Background: Bipolar disorders are amongst the most common, severe and chronic mental health conditions, often associated with unpredictable illness trajectories. Temperament, as a relatively stable however underutilized affective trait in clinical practice, has been proposed as a potential modifier of illness course in BD. The current study sought to classify and examine temperament as a predictive or distinguishing factor in the course of bipolar illness. Methods: Ninety-one initially euthymic individuals with BD I or II were followed for a period of two years. Latent profile analysis (LPA) was conducted based on self-reported temperament (TEMPS-A scale) and used to identify temperament-based subgroups. Longitudinal illness burden was measured using an adapted Morbidity Index (MI), and stepwise multiple regression was performed to test whether temperament classes predicted MI scores. Results: The LPA identified three distinct classes: moderate-reactive, reactive-instable, and resilient-hyperthymic temperament. Individuals in the moderate-reactive (vs. the resilient-hyperthymic) class showed significantly higher MI scores, indicating a worse longitudinal illness course. Conclusion: Temperament classes – reflecting combinations of traits rather than single dimensions alone – may serve as clinically meaningful predictors of illness burden in BD. The potential improvement of incorporating temperament profiles into clinical assessment to identify high-risk patients and guide more tailored treatment approaches is discussed.