© 2018 John Wiley & Sons A/S. Published by John Wiley & Sons Ltd Introduction: Depressive episodes are often prevalent among patients with bipolar disorder, but little is known regarding the differential patterns of development over time. We aimed to determine and characterize trajectories of depressive symptoms among adults with bipolar disorder during 6 months of systematic treatment. Methods: The pragmatic clinical trial, Bipolar Clinical Health Outcomes Initiative in Comparative Effectiveness (CHOICE), randomized 482 outpatients with bipolar disorder to lithium or quetiapine. Depressive symptoms were rated at up to 9 visits using the Montgomery-Asberg Depression Rating Scale (MADRS). Growth mixture modeling was utilized to identify trajectories and multinomial regression analysis estimated associations with potential predictors. Results: Four distinct trajectories of depressive symptoms were identified. The responding class (60.3%) with a rapid reduction and subsequent low level; the partial-responding class (18.4%) with an initial reduction followed by an increase during the remaining weeks; the fluctuating class (11.6%) with a fluctuation in depressive symptoms; and the non-responding class (9.7%) with sustained moderate-severe depressive symptoms. Bipolar type I predicted membership of the non-responding class and randomization to quetiapine predicted membership of either the responding or the non-responding class. Conclusion: Approximately 30% experienced a partial or fluctuating course, and almost 10% had a chronic course with moderate-severe depression during 6 months. Patients diagnosed with bipolar type 1 had higher risk of being categorized into a class with a worse outcome. While no differences in average overall outcomes occurred between the lithium and quetiapine groups, trajectory analysis revealed that the lithium group had more variable courses.