Abstract MP24: Comparison of Predictive Ability of Dietary and Lifestyle Scores For 10-Year Risk of Cardiovascular Mortality in a Pan-European Cohort Study

Introduction: Diet quality indexes and lifestyle indexes (which also include other lifestyle characteristics such as smoking and obesity) have recently received increased attention in disease prevention. Hypothesis: We aimed to investigate the comparative predictive performance of a comprehensive list of dietary and lifestyle indexes in relation to cardiovascular (CVD) mortality. Methods: We applied these indexes to men and women from 10 European countries participating in the European Prospective Investigation into Cancer and Nutrition (EPIC) study and examined their association with 10-year CVD mortality risk. We computed 10 dietary indexes and 2 diet and lifestyle indexes and calculated quartiles. Cox proportional hazard models stratified by age and study centre, adjusted for age, BMI, energy intake, smoking status, physical activity and educational level were fit to estimate Hazard Ratios (HR) and 95% CI. Harrell’s C-statistic, a discrimination measure of predictive performance, was calculated for each model. Results: After 10 years of follow up, 3761 CVD deaths were observed among 451 256 participants. All dietary indexes, except one, were significantly associated with CVD mortality with HR ranging from 0.75 to 0.84 for the fully adjusted model when comparing top vs bottom quartile (Table 1). Stronger effect size was observed for the diet and lifestyle indexes, particularly the Healthy Lifestyle Index (HLI). Discrimination of the full models was high and did not vary between scores. We found no heterogeneity in HRs across countries for most scores, except a modest heterogeneity for Mediterranean diet scores (I 2 =48%) and HLI (75%); however, heterogeneity across countries of the C-statistics was high for all scores (I 2 = 87%). Conclusion: Our results show that diet quality as a whole, or a cluster of lifestyle behaviours including diet, are consistently associated with a reduction of 10-year CVD mortality risk and that models comprising only age, sex and lifestyle risk factors could serve as predictors of CVD mortality.