Since rs1229984 shows extreme stratification in different populations worldwide (40), and alcohol drinking follows cultural patterns (41), there is a possibility of residual confounding by population substructure, even if the sample analyzed only included women reporting white/European ethnicity. To assess this, we conducted sensitivity analyses by further excluding women following two more conservative definitions of ‘white ethnicity’: one based on the woman's reported parental ethnicity, and her own; another based on a set of ancestry-informative genetic markers. Finally, we repeated the analyses after excluding all women reporting to be of Jewish religious faith, as a proxy for Jewish ancestry/ethnicity. This was done since the prevalence of the A allele is much higher in populations from West Asia including Jews from a variety of regions (74), who might also drink less than other individuals of European ancestry that would equally describe themselves as ‘white’.
