A mega-analysis of functional connectivity and network abnormalities in youth depression

Major depressive disorder (MDD) represents the leading cause of mental health disability for young people worldwide but remains poorly understood. Previous neuroimaging research has indicated alterations in the connectivity of brain circuitry in youth MDD; however, findings have been inconsistent. This may relate to limitations in sample size and sample and methodological heterogeneity. In an effort to delineate robust neurobiological markers of youth MDD, we conducted a data-driven, connectome-wide mega-analysis of resting-state functional connectivity in 810 young individuals across 7 independent cohorts with a cross-sectional and case-control design. Compared with healthy comparison individuals (n = 370), youth MDD (n = 440) was associated with significant alterations in connectivity of densely connected brain areas (hubs), anchored in the default mode and dorsal and ventral attention networks. Critically, functional connectivity within these networks was significantly associated with depression symptom severity (r = –0.46 for hypoconnected regions and r = 0.53 for hyperconnected regions; both P values < 0.001), indicating the clinical relevance of functional connectivity alterations. Further, machine-learning analyses demonstrated that individual diagnostic status (AUC = 73.1%) and clinical severity (r = 0.14, P = 0.008) could be predicted on the basis of functional connectivity alone in unseen data using leave-one-site-out cross-validation. Together, our work represents an important first step toward robust characterization of the neurobiological basis of youth depression. We demonstrate the clinical relevance of brain connectivity in youth depression and highlight a critical role of functional hub regions, especially those localized to the default mode and dorsal and ventral attention networks in youth MDD. This mega-analysis of brain resting-state functional connectivity in young individuals with major depressive disorder scanned at six sites across four countries identified hub regions of the attentional and default mode networks as predictors of depression severity.