We hypothesized that each copy of the risk allele carried by the mother would be associated with reduced offspring birth weight, via its association with smoking quantity, in women who smoked during pregnancy. To test this, we selected subjects who said that they smoked during the first and/or the third trimester of pregnancy (excluding multiple births and babies born before 36 full weeks' gestation). We performed linear regression of birth weight on the number of maternal risk alleles carried, with sex, gestational age and study as covariates. We then repeated this analysis including first trimester smoking quantity as an additional covariate. We additionally used a triangulation approach to estimate the expected effect size of the association between maternal genotype and offspring birth weight. We hypothesized that this effect size would be determined by (A) the effect size of the association between maternal genotype and smoking quantity and (B) the effect size of the association between smoking quantity and birth weight (see Supplementary Material, Fig. S1). We estimated (A) and (B) in the 1649 mothers with data available on offspring birth
