Figure 1. Multivariate hazard ratios for coronary heart disease across quintiles of BMI (a,b,c) and waist circumference (d,e,f).

Country-specific HRs were estimated from Prentice-weighted Cox proportional hazards models, and 95%CI estimated with robust variance, to take into account the case-cohort design. HRs were combined by multivariate random-effects meta-analysis across 8 countries and accompanied by a group-specific confidence interval (allowing a confidence interval to be attributed to the reference category). Age was used as the primary time variable, and analyses were stratified by sex and centre. n=17,733 (7,637 cases)

Model 1 (a and d). Adjusted for age, smoking, physical activity, Mediterranean diet score, energy and alcohol intake, educational level

Model 2. Model 1 + adjusted for WC (for BMI, b) or BMI (for WC, e)

Model 3 (c and f). Adjusted for age, smoking status, systolic blood pressure, total cholesterol, HDL cholesterol, history of diabetes

Abbreviations: BMI, body mass index; WC, waist circumference, CHD, coronary heart disease