We think there are several strategies to reduce the heterogeneity in depression. First, examination of the full range of variation in depression (e.g., depressive symptoms), rather than dichotomizing the phenotype (cases and controls), could be a statistically more powerful approach to identify variants associated with depression.151 This would be consistent with evidence that the diagnostic threshold for MDD has been artificially imposed on a continuity of depression risk.149 Second, more data-driven approaches to examine shared features or subtypes of depression through use of latent class analysis152 may also prove helpful. Prior studies applying such methods in both adolescents and adults have found distinct subtypes that differ based on severity, symptoms, and episode length.153,154 Examination of these subtypes in a genetic association study may help to identify variants that are common across or unique to specific subtypes. Third, another strategy would be to continue efforts to examine phenotypes thought to be more proximal to a genetic substrate than are clinically-defined categories.155 Putative “intermediate” or “endophenotypes” related to depression include emotion-based attention biases,156,157 impaired reward function,158 and deficits in domains of executive
