Across the thousands of markers included in GWASs and targeted array studies, it is likely that some will be in LD with causal variants that are associated with the exposure of interest. The fact that few gene-environment interactions have been reported suggests that the magnitude of the bias is small. Indeed, under a typical scenario for current GWASs in which marker and causal variant have the same MAF of 0.2, their correlation is r2 = 0.8, and the causal variant has odds ratio 1.1 with both disease and a binary exposure, then similar calculations to those in Table 2 give the interaction odds ratio for the marker as 1.000433. More than a million cases and controls would be needed to detect this effect with 80% power at p < 0.05.28