It is unlikely that genotype data missing for 6000 women would introduce selection bias. Empirical evidence suggests that genotype frequencies are the same for general-population versus selected samples in the UK (the British 1958 birth cohort versus British blood donors from the Wellcome Trust Case–Control Consortium) (38), and that genotypes display no more correlation with socioeconomic, lifestyle and physiological phenotypes than expected because of chance (39). Moreover, based on our data we found a lack of association of genotypes with socioeconomic and highly socially-patterned factors such as smoking.
