Eight RCTs (719 participants, range 25 to 179) were included in the review. The authors reported that all RCTs had a high risk of bias: allocation sequence generation was reported in three RCTs, allocation concealment in three RCTs, intention-to-treat analysis in three RCTs and blinding in three RCTs; none of the studies reported on all the measures. One study was considered to be free of selective outcome measure reporting.
Depressive symptoms at cessation of treatment (five RCTs): Trials that used the Hamilton Rating Scale for Depression reported a significant reduction for depressive symptoms for cognitive therapy compared with treatment as usual at the end of therapy (MD -2.15, 95% CI -3.70 to -0.60; four RCTs, one RCT contributed two data sets, fixed-effects model). There was no evidence of statistical heterogeneity (I2=0%) and overall results did not differ when the random-effects model was used. However overall results using Beck Depression Inventory found no significant difference between cognitive therapy and care as usual at the end of therapy (MD -4.85, 95% CI -12.08 to 2.39; four RCTs, random-effects model). There was significant evidence of statistical heterogeneity for this analysis (I2=89%). Sensitivity analysis that removed one trial removed statistical heterogeneity (I2=0%), but the result was still not significant with random-effects and with fixed-effect models (MD -1.57, 95% CI -4.30 to 1.16; three trials). Trial sequential analysis reported that insufficient data had been obtained to decide whether cognitive therapy was superior to treatment as usual.
Follow-up (two RCTs): One study reported an increased probability of remission at 12 months compared to community care. One RCT found no significant differences between groups.
Adverse events (one RCT): Two participants dropped out of the control group.
Quality of life (one RCT): There was no significant difference between groups at cessation of treatment.
Participants without remission (two RCTs): There were no significant differences between groups for risk of no remission.
Results of subgroup analyses were reported.