Framingham risk score validation
Pro Research Analysisby 
Searched over 200M research papers
Framingham Risk Score Validation: Discrimination and Calibration Across Populations
Framingham Risk Score Performance in European and Asian Populations
Multiple studies have assessed the validity of the Framingham Risk Score (FRS) in predicting cardiovascular disease (CVD) risk across diverse populations. In a German cohort, the 30-year FRS demonstrated good discrimination (AUC: 78.4 for women, 74.9 for men) but tended to overestimate actual CVD risk, even after recalibration. Refitting the model did not significantly improve its performance, but the FRS30y was still considered adequate for long-term risk prediction and useful for risk communication, especially among younger individuals .
Similarly, in a multiethnic Asian population, the general FRS showed moderate discrimination (AUC: 0.63) and good calibration, with reasonable accuracy for men but some overestimation for women. The FRS was found to be a practical alternative in the absence of local risk charts . Another Malaysian study found that the FRS and other models like the Revised Pooled Cohort Equations (RPCE) had good discrimination (AUC: 0.750 for FRS) but poor calibration, with most models overestimating 10-year CVD risk by 3% to 1430% .
Validation in Specific Risk Domains: Hypertension and Stroke
The Framingham Hypertension Risk Score has also been validated in several non-US populations. In the UK’s Whitehall II Study, the score showed good discrimination (C statistic: 0.80) and calibration, with excellent agreement between predicted and observed hypertension incidences . In Brazil, the Framingham Hypertension Risk Score performed acceptably, with good discrimination and calibration, though it slightly overestimated risk compared to a locally derived model . In a Middle Eastern cohort, the score also performed well (c-index: 0.81), though it initially underestimated risk, which was corrected with a simple model revision .
For stroke risk, a machine learning approach using Framingham data revealed that traditional linear models may miss important non-linear relationships. The non-linear Framingham Stroke Risk Score improved prediction accuracy, especially in multi-ethnic populations, suggesting that more advanced modeling could enhance risk stratification .
Framingham Risk Score in Women and Health Economic Modeling
In Australian women, the Framingham model showed good discrimination (AUC > 0.85) and calibration for predicting 10-year cardiovascular mortality, though it overestimated deaths by more than 15%. Lower treatment thresholds were suggested to improve identification of high-risk individuals .
A systematic review of health economic studies in Europe found that while the FRS generally provided satisfactory accuracy, it tended to underestimate non-fatal myocardial infarctions and sometimes overestimated risk in European populations. The review highlighted the need for further validation and local adaptation when using FRS in economic modeling .
Conclusion
The Framingham Risk Score is a widely used tool for predicting cardiovascular and hypertension risk. Across various populations—including European, Asian, Middle Eastern, and Australian cohorts—the FRS generally demonstrates good discrimination and reasonable calibration, though it often overestimates risk, especially in non-US populations. Local recalibration or revision can improve accuracy, and advanced modeling techniques may further enhance predictive performance. Despite its limitations, the FRS remains a valuable tool for risk assessment and communication, particularly where local risk models are unavailable Rospleszcz2022Chia2015Kivimäki2009+6 MORE.
Sources and full results
Most relevant research papers on this topic