Major depressive disorder (MDD) is uncommon in childhood, but becomes increasingly prevalent during adolescence. By the age of 18, about 15% of adolescents will have experienced at least one episode of MDD, with females twice as likely than males to have suffered an episode. This developmental surge in depression is especially high among teens who have a parent with a history of MDD, with close to half developing the disorder by the end of adolescence. Despite these epidemiological findings, and the range of negative downstream consequences linked to MDD, there are strikingly little data on the neural and behavioral abnormalities that confer risk for future depression onset in youth. The ability to prospectively predict MDD prior to its onset would have important clinical implications for the early identification of ? and targeted deployment of interventions for ? at-risk youth, which is strongly aligned with the NIMH Strategic Plan. To address these gaps, adolescents ages 12-15 at increased risk of MDD onset by virtue of a parental history of MDD, as well as a control sample with no parental history of depression, will complete baseline neural (fMRI) and behavioral assessments of replicated endophenotypes of MDD (neuroticism, anhedonia, cognitive control deficits). Growing evidence and our preliminary data suggest that these endophenotypes are relatively stable trait-like risk markers, have non-overlapping neural substrates, and precede and prospectively predict depression onset. The project has three aims. First, we will evaluate the neural correlates of these three endophenotypes in an adolescent sample (n = 148), half of whom are at elevated risk of MDD (Aim 1). Second, during a 24-month follow-up phase, participants will be contacted by phone every 6 months and administered measures to assess changes in symptoms. Analyses will test whether behavioral and neural endophenotype measures prospectively predict onset of depressive symptoms during the follow-up phase. Importantly, to evaluate incremental predictive validity, we will test whether each endophenotype measure predicts future depressive symptoms above and beyond relevant clinical, familial/demographic and developmental variables previously linked with risk of future depression (Aim 2). Third, we will test whether multivariate machine learning models incorporating behavioral and neural endophenotype markers, as well as clinical, familial/demographic, and developmental characteristics, can be used to predict subject-specific risk of future depression onset with sufficiently high sensitivity and specificity to be clinically useful (Aim 3). Critically, recent advances in machine learning allow for the development of algorithms predicting risk at the individual level, as well as the integration of numerous predictors rather than relying on single variables that may, in isolation, have limited clinically-useful predictive value. Collectively, results are expected to advance our ability to predict the onset of depressive symptoms and, ultimately, to inform the development of a freely available, web-based risk calculator for predicting subject-specific depression risk.
The prevalence of depression increases dramatically during adolescence, particularly for females. Using machine learning algorithms, the overarching goal of the project is to identify neural and behavioral markers that ? in combination with clinical and demographic characteristics ? reliably predict risk of developing depression during adolescence. The ability to prospectively predict depression prior to its initial onset would have important clinical implications for the early identification of ? and targeted deployment of interventions for ? at-risk youth.