To minimise confounding and clustering effects, the sample for analysis (see online Figure-Flow Chart) was restricted to women of white-European ethnicity (reflecting variations in alcohol dehydrogenase allele frequency across different populations [20]) and to children from singleton births (n = 13,171). Using postal questionnaires, information on alcohol consumption at 18 weeks gestation was gathered on 93 % (n = 12,257) of their mothers. Of these respondents, 65 % (n = 7,965) provided further information on alcohol consumption at 32 weeks gestation. This discrepancy reflects 3,378 (28 %) mothers who completed the initial version of the 32-week questionnaire where these questions were not asked and non-response in 914 (7 %) mothers. As previously shown, there was no sampling bias in the mothers not asked these questions [18]. However, maternal correlates of not providing this information when asked included the consumption of ≥4 drinks in a day on at least one occasion between 14 and 18 weeks gestation, smoking and use of cannabis in the early part of the second trimester, lower maternal age and level of education, depression, being unmarried, and living in rented housing.
