Objective: The authors examined the comparative risks of switches in mood polarity into hypomania or mania during acute and continuation trials of adjunctive antidepressant treatment of bipolar depression. Method: One hundred fifty-nine patients with bipolar I disorder or bipolar II disorder participated in a total of 228 acute (10-week) randomized trials of bupropion, sertraline, or venlafaxine as an adjunct to a mood stabilizer. Patients in 87 of these trials entered continuation treatment for up to 1 year. Antidepressant response and the occurrence of subthreshold brief hypomania (emergence of brief hypomania [at least 1 but <7 days] or recurrent brief hypomania) and threshold switches (emergence of full-duration hypomania [≥7 days] or mania) were blindly assessed by using clinician-rated daily reports of mood-associated dysfunction on the National Institute of Mental Health Life Chart Method. Results: Threshold switches into full-duration hypomania and mania occurred in 11.4% and 7.9%, respectively, of the acute treatment trials and in 21.8% and 14.9%, respectively, of the continuation trials. The rate of threshold switches was higher in the 169 trials in patients with bipolar I disorder (30.8%) than the 59 trials in patients with bipolar II disorder (18.6%). The ratio of threshold switches to subthreshold brief hypomanias was higher in both the acute (ratio=3.60) and continuation trials (ratio=3.75) of venlafaxine than in the acute and continuation trials of bupropion (ratios=0.85 and 1.17, respectively) and sertraline (ratios=1.67 and 1.66, respectively). In only 37 (16.2%) of the original 228 acute antidepressant trials, or in only 23.3% of the patients, was there a sustained antidepressant response in the continuation phase in the absence of a threshold switch. Conclusions: Adjunctive treatment with antidepressants in bipolar depression was associated with substantial risks of threshold switches to full-duration hypomania or mania in both acute and long-term continuation treatment. Of the three antidepressants included in the study, venlafaxine was associated with the highest relative risk of such switching and bupropion with the lowest risk.