OBJECTIVE: Antidepressant safety and efficacy remain controversial for the treatment of bipolar depression. The present study utilized data from the National Institute of Mental Health Systematic Treatment Enhancement Program for Bipolar Disorder (STEP-BD) to examine the prevalence and clinical correlates of self-reported switch into mania/hypomania during antidepressant treatment. METHOD: Antidepressant treatment histories were examined from intake assessments for the first 500 subjects enrolled into the STEP-BD between November 1999 and November 2000. Affective switch was defined as a report of mania, hypomania, or mixed episodes within the first 12 weeks of having started an antidepressant. Demographic and clinical characteristics were compared for subjects with or without a history of acute switch during antidepressant treatment. RESULTS: Among the 338 subjects with prior antidepressant treatment and complete data on switch event outcomes, 44% reported at least 1 such occurrence. Patients with a shorter duration of illness (odds ratio [OR] = 1.02, 95% CI = 1.01 to 1.04) and a history of multiple antidepressant trials (OR = 1.73, 95% CI = 1.38 to 2.16) were more likely to report a history of switch than other patients. A significantly increased risk for affective polarity switch was identified in patients who had ever switched to mania/hypomania while taking tricyclic antidepressants (OR = 7.80, 95% CI = 1.56 to 28.9), serotonin reuptake inhibitors (OR = 3.73, 95% CI = 1.98 to 7.05), or bupropion (OR = 4.28, 95% CI = 1.72 to 10.6). Switch was less common during treatment with electroconvulsive therapy or monoamine oxidase inhibitors than other antidepressants. CONCLUSIONS: Antidepressants are associated with the potential risk for treatment-emergent mania or hypomania, particularly in bipolar patients with short illness duration, multiple past antidepressant trials, and past experience of switch with at least one antidepressant.