Attrition from cohort studies may result in biased estimates of socioeconomic inequalities, and the degree of bias may worsen as participation rates decrease.4 However, it is often argued that representativeness is not necessary in studies of this kind,5–9 although this is not universally accepted.10 In particular for genetic variants, where conventional confounding is low,11 it has been argued even by those concerned about selection bias that any problems associated with a lack of representativeness may be modest.10,12 Here we ask: what is the impact of selection bias on the results obtained from these studies? We take the perspective that selection bias can amount to conditioning on a collider (i.e. conditioning on a variable that is independently influenced by two other variables).
