Thirty-four RCTs (n=2,055) were included in the review. Attrition rates ranged from zero to 78%. Study quality was low in five RCTs (score of 5), moderate in 24 (score between 6 and 11) and high in five RCTs (score 12 to 15). Six RCTs met criteria required by the American Psychological Association’s Division 12 Task Force. Follow-up ranged from post intervention up to 24 months.
Behavioural activation versus control (16 RCTs, n=453): At post treatment, behavioural activation interventions showed a significantly greater beneficial effect compared with controls using a random effects model (0.87, 95% CI 0.60 to 1.14, I2=43%; 16 RCTs). Results remained significant for subgroup analysis for pleasant activities, self-control and contextual approaches; only one study assessed contextual approaches. There was some evidence of publication bias and the trim and fill method suggested that four studies were missing. The effects remained significant at one to three month follow-up with a fixed-effect model, but not when a random effects model was used (I2=77%, five RCTs) and were not significant at seven to 12 months (I2=73%, two RCTs).
Behavioural activation versus CBT/CT (15 RCTs, n=536): At post treatment there were no statistically significant differences between treatment groups and this remained the same in subgroup analysis. Sensitivity analysis did not alter the findings significantly. There was some evidence of publication bias, but the trim and fill method suggested no studies were missing. Results were similar between one and 24 months follow-up and remained non-significant.
Behavioural activation versus psychotherapy or other interventions (17 RCTs, n=533): At post treatment, behavioural activation interventions showed a medium effect size compared with other treatment (0.31, 95% CI 0.06 to 0.55, I2=58%; 17 RCTs). Subgroup analysis showed similar findings for pleasant activities, but results were no longer significantly different for self-control and behavioural activation treatment for depression (although the authors reported that the effect sizes were medium to large). There was no evidence of selection bias. Follow-up between one and 12 months showed no statistically significant differences between treatment groups.
Linear regression or Pearson’s correlation coefficient did not indicate that the study characteristics assessed influenced the findings (as reported in the review). Other results were reported in the review.