Given the strong effect of MUC5B on incidence, it is possible that it dominates the estimation of index event bias and that our approach over- or under-corrects the bias in MUC5B itself (Fig. 3). We therefore repeated the analysis after conditioning the survival SNP effects on rs35705950 genotype. The results were very similar, with the regression of survival effects on incidence effects now giving a coefficient of −0.028, decreasing to −59.79 (95% CI: −59.87 to −10.52) after correcting for regression dilution.
