Project Summary/Abstract Major depressive disorder (MDD) affects >300 million people worldwide. It is a leading contributor to disability and suicide, and thus a cross-cutting risk factor for many adverse life and health outcomes. It is heritable, and genome-wide association recently been informative. However, nearly all current MDD samples are not enriched in individuals with the highest clinical severity (i.e., the extreme tail of the phenotype distribution), a critical weakness for clinical prediction. We propose to focus on “phenotype extreme MDD”. We will define these individuals empirically on a population scale over years of follow-up in order to capture individuals with markedly worse MDD clinical features (e.g., treatment-resistance, dense patterns of treatment, psychosis) and poor outcomes (e.g., poor social function, disability, suicide). Cases with phenotype extreme MDD disproportionally contribute to the global burden of MDD. We show that we can identify these individuals and preliminary data suggest these individuals have a greater inherited burden of MDD risk alleles. We will address an additional weakness in the field via multiple, highly powered layers of replication in independent cohorts. We need to know quickly whether a promising model can replicate and generalize, and we have built the infrastructure for this. In Aim 1, we will empirically identify “phenotype extreme MDD” in a training set of ⅓ of the Swedish population with replication in independent samples (the other ⅔ from Sweden and harmonized datasets from Denmark and Norway) and then generalization to independent samples from the UK (Generation Scotland, UK Biobank), and the US (PsycheMERGE). In Aim 2, we will validate the empirical phenotype extreme MDD definition using genomic data in the Aim 1 populations (i.e., pedigree- and SNP-heritability, contrast with other MDD definitions, evaluate whether individuals with phenotype extreme MDD carry higher genetic risk scores for MDD). In Aim 3, we will develop clinically useful prediction algorithms for extreme MDD: can we predict at first presentation who will subsequently develop phenotype extreme MDD? We will have exceptional statistical power for all Aims. Successful completion of these aims will enable our transformative, tertiary-preventive intention of valid and clinically useful prediction of the subsequent development of phenotype extreme MDD early in a person’s treatment history. This is foundational to achieve the overarching translational goal of deploying these models on national scales in order to improve the health of MDD patients who are most severely ill.

Project Narrative Major depressive disorder (MDD) is a common cause of disability, suicide, and adverse life/health outcomes. We propose to identify the subset of people with MDD who are especially impacted by MDD. We will do this is a highly rigorous way via very similar analyses in large samples from the US, UK, Sweden, Denmark, and Norway. We will use all available data (treatment history, other medical and psychiatric problems, and genetic information). Our goal is to predict those who are likely to develop the worst forms of MDD early in their illness. Successful predictions will enable us to craft interventions in future studies to improve the health of MDD patients who are the most severely ill.