Objectives: Although cognitive behaviour therapy (CBT) and pharmacotherapy are equally effective in the acute treatment of adult depression, it is not known how they compare across the longer term. In this meta-analysis, we compared the effects of acute phase CBT without any subsequent treatment with the effects of pharmacotherapy that either were continued or discontinued across 6-18 months of follow-up. Design: We conducted systematic searches in bibliographical databases to identify relevant studies, and conducted a meta-analysis of studies meeting inclusion criteria. Setting: Mental healthcare. Participants: Patients with depressive disorders. Interventions: CBT and pharmacotherapy for depression. Outcome measures: Relapse rates at long-term follow-up. Results: 9 studies with 506 patients were included. The quality was relatively high. Short-term outcomes of CBT and pharmacotherapy were comparable, although drop out from treatment was significantly lower in CBT. Acute phase CBT was compared with pharmacotherapy discontinuation during follow-up in eight studies. Patients who received acute phase CBT were significantly less likely to relapse than patients who were withdrawn from pharmacotherapy (OR=2.61, 95% CI 1.58 to 4.31, p<0.001; numbersneeded- to-be-treated, NNT=5). The acute phase CBT was compared with continued pharmacotherapy at follow-up in five studies. There was no significant difference between acute phase CBT and continued pharmacotherapy, although there was a trend (p<0.1) indicating that patients who received acute phase CBT may be less likely to relapse following acute treatment termination than patients who were continued on pharmacotherapy (OR=1.62, 95% CI 0.97 to 2.72; NNT=10). Conclusions: We found that CBT has an enduring effect following termination of the acute treatment. We found no significant difference in relapse after the acute phase CBT versus continuation of pharmacotherapy after remission. Given the small number of studies, this finding should be interpreted with caution pending replication.