Harlap and Shiono discussed the fact that it was likely that women entering the cohort early (i.e. seeking antenatal care) constituted a selected group with a higher risk of miscarriage, which might affect their alcohol con- sumption as well and thus bias the results