For a biallelic SNP, a genetic interaction (either GxG or GxE) on a quantitative phenotype will lead to differences in phenotypic distribution across the three genotypes, leading to differences in phenotypic variance (scale).13 In light of this, Levene’s scale test of equality of variance14 has been proposed as a method of prioritizing SNPs for subsequent GxG and GxE studies,13,15 in contrast to the standard mean (location) test (i.e., testing for phenotypic differences in mean across genotypes). The advantage of using variance testing to incorporate GxG or GxE is that exposures need not be specified, and the enormous multiple testing burden of formally examining all possible pair-wise interactions is removed. Variance-only testing, however, has limited power to detect SNPs displaying main effects only.