Terapia Grupal en TAS

Journal of Anxiety Disorders 39 (2016) 44–64
Contents lists available at ScienceDirect
Journal of Anxiety Disorders
Review
Efficacy of group psychotherapy for social anxiety disorder: A

meta-analysis of randomized-controlled trials
Sarah Barkowski a , Dominique Schwartze a , Bernhard Strauss a , Gary M. Burlingame b ,
Jürgen Barth c , Jenny Rosendahl a,∗
a
Jena University Hospital, Institute of Psychosocial Medicine and Psychotherapy, Stoystrasse 3, 07740 Jena, Germany
b
Brigham Young University, Department of Psychology, 238 TLRB, Provo, UT 84602, USA
c
University Hospital Zurich and University of Zurich, Institute for Complementary and Integrative Medicine, Rämistrasse 100, 8091 Zurich, Switzerland
a r t i c l e i n f o a b s t r a c t
Article history: Group psychotherapy for social anxiety disorder (SAD) is an established treatment supported by find-
Received 13 May 2015 ings from primary studies and earlier meta-analyses. However, a comprehensive summary of the
Received in revised form 15 January 2016 recent evidence is still pending. This meta-analysis investigates the efficacy of group psychotherapy for
Accepted 8 February 2016
adult patients with SAD. A literature search identified 36 randomized-controlled trials examining 2171
Available online 10 February 2016
patients. Available studies used mainly cognitive-behavioral group therapies (CBGT); therefore, quanti-
tative analyses were done for CBGT. Medium to large positive effects emerged for wait list-controlled
Keywords:
trials for specific symptomatology: g = 0.84, 95% CI [0.72; 0.97] and general psychopathology: g = 0.62,
Social anxiety disorder
Group psychotherapy
95% CI [0.36; 0.89]. Group psychotherapy was also superior to common factor control conditions in alle-
Randomized controlled trials viating symptoms of SAD, but not in improving general psychopathology. No differences appeared for
Format equivalence direct comparisons of group psychotherapy and individual psychotherapy or pharmacotherapy. Hence,
Meta-analysis group psychotherapy for SAD is an efficacious treatment, equivalent to other treatment formats.
© 2016 Elsevier Ltd. All rights reserved.
Contents
1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
2. Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
2.1. Selection of studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
2.2. Data collection and management . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
2.2.1. Coding of outcomes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
2.2.2. Coding of moderators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
2.2.3. Coding of risk of bias . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
2.3. Summary measures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
2.4. Data synthesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
3. Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
3.1. Study inclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
3.2. Characteristics of studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
3.3. Risk of bias within studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
3.4. Group psychotherapy vs. wait list control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
3.4.1. Specific symptomatology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
3.4.2. General psychopathology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
3.5. Group psychotherapy vs. active treatment conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
3.5.1. Specific symptomatology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
3.5.2. General psychopathology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
3.6. Inclusion of non-CBT studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
∗ Corresponding author. Fax: +49 3641 936546.

E-mail address: jenny.rosendahl@med.uni-jena.de (J. Rosendahl).
http://dx.doi.org/10.1016/j.janxdis.2016.02.005
0887-6185/© 2016 Elsevier Ltd. All rights reserved.
S. Barkowski et al. / Journal of Anxiety Disorders 39 (2016) 44–64 45
3.7. Follow-up . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
3.8. Drop-out . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
4. Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
4.1. Limitations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
4.2. Implications for clinical practice and research . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
4.3. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
Funding . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
Appendix A . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
Appendix B . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
Appendix C . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
1. Introduction elicit, as well as from the experience of mutual support (Tefikow

et al., 2012; Yalom & Leszcz, 2005). Therefore, this format of therapy
Social anxiety disorder (SAD) is one of the most common men- administration for SAD patients has received some attention, with
tal disorders in Western countries, showing a lifetime prevalence usually positive appraisal of group therapy in literature reviews
between 7% and 13% (Furmark, 2002). Typically, it occurs early in (Burlingame et al., 2004; Heimberg, 2002).
life and turns into a chronic condition if no treatment is sought Furthermore, several meta-analyses on psychological treatment
(Wittchen & Fehm, 2003). The core symptomatology is character- for SAD included group treatment in their analyses (Acarturk,
ized by a persistent fear and avoidance of social and performance Cuijpers, van Straten, & de Graaf, 2009; Aderka, 2009; Mayo-Wilson
situations (see DSM-5 criteria, American Psychiatric Association, et al., 2014; Powers, Sigmarsson, & Emmelkamp, 2008) and there
2013). This often leads to considerable impairment because it inter- is one meta-analysis that specifically focused on the controlled
feres with social and occupational participation (Dahl & Dahl, 2010; efficacy of CBGT (Wersebe, Sijbrandij, & Cuijpers, 2013). While
Fehm, Pelissolo, Furmark, & Wittchen, 2005). Consequently, SAD is group psychotherapy was consistently evaluated as being supe-
also associated with high economic expenses primarily stemming rior to wait list control groups with effect sizes ranging from 0.64
from indirect non-medical costs such as lower productivity and (Wersebe et al., 2013) to 0.92 (Mayo-Wilson et al., 2014), the evi-
absenteeism (Acarturk, Smit et al., 2009; Lipsitz & Schneier, 2000). dence on comparisons to active control conditions and alternative
Therefore, providing efficient treatment options should be of high treatment approaches is still scarce and provides an unclear pic-
priority (Egger et al., 2015). ture. Aderka (2009) found higher effect sizes for studies treating
Psychotherapy in a group format constitutes one of the patients in an individual format as compared to the group format,
treatments available for individuals suffering from SAD. Most while other meta-analyses did not detect significant differences
prominent in the research literature is cognitive-behavioral group (Acarturk, Cuijpers et al., 2009; Powers et al., 2008). Wersebe et al.
therapy (CBGT), usually consisting of exposure, cognitive restruc- (2013) reported significant positive effects of group therapy in
turing and homework assignment. Individual components of the comparison to active control groups such as placebo and treatment-
cognitive-behavioral therapy (CBT) approach, such as exposure, as-usual.
are administered separately as well. The CBGT model suggested The most recent review by Mayo-Wilson et al. (2014) regarded
by Heimberg and colleagues (Heimberg, 1991; Heimberg & Becker, CBGT in a network meta-analysis and found that it is superior
2002) in particular, has been investigated in a number of con- to wait list control groups, while no differences to psychological
trolled clinical trials and has gained widespread acceptance as the placebo or other forms of psychological treatment were detected.
gold standard group treatment for SAD Burlingame, MacKenzie, Consequently, individual CBT, which, in contrast to group therapy,
& Strauss, 2004; Hofmann & Bögels, 2006). In addition, the cog- was superior to psychological placebo and some other psychologi-
nitive model by Clark and Wells (1995) has been adapted to cal interventions, was recommended as first-line treatment in the
the group setting (e.g., Stangier, Heidenreich, Peitz, Lauterbach, study (as well as in the NICE clinical guidelines; NICE, 2013). How-
& Clark, 2003) as have non-CBT approaches, such as interper- ever, it should be considered that no statistical difference between
sonal psychotherapy (Huang & Liu, 2011) and psychodynamic the two delivery formats of CBT was detected.
therapy (Gawlytta, Tefikow, & Strauss, 2013; Knijnik, Kapczinski, Although these results from comprehensive meta-analyses
Chachamovich, Margis, & Eizirik, 2004). exist, there is a paucity of meta-analyses on SAD, which specifically
These options present important treatment alternatives to indi- aim to evaluate treatment delivered in a group format. These would
vidual psychotherapy, as group psychotherapy can potentially be beneficial considering that comprehensive reviews usually have
provide a more economic way of treatment (c.f. the studies of Otto, to limit their analyses and therefore leave some specific questions
Pollack, and Maki, 2000, and Roberge, Marchand, Reinharz, and unanswered. For example, the aim of the network meta-analysis
Savard, 2008, related to panic disorders). Because the time needed by Mayo-Wilson et al. (2014) was to evaluate and compare com-
per patient is reduced, group psychotherapy can facilitate access to monly used treatment options. However, the effect of the specific
treatment for a greater number of individuals. delivery format was only investigated for full CBT treatments, while
Although the specific characteristics of SAD seem to conflict effects of other psychotherapeutic approaches (e.g., exposure ther-
with the group environment (i.e., groups elicit anxiety and are apy alone and psychodynamic therapy) were not differentiated by
therefore avoided), additional therapeutic factors are introduced format. Additionally, all inferences made on the differential effects
by treating patients in the specific setting. These include for exam- of group psychotherapy and other active treatments (Acarturk,
ple in-session exposure to social situations and the ability to work Cuijpers et al., 2009; Aderka, 2009; Mayo-Wilson et al., 2014;
with a wider range of interpersonal behavior (compare Heimberg Powers et al., 2008) were derived from information that stemmed
& Becker, 2002; Sank & Shaffer, 1984; Tefikow, Bormann, & Strauss, at least partly from between-study comparisons and might be con-
2012). Socially anxious patients might especially benefit from the founded by differing study characteristics.
sense of belonging and self-validation that well-functioning groups
46 S. Barkowski et al. / Journal of Anxiety Disorders 39 (2016) 44–64
It should also be noted that only two reviews considered het- d. Comparison to wait list or an active treatment or control group
erogeneity between group psychotherapy studies. However, one of (common factor control group, individual psychotherapy, or
them included studies published before 2007 only, leaving out any pharmacotherapy). Analogous to Baskin, Callen Tierney, Minami,
recent studies (Acarturk, Cuijpers et al., 2009) while the other was and Wampold’s (2003) definition of placebo treatment, com-
limited to studies comparing CBGT to a control condition (Wersebe parison conditions were considered as “common factor control
et al., 2013). groups” if they were “used in an attempt to control for psy-
Therefore, our aim was to provide a comprehensive review of chotherapy’s common factors by providing a treatment devoid of
the efficacy of group psychotherapy for SAD, concentrating on the specific ingredients” (Baskin et al., 2003; p. 975). This included,
specific format of treatment delivery. We raise questions tailored for example, supportive group treatments or relaxation groups.
to the group format by exploring the influence of specific setting Self-help and internet-based interventions were not eligible as
variables and the treatment type, and discussing heterogeneity comparison conditions.
between group psychotherapy studies in general. By aligning our e. Patient assignment to treatment condition was random. Cluster
inclusion criteria to the specific research question, we expect to randomization was eligible, as well as consecutive assignment if
be able to include more ample evidence on the particular therapy the first assignment was random.
option. f. The study reported outcomes of SAD symptomatology or general
Standardized mean differences will be provided for wait list- psychopathology, such as depression or anxiety for at least one
controlled studies as well as comparisons to active treatment time point after treatment completion.
alternatives such as common factor control groups, individual psy-
chotherapy, and pharmacotherapy. Between-study heterogeneity In the initial study protocol, we additionally intended to inves-
and publication bias will be carefully considered, and explorative tigate economic outcomes and to compare group psychotherapy to
moderator analyses focusing on methodological variables and spe- standard care (treatment-as-usual). However, none of the studies
cific characteristics of the group psychotherapy will be conducted. provided the required data.
Objectives, inclusion criteria, and methods were pre-specified in The databases CENTRAL, Medline (Ovid), PsycINFO, and Web of
a review protocol (Rosendahl, Barkowski, Schwartze, Koranyi, & Science were searched according to a search strategy that specified
Strauss, 2013). terms referring to the patient population (e.g., phobi*, social* anxi*),
treatment (e.g., group therap*, group-delivered), and study design
(e.g., random*, control group$1). The search strategy was developed
2. Methods
with consideration of validated search strategies for retrieving ran-
domized controlled trials (Lefebvre, Manheimer & Glanville, 2011).
2.1. Selection of studies
The database searches were last updated on August 19, 2015. It
was complemented by manual searches of the reference sections
All randomized-controlled trials were considered for inclusion,
of recent reviews and meta-analyses published on the efficacy of
regardless of publication status and language. Study inclusion cri-
psychological interventions for anxiety disorders, as well as by
teria were specified according to the PICOS guideline,1 proposed by
searching references of included primary studies. In order to iden-
the PRISMA statement for preferred reporting items for systematic
tify unpublished studies, we used the ProQuest Dissertations and
reviews and meta-analyses (Moher, Alessandro, Tetzlaff, & Altman,
Theses Full Text Database and contacted the authors of relevant
2009). The following eligibility criteria were applied:
studies asking for information about unpublished data. The final
search strategy for Medline as well as the list of manually searched
a. Publication date from 1990 onwards. The criterion was adopted reviews and meta-analyses can be retrieved from the international
in order to ensure the use of current treatment approaches prospective register of systematic reviews, PROSPERO.2
and state-of-the-art methodology and has been applied in pre- Inclusion of studies was conducted in a two-step process, exe-
vious meta-analyses (Westen & Morrison, 2001). Furthermore, cuted by one of the two raters (DS, SB). Studies that were eligible
Burlingame et al. (2004), reported in a systematic review of small according to the information in the title and the abstract were read
group treatment that in this time span an increasing number of in full text before the final decision on inclusion. Discussions in
group psychotherapy studies related to specific disorders was team meetings (with BS and JR) resolved unclear cases.
published and that these studies were methodologically more
rigorous than previously conducted studies. 2.2. Data collection and management
b. Sample of adult patients with a primary diagnosis of SAD accord-
ing to DSM or ICD criteria. Populations of mixed anxiety diagnoses Information was retrieved according to a coding manual
were eligible if the results for SAD patients were reported sepa- (Appendix A, Table A1) including variables on characteristics of the
rately. study, the participants, the intervention condition, the comparison
c. Group treatment, based on a formal change theory postulating condition and the outcome, as well as data necessary to calculate
psychological principles as the agent of change. One or more ther- effect sizes and the rating for methodological quality of the stud-
apists led groups of at least three patients, providing five or more ies. The variables describe the data on different levels of the study
sessions. Only interventions aiming at changing the primary psy- design. We refer to the study level with a lower case n, to the level
chopathology of SAD were considered. We excluded conditions of comparisons between a group psychotherapy and a comparison
in which group psychotherapy was provided in combination with condition with a lower case k and to the level of single experi-
study-administered concurrent individual or pharmacological mental conditions (i.e., group psychotherapies and/or comparison
treatment. However, it was not required that studies prohibited conditions) with a lower case i.
patients to take psychopharmacological medication prescribed In a pilot phase, five studies were double-coded independently
outside the study regimen. by the first and second author (DS, SB). An analysis of inter-rater
agreement on these studies yielded a mean Cohen’s Kappa of
1
The PICOS guideline defines participants, interventions, comparisons,
2
outcomes, and study design as important domains to be considered for designing Available from http://www.crd.york.ac.uk/PROSPERO/display record.
inclusion criteria and addressing important questions in the review. asp?ID=CRD42013004419.
= 0.88 for categorical variables with a range between = 0.66 and implementation quality was given if studies ensured treatment
= 1.0 and a mean single-score intra-class correlation coefficient completion according to the protocol, either by providing specific
of ICC = 0.98 for metric variables with a range between ICC = 0.67 training and regular supervision or by conducting adherence checks
and ICC = 1.0. These suggest good to excellent interrater agreement for fidelity to a manual.
(Altman, 1991; Cicchetti, 1994). Reasons for discrepancies were dis- We assigned categories to each of the risk of bias variables, indi-
cussed, a consensus found and, where necessary, the instructions cating whether a study had a low risk of bias (i.e., implemented
in the coding manual were adapted. Data of all further studies were adequate measures to minimize the risk of bias) or a high risk of
obtained by a single coder (DS or SB). Ambiguous cases (i.e., cases bias on the respective variable. Studies that did not give sufficient
to which the pre-specified criteria did not apply) were brought for- information to make an informed decision were rated unclear. Two
ward by the respective coder and were discussed between the two separate independent coders (a student assistant at master’s level
coders. If no consensus was found, a decision was made in team and JR) rated methodological quality in duplicate for all studies.
meetings (with BS and JR). If none of the coders was familiar with Consensus across all studies was excellent with a mean Cohen’s
the publication language, we consulted a translator and conducted Kappa of = 0.90 and a range between = 0.80 and = 1.0. Risk of
the rating cooperatively. Since no translator for Farsi was available bias was examined within moderator analyses and subgroup analy-
on the expense of reasonable resources, one study was excluded ses differentiating between studies without any judgement of high
(Dadashzadeh, Yazdandoost, Gharraee, & Asgharnejad farid, 2012). risk of bias and all other studies.
2.2.1. Coding of outcomes 2.3. Summary measures

As a primary outcome domain, we were interested in the
specific symptomatology of SAD. Outcome measures were sum- Between-group effect sizes Hedges’ g (Hedges, 1981) were cal-
marized under this domain provided they assessed symptoms that culated within the Comprehensive Meta-Analysis software (CMA,
are distinctive of SAD, such as fear of public speaking (compare Biostat, Inc., Version 2) for reported outcomes that matched one of
Appendix B, Table B1 for allocation of single measures to domains). the selected outcome domains. Raw population data (M, SD, N) were
As a secondary outcome domain, general psychopathology was preferred over test results (p-value, F-value, etc.) and, where pos-
investigated. This domain addresses all outcomes measuring psy- sible, pre-test data were gathered and implemented in the effect
chological symptoms not specific to the diagnosis of SAD, such as size calculation. If available, intention-to-treat data were chosen
depression and general anxiety. All outcomes derived from assess- over completer data. The magnitude of Hedges’ g was interpreted
ments with published measures were included in the analyses, within the same framework as Cohen’s d, regarding 0.20, 0.50, and
regardless of whether they were self- or observer-reported. One of 0.80 as small, medium, and large effect sizes, respectively (Cohen,
the two raters (DS or SB) and JR double-coded the type of extracted 1992). Positive effect sizes indicate a superiority of the group psy-
outcome and the type of statistical parameters. Considered time chotherapy, while negative effect sizes suggest a superiority of
points were post-treatment (first assessment after the last ses- the comparison condition. Whenever effect size information was
sion and within one week of program termination), short-term missing but a reasonable approximation was possible (e.g., by mea-
follow-up (≤6 months after the intervention), mid-term follow-up suring means and standard deviations from figures or estimating
(6–18 month after the intervention), and long-term follow-up (≥18 post-treatment sample sizes from pre-treatment sample sizes),
months after the intervention). approximated values were used. Furthermore, all non-significant
effects reported without statistical parameters were entered into
2.2.2. Coding of moderators the analyses by setting effect sizes to zero. We employed this
For moderator analyses, we focused on the effect of various approach to ensure consideration of all available information.
characteristics of group psychotherapy on treatment outcome. Since the majority of studies reported multiple outcomes per
We planned to differentiate between the psychotherapeutic outcome domain, effect sizes were subsequently aggregated within
approaches. Similar to a dismantling procedure, full CBGT was com- domains for each unit of analysis (i.e., comparison of group psy-
pared to group exposure alone. This distinction has already been chotherapy and comparison condition) using the “agg”-function
made in previous meta-analyses (e.g., Feske & Chambless, 1995) provided by the “MAd” package version 0.8-1 (Del Re & Hoyt, 2014)
since the two treatments are the most frequently evaluated in for R statistical computing language and environment version 3.1.1
the relevant literature. Number of group psychotherapy sessions, (R Core Team, 2014). The agg-function accounts for between-
length of group psychotherapy sessions, number of sessions per outcome correlations within one study and estimates the variance
week, and the number of patients per group were investigated as of the aggregate effect size accordingly (Borenstein, 2009; Gleser &
continuous variables. Further categorical moderator variables were Olkin, 1994; cited by Del Re and Hoyt, 2014). If correlations between
therapist profession (professional vs. student), type of leadership outcomes were not given by study authors, the correlations were
(single-led vs. co-led), manualization of the treatment (manualized taken from validation studies or were set to 0.50 (for further infor-
with adherence check vs. manual- or model-based but no adher- mation on this approach, compare Wampold et al., 1997). Thus, only
ence check), and presence of an individual preparatory session (yes one effect size estimate per comparison was added to the analysis
vs. no). of a specific outcome domain.
2.2.3. Coding of risk of bias 2.4. Data synthesis

Studies were evaluated for presence of a risk of bias threaten-
ing internal validity using the Cochrane Risk of Bias Tool (Higgins, All eligible comparisons from included studies were considered
Altman, & Sterne, 2011). We assessed risks of selection bias (ade- for analysis. If a study provided for multiple comparisons and one
quate randomization procedure and concealment of allocation), of the reported experimental groups was shared between several
reporting bias (complete outcome reporting), detection bias (blind- comparisons, we divided the sample size of this group by the num-
ing of outcome assessors for observer-reported outcomes), and ber of times it was included (Higgins, Deeks, & Altman, 2011). All
attrition bias (handling incomplete outcome data). In addition, bias other statistical parameters remained constant.
introduced by the implementation quality of the group therapy We computed pooled standardized mean differences separately
condition was evaluated according to a rating proposed by Barth for comparisons to wait list control groups and for comparisons
et al. (2011) for quality ratings in psychotherapy studies. A high to active comparison conditions. The latter was further stratified
by type of comparison (common factor control group, individual initial screening of the title and the abstract, a full-text assessment
psychotherapy, and pharmacotherapy). was conducted for 610 articles, which resulted in the inclusion of 36
Analyses were conducted in R, version 3.1.1 (R Core Team, studies fulfilling all pre-specified inclusion criteria. These studies
2014) for all subgroups and outcome categories, using the were reported on in 40 articles altogether. As the database search
package “metafor,” version 1.9-4 (Viechtbauer, 2010b). We per- was designed to extract studies on anxiety disorders in general, 19
formed all meta-analyses according to the random-effects model studies related to panic disorder with or without agoraphobia or
(Hedges & Vevea, 1998) and estimated heterogeneity with the generalized anxiety disorder were eliminated in the last step and
DerSimonian–Laird-method (DerSimonian & Laird, 1986). Pooled are listed accordingly within the study flow (Fig. 1).
standardized mean differences Hedges’ g are reported along with
their 95%-confidence intervals and the p-value. Heterogeneity was
tested with the Q-statistic and quantified using I2 , which gives
the amount of between-study variability that cannot be explained 3.2. Characteristics of studies
by chance alone. I2 values of 25%, 50% and 75% are commonly
interpreted as low, moderate and high amounts of heterogene- Table 1 contains a summary of the characteristics of all included
ity, respectively (Borenstein, Hedges, Higgins, & Rothstein, 2009; studies. Besides n = 33 studies published in journals, three unpub-
Higgins, Thompson, Deeks, & Altman, 2003). Effect sizes of single lished dissertations were included (Filion-Rosset, 2004; Heideman,
comparisons were considered outliers, and were eliminated from 2008; McDougall, 1999). Included studies reported on k = 50 dif-
further analyses if the effect size’s confidence interval did not over- ferent comparisons of a group psychotherapy to a comparison
lap with the confidence interval of the pooled standardized mean condition, composed of i = 45 group psychotherapies and i = 41
difference (see Cuijpers et al., 2014). comparison conditions. The majority of group psychotherapies
Publication bias was assessed for specific symptomatology via (i = 36, 80.0%) followed the full cognitive-behavioral regimen,
visual examination of the funnel plot, looking for asymmetry that usually incorporating exposure, cognitive restructuring, and home-
might suggest specific omission of non-significant results. Egger’s work assignment. Some groups added supplemental components
regression test was executed for each funnel plot to statistically such as social skills training (Davidson et al., 2004), atten-
analyze the relationship between study effect size and standard tion retraining (Rapee, Abbott, Baillie, & Gaston, 2007), imagery
error (Egger, Smith, Schneider, & Minder, 1997). When funnel plot (McDougall, 1999) or mindfulness and acceptance (Kocovski,
asymmetry was detected, a trim-and-fill analysis was carried out Fleming, Hawley, Huta, & Antony, 2013). Most of these treat-
that added potentially missing studies to the meta-analytic model ments (i = 13) were guided by the treatment protocol developed
(Duval & Tweedie, 2000). by Heimberg and colleagues or adapted versions (Heimberg, 1991;
In order to test whether our findings were biased by decisions Heimberg & Becker, 2002). Two studies (4.4%) followed an adapted
taken during the meta-analytic process, we conducted a number of version of the cognitive therapy model by Clark and Wells (1995).
sensitivity analyses (i.e., meta-analyses on selected subsamples of Other treatment approaches were exposure alone (i = 6, 13.3%),
effect sizes) for specific symptomatology. Two of these were per- social skills training alone (i = 1), interpersonal psychotherapy
formed on subsamples of outcome measures, restricting included (i = 1), and psychodynamic psychotherapy (i = 1, 2.2% each). Further
outcomes to standard self-report measures or non-estimated effect information on characteristics of the group psychotherapies can be
sizes (i.e., effect sizes that were not calculated from approximated found in Table 2. As most of them followed a cognitive-behavioral
parameters or zero effects), respectively. Standard self-report mea- approach or used cognitive-behavioral elements as main ther-
sures for SAD symptomatology were chosen according to relevant apeutic technique, we conducted the statistical meta-analyses
reviews on the subject (Antony, 1997; Hart, Jack, Turk, & Heimberg, including CBT approaches only. However, the two additional stud-
1999: SADS, SPS, SIAS, FQ-SP, SPAI, and (B)FNE; abbreviations ies (one interpersonal and one psychodynamic) were included in
defined in Table B1). Further sensitivity analyses were confined to the review nevertheless and are discussed and analyzed in Section
studies ensuring adherence to the treatment manual, studies ban- 3.6.
ning concurrent psychopharmacological medication treatment and The majority of comparison conditions were wait list control
studies with professional non-student group leaders. groups (i = 24, 58.5%), followed by common factor control groups
We investigated the influence of potential effect modifiers in (i = 6, 14.6%), pharmacotherapies (i = 6, 14.6%), and individual psy-
mixed-effects meta-regression analyses for continuous variables chotherapies (i = 5, 12.2%). Usage of treatment-as-usual was minor
(Viechtbauer, 2010a). In these models, moderator variables are and considerably heterogeneous. Therefore, we allocated the two
entered into regression analyses as potential predictors of the such-labeled comparison conditions consistent with their content
effect size Hedges’ g, and their ability to explain heterogeneity is to pharmacotherapy (Mörtberg, Clark, Sundin, & Aberg Wistedt,
investigated. The models provide beta coefficients and a test of 2007) and common factor control groups (Cottraux et al., 2000),
moderator significance on the Q-statistic. The amount of variance respectively.
explained by the moderator is reported as R2 . Dichotomous mod- Across all trials, 1187 patients were treated with group psy-
erators as well as subgroup analyses were investigated by testing chotherapy and 984 were assigned to a comparison condition. On
for significant differences in effect sizes between two subsets of the average, the percentage of male and female patients was evenly dis-
studies. Again, a mixed-effects model was used by fitting random- tributed (Mdn = 50% female, Interquartile range (IQR) = 14.13) and
effects models separately for the subgroups and comparing the the distribution of mean age across all studies showed a median
resulting effect size estimates within a fixed effects model with of 34 (IQR = 5, n = 35). The most frequent exclusion criteria were a
moderators. diagnosis of psychosis or substance abuse, followed by a diagnosis
of bipolar disorder or depression, suicidality and concurrent psy-
chotherapeutic or medication treatment. Usual comorbidities were
3. Results another anxiety or mood disorder (reported by n = 12 studies, each).
Seven studies reported on comorbid avoidant personality disorder
3.1. Study inclusion and eight on comorbid substance abuse, with one study includ-
ing exclusively patients with SAD and comorbid alcohol problems
A total of 5894 references were identified by means of database (Heideman, 2008). Only one study (Anderson et al., 2013) reported
searches and another 220 references from manual search. After an absence of any comorbidity for the majority of patients (76%).
Table 1
Study characteristics of included studies.
Study Country Population Group treatment n No. of No. of ses- Comparison n Drop-out Recruitment Diagnostic Concurrent Outcome Assessment
patients sions/length treatment (%) strategy instrument medication category times
per group in minutes
Alden Canada Generalized Interpersonal 31 5–7 12/120 WL 25 10.7 Referred/ ADIS-IV Yes, if Spec Post
and SAD cognitive- recruited stable
Taylor behavioral group
(2011) treatment
Anderson USA SAD Exposure group 39 Up to five 8/n.r. a) WL 28 22.7 Referred/ SCID for Yes, if Spec, Gen Post
et al. therapy (Hofmann, recruited DSM-5 stable
(2013) 2004)
b) IP: virtual 30
reality exposure
Bjornsson USA SAD CBGT (shorter 22 5–7 8/120 ComFC: 23 13.3 Recruited SCID No Spec, Gen Post
et al. version of the non-specific
S. Barkowski et al. / Journal of Anxiety Disorders 39 (2016) 44–64

(2011) Heimberg model) elements of
group
psychotherapy
(Yalom & Leszcz,
2005), designed
to be both
credible and
structurally
equivalent
Blanco USA SAD CBGT (Heimberg 34 4–6 12/150 Med: phenelzine 35 30.5 Referred/ n.r. No Spec, Gen Post
et al. model) recruited
(2010)
Cottraux France SAD Group social skills 32 4–6 6/120 ComFC: 35 17.9 Recruited Structured No Spec, Gen Post
et al. training, supportive interview
(2000) administered after therapy,
6 weeks baseline practiced “as
treatment of usual”, after 6
individual weeks of the
cognitive therapy same baseline
treatment
Davidson USA Generalized Comprehensive 60 5–6 14/n.r. Med: fluoxetine 57 28.5 Recruited SCID No Spec, Gen Post
et al. SAD CBGT with
(2004) exposure, cognitive
and restructuring and
Ledley social skills
et al. training
(2005)
D’El Rey Brasil Generalized CBGT (Heimberg 16 ca. 5 12/120 WL 16 3.1 Recruited SCID-I No Spec, Gen Post
et al. SAD model, adapted)
(2008)
Dogaheh Iran SAD CBGT (adapted 14 n.r. 12/n.r. IP: individual 14 21.4 Recruited Clinical n.r. Spec, Gen Post
et al. mainly from cognitive- interview
(2011) Heimberg model) behavioral
treatment
Erickson Canada Mixed CBGT for 14a 9–13 11/120 WL 14a 42.1 Referred SCID-IV Yes Gen Post
et al. anxiety diagnostically
(2007) diagnoses heterogeneous
groups
49
50
Table 1 (Continued)
Filion- Canada SAD CBGT 32 n.r. 13/120 WL 36 34.6 Recruited SCID Yes, if Spec, Gen Post
Rosset stable
(2004)
Furmark Sweden SAD CBGT (Heimberg 6 6 8/180 a) Med: 6 0.0 Recruited SCID No Spec Post
et al. model) citalopram
(2002)
b) WL 6

Gruber USA SAD a) CBGT (Heimberg 18 n.r. 12/150 WL 18 14.8 Recruited ADIS-R Yes, if Spec, Gen Post
et al. model) stable
(2001)
b) Short CBGT 18 n.r. 8/150
facilitated by a
hand-held
computer
Heideman USA SAD and CBGT (Heimberg 9 n.r. 6/60-75 WL 9 22.2 Recruited ADIS-IV, Yes, if Spec Post
(2008) alcohol model), after SCID-I stable
problems attending one
individual BASICS
feedback session
for drinking
behavior
Heimberg USA SAD CBGT 25 4–7 12/120 ComFC: 24 18.4 Recruited ADIS n.r. Spec, Gen Post,
et al. educational- 6-month
(1990, supportive fu, 5-year
1993) psychotherapy fu
comparison
group, credible
placebo control
Heimberg USA SAD CBGT 36 5–7 12/150 a) Med: 31 19.5 Referred/ ADIS-R n.r. Spec, Gen Post
et al. phenelzine recruited
(1998)
b) ComFC: 33
educational-
supportive group
therapy
Hofmann USA SAD a) CBGT (Heimberg 30a 5–7 12/n.r. WL 30a N.r. Referred ADIS-IV-L n.r. Spec Post
(2004) model)
b) Exposure group 30a 5–7 12/n.r.
therapy
Hope USA SAD a) CBGT (Heimberg 18 6–7 12/120-150 WL 11 17.5 Referred ADIS-R Yes, if Spec Post
et al. model) stable
(1995)
b) Exposure alone 11 n.r. 12/120-150
(developed for the
study to match
CBGT)
Huang China SAD a) Group 15 15 8/120 WL 15 6.7 Recruited n.r. n.r. Spec Post
and Liu interpersonal
(2011) psychotherapy
b) CBGT 15 15 8/120

Knijnik Brazil Generalized Psychodynamic 15a 7–8 12/90 ComFC: credible 15a 25.0 Referred/ SCID-I No Spec, Gen Post
et al. SAD group therapy placebo control recruited
(2004) group (Heimberg
et al., 1990)
Kocovski Generalized a) Mindfulness and 53 n.r. 12/120 WL 31 27.0 Recruited SCID Yes, if Spec, Gen Post
et al. SAD acceptance based stable
(2013) group therapy
b) CBGT 53 n.r. 12/120
Mácia- Spain Generalized CBGT (Heimberg 17 8–9 12/150 WL 16 0.0 Recruited ADIS-IV-L No Spec Post
Antòn SAD model)
et al.
(2012)
McDougall Canada SAD a) Cognitive- 20 3–8 10/150 WL 17 27.1 Referred/ ADIS-IV Yes Spec, Gen Post
(1999) behavioral plus recruited
imagery group
treatment
b) CBGT 22 3–8 10/150
Mörtberg Sweden SAD Intensive group 35 6–7 16/a 3-h a) IP: individual 32 28.0 Recruited SCID No Spec, Gen 4-, 8-,
et al. cognitive morning cognitive therapy 12-month
(2007, treatment (adapted session and (Clark & Wells, fu, 5-year
2011) from Clark & Wells, a 2-h 1995) fu
1995) afternoon
session
b) Med: flexible, 33 4-, 8-,
predominantly 12-month
SSRIs fu
Mörtberg Sweden SAD Intensive CBGT 13 6–8 16/154 WL 13 7.7 Referred SCID, Yes Spec, Gen Post, 3-,
et al. SCID-II 6-month fu
(2006)
Newman USA SAD Performance-based 18 6 8/120 WL 18 8.3 Recruited SCID No Spec, Gen Post
et al. exposure
(1994) treatment, speech
phobia manual
51
52
Table 1 (Continued)

Nortje South Generalized a) Cognitive 15 7–8 12/150 WL 14 0.0 n.r. ADIS-IV Yes, if Spec, Gen Post,
et al. Africa SAD restructuring plus stable 3-month fu
(2008) exposure (CR + E)
combined in CBGT
(Heimberg model)
b) Exposure alone 15 7–8 12/150
(EA), taken from
CBGT (Heimberg
model)
Olivares Spain Generalized Intervention in 10 5 12/90 WL 10 0.0 Recruited ADIS-IV-L No Spec Post, 6-,
et al. SAD adolescents with 12-month
(2009) social phobia fu
(CBGT)
Otto, USA SAD CBGT (Heimberg 20 3–8 12/150 Med: 25 33.3 Referred/ SCID for No Spec, Gen Post
Pollack, model) clonazepam recruited DSM-III-R
Gould
et al.
(2000)
Pishyar Australia SAD Standard CBGT, 16 n.r. N.r. WL 16 0.0 Recruited Telephone n.r. Spec, Gen Post
et al. developed for the screening
(2008) study interview
and
follow-up
assessment
Rapee Australia SAD Standard group 59 n.r. 10/120 WL 52 21.0 Referred/ ADIS-IV Yes, if Spec Post
et al. treatment; recruited stable
(2007) components
included cognitive
restructuring of
negative evaluation
beliefs, exposure to
feared social
situations, realistic
feedback of social

performance, and
attention training.
Rapee Australia SAD a) Enhanced 71 Approx. 6 12/120 ComFC: 58 17.9 Referred/ ADIS-IV Yes, if Spec Post
et al. treatment (CBGT non-specific recruited stable
(2009) with additional stress
treatment management
techniques: e.g.
performance
feedback, attention
retraining)
b) Standard 66 Approx. 6 12/120
treatment (CBGT,
purely cognitive
restructuring and
in vivo exposure,
loosely based on
other models, such
as Heimberg)
Salaberría Spain SAD a) Self-exposure 24 n.r. 8/150 WL 23 15.5 Referred/ ADIS-R n.r. Spec, Gen 4-month fu
Irízar and in vivo recruited
Echeburúa
Odriozola
(1995)
b) Self-exposure 24 n.r. 8/150
and cognitive
restructuring
following Beck
Schmidt USA Mixed False Safety 17 4–6 10/120 WL 16 4.2 Self- SCID Yes, if Spec Post
et al. anxiety Behavior presented stable
(2012) diagnoses Elimination
Therapy (F-SET)
53
54
Table 1 (Continued)

Scholing Netherlands Generalized CBGT with three 36 5–7 14/150 IP: cognitive- 37 19.2 Referred/ ADIS-R Yes Spec, Gen Post, 3-,
and SAD treatment options: behavioral recruited 18-month
Emmelkamp exposure alone, individual fu
(1993, cognitive therapy therapy with the
1996) and exposure same three
separated in two treatment
blocks, integrated options
cognitive-
behavioral
treatment
Stangier Germany SAD Group cognitive 26 4–7 15/120 a) IP: individual 24 8.5 Referred/ SCID for No Spec, Gen Post,
et al. therapy based on cognitive therapy recruited DSM-5, 6-month fu
(2003) the model and (Clark & Wells, SCID-II
approach of Clark 1995)
and Wells, 1995
b) WL 21
Wong China SAD CBGT (Heimberg 17 8–9 10/150 WL 17 0.0 Referred/ n.r. Yes Spec Post
and Sun model, adapted for recruited
(2006) improving cultural
relevance)
Note: SAD = social anxiety disorder, n = sample size (randomized if not stated otherwise), CBGT = cognitive-behavioral group therapy, * = evaluated were the data available for our analyses, not if any results on an ITT-based
sample were provided, a = completer sample, WL = wait list, ComFC = common factor control group, IP = individual psychotherapy, Med = medication treatment, SCID = Structured Clinical Interview for DSM, ADIS = Anxiety Disorder
Interview Schedule for DSM, Spec = specific psychopathology, Gen = general psychopathology, Post = post treatment, fu = follow-up.
Fig. 1. Flow chart displaying the process of study inclusion.
Table 2
Characteristics of the group treatment (total i = 45 group treatments).
Median IQR i
No. of patients per group 6 1.4 34

No. of sessions 12 2.5 44
Length of sessions (min) 120 30 39
No. of sessions per week 1 0 41
Category k %
Type of treatment Cognitive-behavioral 36 80.0

Exposure 6 13.3
Interpersonal 1 2.2
Psychodynamic 1 2.2
Social skills training 1 2.2
Therapist profession Students (graduate and undergraduate) 7 15.6

Professional or mixed 31 68.9
n.r. 7 15.6
Leadership Single-led 6 13.3

Co-led 35 77.8
n.r. 4 8.9
Individual session Yes 8 17.8

No/n.r. 37 82.2
Manualization Manualized and adherence checked 28 62.2

Manual-/model-based 17 37.8
Note: n.r. = not reported, IQR = interquartile range, i = number of treatment groups.
3.3. Risk of bias within studies 3.4. Group psychotherapy vs. wait list control
A lack of reporting impeded the evaluation of risk of bias espe- 3.4.1. Specific symptomatology
cially for selection bias and reporting bias where more than half of A meta-analysis of 28 comparisons of group psychotherapy
the studies did not allow for an informed decision. There was little versus wait list control groups revealed an average effect size of
evidence for an implementation bias to be present in the included g = 1.05 (95% CI [0.81; 1.29], p < 0.001). However, heterogeneity
studies with most of the studies fulfilling the requirements for low was high with I2 = 72.9% (Q = 99.79, df = 27, p < 0.001). After elim-
risk of bias. However, we observed a high level of attrition bias; ination of three outliers (D’El Rey, Lacava, Cejkinski, & Mello, 2008;
only a minority of studies reported on intention-to-treat samples Olivares, Rosa-Alcázar, Olivares-Olivares, & Rosa-Alcázar, 2009;
(Fig. 2). Pishyar, Harris, & Menzies, 2008; g > 2.50), the effect size was
3.4.1.2. Clinical and methodological diversity. Because statistical

heterogeneity was zero for comparisons to wait list controls, mod-
erator analyses were not indicated from a statistical viewpoint.
Nevertheless, some important clinical and methodological diver-
sity was assumed between studies, and regarded in subgroup
analyses. Subgroup analyses on differences between full CBGT
(g = 0.83 [0.70; 0.97], p < 0.001, k = 20, Q = 13.27, df = 19, p = 0.825,
I2 = 0.0%) and exposure group treatment alone (g = 0.91 [0.61; 1.20],
p < 0.001, k = 5, Q = 2.33, df = 4, p = 0.676, I2 = 0.0%) revealed no signif-
icant difference (z = −0.46; p = 0.648). Studies without high risk of
bias on any of the risk of bias variables showed a significantly larger
treatment effect (z = 2.06, p = 0.040; g = 1.03 [0.81; 1.25], p = < 0.001,
k = 6, Q = 1.97, df = 5, p = 0.854, I2 = 0.0%) compared to the remain-
ing studies (g = 0.76 [0.61; 0.90], p ≤ 0.001, k = 19, Q = 9.61, df = 18,
p = 0.944, I2 = 0.0%).
Fig. 2. Distribution of risk of bias in included studies.

3.4.2. General psychopathology
Fourteen studies compared the effect of group psychotherapy
reduced to g = 0.84 (95% CI [0.72; 0.97], p < 0.001, k = 25). For results with wait list control groups regarding general symptomatology.
of included studies see Fig. 3. Heterogeneity of the model with- In line with the results on specific symptomatology, a large rela-
out outliers was non-significant, with I2 = 0.0% (Q = 15.80, df = 24, tive effect of group psychotherapy emerged (g = 0.80 [0.43; 1.17],
p = 0.895). p < 0.001, k = 14). Again, heterogeneity was high (Q = 43.36, df = 13,
p < 0.001, I2 = 70.0%) and one outlier was eliminated from further
3.4.1.1. Publication bias and sensitivity analyses. Egger’s regression analyses (Pishyar et al., 2008; g = 3.24). Analyses on the reduced
test for funnel plot asymmetry did not indicate missing studies sample produced a moderate treatment effect of g = 0.62 (95%
(t = −0.72, df = 23, p = 0.477) and neither did visual examination of CI [0.36; 0.89], p < 0.001, k = 13) and a considerably smaller het-
the funnel plot (Appendix C, Fig. C1). Furthermore, the robustness erogeneity (Q = 19.42, df = 12, p = 0.079, I2 = 38.2%). None of the
of results was tested against the influence of decisions undertaken pre-specified moderator variables of treatment setting or risk of
during the meta-analytic process (i.e., study inclusion, selection bias explained further heterogeneity. However, eliminating D’El
of outcomes, effect size calculation). Deductions from results did Rey et al. (2008) (g = 2.12) from the analyses reduced heterogeneity
not change after confining sensitivity analyses to results from in the remaining sample to I2 = 0.0% (Q = 6.71, df = 11, p = 0.22) and
non-estimated effect sizes, group psychotherapies where manual resulted in a standardized mean difference of g = 0.50 (95% CI [0.29;
adherence was adequately ensured and reported, or group psy- 0.71], p < 0.001, k = 12).
chotherapies led by non-student professional therapists (compare Analyses on this sample (k = 12) were stratified by type of treat-
Table 3). Slightly smaller effects resulted from standard self-report ment and high versus low risk of bias. No significant subgroup
scales and for comparisons of studies not allowing for concurrent differences occurred (type of treatment: z = −0.14, p = 0.890; risk
medication, but significance was maintained. of bias: z = 1.38, p = 0.167). Nevertheless, treatment effects were
somewhat larger for exposure treatments (g = 0.54 [−0.00; 1.07],
Fig. 3. Forest plot of study effect sizes and overall standardized mean difference for comparisons to wait list.
Table 3
Statistics of the meta-analyses and sensitivity analyses on specific symptomatology.
Comparisons to wait list control Comparisons to active treatment groups
Heterogeneity Heterogeneity
2
k g 95% CI (g) p(Q) I (%) k g 95% CI(g) p(Q) I2 (%)
Overall 28 1.05 0.81; 1.29 <0.001 72.9 15 0.06 −0.19; 0.31 0.656 76.5
Without outlier 25 0.84 0.72; 0.97 0.895 0.0 14 0.15 −0.03; 0.33 0.012 52.1
Without outlier, including non-CBT 26 0.86 0.74; 0.98 0.774 0.0 15 0.15 −0.02; 0.32 0.097 48.5
Sensitivity analyses
Non-estimated ES 19 0.79 0.65; 0.93 0.241 17.5 8 0.14 −0.12; 0.39 0.022 57.1
Standard self-report scales 20 0.70 0.55; 0.85 0.982 0.0 12 0.01 −0.20; 0.21 0.043 45.6
Manual adherence ensured and reported 17 0.85 0.70; 1.00 0.732 0.0 9 0.22 0.06; 0.38 0.263 20.2
No concurrent psychotropic medication 4 0.73 0.41; 1.05 0.531 0.0 7 −0.09 −0.30; 0.12 0.228 26.3
Professional non-student therapists 20 0.86 0.72; 1.00 0.862 0.0 8 0.18 −0.08; 0.44 0.007 64.2
Note: k = number of comparisons included in the analysis, g = Hedges’ g, 95% CI(g) = 95% confidence interval for g, p(Q) = level of significance for the Q-statistic, I2 = proportion
of heterogeneity not explained by within-study variance, CBT = cognitive-behavioral therapy, ES = effect size.
p = 0.052, k = 2, Q = 0.68, df = 1, p = 0.410, I2 = 0.0%) compared to full isons to common factor control groups (Q = 4.64, df = 5, p = 0.461,
CBGT (g = 0.49 [0.27; 0.72], p < 0.001, k = 10, Q = 6.01, df = 9, p = 0.739, I2 = 0.0%).
I2 = 0.0%), and for studies without high risk of bias (g = 0.93 [0.29;
1.57], p = 0.005, k = 2, Q = 0.08, df = 1, p = 0.777, I2 = 0.0%) compared 3.5.1.3. Clinical and methodological diversity. Given the limited data
to the remaining studies (g = 0.45 [0.23; 0.67], p < 0.001, k = 10, in the stratified samples, no further moderator analyses were
Q = 4.72, df = 9, p = 0.858, I2 = 0.0%). conducted. However, in order to address issues of clinical and
methodological diversity, results for subgroups of full CBGT stud-
ies and studies without judgment of high risk of bias are reported.
3.5. Group psychotherapy vs. active treatment conditions
We provide results for only one subgroup instead of contrasting the
two opposing groups since the number of studies per subgroup is
3.5.1. Specific symptomatology
too small to allow for informative subgroup analyses.
Fifteen comparisons of group psychotherapy to an active treat-
With regard to the subgroup of full CBGT studies, unchanged
ment condition reported outcomes specific to the diagnosis of social
results were obtained for comparisons to both common factor con-
phobia. A meta-analysis across all these comparisons resulted
trol groups (g = 0.31 [0.11; 0.51], p = 0.002, k = 5, Q = 4.62, df = 4,
in a standardized mean difference of g = 0.06, not significantly
p = 0.328, I2 = 13.4%), and individual psychotherapy (g = 0.26 [−0.36;
different from zero (95% CI [−0.19; 0.31], p = 0.656). High hetero-
0.88], p = 0.407, k = 3, Q = 6.13, df = 2, p = 0.047, I2 = 67.4%). The analy-
geneity was observed (Q = 59.61, df = 14, p < 0.001, I2 = 76.5%) and
sis relative to pharmacotherapy was already exclusively comprised
one negative outlier was identified (Heimberg et al., 1998; CBGT
of full CBGT studies.
vs. pharmacotherapy; g = −1.48). The standardized mean differ-
The subgroup of studies without high risk of bias showed consis-
ence across the remaining studies was g = 0.15 (95% CI [−0.03;
tent results with respect to comparisons to common factor control
0.33], p = 0.097). Heterogeneity for the model remained moder-
groups (g = 0.49 [0.23; 0.74], p < 0.001, k = 2, Q = 0.10, df = 1, p = 0.746,
ate at I2 = 52.1% (Q = 27.13, df = 13, p = 0.012). For results of primary
I2 = 0.0%) and comparisons to individual psychotherapy (g = 0.23
studies, see Fig. 4.
[-0.20; 0.67], p = .296, k = 1). The two comparisons to pharmacother-
apy were significantly in favor of pharmacotherapy (g = −0.36
3.5.1.1. Publication bias and sensitivity analyses. Egger’s regression [−0.66; −0.07], p = 0.017, k = 2, Q = 0.08, df = 1, p = 0.779, I2 = 0.0%).
test did not indicate publication bias (t = 0.54, df = 12, p = 0.598) and
there were no studies added in trim-and-fill analyses (Appendix 3.5.2. General psychopathology
C, Fig. C2). Sensitivity analyses are displayed in Table 3. For two The pooled standardized mean difference for comparisons to
subgroups of comparisons, confining analyses to the new sample active treatments on general psychopathology was negative but
changed the result in a meaningful way. Comparisons to group non-significant (g = −0.34 [−0.86; 0.17], p = 0.190, k = 10). Hetero-
treatments for which adherence to the manual was checked yielded geneity was high at I2 = 88.9% (Q = 81.16, df = 9, p < 0.001). The
a significant positive effect size. In contrast, comparisons from stud- pattern of results did not change after removing one study iden-
ies that did not allow for concurrent medication treatment showed tified as negative outlier (Heimberg et al., 1998; CBGT vs. common
a negative effect size, albeit still non-significant. factor control group; g = −3.55). The new standardized mean dif-
ference lay at g = −0.06 (95% CI [−0.38; 0.27], p = 0.726, k = 9),
3.5.1.2. Subgroup analyses stratified by type of comparison condition. and heterogeneity dropped to a moderate level (Q = 28.65, df = 8,
Since a number of different comparison conditions were collapsed p < 0.001, I2 = 72.1%).
under the label of active treatment groups, subgroup analyses With regard to general psychopathology, there were no signif-
were computed. Group psychotherapy was significantly superior to icant effects for comparisons of any of the subgroups to group
treatment in a common factor control group (g = 0.31 [0.14; 0.48], psychotherapy (common factor control group: g = 0.29 [−0.06;
p < 0.001, k = 6), while no difference to pharmacotherapy (g = −0.15 0.64], p = 0.100, k = 2; individual psychotherapy: g = 0.15 [−0.18;
[−0.44; 0.15], p = 0.335, k = 4) or individual psychotherapy (g = 0.23 0.47], p = 0.378, k = 3; pharmacotherapy: g = −0.41 [−0.95; 0.13],
[−0.17; 0.63], p = 0.261, k = 4) was found (Fig. 4). Heterogene- p = 0.140, k = 4). High heterogeneity only emerged for comparisons
ity reached a moderate but non-significant level for comparisons to pharmacotherapy (Q = 15.19, df = 3, p = 0.002, I2 = 80.3%), while it
to individual psychotherapy (Q = 6.18, df = 3, p = 0.103; I2 = 51.4%), dropped to I2 = 0% for common factor control groups (Q = 0.01, df = 1,
while dropping to a low level for comparisons to pharmacotherapy p = 0.922) and to I2 = 1.7% for individual psychotherapy (Q = 2.03,
(Q = 4.99, df = 3, p = 0.173, I2 = 39.8%) and even to zero for compar- df = 2, p = 0.362).
Fig. 4. Forest plot of study effect sizes and overall standardized mean differences for comparisons to active treatments stratified by comparison condition.
3.6. Inclusion of non-CBT studies group revealed drop-out rates of 17.1% [13.5%; 20.7%] for group
psychotherapy (i = 41), 14.5% [8.6%; 20.3%] for individual psy-
Analyses were repeated under inclusion of the two non-CBT chotherapy (i = 5), 29.5% [18.4%; 40.5%] for pharmacotherapy (i = 6),
studies. One study that investigated interpersonal group psy- 15.8% [7.2%, 24.3%] for common factor control groups (i = 5), and
chotherapy (Huang & Liu, 2011) provided for wait list-controlled 7.8% [4.5%; 11.0%] for wait list control groups (i = 22).
effect sizes for specific symptomatology (g = 1.69 [0.83; 2.55]). A Direct comparisons of drop-out rates in the group psychother-
study of psychodynamic group psychotherapy (Knijnik et al., 2004) apy condition and the comparison condition within the same
supplied data on both outcome domains for comparisons to a studies resulted in non-significant relative risks for comparisons
common factor control group (specific symptomatology: g = 0.10 to wait list (RR = 1.28 [0.95; 1.71], p = 0.101, k = 30), common factor
[−0.57; 0.76]; general psychopathology g = −0.24 [−0.93; 0.46]). control groups (RR = 1.05 [0.58; 1.90], p = 0.872, k = 5), and individ-
The confidence intervals of these effect sizes overlapped with ual psychotherapy (RR = 1.50 [0.91; 2.47], p = 0.108; k = 5). A trend
those of the standardized mean differences across all other stud- in favor of group psychotherapy emerged for comparisons to phar-
ies. Results did not change after inclusion of the two studies to the macotherapy (RR = 0.76 [0.55; 1.05], p = 0.095; k = 6).
meta-analyses (Table 3).
3.7. Follow-up 4. Discussion
Eight studies provided follow-up data. Group psychotherapy In the present meta-analysis, group psychotherapy obtained
remained consistently superior to wait list control groups over all large positive effects compared to wait list control groups
outcome categories at short-term follow-up. (Table 4). Only one and demonstrated equivalence to alternative treatment options.
study provided for data on specific symptomatology at mid-term The results were consistent across different outcome domains,
follow-up (Olivares et al., 2009; g = 3.91 [2.66; 5.15]). Comparisons except for comparisons to common factor controls, where
to active treatment groups did not gain significance at either of the group psychotherapy was superior in alleviating the specific
follow-up time points or outcome categories. Results on long-term SAD symptomatology, but did not differ for effects on general
follow-up were reported by only two studies and did not indicate psychopathology. The robustness of the results was generally con-
differential effects for group psychotherapy and individual psy- firmed by sensitivity analyses and absence of publication bias.
chotherapy or common factor control groups (results are available The effect size that we obtained for wait list-controlled trials of
upon request). group psychotherapy is similar to that of efficacy measures from
other meta-analyses examining psychotherapeutic treatments for
3.8. Drop-out SAD in general (Powers et al., 2008: g = 0.84), as well as to those
including only CBGT (Mayo-Wilson et al., 2014: g = 0.92).
The weighted drop-out rate3 in a meta-analysis across all
Our analyses extend the existing knowledge on the differential
included experimental groups was 15.2% (95% CI [12.6%; 17.7%],
effects of group psychotherapy compared to other treatment for-
i = 79). Separate subgroup analyses for the type of experimental
mats, providing results that were exclusively derived from direct
comparisons within studies. Generally, they are in accordance with
findings from Mayo-Wilson et al. (2014), who found non-significant
3
Drop-out was conceptualized by contrasting randomized and post-treatment effects for comparisons of CBGT and other psychotherapeutic and
completer sample sizes. More patients may have finished the treatment without pharmacologic treatment options in their network meta-analysis.
providing data at post-assessment or some may have provided post-assessment
data without finishing the treatment. This approximation was necessary as studies
While, in their analyses no differences were detected in relation to
defined drop-out in various ways and often did not differentiate between treatment “psychological placebo,” our results suggest that group psychother-
and assessment drop-out. apy is superior to common factor controls in reducing specific
Table 4
Meta-analyses on follow-up data for specific symptomatology and general psychopathology stratified by comparison conditions.
Short-term follow-up Mid-term follow-up
Heterogeneity Heterogeneity
2
k g 95% CI(g) p(Q) I (%) k g 95% CI(g) p(Q) I2 (%)
Wait list Specific 7 1.09 0.63; 1.55 0.009 65.1 1 3.91 2.66; 5.15
General 6 0.73 0.44; 1.02 0.625 0.0 –
Active comparisons Specific 5 −0.01 −0.37; 0.35 0.037 60.9 3 −0.06 −0.55; 0.43 0.029 71.7
General 5 −0.10 −0.36; 0.15 0.678 0.0 3 −0.20 −0.69; 0.29 0.070 62.5
Common factor control Specific 1 0.43 −0.07; 0.93 –
General 1 −0.09 −0.74; 0.57 –
Individual therapy Specific 3 −0.24 −0.72; 0.23 0.058 65.0 2 −0.22 −0.89; 0.44 0.038 76.9
General 3 −0.18 −0.53; 0.17 0.239 30.2 2 −0.35 −1.03; 0.33 0.048 74.4
Pharmacoththerapy Specific 1 0.15 −0.32; 0.62 1 0.22 −0.18; 0.63
General 1 0.00 −0.58; 0.58 1 0.08 −0.42; 0.58
Note: k = number of comparisons included in the analysis, g = Hedges’ g, CI(g) = 95% confidence interval for g, p(Q) = level of significance for the Q-statistic, I2 = proportion of
heterogeneity not explained by within-study variance.
symptomatology. Explanations for this divergence could be the dif- 2014; Powers et al., 2008) for SAD. Evidence for further treat-
ferent sample of included studies or the different methodology ment approaches was scarce, which reveals the need for more
between network meta-analyses and traditional meta-analyses. well-controlled efficacy studies. It would be informative to com-
Network meta-analyses have an advantage over traditional meta- pare CBGT and group exposure treatments to non-CBT treatment
analyses because the evidence base is larger since they allow for approaches. However, our sample did not contain enough data
comparisons across studies (Mayo-Wilson et al., 2014). However, to allow for meaningful analyses. Even so, single study results
different characteristics between studies and comparisons are at from studies administering alternative treatment approaches did
risk of biasing the effect (Jansen & Naci, 2013). In contrast, our not show noticeably different effects and including them in over-
results are in accordance with moderate positive effect sizes found all analyses did not change the findings but instead marginally
by Wersebe et al. (2013). Non-significant effects in comparison to reduced heterogeneity.
common factor control groups emerged for outcomes of general With regard to the design of future primary studies, it should
psychopathology in our analyses. This might be due to charac- be noted that both the wait list-controlled treatment effect and the
teristics of the comparison conditions, which have demonstrated effect relative to active comparisons was reduced if only studies
considerable effects in previous meta-analyses (Baskin et al., 2003; were included that prohibited concurrent psychotropic medica-
Wampold, Minami, Callen Tierny, Baskin, & Bhati, 2005). tion. Nevertheless, this change was too minor to influence the
Compared to previous meta-analyses that investigated group deductions made from the results. Information from primary stud-
psychotherapy for SAD, our results are based on an especially wide ies on the actual medication use per experimental condition was
spectrum of evidence, since studies were included regardless of scarce and did not allow for any informed conclusions. There-
any treatment approach, comparison condition, or publication lan- fore, it remains unclear how the effect might have emerged and
guage. We were able to examine data from 18 additional studies whether it was meaningful. This is a question that should be
that were not included in the meta-analysis by Mayo-Wilson et al. addressed in future research. There is one study on treatment of
(2014). At the same time, ten group psychotherapy studies included complicated grief that reports treatment-specific effects of con-
in the network meta-analysis were not eligible for our analyses due current medication with higher response and completion rates
to missing or non-eligible comparison conditions. We also added for patients receiving concurrent medication while showing dif-
26 studies to the sample analyzed by Wersebe et al. (2013). ferent effect magnitudes for complicated grief targeted therapy
Additionally, our study considers several factors of risk of bias and interpersonal psychotherapy (Simon et al., 2008). Similar
and characteristics of the group psychotherapy with respect to treatment-specific effects might have been relevant in our included
their influence on the relative treatment effect. Surprisingly, stud- studies.
ies without high risk of bias consistently reported larger effect sizes The effect was also marginally reduced if only standard self-
than all other studies. This result suggests that our analyses are report scales were considered, but again this did not change
unlikely to be biased by low quality studies. However, it is in con- deductions. Lower effects for self-report scales have been reported
tradiction to most findings on the effects of methodological quality, in the literature (Feske & Chambless, 1995).
which usually report smaller effect sizes for high quality studies Findings of higher attrition within group psychotherapies com-
(Cuijpers, van Straten, Bohlmeijer, Hollon, & Andersson, 2010; Jüni, pared to the individual format reported in a previous meta-analysis
Altman, & Egger, 2001). Nevertheless, a specific direction of the bias on SAD (Aderka, 2009), were not confirmed in our sample of
was not universally confirmed and usually depends on the type included studies. This is in line with reviews related to drop outs
of quality assessment and the assessed sample of included stud- in different psychotherapy settings (Swift & Greenberg, 2012).
ies (Jüni et al., 2001). Larger effects for high quality studies might Instead, a trend for higher drop-out in pharmacotherapy was found
be explained, assuming that high methodological quality is paired and indicates less treatment acceptance for these conditions, pos-
with high intervention quality. However, all conclusions on the sibly due to the side effects of psychopharmacological medication.
impact of methodological quality should be treated with caution, Together with reported higher relapse rates for pharmacotherapy
considering that the reporting quality for the relevant variables was (Liebowitz et al., 1999), this trend should be considered when inter-
poor. preting the somewhat higher post-treatment effect sizes of the
No differential effects between CBGT and group exposure condition. However, power was probably an issue within these
therapy were detected. These findings are in line with earlier analyses. Therefore, analyses on larger study pools of direct com-
meta-analyses, which found that exposure produced equivalent parisons should be conducted before conclusions can be drawn.
outcomes to full CBT (Feske & Chambless, 1995; Mayo-Wilson et al.,
4.1. Limitations that were not found, unpublished, or published after completion of
our final search. One study (Dadashzadeh et al., 2012) could not be
A number of limitations are worth noting when interpreting the included due to limited resources for translation.
results from the present study. First, although we wanted to include
group psychotherapy studies regardless of the theoretical orienta- 4.2. Implications for clinical practice and research
tion, the sample is mainly based on CBT approaches (i.e., 80.0% use
full CBGT, 95.6% have CBT elements as main therapeutic technique) Currently, the 2013 NICE clinical guideline 159 for SAD discour-
and main analyses were therefore restricted to this approach. This ages treatment providers from administering group psychotherapy
is a common problem within efficacy research (Burlingame, 2010) instead of individual psychotherapy and states that “it is less clin-
and further primary studies on non-CBT interventions are needed ically and cost effective than individual CBT” (p. 19). However,
for conclusions about differential effects. in accordance with findings from other meta-analyses (Acarturk,
Second, we decided to exclude statistical outliers from subse- Cuijpers et al., 2009; Powers et al., 2008; Wersebe et al., 2013) our
quent analyses even though the introduced heterogeneity could results advocate that group psychotherapy for patients with SAD
be meaningful. This procedure was adopted because heterogeneity should be considered as a valid treatment option. To date, most
was largely due to these single studies, and there was a high likeli- evidence acknowledges an equivalent efficacy of group psychother-
hood of getting chance findings in moderator analyses. This is less apy compared to treatment alternatives (Burlingame et al., personal
problematic for wait list-controlled trials, where all outliers were communication, June 5, 2015).
positive, and thus the estimate without outliers provided a more Additionally, group psychotherapy can provide an economic
conservative value. However, for comparisons to active treatment way of treating a greater number of patients, which is impor-
groups, a negative outlier was identified and eliminating the com- tant especially as the prevalence of anxiety disorders increases
parison from the analysis changed the effect size in favor of group (Kessler & Greenberg, 2002) and public financing of psychoso-
psychotherapy. Notably, this change was not pronounced enough cial interventions gets increasingly difficult in some countries.
to alter deductions from the results. Randomized-controlled studies that explicitly investigate eco-
Third, we could not investigate all of the pre-specified mod- nomic outcomes of group psychotherapy for SAD would be
erator variables due to limited variability between studies and desirable.
generally low statistical heterogeneity. Additionally, reporting on The data reveal a considerable lack of randomized-controlled
methodological quality was poor. Therefore, results of our analy- studies that examined approaches other than CBGT. Given the sub-
ses rely on uncertainty as to the quality of some studies, where stantial amount of patients that do not recover after CBGT (compare
poor reporting might not always accurately reflect poor qual- Knijnik et al., 2008; Stangier et al., 2003), developing and evalu-
ity. Although heterogeneity was low, the consideration of further ating alternative treatment approaches such as interpersonal or
moderators might be beneficial. Based upon the observation that psychodynamic psychotherapy should be of high priority.
the diagnosis of SAD is often described as heterogeneous (e.g.,
Hofmann, 2010), with differentiation of a generalized subtype and 4.3. Conclusions
proximity to avoidant personality disorder, taking into account
diagnostic aspects might be promising. Group psychotherapy for SAD is efficacious and shows equiva-
Fourth, heterogeneity remained even after moderator and sub- lent effect sizes to other treatment formats. However, additional
group analyses for some analyses, i.e., comparisons to individual evidence and clarification is needed especially for non-CBT
psychotherapy and pharmacotherapy, where only small study approaches and direct comparisons to alternative formats. Further
pools were available. Heterogeneity for these analyses relied analyses should investigate the differential efficacy of the vari-
heavily on the analyzed outcome domain, possibly due to the dif- ous approaches of group psychotherapy as well as the differential
ferent subsamples of studies. High heterogeneity remained for suitability for distinct patient populations, such as those with gen-
comparisons to pharmacotherapy when regarding general psy- eralized SAD or comorbid avoidant personality disorder.
chopathology and moderate but non-significant heterogeneity for
comparisons to individual psychotherapy when regarding specific Funding
symptomatology. In these cases, results between single stud-
ies were contradicting with respect to the observed direction of The work was funded by the German Federal Ministry of Edu-
the effect. One possible explanation could be an effect of biased cation and Research (FKZ 01KG1216). The funding source was not
researcher allegiance that has previously been discussed within involved in content-related decisions that led to the preparation of
the context of head-to-head comparisons (Luborsky et al., 1999; this article.
Munder, Gerger, Trelle, & Barth, 2011). Another possible reason
for the divergence might be that group psychotherapies that are Acknowledgements
adapted from individual treatment approaches might lack a trans-
position accounting for the use of specific therapeutic factors in a We kindly thank Anna Lindner and Rahel Klatte for their help
group, thus resulting in poorer outcomes (Strauss & Burlingame, with the coding of methodological quality and the literature search,
2012). It would be of interest to regard the impact of these char- and Xin Chang, Stefanie Kern, and Swetlana Philipp for their support
acteristics as statistical moderators. However, the small study pool with translations.
did not allow for further analyses.
Finally, although a comprehensive search strategy was con-
ducted, we may have missed some existing evidence from studies
Appendix A Appendix C
Table A1
Data coding manual.
Information Variables
Study characteristics Author; publication year; publication status; language;
country of study performance
Participant Sample size; age; gender; ethnicity; previous treatment or
characteristics hospitalization; concurrent treatment; recruitment method
(referred, recruited etc.); diagnostic manual (DSM, ICD
versions); diagnostic instrument; duration of disorder;
comorbidity; exclusion criteria
Intervention group Theoretical orientation (CBT, psychodynamic, interpersonal
characteristics etc.); group composition regarding gender and disorder
(homogeneous or mixed); treatment setting (inpatient,
outpatient, day treatment); treatment location (university
counseling center, state hospital, private practice etc.),
duration (number and length of sessions) and frequency of
treatment (number of sessions per week); number of patients
per group; manualization of treatment; group membership
(open, closed, etc.); therapists profession (graduate level,
doctoral level, professional level etc.); group leadership
(single-led, co-led); booster sessions (yes/no); preparatory
individual sessions (yes/no)
Control group Type of control group (wait list, common factor control, Fig. C1. Funnel plot for effect sizes and their standard errors regarding specific
characteristics treatment as usual, individual therapy, pharmacotherapy); symptomatology compared to wait list control groups.
bona fide treatment
Outcome Type of analyses (ITT or completer); length of follow-up; type
characteristics of outcome measure (specific symptomatology, general
psychopathology, remission/response); outcome inventory;
outcome form (continuous, dichotomous); outcome source
(self-reported, observer-reported)
Effect size related Statistical parameters (M, SD, p-values, t-values, F-values,
information sample size etc.); effect direction, estimation by approximation
Study quality Sequence generation; allocation concealment; selective
reporting; implementation quality; blinding of outcome
assessment, handling of incomplete outcome data
Appendix B
Table B1
Allocation of outcome measures to domains.
Specific symptomatology General psychopathology
Measures ADIS-IV Fear/Avoidance, BAIo , BDIp , FQ-Depression

ADIS-IV Severity, BSPSa , CGIb , Anxiety, FQ-Disability, HAMAq ,
FNE/BFNEc , FQd -SP, FQ-Main HAMDr , SCL-90s , STAIt
Fear, IASe , LSASf , PRCSg , SADSh , Fig. C2. Funnel plot for effect sizes and their standard errors regarding specific
SIASi , SPAIj , SPINk , SPSl , SPSQm , symptomatology compared to active treatment groups.
SPWSSn , Social Phobic
Disorders Severity and Change
Form, study-specific composite References 4
scores of SAD symptomatology
Note: If not specified otherwise, all subscales of the measure were included. Acarturk, C., Cuijpers, P., van Straten, A., & de Graaf, R. (2009). Psychological
a treatment of social anxiety disorder: a meta-analysis. Psychological Medicine,
Brief Social Phobia Scale.
b 39, 241–254. http://dx.doi.org/10.1017/S0033291708003590
Clinical Global Impression Scale.
c
Acarturk, C., Smit, F., de Graaf, R., van Straten, A., ten Have, M., & Cuijpers, P. (2009).
(Brief) Fear of Negative Evaluation Scale. Economic costs of social phobia: a population-based study. Journal of Affective
d
Fear Questionnaire. Disorders, 115, 421–429. http://dx.doi.org/10.1016/j.jad.2008.10.008
e
Interaction Anxiety Scale. Aderka, I. M. (2009). Factors affecting treatment efficacy in social phobia: the use
f
Liebowitz Social Anxiety Scale. of video feedback and individual vs. group formats. Journal of Anxiety Disorders,
g
Personal Report of Confidence as a Speaker. 23(1), 12–17. http://dx.doi.org/10.1016/j.janxdis.2008.05.003
h
Social Anxiety Disorder Scale. *Alden, L. E., & Taylor, C. T. (2011). Relational treatment strategies increase social
i approach behaviors in patients with generalized social anxiety disorder.
Social Interaction Anxiety Scale.
j
Social Phobia and Anxiety Inventory. Journal of Anxiety Disorders, 25(3), 309–318. http://dx.doi.org/10.1016/j.
k
Social Phobia Inventory. janxdis.2010.10.003
l
Social Phobia Scale. Altman, D. G. (1991). Practical statistics for medical research. London: Chapman and
m
Social Phobia Screening Questionnaire. Hall.
n American Psychiatric Association. (2013). Diagnostic and statistical manual of
Social Phobia Weekly Summary Scale.
o mental disorders (5th ed.). Arlington, VA: American Psychiatric Publishing.
Beck Anxiety Inventory.
p
*Anderson, P. L., Price, M., Edwards, S. M., Obasaju, M. A., Schmertz, S. K., Zimand,
Beck Depression Inventory. E., & Calamares, M. R. (2013). Virtual reality exposure therapy for social anxiety
q
Hamilton Anxiety Rating Scale. disorder: a randomized controlled trial. Journal of Consulting and Clinical
r
Hamilton Depression Rating Scale. Psychology, 81(5), 751–760. http://dx.doi.org/10.1037/t03814-000
s
Symptom Checklist 90.
t
State Trait Anxiety Inventory.
4
Studies marked with an asterisk (*) were included in the review.
Antony, M. M. (1997). Assessment and treatment of social phobia. Canadian Journal Fehm, L., Pelissolo, A., Furmark, T., & Wittchen, H.-U. (2005). Size and burden of
of Psychiatry, 42, 826–834. social phobia in Europe. European Neuropsychopharmacology, 15, 453–462.
Barth, J., Gerger, H., Munder, T., Rabung, S., Rosendahl, J., & Tefikow, S. (2011). http://dx.doi.org/10.1016/j.euroneuro.2005.04.002
Reporting meta-analyses of psychotherapeutic effectiveness: PPRISMA. In Feske, U., & Chambless, D. L. (1995). Cognitive behavioral versus exposure only
42nd International Meeting of the Society for Psychotherapy Research. treatment for social phobia: a meta-analysis. Behavior Therapy, 26(4), 695–720.
Baskin, T. W., Callen Tierney, S., Minami, T., & Wampold, B. E. (2003). Establishing http://dx.doi.org/10.1016/S0005-7894(05)80040-1
specificity in psychotherapy: a meta-analysis of structural equivalence of *Filion-Rosset, J. (2004). Cognitive changes following cognitive-behavioral therapy
placebo controls. Journal of Consulting and Clinical Psychology, 71(6), 973–979. for social phobia. (65), ProQuest Information & Learning, US. Retrieved from
http://dx.doi.org/10.1037/0022-006X.71.6.973 http://search.ebscohost.com/login.aspx?direct=true&db=psyh&AN=2004-
*Bjornsson, A. S., Bidwell, L. C., Brosse, A. L., Carey, G., Hauser, M., Mackiewicz 99022-109&site=ehost-live. Available from EBSCOhost psych database.
Seghete, K. L., & Craighead, W. E. (2011). Cognitive-behavioral group therapy Furmark, T. (2002). Social phobia: overview of community surveys. Acta
versus group psychotherapy for social anxiety disorder among college Psychiatrica Scandinavica, 105, 84–93. http://dx.doi.org/10.1034/j.1600-0447.
students: a randomized controlled trial. Depression & Anxiety, 28(11), 2002.1r103.x
1034–1042. http://dx.doi.org/10.1002/da.22180 *Furmark, T., Tillfors, M., Marteinsdottir, I., Fischer, H., Pissiota, A., Langstrom, B., &
*Blanco, C., Heimberg, R. G., Schneier, F. R., Fresco, D. M., Chen, H., Turk, C. L., & Fredrikson, M. (2002). Common changes in cerebral blood flow in patients
Liebowitz, M. R. (2010). A placebo-controlled trial of phenelzine, cognitive with social phobia treated with citalopram or cognitive-behavioral therapy.
behavioral group therapy, and their combination for social anxiety disorder. Archives of General Psychiatry, 59(5), 425–433. http://dx.doi.org/10.1001/
Archives of General Psychiatry, 67(3), 286–295. http://dx.doi.org/10.1001/ archpsyc.59.5.425
archgenpsychiatry.2010.11 Gawlytta, R., Tefikow, S., & Strauss, B. (2013). Psychodynamic short-term group
Borenstein, M., Hedges, L. V., Higgins, J. P. T., & Rothstein, H. R. (2009). Introduction treatment for social phobia: description of the treatment concept and a case
to meta-analysis. New York: Wiley. study. Psychotherapie, Psychosomatik und Medizinische Psychologie, 63(1),
Burlingame, G. M. (2010). Small group treatments: recent advances and issues. 58–64. http://dx.doi.org/10.1055/s-0032-1330010
Psychotherapy Research, 20, 1–7. *Gruber, K., Moran, P. J., Roth, W. T., & Taylor, C. B. (2001). Computer-assisted
Burlingame, G. M., MacKenzie, K. R., & Strauss, B. (2004). Small group treatment: cognitive behavioral group therapy for social phobia. Behavior Therapy, 32(1),
evidence for effectiveness and mechanisms of change. In: M. J. Lambert (Ed.), 155–165. http://dx.doi.org/10.1016/s0005-7894(01)80050-2
Bergin and Garfieldı́s handbook of psychotherapy and behavior change (5th ed., Hart, T. A., Jack, M. S., Turk, C. L., & Heimberg, R. G. (1999). Issues for the
pp. 647–696). New York: John Wiley & Sons. measurement of social phobia. In: H. G. M. Westenberg, & J. A. den Boer (Eds.),
Cicchetti, D. V. (1994). Guidelines, criteria, and rules of thumb for evaluating Social phobia: recent trends and progress (pp. 133–155). Amsterdam: Syn-Thesis
normed and standardized assessment instruments in psychology. Psychological Publishers.
Assessment, 6(4), 284–290. http://dx.doi.org/10.1037/1040-3590.6.4.284 Hedges, L. (1981). Distribution theory for Glass’s estimator of effect size and
Clark, D. M., & Wells, A. (1995). A cognitive model of social phobia. In: R. G. related estimators. Journal of Educational Statistics, 6, 107–128. http://dx.doi.
Heimberg, M. Liebowitz, D. Hope, & F. Schneier (Eds.), Social phobia: diagnosis, org/10.3102/10769986006002107
assessment, and treatment (pp. 69–93). New York: Guilford. Hedges, L. V., & Vevea, J. L. (1998). Fixed- and random-effects models in
Cohen, J. (1992). A power primer. Psychological Bulletin, 112(1), 155–159. http://dx. meta-analysis. Psychological Methods, 3(4), 486–504. http://dx.doi.org/10.
doi.org/10.1037/0033-2909.112.1.155 1037/1082-989X.3.4.486
*Cottraux, J., Note, I., Albuisson, E., Yao, S. N., Note, B., Mollard, E., & Coudert, A. J. *Heideman, P. W. (2008). Combining cognitive behavioral therapy with an alcohol
(2000). Cognitive behavior therapy versus supportive therapy in social phobia: intervention to reduce alcohol problems among socially anxious college
a randomized controlled trial. Psychotherapy & Psychosomatics, 69(3), 137–146. students. (69), ProQuest Information & Learning, US. Retrieved from http://
http://dx.doi.org/10.1159/000012382 search.ebscohost.com/login.aspx?direct=true&db=psyh&AN=2009-99080-
Cuijpers, P., Sijbrandij, M., Koole, S. L., Huibers, M., Berking, M., & Andersson, G. 288&site=ehost-live. Available from EBSCOhost psyh database.
(2014). Psychological treatment of generalized anxiety disorder: a Heimberg, R. G. (1991). A manual for conducting cognitive-behavioral group therapy
meta-analysis. Clinical Psycholog Review, 34, 130–140. http://dx.doi.org/10. for social phobia. (2nd ed.). Unpublished manuscript available from the Center
1016/j.cpr.2014.01.002 for Stress and Anxiety Disorders, Albany, NY 12205.
Cuijpers, P., van Straten, A., Bohlmeijer, E., Hollon, S. D., & Andersson, G. (2010). Heimberg, R. G. (2002). Cognitive-behavioral therapy for social anxiety disorder:
The effects of psychotherapy for adult depression are overestimated: a current status and future directions. Biological Psychiatry, 51, 101–108. http://
meta-analysis of study quality and effect size. Psychological Medicine, 40, dx.doi.org/10.1016/S0006-3223(01)01183-0
211–223. http://dx.doi.org/10.1017/S0033291709006114 Heimberg, R. G., & Becker, R. E. (2002). Cognitive-behavioral group therapy for social
Dadashzadeh, H., Yazdandoost, R., Gharraee, B., & Asgharnejad farid, A. (2012). phobia: Basic mechanisms and clinical strategies. New York: Guilford Press.
Effectiveness of cognitive-behavioral group therapy and exposure therapy on *Heimberg, R. G., Dodge, C. S., Hope, D. A., Kennedy, C. R., Zollo, L. J., & Becker, R. E.
interpretation bias and fear of negative evaluation in social anxiety disorder. (1990). Cognitive behavioral group treatment for social phobia: comparison
Iranian Journal of Psychiatry and Clinical Psychology, 18(1), 40–51. with a credible placebo control. Cognitive Therapy and Research, 14(1), 1–23.
Dahl, A. A., & Dahl, C. F. (2010). Are there gender differences in impairment http://dx.doi.org/10.1007/bf01173521
associated with high social anxiety? A community-based study. Journal of *Heimberg, R. G., Liebowitz, M. R., Hope, D. A., Schneier, F. R., Holt, C. S., Welkowitz,
Anxiety Disorders, 24(5), 487–493. http://dx.doi.org/10.1016/j.janxdis.2010.03. L. A., & Klein, D. F. (1998). Cognitive behavioral group therapy vs phenelzine
005 therapy for social phobia: 12-week outcome. Archives of General Psychiatry,
*Davidson, J. R., Foa, E. B., Huppert, J. D., Keefe, F. J., Franklin, M. E., Compton, J. S., & 55(12), 1133–1141. http://dx.doi.org/10.1001/archpsyc.55.12.1133
Gadde, K. M. (2004). Fluoxetine, comprehensive cognitive behavioral therapy, *Heimberg, R. G., Salzman, D. G., Holt, C. S., & Blendell, K. A. (1993).
and placebo in generalized social phobia. Archives of General Psychiatry, 61(10), Cognitive-behavioral group treatment for social phobia: effectiveness at 5-year
1005–1013. http://dx.doi.org/10.1001/archpsyc.61.10.1005 follow-up. Cognitive Therapy and Research, 17(4), 325–339. http://dx.doi.org/
Del Re, A. C., & Hoyt, W. T. (2014). MAd: Meta-Analysis with Mean Differences. R 10.1007/bf01177658
package version 0. 8-2. Available from http://cran.r-project.org/web/packages/ Higgins, J. P. T., Altman, D. G., Sterne, J. A. C., Higgins, J. P. T., & Green, S. (Eds.).
MAd. (2011). Assessing risk of bias in included studies. In: Cochrane handbook for
*D’El Rey, G. J. F., Lacava, J. P. L., Cejkinski, A., & Mello, S. L. (2008). systematic reviews of interventions version 5.1.0. The Cochrane Collaboration
Cognitive-behavioral group treatment in social phobia: 12-week outcome. [chapter 8] (updated March 2011). Available from www.cochrane-handbook.
Revista de Psiquiatria Clinica, 35(2), 79–83. http://dx.doi.org/10.1590/S0101- org
60832008000200006 Higgins, J. P. T., Deeks, J. J., & Altman, D. G. (2011). Special topics in statistics. In: J. P.
DerSimonian, R., & Laird, N. (1986). Meta-analysis in clinical trials. Controlled T. Higgins, & S. Green (Eds.), Cochrane handbook for systematic reviews of
Clinical Trials, 7(3), 177–188. http://dx.doi.org/10.1016/0197-2456(86)90046-2 interventions. Version 5.1.0. The Cochrane Collaboration (updated March 2011).
*Dogaheh, E. R., Mohammadkhani, P., & Dolatshahi, B. (2011). Comparison of group Higgins, J. P. T., Thompson, S. G., Deeks, J. J., & Altman, D. G. (2003). Measuring
and individual cognitive-behavioral therapy in reducing fear of negative inconsistency in meta-analyses. British Medical Journal, 327, 557–560. http://
evaluation. Psychological Reports, 108(3), 955–962. http://dx.doi.org/10.2466/ dx.doi.org/10.1136/bmj.327.7414.557
02.21.PR0.108.3.955-962 *Hofmann, S. G. (2004). Cognitive mediation of treatment change in social phobia.
Duval, S., & Tweedie, R. (2000). Trim and fill: a simple funnel-plot-based method Journal of Consulting & Clinical Psychology, 72(3), 393–399. http://dx.doi.org/10.
for testing and adjusting for publication bias in meta-analysis. Biometrics, 56, 1037/0022-006X.72.3.392
455–463. http://dx.doi.org/10.1111/j.0006-341X.2000.00455.x Hofmann, S. G. (2010). Recent advances in the psychosocial treatment of social
Egger, M., Smith, G. D., Schneider, M., & Minder, C. (1997). Bias in meta-analysis anxiety disorder. Depression and Anxiety, 27, 1073–1076. http://dx.doi.org/10.
detected by a simple, graphical test. British Medical Journal, 315, 629–634. 1002/da.20771
http://dx.doi.org/10.1136/bmj.315.7109.629 Hofmann, S. G., & Bögels, S. M. (2006). Recent advances in the treatment of social
Egger, N., Konnopka, A., Beutel, M. E., Herpertz, S., Hiller, W., Hoyer, J., & König, phobia: Introduction to the special issue. Journal of Cognitive Psychotherapy, 20,
H.-H. (2015). Short-term cost-effectiveness of psychodynamic therapy and 3–5. http://dx.doi.org/10.1891/jcop.20.1.3
cognitive-behavioral therapy in social anxiety disorder: results from the *Hope, D. A., Heimberg, R. G., & Bruch, M. A. (1995). Dismantling
SOPHO-NET trial. Journal of Affective Disorders, 180, 21–28. http://dx.doi.org/10. cognitive-behavioral group therapy for social phobia. Behaviour Research &
1016/j.jad.2015.03.037 Therapy, 33(6), 637–650. http://dx.doi.org/10.1016/0005-7967(95)00013-N
*Erickson, D. H., Janeck, A. S., & Tallman, K. (2007). A cognitive-behavioral group for *Huang, H.-L., & Liu, X.-M. (2011). Effects of group interpersonal psychotherapy
patients with various anxiety disorders. Psychiatric Services, 58(9), 1205–1211. and group cognitive behavioral therapy on social anxiety in college students.
http://dx.doi.org/10.1176/appi.ps.58.9.1205 Chinese Mental Health Journal, 25(5), 324–327.
Jansen, J. P., & Naci, H. (2013). Is network meta-analysis as valid as standard *Otto, M. W., Pollack, M. H., Gould, R. A., Worthington, J. J., 3rd, McArdle, E. T., &
pairwise meta-analysis? It all depends on the distribution of effect modifiers. Rosenbaum, J. F. (2000). A comparison of the efficacy of clonazepam and
BMC Medicine, 11, 159–167. http://dx.doi.org/10.1186/1741-7015-11-159 cognitive-behavioral group therapy for the treatment of social phobia. Journal
Jüni, P., Altman, D. G., & Egger, M. (2001). Systematic reviews in health care. of Anxiety Disorders, 14(4), 345–358. http://dx.doi.org/10.1016/S0887-
Assessing the quality of controlled clinical trials. British Medical Journal, 323, 6185(00)00027-X
42–46. http://dx.doi.org/10.1136/bmj.323.7303.42 Otto, M. W., Pollack, M., & Maki, K. (2000). Empirically supported treatments for
Kessler, R. C., & Greenberg, P. E. (2002). The economic burden of anxiety and stress panic disorder. Journal of Consulting and Clinical Psychology, 68(4), 556–563.
disorders. In: K. L. Davis, D. Charney, J. T. Coyle, & C. Nemeroff (Eds.), http://dx.doi.org/10.1016/j.psc.2009.05.005
Neuropsychopharmacology: the fifth generation of progress (pp. 981–992). *Pishyar, R., Harris, L. M., & Menzies, R. G. (2008). Responsiveness of measures of
Philadelphia: Lippincott, Williams, and Wilkins. attentional bias to clinical change in social phobia. Cognition & Emotion, 22(7),
Knijnik, D. Z., Blanco, C., Salum, G. A., Moraes, C. U., Mombach, C., Almeida, E., & 1209–1227. http://dx.doi.org/10.1080/02699930701686008
Eizirik, C. L. (2008). A pilot study of clonazepam versus psychodynamic group Powers, M. B., Sigmarsson, S. R., & Emmelkamp, P. M. G. (2008). A meta-analytic
therapy plus clonazepam in the treatment of generalized social anxiety review of psychological treatments for social anxiety disorder. International
disorder. European Psychiatry, 23, 567–574. http://dx.doi.org/10.1016/j.eurpsy. Journal of Cognitive Therapy, 1(2), 94–113. http://dx.doi.org/10.1680/ijct.2008.
2008.05.004 1.2.94
*Knijnik, D. Z., Kapczinski, F., Chachamovich, E., Margis, R., & Eizirik, C. L. (2004). *Rapee, R. M., Abbott, M. J., Baillie, A. J., & Gaston, J. E. (2007). Treatment of social
Psychodynamic group treatment for generalized social phobia. Revista phobia through pure self-help and therapist-augmented self-help. British
Brasileira de Psiquiatria, 26(2), 77–81. http://dx.doi.org/10.1590/S1516- Journal of Psychiatry, 191, 246–252. http://dx.doi.org/10.1192/bjp.bp.106.
44462004000200003 028167
*Kocovski, N. L., Fleming, J. E., Hawley, L. L., Huta, V., & Antony, M. M. (2013). *Rapee, R. M., Gaston, J. E., & Abbott, M. J. (2009). Testing the efficacy of
Mindfulness and acceptance-based group therapy versus traditional cognitive theoretically derived improvements in the treatment of social phobia. Journal
behavioural group therapy for social anxiety disorder: a randomized of Consulting and Clinical Psychology, 77(2), 317–327. http://dx.doi.org/10.1037/
controlled trial. Behaviour Research and Therapy, 51, 889–898. http://dx.doi.org/ a0014800
10.1016/j.brat.2013.10.007 R Core Team. (2014). R: a language and environment for statistical computing.
*Ledley, D. R., Huppert, J. D., Foa, E. B., Davidson, J. R. T., Keefe, F. J., & Potts, N. L. S. Vienna, Austria: R Foundation for Statistical Computing. Available from
(2005). Impact of depressive symptoms on the treatment of generalized social https://www.r-project.org
anxiety disorder. Depression and Anxiety, 22(4), 161–167. http://dx.doi.org/10. Roberge, P., Marchand, A., Reinharz, D., & Savard, P. (2008). Cognitive-behavioral
1002/da.20121 treatment for panic disorder with agoraphobia. Behavior Modification, 32,
Lefebvre, C., Manheimer, E., & Glanville, J. (2011). Searching for studies. In: J. P. T. 333–351. http://dx.doi.org/10.1177/0145445507309025
Higgins, & S. Green (Eds.), Cochrane handbook for systematic reviews of Rosendahl, J., Barkowski, S., Schwartze, D., Koranyi, S., & Strauss, B. (2013).
interventions. Version 5.1.0. The Cochrane Collaboration [chapter 6] (updated Systematic review and meta-analyses on the efficacy of small group treatment for
March 2011). mental disorders PROSPERO: CRD42013004419. Available from http://www.crd.
Liebowitz, M. R., Heimberg, R. G., Schneier, F. R., Hope, D. A., Davies, S., & Klein, D. F. york.ac.uk/PROSPERO/display record.asp?ID=CRD42013004419.
(1999). Cognitive-behavioral group therapy versus phenelzine in social *Salaberría Irízar, K., & Echeburúa Odriozola, E. (1995). Tratamiento psicológico de
phobia: long term outcome. Depression and Anxiety, 10, 89–98. http://dx.doi. la fobia social: un estudio experimental. Análisis y Modificación de Conducta,
org/10.1002/(SICI) 1520-6394(1999)10:3<89:AID-DA1>3.0.CO;2-5 21(76), 151–179.
Lipsitz, J. D., & Schneier, F. R. (2000). Social phobia: epidemiology and cost of Sank, L. I., & Shaffer, C. S. (1984). A therapist’s manual for cognitive behavior therapy
illness. Pharmacoeconomics, 18(1), 23–32. in groups. New York: Basic Books.
Luborsky, L., Diguer, L., Seligman, D. A., Rosenthal, R., Krause, E. D., Johnson, S., & *Schmidt, N. B., Buckner, J. D., Pusser, A., Woolaway-Bickel, K., Preston, J. L., & Norr,
Schweizer, E. (1999). The researcher’s own therapy allegiances: a “wild card” A. (2012). Randomized controlled trial of false safety behavior elimination
in comparisons of treatment efficacy. Clinical Psychology: Science and Practice, therapy: a unified cognitive behavioral treatment for anxiety psychopathology.
6, 95–106. http://dx.doi.org/10.1093/clipsy.6.1.95 Behavior Therapy, 43(3), 518–532. http://dx.doi.org/10.1177/
*Mácia-Antòn, D., Olivares-Olivares, P. J., & Amoros-Boix, M. (2012). Detection and 0145445514553094
cognitive-behavioral therapy in young adults with generalized social phobia. *Scholing, A., & Emmelkamp, P. M. (1993). Exposure with and without cognitive
Anales De Psicologia, 28(1), 19–27. therapy for generalized social phobia: effects of individual and group
Mayo-Wilson, E., Dias, S., Mavranezouli, I., Kew, K., Clark, D. M., Ades, A. E., & Pilling, treatment. Behaviour Research & Therapy, 31(7), 667–681.
S. (2014). Psychological and pharmacological interventions for social anxiety *Scholing, A., & Emmelkamp, P. M. (1996). Treatment of generalized social phobia:
disorder in adults: a systematic review and network meta-analysis. Lancet results at long-term follow-up. Behaviour Research & Therapy, 34(5–6),
Psychiatry, 1, 368–376. http://dx.doi.org/10.1016/S2215-0366(14)70329-3 447–452. http://dx.doi.org/10.1016/0005-7967(96)00010-1
*McDougall, D. L. (1999). The group treatment of social phobia with Simon, N. M., Shear, M. K., Fagliolini, A., Frank, E., Zalta, A., Thompson, E. H., &
cognitive-behavioural therapy and imagery. Dissertation Abstracts Silowash, R. (2008). Impact of concurrent naturalistic pharmacotherapy on
International, 61(4-b), 2211. Retreived from http://onlinelibrary.wiley.com/o/ psychotherapy for complicated grief. Psychiatry Research, 159, 31–36. http://dx.
cochrane/clcentral/articles/118/CN-00710118/frame.html doi.org/10.1016/j.psychres.2007.05.011
*Mörtberg, E., Clark, D. M., & Bejerot, S. (2011). Intensive group cognitive therapy *Stangier, U., Heidenreich, T., Peitz, M., Lauterbach, W., & Clark, D. M. (2003).
and individual cognitive therapy for social phobia: sustained improvement at Cognitive therapy for social phobia: individual versus group treatment.
5-year follow-up. Journal of Anxiety Disorders, 25(8), 994–1000. http://dx.doi. Behaviour Research & Therapy, 41(9), 991–1007. http://dx.doi.org/10.1016/
org/10.1016/j.janxdis.2011.06.007 S0005-7967(02)00176-6
*Mörtberg, E., Clark, D. M., Sundin, O., & Aberg Wistedt, A. (2007). Intensive group Strauss, B., & Burlingame, G. M. (2012). Gruppenpsychotherapieforschung und
cognitive treatment and individual cognitive therapy vs. treatment as usual in Wirksamkeitsnachweise von Gruppenbehandlungen. In: B. Strauss, & D.
social phobia: a randomized controlled trial. Acta Psychiatrica Scandinavica, Mattke (Eds.), Gruppenpsychotherapie—Lehrbuch für die Praxis (pp. 193–211).
115(2), 142–154. http://dx.doi.org/10.1111/j.1600-0447.2006.00839.x Heidelberg: Springer.
*Mörtberg, E., Karlsson, A., Fyring, C., & Sundin, O. (2006). Intensive Swift, J. K., & Greenberg, R. P. (2012). Premature discontinuation in adult
cognitive-behavioral group treatment (CBGT) of social phobia: a randomized psychotherapy: a meta-analysis. Journal of Consulting and Clinical Psychology,
controlled study. Journal of Anxiety Disorders, 20(5), 646–660. http://dx.doi.org/ 80(4), 547–559. http://dx.doi.org/10.1037/a0028226
10.1016/j.janxdis.2005.07.005 Tefikow, S., Bormann, B., & Strauss, B. (2012). Gruppentherapie der Sozialphobie.
Moher, D., Alessandro, L., Tetzlaff, J., & Altman, D. G. (2009). Preferred reporting In: B. Strauss, & D. Mattke (Eds.), Gruppenpsychotherapie—Lehrbuch für die
items for systematic reviews and meta-analyses: the PRISMA statement. PLoS Praxis (pp. 243–258). Heidelberg: Springer.
Medicine, 6(7), e1000097. http://dx.doi.org/10.1371/journal.pmed.1000097 Viechtbauer, W. (2010). Conducting meta-analyses in R with the metaphor
Munder, T., Gerger, H., Trelle, S., & Barth, J. (2011). Testing the allegiance bias package. Journal of Statistical Software, 36(3), 1–48.
hypothesis: a meta-analysis. Psychotherapy Research, 21(6), 670–684. http://dx. Viechtbauer, W. (2010b). metafor: meta-analysis package for R. R package version
doi.org/10.1080/10503307.2011.602752 1.4-0. Available from http://CRAN.R-project.org/package=metafor.
*Newman, M. G., Hofmann, S. G., Trabert, W., Roth, W. T., & Taylor, C. B. (1994). Wampold, B. E., Minami, T., Callen Tierney, S., Baskin, T. W., & Bhati, K. S. (2005).
Does behavioral treatment of social phobia lead to cognitive changes. Behavior The placebo is powerful: estimating placebo effects in medicine and
Therapy, 25(3), 503–517. http://dx.doi.org/10.1016/s0005-7894(05)80160-1 psychotherapy from randomized clinical trials. Journal of Clinical Psychology,
NICE (2013). Social anxiety disorder: recognition, assessment and treatment. NICE 61(7), 835–854. http://dx.doi.org/10.1002/jclp.20129
clinical guideline 159. Available at guidance.nice.rorg.uk/cg159 [NICE guideline]. Wampold, B. E., Mondin, G. W., Moody, M., Stich, F., Benson, K., & Ahn, H. (1997). A
*Nortje, C., Posthumus, T., & Moller, A. T. (2008). Comparison of integrated meta-analysis of outcome studies comparing bona fide psychotherapies:
cognitive restructuring plus exposure with exposure alone in group treatment empirically, ‘all must have prizes.’. Psychological Bulletin, 122(3), 203–215.
of generalised social anxiety disorder. South African Journal of Psychology, 38(4), http://dx.doi.org/10.1192/pb.bp.109.025643
647–658. http://dx.doi.org/10.1177/008124630803800405 Wersebe, H., Sijbrandij, M., & Cuijpers, P. (2013). Psychological group-treatments
*Olivares, J., Rosa-Alcázar, A. I., Olivares-Olivares, P. J., & Rosa-Alcázar, Á. (2009). of social anxiety disorder: a meta-analysis. PLoS One, 8(11), e79034. http://dx.
Treatment of young adults with generalized social phobia. International Journal doi.org/10.1371/journal.pone.0079034
of Hispanic Psychology, 1(2), 1–14. http://dx.doi.org/10.4067/S0718- Westen, D., & Morrison, K. (2001). A multidimensional meta-analysis of treatments
48082012000300006 for depression, panic, and generalized anxiety disorder: an empirical
examination of the status of empirically supported therapies. Journal of
Consulting and Clinical Psychology, 69(6), 875–899. http://dx.doi.org/10.1037// *Wong, D. F. K., & Sun, S. Y. K. (2006). A preliminary study of the efficacy of group
0022-006X.69.6.875 cognitive-behavioural therapy for people with social anxiety in Hong Kong.
Wittchen, H. U., & Fehm, L. (2003). Epidemiology and natural course of social fears Hong Kong Journal of Psychiatry, 16(2), 50–56.
and social phobia. Acta Psychiatrica Scandinavica, 108(Suppl. 417), 4–18. http:// Yalom, I. D., & Leszcz, M. (2005). The theory and practice of group psychotherapy.
dx.doi.org/10.1034/j.1600-0447.108.s417.1.x New York: Basic Books.

Terapia Grupal en TAS

Uploaded by

Copyright:

Available Formats

Terapia Grupal en TAS

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Terapia Grupal en TAS

Uploaded by

Copyright:

Available Formats

Journal of Anxiety Disorders 39 (2016) 44–64

Contents lists available at ScienceDirect

Journal of Anxiety Disorders

Efﬁcacy of group psychotherapy for social anxiety disorder: A

∗ Corresponding author. Fax: +49 3641 936546.

1. Introduction elicit, as well as from the experience of mutual support (Teﬁkow

2.2.1. Coding of outcomes 2.3. Summary measures

2.2.3. Coding of risk of bias 2.4. Data synthesis

S. Barkowski et al. / Journal of Anxiety Disorders 39 (2016) 44–64

S. Barkowski et al. / Journal of Anxiety Disorders 39 (2016) 44–64

S. Barkowski et al. / Journal of Anxiety Disorders 39 (2016) 44–64

S. Barkowski et al. / Journal of Anxiety Disorders 39 (2016) 44–64

S. Barkowski et al. / Journal of Anxiety Disorders 39 (2016) 44–64

S. Barkowski et al. / Journal of Anxiety Disorders 39 (2016) 44–64

Fig. 1. Flow chart displaying the process of study inclusion.

No. of patients per group 6 1.4 34

Type of treatment Cognitive-behavioral 36 80.0

Therapist profession Students (graduate and undergraduate) 7 15.6

Leadership Single-led 6 13.3

Individual session Yes 8 17.8

Manualization Manualized and adherence checked 28 62.2

3.4.1.2. Clinical and methodological diversity. Because statistical

Fig. 2. Distribution of risk of bias in included studies.

Comparisons to wait list control Comparisons to active treatment groups

3.7. Follow-up 4. Discussion

Short-term follow-up Mid-term follow-up

Speciﬁc symptomatology General psychopathology

Measures ADIS-IV Fear/Avoidance, BAIo , BDIp , FQ-Depression

You might also like