A tool that overestimates risk in one population might lead to unnecessary treatment
Two of medicine's most trusted cardiovascular risk calculators—PREVENT, born from American data, and SCORE2, shaped by European populations—have now been tested against the full breadth of human diversity, across 6.4 million people in nearly fifty countries. A landmark study in Nature Medicine confirms that both tools reliably distinguish those most likely to suffer heart attacks or strokes from those who are not, offering clinicians a dependable compass in the daily work of prevention. Yet the same study reveals that universality is not uniformity: both tools tend to overestimate risk in Asian populations, reminding us that a map drawn in one landscape may need redrawing before it guides travelers in another.
- Cardiovascular disease kills more people than any other cause on Earth, and the urgency of identifying who is most at risk—before a heart attack or stroke arrives—has never been greater.
- Until now, PREVENT and SCORE2 were trusted largely within the populations that built them, leaving clinicians in Asia, Africa, and the Middle East uncertain whether these equations could be trusted with their patients.
- A coalition spanning 18 randomized trials and 44 observational cohorts, tracking nearly 300,000 cardiovascular events, put both tools through the most rigorous global stress test they have ever faced.
- Both calculators passed with moderate-to-strong discrimination scores—PREVENT at 0.702, SCORE2 at 0.683—and PREVENT outperformed older American tools still in clinical circulation, signaling a generational shift in the standard of care.
- A fault line emerged nonetheless: both tools overestimate risk in Asian populations, and the data from underrepresented regions remains too thin to draw firm conclusions, leaving a critical gap in global equity of care.
- Researchers are now calling for population-specific calibration, deeper validation across the Global South, and the integration of biomarkers like albuminuria to bring the tools' precision in line with the full diversity of human cardiovascular risk.
Two of the world's most widely used cardiovascular risk calculators have just passed their biggest global test. Researchers analyzed data from 6.4 million people across nearly fifty countries and found that PREVENT—developed by the American Heart Association—and SCORE2, its European counterpart, can reliably identify patients heading toward heart disease or stroke regardless of where they live. The validation, published in Nature Medicine, matters because these tools guide treatment decisions for millions of people every day, yet most prior evidence came from the specific populations where they were originally built.
The study assembled data from 18 randomized controlled trials and 44 observational cohorts, tracking hundreds of thousands of cardiovascular events over a mean follow-up of five years. The two tools define cardiovascular disease slightly differently—PREVENT includes heart failure alongside heart attacks and strokes, while SCORE2 focuses on fatal and nonfatal cardiac events and cardiovascular death—yet both performed with similar reliability across regions and clinical settings. PREVENT achieved a C-statistic of 0.702 and SCORE2 reached 0.683, placing both in the moderate-to-strong range of predictive accuracy. PREVENT showed particularly strong discrimination for heart failure events, reaching 0.78.
Yet the study also exposed important limitations. Both tools modestly overestimated overall cardiovascular risk, with signals of overprediction emerging in Asian populations and other underrepresented groups. A tool that overestimates risk may lead to unnecessary treatment; one that underestimates may leave vulnerable patients without the intervention they need. Researchers found that adding albuminuria—a marker of kidney damage—improved PREVENT's accuracy, particularly in patients with diabetes or kidney disease. When compared directly with older American prediction tools still in clinical use, PREVENT showed better calibration, suggesting it may be ready to replace an outdated standard.
The findings give clinicians meaningful confidence that both tools can guide prevention strategies across diverse global settings. But they also define the next frontier: local calibration. A risk calculator built on European or American hearts may need adjustment before it can serve patients in Manila, Mumbai, or Nairobi with equal precision. The study calls for population-specific tuning, further validation across Africa, Asia, and the Middle East, and the integration of additional biomarkers. The tools work globally, the research shows—but making them work equally well everywhere will take more work.
Two of the world's most widely used cardiovascular risk calculators have just passed their biggest test yet. Researchers analyzed data from 6.4 million people across dozens of countries and found that PREVENT—developed by the American Heart Association—and SCORE2, its European counterpart, can reliably identify which patients are heading toward heart disease or stroke, regardless of where they live. The validation, published in Nature Medicine, matters because these tools guide treatment decisions for millions of people every single day, yet until now, most evidence supporting them came from the specific populations where they were originally built.
Cardiovascular disease remains the leading cause of death globally, and prevention hinges on identifying risk early. A doctor needs to know: which patient in the waiting room needs aggressive cholesterol or blood pressure treatment right now, and which one can wait? The PREVENT and SCORE2 equations attempt to answer that question by weighing factors like age, blood pressure, cholesterol, smoking history, and other markers into a single prediction. But here's the catch: PREVENT was developed using American data, SCORE2 using European data. When a tool is tested only in the population it was built from, no one knows if it works equally well in Asia, Africa, or the Middle East. That uncertainty has limited their global adoption.
The new study assembled an unusual coalition: data from 18 randomized controlled trials and 44 observational cohorts, spanning nearly 50 countries and including general populations, patients with chronic kidney disease, and electronic health records. Over a mean follow-up of five years, researchers tracked 293,737 cardiovascular events using PREVENT's definition and 258,086 using SCORE2's. The two tools define CVD slightly differently—PREVENT includes heart failure alongside heart attacks and strokes, while SCORE2 focuses on fatal and nonfatal heart attacks, strokes, and cardiovascular death—yet both performed with similar reliability across the board.
PREVENT achieved a C-statistic of 0.702, a measure of how well a tool distinguishes between people who will and won't have a heart attack or stroke. SCORE2 hit 0.683. For context, a score of 0.5 means the tool is guessing; 1.0 means perfect prediction. Both tools landed in the moderate-to-strong range, and PREVENT showed particularly impressive discrimination for heart failure events, reaching 0.78. The consistency across regions and trial settings was striking. Discrimination did decline in higher-risk populations, but researchers attribute this to the complexity of sicker patients rather than flaws in the equations themselves.
Yet the study also exposed cracks. Both tools modestly overestimated overall CVD risk, with SCORE2 showing more overprediction than PREVENT. More troubling: signals of overprediction emerged in Asian populations and other underrepresented groups, though limited data made firm conclusions difficult. This is the study's most important finding for the future. A tool that overestimates risk in one population might lead to unnecessary treatment; one that underestimates might miss people who need help. The researchers found that adding albuminuria—a marker of kidney damage—improved PREVENT's accuracy, especially in patients with diabetes or kidney disease. Adding HbA1c, a measure of blood sugar control, helped less.
When compared head-to-head with older American risk prediction tools still in use, PREVENT showed better calibration, meaning its predictions more closely matched what actually happened. This suggests it may be replacing an outdated standard. The findings give clinicians confidence that both PREVENT and SCORE2 can be used across diverse settings to identify high-risk patients and guide prevention strategies. But they also expose the next frontier: these tools need local tuning. A risk calculator built on European hearts may need adjustment when applied to hearts in Manila or Mumbai. The study calls for population-specific calibration, further validation in Africa, Asia, and the Middle East, and integration of additional biomarkers to sharpen predictions. The tools work globally, the research shows. But making them work equally well everywhere will take more work.
Notable Quotes
These tools need to be continually refined to improve regional adaptations, particularly for populations across Asia, Africa, and the Middle East.— Study findings in Nature Medicine
The Hearth Conversation Another angle on the story
Why does it matter that these tools work across 6.4 million people instead of just the populations they were built in?
Because a doctor in Singapore or Lagos needs to know if a risk calculator developed in Boston or Amsterdam actually applies to their patients. If you validate a tool only in the population it was built from, you're essentially asking a test question to students who studied that exact test. You don't know if it transfers.
But both tools overestimated risk, especially in Asia. Doesn't that undermine the findings?
Not really. Overprediction is actually revealing. It tells us the tools work—they're identifying patterns—but they're calibrated to one population's baseline. It's like a scale that's consistently five pounds heavy. You can still use it to tell if someone gained weight; you just need to adjust the reading.
What's the practical difference between PREVENT and SCORE2 for a patient sitting in a doctor's office?
They define cardiovascular disease slightly differently. PREVENT includes heart failure; SCORE2 doesn't. PREVENT performed better overall and showed stronger discrimination for heart failure specifically. But both got the job done across regions. A doctor might choose based on what conditions matter most for their patient population.
The study mentions albuminuria improved predictions. Why does that matter?
Albuminuria—protein in the urine—signals kidney damage, which is a powerful predictor of heart disease. It's especially predictive in people with diabetes or existing kidney disease. If you add that one marker, you catch more high-risk people, particularly those the basic equations might miss.
What happens next? Are these tools suddenly better?
No. The validation proves they work globally, but it also exposes what's missing. The tools need to be recalibrated for different regions, especially Asia and Africa where the overprediction signals appeared. And researchers need to test them in populations that barely appear in the current data. That's the real work ahead.