Thirteen RCTs were included; a total of 1,624 participants were randomised to PST (n=588), other active interventions (n=491) or control conditions (n=545). Between the text and the tables there was some discrepancy in the numbers of studies and participants included.
Study quality was variable. Only 5 studies conducted ITT analysis: the rest analysed completers only. Only 3 studies reported using allocation concealment, none described blinding of the participants and 6 reported blinding of the assessors. The drop-out rate ranged from nil to 43%.
PST versus control conditions.
When the comparisons of PST versus control conditions were pooled (12 studies, n=1,053), the mean post-treatment ES was 0.34 (95% confidence interval, CI: 0.02, 0.48) using a fixed-effect model and 0.83 (95% CI: 0.45, 1.21) using a random-effects model. This suggested that PST has a moderate-to-large short-term effect on depressive symptoms, though the results varied according to the method of analysis used. Heterogeneity was high (Q statistic, p<0.001; I2 = 82.8%). Subgroup analysis showed significantly higher ES for group interventions and for studies using SPST. The ES were lower for studies of participants meeting the criteria for major depression, studies using ITT analysis and studies in which the control group had a pill placebo. Heterogeneity was moderate to high (I2 = 50% or above) amongst the studies selected for subgroup analysis.
Sensitivity analyses did not materially affect any of these results.
PST versus active comparators.
When PST was compared directly with other psychological treatments (5 studies) and with antidepressants (4 studies), a significant difference was found between the groups only in the fixed-effect model (ES 0.29, 95% CI: 0.07, 0.50) comparing PST with other psychological treatments. Heterogeneity was moderately high (I2 = 59.5%).
Further improvements were noted in the 5 studies that reported 3- to 12-month outcomes, but there were too few studies to draw any definite conclusions.