A huge multinational analysis suggests two leading cardiovascular risk tools can help clinicians identify high-risk patients across regions, but local calibration remains key to making prevention more precise.
Study: Multinational validation of the PREVENT and SCORE2 cardiovascular risk equations in 6.4 million subjects. Image credit: Dragon Claws / Shutterstock
cardiovascular disease (CVD) Risk calculators guide treatment decisions for millions of people around the world every day. Most of these, however, have been developed for specific populations and regions. A large multinational analysis has been accepted for journal publication Nature Medicine supports the wider application of two main forecasting tools. These include the American Heart Association (Aha)’s Cardiovascular Disease Risk Prediction EVENT (PREVENT) equations and Europe’s Systematic Coronary Risk Evaluation 2 (RATING 2) algorithm. Their generally good performance in study populations suggests that clinicians can use them to identify and stratify high-risk individuals in various clinical settings.
PREVENT and SCORE2 Risk Prediction Background
Cardiovascular diseases continue to cause widespread morbidity and mortality worldwide. In preventive care areas, CVD Risk prediction tools could help health care providers identify people who are most likely to benefit early from cholesterol- and blood pressure-lowering interventions. Therefore, tools could enable faster response and improve resource allocation, thereby reducing CVD burden on individuals and health care systems.
PREVENT and RATING 2 are among the most widely used models today. Researchers have developed these tools using large local datasets, and major clinical guidelines now recommend their use in routine care. However, they have been mostly validated in populations of origin, leaving uncertainty about their accuracy in more diverse populations.
Multinational CVD Validation Study Design
In the present study, the researchers evaluated its performance Aha‘small PREVENT and Europe RATING 2 cardiovascular risk equations using data from 18 randomized controlled trials (RCTs) and 44 observational study cohorts within it CKD Forecasting Consortium. The study included more than 6.4 million people for PREVENT and 5.4 million for RATING 2. None of the participants had a previous history CVD diagnosis. The study included people from Europe, North America and Asia-Pacific and other regions, with RCTs enrolling participants from nearly 50 countries.
The team evaluated discrimination, followed by calibration. Discrimination refers to the ability of the models to distinguish between the humans that were developed CVD and those who didn’t. They then investigated whether the predicted risks matched the observed outcomes (calibration). The researchers used Harrell’s C-statistics and calibration slope analyzes on each study and stratified results by region. They also appreciated the short term CVD risk for 1 to 9 years using scaling factors derived from the PREVENT algorithms.
The analysis included cohorts of the general population, CKD-specific cohorts, cohorts based on electronic health records and multinational randomized trials. Randomized trials evaluated modern cardiovascular, renal and metabolic therapies. These included glucagon-like peptide-1 receptor agonists (GLP-1RA), sodium-glucose cotransporter-2 inhibitors (SGLT2i), renin-angiotensin system (RAS) mineralocorticoid receptor blockers and non-steroidal antagonists (nsMRA).
In addition, the team investigated whether metabolic and renal markers such as glycated hemoglobin were incorporated (HbA1c) and albuminuria could improve prediction accuracy. In sensitivity analyses, they compared PREVENTit is atherosclerotic CVD (ASCVD) equation with commonly used group cohort equations (PCE).
PREVENT and SCORE2 performance findings
The researchers followed the participants for an average of about five years. During this monitoring period, they documented 293,737 observed CVD events that use the PREVENT definition and 258,086 using RATING 2 among more than six million participants worldwide. Events recorded by PREVENT included fatal and non-fatal ASCVD and heart failure, while RATING 2 focused on myocardial infarction, stroke and cardiovascular death. Despite these differences, both equations showed similar reliable performance in observational cohorts and multinational randomized trials.
THE PREVENT equations showed moderate to strong discrimination, with a median C statistic of 0.702 for CVD prophecy. RATING 2 achieved a comparable C-statistic of 0.683. PREVENT it also performed well in forecasting ASCVD only (0.695) and showed particularly strong discrimination for HF events (0.78). These models showed generally consistent results across regions and multinational trials. However, discrimination was reduced in higher-risk populations, likely due to differences in patient risk profiles and case-mix heterogeneity rather than limitations of the models themselves.
Both tools were moderately overestimated overall CVD risk, in particular RATING 2. Signals of overprediction emerged in Asian and other underrepresented populations, although limited data limited conclusions. However, calibration remained robust in rigorously validated multinational trials CVD results. Adding albuminuria improved PREVENTpredictive performance, especially in high-risk populations with diabetes or CKDwhile HbA1c brought minor improvements. Compared to the older ones PCE estimates used in the United States (US), PREVENT also consistently demonstrated better calibration.
Implications of the Global Cardiovascular Disease Risk Tool
The findings strengthen the evidence for use PREVENT and RATING 2 for reliable stratification of high-risk individuals CVD in various geographic and clinical settings, while emphasizing the need for local calibration and further validation in underrepresented populations. These tools can be especially valuable in primary care, where proactive decisions can be most life-changing. The findings also suggest that PREVENT can outlast the older ones US risk prediction tools because of its better calibration and broader integration of cardiovascular, renal, and metabolic risk factors. Assessment of additional biomarkers such as albuminuria could help clinicians more accurately identify individuals at higher CVD risk, especially those with diabetes or CKD.
These tools need to be continually refined to improve regional adaptations, particularly for populations across Asia, Africa and the Middle East. Future efforts should focus on specific population calibration, further validation in underrepresented regions, and clearer treatment thresholds for emerging cardiometabolic therapies. Incorporation of available biomarkers could improve the accuracy and accessibility of CVD risk prediction worldwide.
