There are some limitations to our study. First, data on smoking behaviour were self-reported. Multiple validation studies using biochemical markers such as cotinine have demonstrated that pregnant women may not admit to smoking (2), so our prevalence figures are likely to be underestimates. In addition, examination of our data suggests that non-responders in the third trimester in the Avon Longitudinal Study of Parents and Children (ALSPAC) study were more likely to be pre-pregnancy smokers than were responders (Supplementary Material, Fig. S2). However, such sources of error in the phenotypic data are likely to result in increased noise and reduced power rather than confounding (19). Importantly, the associations we observed are unlikely to be due to bias in self-reporting of smoking cessation. Any bias in our study of pregnant women is more likely to be towards a null result than a false-positive result. This is because while pregnant smokers may falsely report that they have quit, the reverse situation, in which pregnant women falsely report smoking, is extremely unlikely. Women who continue to smoke in pregnancy are expected to have a
