An open question is whether GWAS meta-analyses really do require fewer subjects with imaging than they do when behavior is the target of study. In 2009, before ENIGMA began, one of its founders noted, “Just because the phenotypes are expensive to collect does not change the power calculations.” (N. Martin, pers. commun., 2009). But more recently, Rose and Donohoe (2013) performed an empirical analysis of effect sizes in genetic studies of cognitive and neuroimaging traits in schizophrenia, and found evidence supporting the efficiency of using imaging traits. However, some evidence does suggest that imaging traits may have intermediate effect sizes when compared to phenotypes theoretically closer or farther away from the underlying biology. The percent variance explained in gene expression GWASs (often called eQTL studies) for the top SNP hits are well above 10 % of the variance in the expression of a particular gene (Stranger et al. 2007). The percentage of trait variance in hippocampal volume explained by the top genetic variant in ENIGMA1 was 0.27 % (Stein et al. 2012) although independent cohorts will be required to estimate
