We assessed Cox and logistic regression models in case-cohort studies by using the entire EPIC-CVD case-cohort and the same 25 SNPs as above. Here, the effect estimates were directionally concordant for both models for 22 of the 25 SNPs (Table 3), of which 21 had the same direction of effect as in the literature (Supplementary Table S8). Like in the subcohort, the effect estimates of the SNPs that were directionally discordant between the models lie close to the null and the SNP that was directionally discordant with the literature had a small effect on CHD in the large consortia (Supplementary Table S8). The logistic regression model was more precise than the Cox model for all of the SNPs often leading to much smaller P-values. For instance, the SNP rs974819 (chr11:g.103660567C>T; an SNP located downstream of PDGFD) had similar effect sizes for the Cox (log(HR)=−0.082) and logistic (log(OR)=−0.080) regression models, but the SEs were 0.031 and 0.027, respectively. This led to a smaller P-value for the logistic regression model (P=0.003 compared with P=0.007). Again, the effect sizes of the logistic regression
