British Journal of Psychiatry, 204(1), 77.
doi.org/10.1192/bjp.bp.113.126284
McGuire, J. F., Piacentini, J., Lewin, A. B., Brennan, E. A., Murphy, T. K., & Storch, E. A.
(2015). A meta-analysis of cognitive behavior therapy and medication for child
obsessive-compulsive disorder: moderators of treatment efficacy, response, and
remission. Depression and Anxiety, 32(8), 580-593. doi.org/10.1002/da.22389
*Merlo, L. J., Storch, E. A., Lehmkuhl, H. D., Jacob, M. L., Murphy, T. K., Goodman, W. K.,
& Geffken, G. R. (2010). Cognitive behavioral therapy plus motivational interviewing
improves outcome for pediatric obsessive-compulsive disorder: A preliminary study.
Cognitive Behaviour Therapy, 39, 24-27. doi.org/10.1080/16506070902831773
Micali, N., Heyman, I., Perez, M., Hilton, K., Nakatani, E., Turner, C., & Mataix-Cols, D.
(2010). Long-term outcomes of obsessive-compulsive disorder: follow-up of 142
children and adolescents. British Journal of Psychiatry, 197(2), 128-134.
doi.org/10.1192/bjp.bp.109.075317
Moncrieff, J., Churchill, R., Drummond, D. C., & McGuire, H. (2001). Development of a
quality assessment instrument for trials of treatments for depression and neurosis.
International Journal of Methods in Psychiatric Research, 10(3), 126-133.
doi.org/10.1002/mpr.108
Morris, S.B., & DeShon, R.P. (2002). Combining effect size estimates inmeta-analysis with
repeated measures and independent-groups designs. Psychological Methods, 7(1),
105125. doi.org/10.1037/1082-989X.7.1.105.
*Neziroglu, F., Yaryura-Tobias, J. A., Walz, J., & McKay, D. (2000). The effect of
fluvoxamine and behavior therapy on children and adolescents with obsessivecompulsive disorder. Journal of Child and Adolescent Psychopharmacology, 10(4),
295-306. doi.org/10.1089/cap.2000.10.295
NICE. (2005). Obsessive-compulsive disorder: Core interventions in the treatment of
obsessive-compulsive disorder and body dysmorphic disorder. Retrieved from
https://www.nice.org.uk/guidance/cg31
O'Kearney, R. T., Anstey, K. J., Von Sanden, C. & Hunt, A.(2006). Behavioural and
cognitive behavioural therapy for obsessive compulsive disorder in children and
adolescents. Cochrane database of systematic reviews(4), CD004856.
doi.org/10.1002/14651858.CD004856.pub2
st, L.-G. (2008). Efficacy of the third wave of behavioral therapies: A systematic review and
meta-analysis. Behaviour Research and Therapy, 46(3), 296-321.
doi.org/10.1016/j.brat.2007.12.005
35

st, L. G. (2014). The efficacy of Acceptance and Commitment Therapy: an updated

systematic review and meta-analysis. Behaviour Research and Therapy, 61, 105-121.
doi.org/10.1016/j.brat.2014.07.018
st, L.-G., Havnen, A., Hansen, B., & Kvale, G. (2015). Cognitive behavioral treatments of
obsessive-compulsive disorder. A systematic review and meta-analysis of studies
published 19932014. Clinical Psychology Review, 40, 156-169.
doi.org/10.1016/j.cpr.2015.06.003
*Pediatric OCD Treatment Study Team (2004). Cognitive-behavior therapy, sertraline, and
their combination for children and adolescents with obsessive-compulsive disorder:
the Pediatric OCD Treatment Study (POTS) randomized controlled trial. JAMA, 292,
1969-1976. doi.org/10.1001/jama.292.16.1969
*Peris, T. S., & Piacentini, J. (2013). Optimizing treatment for complex cases of childhood
obsessive compulsive disorder: A preliminary trial. Journal of Clinical Child and
Adolescent Psychology, 42, 1-8. doi.org/10.1080/15374416.2012.673162
Piacentini, J., Bergman, R. L., Keller, M., & McCracken, J. (2003). Functional impairment in
children and adolescents with obsessive-compulsive disorder. Journal of Child and
Adolescent Psychopharmacology, 13, (supplement 1), 61-69.
doi.org/10.1089/104454603322126359
*Piacentini, J., Bergman, R. L., Chang, S., Langley, A., Peris, T., Wood, J. J., & McCracken,
J. (2011). Controlled comparison of family cognitive behavioral therapy and
psychoeducation/relaxation training for child obsessive-compulsive disorder. Journal
of the American Academy of Child & Adolescent Psychiatry, 50(11), 1149-1161.
doi.org/10.1016/j.jaac.2011.08.003
Rasmussen, S. A. (1990). Epidemiology of obsessive compulsive disorder. Journal of Clinical
Psychiatry, 51(Suppl S), 10-13. http://www.ncbi.nlm.nih.gov/pubmed/2404965
*Reynolds, S. A., Clark, S., Smith, H., Langdon, P. E., Payne, R., Bowers, McIlwham, H.
(2013). Randomized controlled trial of parent-enhanced CBT compared with
individual CBT for obsessive-compulsive disorder in young people. Journal of
Consulting and Clinical Psychology, 81, 1021-1026. doi.org/10.1037/a0034429
*Riddle, M. A., Scahill, L., King, R. A., Hardin, M. T., Anderson, G. M., Ort, S. I., Cohen,
D. J. (1992). Double-blind, crossover trial of fluoxetine and placebo in children and
adolescents with obsessive-compulsive disorder. Journal of the American Academy of
Child & Adolescent Psychiatry, 31, 1062-1069. doi.org/10.1097/00004583199211000-00011
*Riddle, M. A., Reeve, E. A., Yaryura-Tobias, J. A., Yang, H. M., Claghorn, J. L., Gaffney,
G., Walkup, J. T. (2001). Fluvoxamine for children and adolescents with obsessivecompulsive disorder: A randomized, controlled, multicenter trial. Journal of the
American Academy of Child & Adolescent Psychiatry, 40, 222-229.
doi.org/10.1097/00004583-200102000-00017
Snchez-Meca, J., Rosa-Alczar, A. I., Iniesta-Seplveda, M., & Rosa-Alczar, . (2014).
Differential efficacy of cognitive-behavioral therapy and pharmacological treatments
for pediatric obsessivecompulsive disorder: A meta-analysis. Journal of Anxiety
Disorders, 28(1), 31-44. doi.org/10.1016/j.janxdis.2013.10.007
Scahill, L., Riddle, M. A., McSwiggin-Hardin, M., Ort, S. I. (1997). Children'sYale-Brown
obsessive compulsive scale: reliability and validity. Journal of the American Academy
of Child & Adolescent Psychiatry 36, 844-852. doi.org/10.1097/00004583199706000-00023
*Simons, M., Schneider, S., & Herpertz-Dahlmann, B. (2006). Metacognitive therapy versus
exposure and response prevention for pediatric obsessive-compulsive disorder.
Psychotherapy and Psychosomatics, 75, 257-264. doi.org/10.1159/000092897
36

Skarphedinsson, G., Hanssen-Bauer, K., Kornr, H., Heiervang, E. R., Landr, N. I.,
Axelsdottir, B., . . . Ivarsson, T. (2014). Standard individual cognitive behaviour
therapy for paediatric obsessivecompulsive disorder: A systematic review of effect
estimates across comparisons. Nordic Journal of Psychiatry, 69(2), 81-92.
doi.org/10.3109/08039488.2014.941395
*Skarphedinsson, G., Weidle, B., Thomsen, P., Dahl, K., Torp, N., Nissen, J., . . . Ivarsson, T.
(2015). Continued cognitive-behavior therapy versus sertraline for children and
adolescents with obsessivecompulsive disorder that were non-responders to
cognitive-behavior therapy: a randomized controlled trial. European Child &
Adolescent Psychiatry, 24(5), 591-602. doi.org/10.1007/s00787-014-0613-0
Stewart, S. E., Geller, D. A., Jenike, M., Pauls, D., Shaw, D., Mullin, B., & Faraone, S. V.
(2004). Long-term outcome of pediatric obsessive-compulsive disorder: a metaanalysis and qualitative review of the literature. Acta Psychiatrica Scandinavica,
110(1), 4-13. doi.org/10.1111/j.1600-0447.2004.00302.x
*Storch, E. A., Bussing, R., Small, B. J., Geffken, G. R., McNamara, J. P., Rahman, O.,
Murphy, T. K. (2013). Randomized, placebo-controlled trial of cognitive-behavioral
therapy alone or combined with sertraline in the treatment of pediatric obsessivecompulsive disorder. Behaviour Research and Therapy, 51, 823-829.
doi.org/10.1016/j.brat.2013.09.007
*Storch, E. A., Caporino, N. E., Morgan, J. R., Lewin, A. B., Rojas, A., Brauer, L.,
Murphy, T. K. (2011). Preliminary investigation of web-camera delivered cognitivebehavioral therapy for youth with obsessive-compulsive disorder. Psychiatry
Research, 189(3), 407-412. doi.org/10.1016/j.psychres.2011.05.047
*Storch, E. A., Geffken, G. R., Merlo, L. J., Mann, G., Duke, D., Munson, M., . . . Goodman,
W. K. (2007). Family-based cognitive-behavioral therapy for pediatric obsessivecompulsive disorder: Comparison of intensive and weekly approaches. Journal of the
American Academy of Child & Adolescent Psychiatry, 46(4), 469-478.
doi.org/10.1097/chi.0b013e31803062e7
*Storch, E. A., Murphy, T. K., Goodman, W. K., Geffken, G. R., Lewin, A. B., Henin, A.,
Geller, D. A. (2010). A preliminary study of d-cycloserine augmentation of cognitivebehavioral therapy in pediatric obsessive-compulsive disorder. Biological Psychiatry,
68(11), 1073-1076. doi.org/10.1016/j.biopsych.2010.07.015
Thulin, U., Svirsky, L., Serlachius, E., Andersson, G., & st, L.-G. (2014). The effect of
parent involvement in the treatment of anxiety disorders in children: A meta-analysis.
Cognitive Behaviour Therapy, 43(3), 185-200.
doi.org/10.1080/16506073.2014.923928
*Turner, C. M., Mataix-Cols, D., Lovell, K., Krebs, G., Lang, K., Byford, S., & Heyman, I.
(2014). Telephone cognitive-behavioral therapy for adolescents with obsessivecompulsive disorder: A randomized controlled non-inferiority trial. Journal of the
American Academy of Child & Adolescent Psychiatry, 53(12), 1298-1307.
doi.org/10.1016/j.jaac.2014.09.012
Watson, H. J., & Rees, C. S. (2008). Meta-analysis of randomized, controlled treatment trials
for pediatric obsessive-compulsive disorder. Journal of Child Psychology and
Psychiatry, 49(5), 489-498. doi.org/10.1111/j.1469-7610.2007.01875.x
*Williams, T. I., Salkovskis, P. M., Forrester, L., Turner, S., White, H., & Allsopp, M. A.
(2010). A randomised controlled trial of cognitive behavioural treatment for obsessive
compulsive disorder in children and adolescents. European Child & Adolescent
Psychiatry, 19, 449-456. doi.org/10.1007/s00787-009-0077-9

37

Zohar, A. H. (1999). The epidemiology of obsessive-compulsive disorder in children and

adolescents. Child and Adolescent Psychiatric Clinics of North America, 8(3), 445460. ncbi.nlm.nih.gov/pubmed/10442225

38

References identified by
literature search: 674

After removal of duplicates:

551 abstracts
Excluded based on abstract: 488

Full-text articles retrieved: 63

Excluded: 29
-OCD not principal diagnosis: 7
-Not a RCT: 5
-Not using CY-BOCS: 4
-Secondary analysis: 6
-Not testing CBT or SRI: 2
-Not treating children: 3
-Various: 2

Included: 34 RCTs

Fig. 1. Flowchart of the inclusion of studies.

39

Table 1
Effect sizes (Hedges g) on CY-BOCS for all OCD RCTs and divided on comparison conditions for posttreatment assessments.
Q-value

I2

5.31d

192.7d

78

0.25-0.80

3.74d

174.7d

83

0.48

0.26-0.70

4.36d

14.1

43

0.80

0.48-1.11

4.97d

0.3

23

0.21

-0.03-0.45

1.71

74.3d

70

CBT vs. SRI

0.22

-0.08-0.52

1.46

4.7

15

CBT vs. Combo

-0.14

-0.41-0.14

-0.97

4.0

CBT1 vs. CBT2

0.03

-0.27-0.32

0.16

12.2

43

CBT vs. Placebo

0.93

0.47-1.38

3.96d

13.6b

71

CBT vs. WLC

1.53

0.95-2.10

5.22d

27.1

78

SRI vs. Placebo

0.51

0.31-0.71

4.98d

10.9

36

Comparison

g-value

95% CI

All studies

42

0.52

0.33-0.72

All CBT

30

0.53

All SRI

Combo
CBT vs. active Tx

z-value

________________________________________________________________________________________
Note: k = number of comparisons. A positive g-value means that the first treatment in the comparison is better
a
b
c
d
and a negative that the second is better. p<0.05, p<0.01, p<0.001, p<0.0001

40

Table 2
Moderator analyses of the overall effect size of OCD RCTs at post-treatment.
Variable

Point est.

Mean age at treatment

34

-0.0098 -4.01

31

0.0012

3 child studies deleted

Pre-treatment CY-BOCS score

34

z-value

p-value
0.001

0.31

0.759

-0.0064 -2.26

0.024

25

-0.0187

-5.36

0.001

0.0280

2.46

0.014

SRI-studies

0.0160

2.89

0.004

Methodology scores CBT

25

CBT-studies
CBT vs. placebo

0.0573 3.30

0.001

Methodology scores SRI

-0.0086 -0.45

0.653

Risk of bias (all studies)

34

0.0038 1.16

0.245

Percent declining participation 28

-0.0011 -3.91

0.001

Percent females

33

0.0054 0.90

0.367

Percent total comorbidity

29

0.0034 1.83

0.067

27

0.0076

Percent on current drug Tx (CBT)

24

0.0007 0.43

0.667

Weeks of treatment

34

-0.0254 -1.95

0.051

33

-0.0166

25

0.0087 0.22

Percent anxiety comorbidity

1 outlier modified

Number of sessions (CBT)

3.90

-1.03

0.302

0.825

Total hours of Tx (CBT)

25

-0.0015 -1.08

0.281

Intensity (CBT)

25

-0.0098 -1.38

0.167

Attrition total

0.001

34

-0.0005 -1.31

0.189

Attrition primary Tx

34

-0.0005 -1.19

0.233

Attrition comparison

34

-0.0005 -1.34

0.179

41

Table 3
Subgroup analyses of the overall effect size of OCD RCTs at post-treatment.
Variable

95% CI

Type of comparison
Active treatment
Passive control 9

33
1.21

0.36
0.18-0.53
0.61-1.81

Type of data analysis

Completer
Intent-to-treat

4
38

0.53
0.52

0.08-0.99
0.31-0.72

Treatment format (CBT)

Individual
Family

20
9

0.42
0.72

0.15-0.68
0.12-1.32

Degree of parental involvement (CBT)

Low
7
Moderate
18
High
7

0.78
0.40
0.66

0.40-1.17
0.05-0.76
0.03-1.28

Assessment of adherence (CBT)

Yes
13
No
20

0.65
0.48

0.14-1.16
0.22-0.74

Qb-value

p-value
7.17

0.007

0.00

0.961

0.81

0.369

2.06

0.356

0.33

0.566

Assessment of competence (CBT)

0.004
0.951
Yes
3
0.57 -0.40-1.54
No
30
0.54
0.27-0.81
_______________________________________________________________________
Note: k = number of comparisons, Qb = Q between subgroups.

42

Table 4
Proportions of response and remission at post-treatment.
Comparison

Mea-

sure
CBT

SRI

CBT+SRI

Placebo

WLC

Propor-

95% CI

z-value*

Q-value

I2

tion

Resp.

20

69.6%

61.7-76.5

4.64d

37.6b

50

Rem.

22

52.7%

46.0-59.3

0.80

41.3b

49

Resp.

12

48.9%

40.7-57.1

-0.27

30.1b

63

Rem.

24.1%

14.2-37.9

-3.45c

0.2

Resp.

66.1%

56.1-74.9

3.09b

2.6

Rem.

49.1%

31.8-66.6

-0.10

8.5

53

Resp.

10

29.3%

23.5-35.9

-5.74d

16.4

45

Rem.

14.7%

7.2-27.9

-4.26d

3.2

Resp.

13.0%

3.3-39.7

-2.51a

0.3

Rem.

10.1%

3.8-24.0

-4.15d

1.0

_______________________________________________________________________________________
Note: k = number of comparisons, *Test if significantly different from 50%, Resp. = response, Rem. = remission.
a
b
c
d
p<0.05, p<0.01, p<0.001, p<0.0001.

43

Table 5
Within-group effect sizes (Hedges g) for all treatment conditions at post-treatment and follow-up
assessments.
k

g-value

95% CI

z-value

Q-value

I2

Post

74

1.35

1.24-1.47

23.18d

611.2d

88

F-up

24

2.66

2.31-3.01

14.75d

62.7d

80

Post

32

2.43

2.17-2.69

18.28d

87.7d

66

F-up

20

2.68

2.29-3.06

13.72d

53.6d

65

Post

16

1.50

1.30-1.70

15.04d

33.0b

58

F-up

1.71

0.44-2.98

2.64b

Post

2.54

1.87-3.22

7.39d

26.4d

73

F-up

3.58

2.21-4.94

5.13d

3.5

43

Placebo*

Post

13

0.84

0.64-1.03

8.22d

39.7b

70

WLC*

Post

0.22

-0.13-0.57

1.22

10.1a

60

Comparison

Time
point

All conditions

CBT

SRI

Comb.

________________________________________________________________________________________
a
b
c
d
Note: k = number of comparisons, p<0.05, p<0.0001, p<0.0001, p<0.0001, * No follow-up assessment.

44