Effect of including nonfatal events in cardiovascular risk estimation, illustrated with data from The Netherlands

European physicians use SCORE risk charts to predict a patient's 10-year risk of cardiovascular diseases (CVD) mortality. We examined whether the inclusion of nonfatal events improved risk estimation and the identification of high-risk persons.

In the EPIC-NL cohort, risk factor data were collected between 1993 and 1997 in 6772 men and 9108 women aged 35-65 years. During 10 years of follow up, 540 total (fatal + nonfatal) CVD events occurred, of which 122 (23%) were fatal. Risk equations were developed using Cox proportional hazard models. Discriminating ability and hazard ratios for CVD risk factors did not differ between the two endpoints. Absolute risks for total CVD were approximately 4-fold higher than for CVD mortality. Using the current 5% CVD mortality threshold or the 22% total CVD threshold for identification of high-risk persons leaves more than 84% of all male and 98% of all female future cases untreated. Of those exceeding these thresholds, 20% and 27% of the men, respectively, and 16% and 19% of women will get a CVD event in the next 10 years. Cut-off points of 2% for CVD mortality, corresponding to 10% for total CVD, will identify high-risk persons of whom approximately 10% will get an event in the next 10 years.

CVD mortality comprises a quarter of all total CVD events. Risk functions and the discriminating ability did not differ between the two endpoints. Cut-off points of 2% for CVD mortality or 10% for total CVD could be considered to identify high-risk persons.