Association studies of such subsequent events are vulnerable to index event bias, whereby biased associations can result from selection of subjects according to their disease status9. This is one of several types of selection bias whose relevance to genetic epidemiology has recently been discussed10,11. Independent causes of disease become correlated when selecting only the cases of disease, creating indirect associations between causes of disease with subsequent events (Fig. 1). A well-known example is the so-called obesity paradox whereby, among individuals with cardiovascular disease (CVD), those with higher body mass index (BMI) tend to survive longer12. A possible explanation is that, if an individual with CVD has a high BMI, they may well have lower levels of other risk factors. If those lower levels tend to increase survival, then increased BMI may be associated with longer survival. In the notation of Fig. 1, BMI plays the role of the SNP G, while X is CVD and Y is survival. It remains controversial whether this paradox is explained by index event bias13.
