Some authors have considered bias when analysing a risk factor for the trait under selection17–20. For example, accepting BMI as a cause of type-2 diabetes, a SNP with a direct effect on type-2 diabetes may have a biased association with BMI when studied within a case/control sample of type-2 diabetes (Fig. 2). This is different from the situation considered here as the trait of interest is a precursor of, and not subsequent to, the index event. This has also been called an index event bias17, but here we reserve the term for when the trait of interest is subsequent to the selection criterion (Fig. 1). Methods are available to adjust analyses of risk factors for selection into case/control studies18,19,21,22, but they do not apply here, when selection bias acts entirely through unobserved confounders.
