Chance is an unlikely explanation of our results. The choice of the candidate variants was informed by a strong biological prior, and for rs1229984, this was in fact an attempt to replicate the association with alcohol intake seen in different contexts. The sample size was large, even by the standards of genetic association studies. The power to detect effect sizes like the ones observed for rs1229984 comparing A allele carriers versus non-carriers was ∼76% for binge drinking during pregnancy and 80% for drinking cessation. Moreover, over-conservative Bonferroni corrections and Bayesian models were used in the first part of the analyses, both accounting for multiple testing. However, the limited phenotypic variation in this sample compared with previous studies (23,24) (approximately 1 versus 3 drinks/week on average for women) could reduce the power to detect associations of alcohol use with other genetic variants with less marked effects.
