Validation of the 30-Year Framingham Risk Score in a German Population-Based Cohort

Abstract

The Framingham Risk Score to predict 30-year risk (FRS30y) of cardiovascular disease (CVD) constitutes an important tool for long-term risk prediction. However, due to its complex statistical properties and the paucity of large population-based cohorts with appropriate data, validation of the FRS30y is lacking. A population-based cohort from Southern Germany (N = 3110, 1516 (48.7%) women) was followed up for a median time of 29.5 [18.7, 31.2] years. Discrimination and calibration were assessed for the original, recalibrated and refitted FRS30y version. During follow up, 620 incident CVD events (214 in women) occurred. The FRS30y showed adequate discrimination (original and recalibrated version: Area under the curve (AUC): 78.4 for women and 74.9 for men) but overestimated actual CVD risk (original version: discordance 45.4% for women and 37.3% for men, recalibrated version: 37.6% and 28.6%, respectively). Refitting showed substantial improvement in neither discrimination nor calibration. The performance of FRS30y is adequate for long-term CVD risk prediction and could serve as an important tool in risk communication, especially for younger audiences.