The aim of this study is to investigate possible clinical predictors of the long-term outcome of mood disorders. We undertook a retrospective assessment of 426 inpatients affected by major depressive disorder (n=182) and bipolar disorder (n=244), with at least two episodes of illness alternating with complete recovery; subjects were affected for an average of 14.43+/-11.34 years and presented an average of 4.4+/-2.1 episodes. Random regression model analysis (http://www.uic.eu/hedeker/mix.html) was used to investigate the longitudinal time course of the illness. A progressive cycle shortening was observed, whereby the more episodes a subject experienced, the shorter the interval was between episodes, up to a plateau frequency of one episode/year on average. Bipolar diagnosis was the strongest predictive factor toward high frequency of episodes; a manic onset among bipolars was associated with an even worse outcome. Gender, education level, family history, duration of the first interval, severity of the first episode, lifetime mean severity and lifetime mean treatment level were not associated with outcome in terms of episode frequency. Our results suggest that recurrent affective disorders recruited in a clinical setting have a marked deteriorating mean time course.