OBJECTIVE: The longitudinal course of 51 patients with treatment- refractory bipolar disorder was examined to assess possible effects of heterocyclic antidepressants on occurrence of manic episodes and cycle acceleration. METHOD: Using criteria established from life charts, investigators rated the patients' episodes of mania or cycle acceleration as likely or unlikely to have been induced by antidepressant therapy. Discriminant function analyses were performed to assess predictors of vulnerability to antidepressant-induced mania or cycle acceleration. Further, the likelihood of future antidepressant- induced episodes in persons who had had one such episode was assessed. RESULTS: Thirty-five percent of the patients had a manic episode rated as likely to have been antidepressant-induced. No variable was a predictor of vulnerability to antidepressant-induced mania. Cycle acceleration was likely to be associated with antidepressant treatment in 26% of the patients assessed. Younger age at first treatment was a predictor of vulnerability to antidepressant-induced cycle acceleration. Forty-six percent of patients with antidepressant-induced mania, but only 14% of those without, also showed antidepressant- induced cycle acceleration at some point in their illness. CONCLUSIONS: Mania is likely to be antidepressant-induced and not attributable to the expected course of illness in one-third of treatment-refractory bipolar patients, and rapid cycling is induced in one-fourth. Antidepressant-induced mania may be a marker for increased vulnerability to antidepressant-induced cycle acceleration. Antidepressant-induced cycle acceleration (but not antidepressant- induced mania) is associated with younger age at first treatment and may be more likely to occur in women and in bipolar II patients.