Maintenance treatments in bipolar disorders and schizophrenia are securely established, and their discontinuation is associated with high but modifiable risk of early relapse. The benefits of long-term antidepressant treatment in major depression and the risks of discontinuing medication at various times after clinical recovery from acute depression are not as well defined. Computerized searching found 27 studies with data on depression risk over time including a total of 3037 depressive patients treated for 5.78 (0-48) months and then followed for 16.6 (5-66) months with antidepressants continued or discontinued. Compared with patients whose antidepressants were discontinued, those with continued treatment showed much lower relapse rates (1.85 vs. 6.24%/month), longer time to 50% relapse (48.0 vs. 14.2 months), and lower 12-month relapse risk (19.5 vs. 44.8%) (all p < 0.001). However, longer prior treatment did not yield lower postdiscontinuation relapse risk, and differences in relapses off versus on antidepressants fell markedly with longer follow-up. Contrary to prediction, gradual discontinuation (dose-tapering or use of long-acting agents) did not yield lower relapse rates. Relapse risk was not associated with diagnostic criteria. More previous illness (particularly three or more prior episodes or a chronic course) was strongly associated with higher relapse risk after discontinuation of antidepressants but had no effect on response to continued treatment; patients with infrequent prior illness showed only minor relapse differences between drug and placebo treatment.