One strategy is to simply be less stringent in the rules for power and multiple testing p value corrections, for exploratory purposes. Typically, a significant treatment × sex interaction is necessary to justify conducting a follow-up analysis stratified by sex. However, uncovering a significant interaction that passes correction for multiple testing in the realm of human genetics is difficult, unless there is a very large sex effect on an association between phenotype and genotype [36]. Thus, simply conducting the analyses allowing for some flexibility (less stringent with p values) often uncovers patterns in the data, and patterns in data can be just as informative as p values derived from association tests of single genetic loci. Reconsider the story of rs17810398 within the DAPL1 gene. Previous GWAS did not uncover an association between rs17810398 and age-related macular degeneration, but when analyses were stratified by sex, the association was highly significant for females (p = 2.6 × 10−8) and not males (p = 0.382) [25]. Significant associations can be lost when combining male and female data into one dataset.
