Lifetime risk predictions for cardiovascular diseases: competing risks analyses on a population-based cohort in Sweden

Abstract Background and aims There are guideline discussions on a lifetime approach to cardiovascular risk. Many of the available risk models estimate the short-term, usually 10-year risk of non-fatal and fatal cardiovascular diseases (CVD) grouped together. We aimed to develop lifetime risk models for non-fatal coronary heart disease, stroke, heart failure and death from CVD and non-CVD. Methods We included 92,915 individuals who had participated in a community-based lifestyle intervention programme at 40, 50 and/or 60 years of age. Their collected data on selected risk factors were linked to register data on hospitalizations and death. Parametric multivariable survival regression with a competing risks approach was employed to model cause-specific hazards, which were translated into cumulative incidence functions to provide the risk of experiencing each event separately. All analyses were performed gender-age wise. For illustrative purposes, “better” and “worse” risk profiles were created by setting three modifiable risk factors to the best and worst levels, respectively. Results Most of the risk factors qualified for inclusion in the regressions. Men had a higher risk of cardiovascular events and the events occurred at a younger age than women. In the created risk profiles, where serum total cholesterol, smoking status and blood pressure were modified, an excessive number of CVD events were observed in the worse profiles. Conclusions Using these models, the lifetime risk of each of the first CVD events can be estimated for different risk factor profiles. Since the predictions are diagnosis specific, the estimates are more accurate.