EPIC-CVD centre (as a categorical variable) and 10 principal components to adjust for ancestry. Age, sex and EPIC centre were used as covariates to account for differences across EPIC centres (including any differences in the distributions of age and sex). We compared the associations from the Cox and logistic modelling approaches in the full EPIC-CVD case-cohort and in the subcohort of EPIC-CVD (as a surrogate cohort study). The Cox model was fitted using the time-on-study time-scale in both settings. Prentice weights and robust SEs were used for the Cox model in the full case-cohort to account for the sampling process. SNPs were aligned to the plus strand of the human genome reference sequence and are displayed on version 19 (Build 37) of the human genome.
