Our study is not without limitations. Cumulative AUDIT scores reflect two distinct constructs: one measuring alcohol consumption and another measuring alcohol-related problems; thus, AUDIT scores may conflate multiple genetic signals (Bergman & Källmén 2002; Shevlin & Smith 2007). When we split AUDIT scores in three domains of consumption (items 1–3; AUDIT-C), dependence (items 4–6; AUDIT-D) and hazardous use (items 7–10; AUDIT-H), we observed higher scores for the domain of alcohol use (AUDIT-C: 2.96 ± 1.95; AUDIT-D: 0.203 ± 0.76; AUDIT-H: 0.665 ± 1.513). Our study may be tagging genetic risk for high quantity/frequency of alcohol consumption, as shown by the high genetic correlation with other alcohol-consumption traits, but may not overlap with other GWAS of AUD. In addition, our study focused on a cohort with relatively low levels of alcohol use; the unexpected positive genetic correlation between AUDIT and educational attainment, and the negative genetic correlation between AUDIT and both BMI/obesity and ADHD, may not generalize to cohorts with higher levels of alcohol use (Goldman et al. 2005) or AUD populations. Indeed, the selection of the 23andMe cohort (e.g. highly
