First, recognizing that effect size is the average effect of a variant across all genetic and environmental backgrounds,52 it may be possible to forgo larger sample sizes to concentrate on more homogenous subsets in which larger effect sizes may be detectable. There has been much debate about the genetic heterogeneity of MDD. Some genetic heterogeneity is consistent with the polygenic model; each affected individual could potentially harbor a different combination of risk variants which could imply a spectrum of phenotypic symptoms, with those presenting similar symptom profiles carrying more similar profiles of risk alleles, consistent with symptom sharing of closely related individuals.53 For this reason, most GWAS for MDD have prioritized genotyping of the less prevalent and more heritable recurrent early onset MDD, but even so genome-wide significant associations have been elusive. Stratification by sex has also not yielded consistent results. Other possibilities certainly exist but phenotypes may be unavailable (for example, accurate delineation of typical versus atypical MDD)54 or prohibitively expensive (for example, magnetic resonance imaging). Use of quantitative scores of severity and reliability may be the best strategy to balance sample size with phenotype definition,55 as MDD symptom profiles are consistent with a quantitative rather than dichotomous liability.56
