Twenty-three RCTs were included. There were 4,209 participants (range 15 to 2,481, most trials included fewer than 100 participants). Eight RCTs reported adequate allocation. Concealment of allocation was not possible or not reported in any of the trials. Blinding was reported in nine trials, was not possible in nine trials as self-report measures were used and was unreported in five trials. Loss to follow-up ranged from 0% to 41%. Fourteen studies used intention-to-treat analyses.
Psychotherapy was significantly more effective compared to controls. There was a large overall effect size of 1.00 (95% confidence interval (CI) 0.57 to 1.44; 15 RCTs with 18 comparisons). Because of the very high heterogeneity (I2=92.5%) an additional analysis was conducted without two possible outliers and this resulted in a moderate overall effect size of 0.42 (95% CI 0.27 to 0.58) and low heterogeneity (I2=25.9%).
There was no difference between effect sizes calculated from the lowest and highest reported results. None of the subgroup analyses significantly altered the findings.
Because there was some indication of asymmetry in the funnel plot, a trim-and-fill analysis was performed using five imputed studies with negative results. The results of this analysis showed a small positive effect size in the psychotherapy group compared to controls (effect size 0.31, 95% CI 0.14 to 0.48).
CBT or interpersonal therapy was significantly more effective when compared to supportive therapy or information sessions (effect size 0.42, 95% CI, 0.14 to 0.69; five RCTs with six comparisons).
Results from the three RCTs (four comparisons) that compared psychotherapies to pharmacotherapy were reported narratively.