Psychological and pharmacological interventions for social anxiety disorder in adults: a systematic review and network meta-analysis

Summary Background Social anxiety disorder—a chronic and naturally unremitting disease that causes substantial impairment—can be treated with pharmacological, psychological, and self-help interventions. We aimed to compare these interventions and to identify which are most effective for the acute treatment of social anxiety disorder in adults. Methods We did a systematic review and network meta-analysis of interventions for adults with social anxiety disorder, identified from published and unpublished sources between 1988 and Sept 13, 2013. We analysed interventions by class and individually. Outcomes were validated measures of social anxiety, reported as standardised mean differences (SMDs) compared with a waitlist reference. This study is registered with PROSPERO, number CRD42012003146. Findings We included 101 trials (13 164 participants) of 41 interventions or control conditions (17 classes) in the analyses. Classes of pharmacological interventions that had greater effects on outcomes compared with waitlist were monoamine oxidase inhibitors (SMD −1·01, 95% credible interval [CrI] −1·56 to −0·45), benzodiazepines (−0·96, −1·56 to −0·36), selective serotonin-reuptake inhibitors and serotonin–norepinephrine reuptake inhibitors (SSRIs and SNRIs; −0·91, −1·23 to −0·60), and anticonvulsants (−0·81, −1·36 to −0·28). Compared with waitlist, efficacious classes of psychological interventions were individual cognitive–behavioural therapy (CBT; SMD −1·19, 95% CrI −1·56 to −0·81), group CBT (−0·92, −1·33 to −0·51), exposure and social skills (−0·86, −1·42 to −0·29), self-help with support (−0·86, −1·36 to −0·36), self-help without support (−0·75, −1·25 to −0·26), and psychodynamic psychotherapy (−0·62, −0·93 to −0·31). Individual CBT compared with psychological placebo (SMD −0·56, 95% CrI −1·00 to −0·11), and SSRIs and SNRIs compared with pill placebo (−0·44, −0·67 to −0·22) were the only classes of interventions that had greater effects on outcomes than appropriate placebo. Individual CBT also had a greater effect than psychodynamic psychotherapy (SMD −0·56, 95% CrI −1·03 to −0·11) and interpersonal psychotherapy, mindfulness, and supportive therapy (−0·82, −1·41 to −0·24). Interpretation Individual CBT (which other studies have shown to have a lower risk of side-effects than pharmacotherapy) is associated with large effect sizes. Thus, it should be regarded as the best intervention for the initial treatment of social anxiety disorder. For individuals who decline psychological intervention, SSRIs show the most consistent evidence of benefit. Funding National Institute for Health and Care Excellence.


I. Priors
The between-trials standard deviation was given a Uniform (0,5) prior.
The within-class mean treatment effects were given vague priors and the within-class variability parameters had priors with a=3.9 and b=0.35 chosen so that the mean of the within class standard deviation is the same as the posterior mean of the between-trial standard deviation (estimated in a previous run of the model without class effects) and the Credible interval can go from approximately half to double that mean.

Botella 2010 21
Not an eligible intervention (intervention for public speaking anxiety only). Butler 1984 22 No usable data (outcomes not reported in a format that can be extracted for meta-analysis).
Carlbring ongoing 24 Ongoing study of internet-based CBT compared with attention training.
Cassin 2011 25 Not an eligible intervention (not a bona fide therapy, outcomes measured immediately after intervention).

Clark ongoing 27
Ongoing study of internet-based cognitive therapy, face-to-face cognitive therapy, and waitlist.

Davidson 2002 29
Could not locate full report. Unclear if this is related to a RCT of venlafaxine.

D'El Rey 2007 30
The reported data are implausible (significant at the level p=0.000002 with 30 particiapnts) and could not be verified.

De Oliviera 2012 31
Interventions not connected (two unique forms of psychotherapy).

Donahue 2009 32
Not an eligible intervention (single dose of quetiapine).

Dogaheh 2011 33
No usable data (study compares group and individual CBT, but only the Brief Fear of Negative Evaluation Scale is reported as a measure of social anxiety).

Furmark 2005 35
Not an eligible intervention (NK1A). Emmelkamp 1985 36 No usable data (paper reports means but not SDs; we did not impute SDs to connect Rational Emotive Therapy and Self-Instructional Training to the network).

Hayes 2006 37
No usable data (intervention for social anxiety and comorbid alcohol misuse; no social anxiety outcomes reported).

Heideman 2008 39
No usable data (intervention for social anxiety and comorbid alcohol misuse; no social anxiety outcomes reported).

Hofmann 2004 40
No usable data (compares two types of group therapy with waitlist; paper reports mediation analyses, but not main effects for each intervention).

Johnston 2011 41
No usable data (outcomes collapsed across social anxiety and other diagnostic groups). Katzenick 1995 42 No usable data (outcomes reported only after crossover).
Krylov 1996 44 Could not locate full report (compares alprazolam and buspirone).
Lee 1997 45 Could not locate full report (compares alprazolam and buspirone).
Lott 1997 46 Not an eligible intervention (short-term study of a beta-blocker).

Marks 2004 47
No usable data (computerised therapy for mixed population; disaggregated data not available). Mersch 1995 48 No usable data (waitlist group were assigned after the first phase of the study; data from two treatments merged for comparison with waitlist).

Mortberg 2006 49
No usable data (reports F-tests only for significant interactions, so these were not extracted).

Muehlbacher 2005 50
The reported data are implausible. The journal and the authors were unable to verify details upon request.

Mulkens 2001 51
Not an eligible intervention (intervention for fear of blushing only).

Norton 2012 53
No usable data. Although primary diagnosis was not related to improvement with a transdiagnostic intervention (n=12) or a diagnosis-specific intervention (n=13), change on the measure of social anxiety non-significantly favoured the specific intervention. Norwegian University of Science and Technology 2011 54 Could not locate full report (compares paroxetine, cognitive therapy, paroxetine with cognitive therapy, and pill placebo). Noyes 1997 55 No usable data (includes figures, but raw scores not given).
Ontiveros 2008 56 Assigned in groups of 10, but method of assignment not reported and not clear from personal correspondence (unclear if random).

Peng 2003 57
Could not locate full report. Unclear if this is a RCT.

Pollack 2013 58
Not an eligible intervention (second-line treatment for sertraline non-responders).

Randall 2001 59
No usable data (paroxetine for social anxiety and comorbid alcohol misuse; no social anxiety outcomes reported).

Schneider 2005 62
No usable data (computerised therapy for mixed population; disaggregated data not available for 24 participants with primary social anxiety disorder). Shaw 1979 63 No usable data (desensitization, flooding, and social skills training; 30 participants with social phoabia). Solyom 1981 64 No usable data (medication for mixed population; disaggregated data not available for 40 participants with agoraphobia or social phobia).
Taylor 1997 65 Not an eligible intervention (compared exposure with or without cognitive restructuring, but explicitly not as restructuring is used in clinical practice).

Taylor 2010 67
Not an eligible intervention (compared two forms of exposure in a laboratory-based interaction to explore mediators of change).
Turner 1994 68 Not an eligible intervention (short-term study of a beta-blocker). Tyrer 1973 69 No usable data (medication for mixed population; disaggregated data not available for 15 particaipnts with social anxiety disorder).

Van Ameringen ongoing 72
Ongoing study of sertraline and placebo.

Yang 1999 76
Could not locate full report (compares systematic desensitization with or without therapist accompaniment).

Yoshinaga ongoing 77
Ongoing study of CBT for sertraline non-responders.