Features preceding diagnosis of bipolar versus major depressive disorders.

BACKGROUND: Better and earlier predictive differentiation of bipolar (BD) vs. unipolar major depressive disorder (UD) diagnoses should improve long-term clinical planning. METHODS: We reviewed randomly selected clinical records of 334 adults diagnosed with DSM-IV-TR BD-I (n=109), BD-II (n=106), and UD (n=119) and compared features preceding major affective episodes or diagnoses, using bivariate, multivariate, and Bayesian methods. RESULTS: We identified antecedents selectively associated with later BD vs. UD in 52.6% vs. 31.1% of subjects in childhood, starting at age 7.4 years, and 60.0% vs. 32.8% in adolescence, with far more features in BD than UD cases (10.3 vs. 4.64/100 person-years; p<0.001). In multivariate modeling, BD-selective factors were: younger at first clinical event > male sex > family BD-history > cyclothymic or hyperthymic temperament > antecedents/person-year. Nonaffective (anxiety, eating, or substance-use) disorders preceded BD vs. UD in 41.4% vs. 28.6% of subjects (p=0.02). By ROC analysis, differential prediction of BD vs. UD was optimal with any ≥ 3 factors/person. LIMITATIONS: The validity and timing of antecedent events and factors identified retrospectively from clinical records could not be verified independently, but information was recorded systematically and consistently by a single mood-disorder expert prior to diagnosis, and extracted by two independent observers. COMMENT: Early clinical features distinguished later BD from UD, often by years. Such prediction should improve treatment-planning and limit risk of mood-switching.