The regression of survival log hazard ratios on incidence log odds ratios gave a coefficient of −0.025. The sign of this coefficient changed under the Hedges–Olkin based adjustment for regression dilution, which is implausible because regression dilution bias is the ratio of two variances28. However, applying our SIMEX based adjustment, the coefficient decreased to −65.63 (95% CI: −65.88 to −5.68). The very wide confidence interval reflects high standard errors on individual SNP effects. Nevertheless, the coefficient is significantly negative, which implies that there are common causes of incidence and prognosis that have the same net direction of effect.
