The Importance of Sleep for Prediction of Depression in Adolescence - ABSTRACT Depression increases dramatically during adolescence, with a prevalence rate increasing to over 25% at age 17 years for major depressive disorder (MDD), and a 3-fold greater risk for females. Depression in later adolescence represents the greatest global burden of disease, is associated with substance use and impairment across multiple domains, and is a major risk for suicide. Although a high-risk period, the neural plasticity during this phase of life provides a window of opportunity for early intervention efforts targeting modifiable behaviors, such as sleep. Complicating intervention is the heterogeneity of depressive symptoms, potentially with different underlying biological pathways, the broader social/environmental context that forms the backdrop for depression risk, and the interactive nature of key risk factors. To account for these challenges and to advance discovery, we conduct individual-level analyses with a large number of adolescents tracked over time, considering neurobiological pathways, contextual factors, and divergence by sex. We focus on modifiable sleep behavior, which shows promise as a prevention and intervention target. Applying our novel deep-learning framework using two public longitudinal datasets of adolescents (National Consortium on Alcohol and NeuroDevelopment in Adolescence [NCANDA; N=624] and the Adolescent Brain Cognitive Development cohort [ABCD; N=11800]), our overall objective is to clarify neuromechanistic and behavioral risk for depression by identifying constellations of measures and their interactions across sociodemographic and environmental factors, modifiable behaviors (e.g. sleep), cognitive function, personality, and neural circuity that forecast depressive symptoms and herald MDD in individual adolescents. Depressive symptoms are grouped according to constructs of the NIMH RDoC domains negative and positive valence. Aim 1 uses a data-driven search across sleep health and brain circuitry measures to characterize depressive symptoms across RDoC constructs and MDD diagnosis, also considering interactions with sociodemographic factors, in individuals in NCANDA (ages 12-17 yrs) and seeks to replicate findings in the large, diverse ABCD cohort (ages: 9-17yrs). Aim 2 focuses on the ABCD cohort, to forecast RDoC constructs and MDD in older adolescents (16-17 yrs) based on the closest prior visit, explicitly examining differences by sex. Analyses are enhanced with personality measures and objective measures of sleep, cognitive assessments, and task-fMRI of emotion and reward. Aim 3 uses all available longitudinal ABCD data to forecast the magnitude of RDoC constructs and MDD in future years to identify trajectories of interactions across risk factors. Our innovative, individual-focused analyses to identify risk factors for depression in developing adolescents ensure rigor and reproducibility, potentially improving depression risk assessment in prevention programs. The outcome is poised to have high public health significance for advancing efforts for early intercession, mitigation, and prevention of this debilitating, life-threatening disorder.
