Fig. 3: Metabolomic predictors of recurrence in remitted recurrent major depressive disorder (rrMDD).

A Kaplan–Meier analysis of latency to recurrence in subjects with remitted recurrent major depressive disorder (rrMDD), Females, B Males. Dotted boundaries indicate the 95% confidence intervals. Metabolic predictors of recurrence in recurrent major depressive disorder. Females: C, E, G. Males: D, F, H. CD Multivariate metabolomic discrimination of subjects with rMDD who experienced recurrence in the next 2.5 years, and those who did not, analyzed by partial least squares discriminant analysis. EF Rank order of top 15 discriminating metabolites by variable importance in projection (VIP) scores. GH Bubble impact plot of pathway alterations. Receiver operator characteristic (ROC) curve analysis of multianalyte diagnostic classifiers for rrMDD. I. The classifier for females used 7 metabolites. J The classifier for males used three metabolites. AUC area under the curve, rdCV repeated double cross validation accuracy. KL 2 × 2 contingency table analysis. Cox proportional hazard analysis of selected metabolites. M Decreased methylcysteine predicted a higher risk of recurrence in females. N Decreased 15-hydroxyeicosatetraenoic acid (15-HETE) predicted a higher risk of recurrence in males. O Decreased monohexosyl ceramide (MHC(d18:1/20:0)) predicted a higher risk of recurrence in females. P Decreased β-carotene predicted a higher risk of recurrence in males. rrMDD subjects were followed prospectively for 2.5 years: n = 42 females (24 with recurrence, 18 no recurrence), 20 males (11 with recurrence, 9 no recurrence).