Validating a two-dimensional bipolar spectrum model integrating DSM-5's mixed features specifier for Major Depressive Disorder

Introduction: The literature on DSM-5's 'Major Depressive Disorder with lifetime mixed features' (MDD-MF) is limited. This study investigated MDD-MF's potential inclusion into a bipolar spectrum. Methods: We recruited 287 patients with Bipolar I disorder (BD-I), BD-II, MDD-MF or 'MDD without lifetime mixed features' (MDD-noMF); most (N = 280) were stabilized for at least one year on medication. Sixteen validators (clinical features, psychiatric family history, temperament, stabilizing treatment) were compared across groups and subjected to trend analyses. Two discriminant function analyses (DFA; primary and secondary), excluding or including, respectively, treatment-related predictors, explored latent dimensions maximizing between group discrimination; mahalanobis distances between group 'centroids' were calculated. Results: Eleven validators differed significantly across groups; nine varied monotonically along a bipolar diathesis gradient with significant linear trends; two peaked at MDD-MF and displayed significant quadratic trends. In the primary DFA, apart from a classic bipolarity dimension, correlating with hospitalizations, early age at onset, lifetime psychosis and lower anxious temperament scores, on which groups ranked along a bipolar propensity gradient, a second dimension was also significant, peaking at BD-II and MDD-F (challenging the classic bipolar ranking), which correlated with lifetime psychiatric comorbidities, suicidality, lower lifetime psychosis rates, female gender, higher cyclothymic and lower depressive temperament scores; MDD-MF was equipoised amidst BD-II and MDD-noMF. After including treatment related predictors (secondary DFA), discrimination improved overall but BD-II and MDD-MF were closest than any other pair, suggesting similar treatment patterns for these two groups at this naturalistic setting. Conclusions: To our knowledge, this is the first time a two-dimensional bipolar spectrum based on classic external validators is proposed, fitting the data better than a unidimensional model. Additional predictors are warranted to improve BD-II/MDD-MF discrimination. (C) 2017 Elsevier Inc. All rights reserved.