OBJECTIVES: There are almost no data addressing antidepressant's long-term prophylaxis use that both establishes the mood stability and delays depressive episodes in patients with bipolar disorder. This study concentrated on the question that whether patients who continued taking antidepressant beyond six-months after depressive remission are less likely to have depressive relapse than patients who discontinued early, with a focus on modern second-generation antidepressant medications. METHODS: A total of 589 bipolar subjects were identified with interested antidepressant use after a depressive remission, followed by at least six-months of continuous enrollment in a national managed care plan between January, 1998 and December, 2002. Duration of pharmacotherapy was defined based on the computerized diagnosis and pharmacy records. A Cox proportional hazard model was developed to predict time from depressive remission to next depressive relapse with continuous antidepressant use either longer or shorter than six-months. Propensity score method with greedy matching was employed in addition to further balance the observed background covariates and baseline disease severity between comparison groups. RESULTS: The Kaplan-Meier estimate of the sample from propensity score matching showed that time to 50% survival with continuation and discontinuation groups were 16.5 months and 6.8 months respectively. The log-rank homogeneity test of survival curves indicated a significant difference (p<0.05). The Cox model identified a significantly lower risk of depressive relapse among those who continued antidepressant treatment beyond six-months after remission than those who discontinued treatment within six-months, with a statistically significant hazard ratio of 0.61 (95% CI: 0.42-0.88). CONCLUSIONS: This study suggests the potential adverse outcome of removing an antidepressant treatment after depressive remission in patients with bipolar disorder. Given the concerns regarding a risk of switching into mania by antidepressant use, an optimal prophylaxis treatment after depressive remission should balance risks between depressive relapse and manic switch.