Ninety RCTs (10,534 participants, 168 effects) were included in the review. Seventy-nine of 168 effects reported blinding to allocation concealment and 53 reported ITT analysis. The mean exercise training adherence rate was 77%.
Exercise training statistically significantly reduced depression (mean effect size 0.30, 95% CI 0.25 to 0.36, NNT=6, I2=47%; 168 effects). Sensitivity analyses did not alter the findings significantly.
Meta-regression indicated that the largest reductions in depression were shown in patients with higher depression symptom scores at baseline and where the primary outcomes assessed in the trials showed significant improvements. A large number of other results on moderator variables were reported in the review.
There was no evidence of significant publication bias.