Validation of the New American Heart Association’s PREVENT Equations

By Michael H. Crawford, MD

Synopsis: An analysis of the accuracy of the new American Heart Association PREVENT Equations for predicting 10-year cardiovascular disease mortality in the National Health and Nutrition Examination Survey population has shown excellent discrimination with only modest underprediction and supports its use vs. the pooled cohort equation, which is the current standard.

Source: Scheuermann B, Brown A, Colburn T, et al. External validation of the American Heart Association PREVENT cardiovascular risk equations. JAMA Netw Open. 2024;7(10):e2438311.

The existing pooled cohort equations from the American Heart Association (AHA)/American College of Cardiology are the current clinical standard for estimating the mortality risk from cardiovascular disease (CVD) but tend to overestimate this risk, especially in men. Thus, there is considerable interest in the new AHA Predicting Risk of Cardiovascular Disease Events (PREVENT) equations.

This report is an external validation study of the fatal CVD risk in the National Health and Nutrition Examination Survey (NHANES) as predicted by the PREVENT equations. It uses the NHANES data from 1999 to 2010 to capture a 10-year follow-up.

NHANES data are collected by interviews, physical examinations, and selected laboratory tests. Mortality data were obtained from the linked National Death Index databases, which include non-CVD mortality or competing risks. The variables used in PREVENT are age, systolic blood pressure (SBP), high-density lipoprotein (HDL)-cholesterol, total cholesterol, estimated glomerular filtration rate (eGFR), smoking history, use of antihypertensive medications, statin use, and a diagnosis of diabetes.

PREVENT then calculates 10-year CVD mortality risk by sex. Comparisons of calculated risk vs. actual CVD mortality were performed by receiver-operator characteristics or the C-statistic (> 0.80 excellent, 0.70 to 0.80 good). Sensitivity analyses were performed for body mass index (BMI), SBP, cholesterol values, and smoking now or ever. After excluding those with missing data or known CVD, there were 24,582 subjects (mean age 45 years, 52% women) who were followed for up to 13 years. A CVD death occurred in 5% and cancer occurred in 4%.

The PREVENT equations showed that every 1% increase in risk was significantly associated with CVD mortality (hazard ratio [HR], 1.69; 95% confidence interval [CI], 1.08-1.09). The C-statistic was 0.89, but there was moderate underfitting of the predicted vs. actual mortality (slope, 1.13 vs. 1.0).

There was no significant influence of race, BMI, HDL cholesterol, or smoking on the results. PREVENT predicted CVD mortality over 10 years more accurately than the pooled cohort equation, with a net reclassification index of 0.09.

The authors concluded that the PREVENT equations showed excellent discrimination with only modest discrepancy in the calibration assessment, which supports their use rather than the pooled cohort equation for clinical risk assessment.

Commentary

The pooled cohort equation has been criticized because it overpredicted the 10-year risk of CVD mortality and may have occasioned overuse of statins and other prevention treatments. PREVENT slightly underpredicts CVD mortality, but it may better detect sex differences and the effect of cardiometabolic disease. The reasons for underprediction are not completely clear, but PREVENT does not include social determinants of health and did not account for the 8% who were cancer survivors. Many cancer therapies increase the risk of vascular disease.

Mortality was determined by ICD-10 codes, which do not include non-fatal CVD events that would lower the predicted risk. Also, ICD-10 codes do not distinguish ischemic stroke from intracerebral hemorrhage, which would dilute the effect of ischemic stroke on CVD mortality risk.

There are other weaknesses of the PREVENT equations. They do not distinguish between current and former smokers. They did not exclude patients with extreme values for SBP, BMI, or cholesterol, but sensitivity analyses showed that these considerations did not significantly affect the results. Also, the results of coronary calcium by computed tomography (CT) scan, if there were any, were not included. However, those with CT scan results likely are a minority of subjects in NHANES.

On the other hand, this is a very large study using a comprehensive database of subjects that reflects the ethnic diversity of the United States and serves as an external validation of the PREVENT equations. We look forward to future analyses of PREVENT that will include non-fatal CVD events.

The PREVENT equation online calculator can be found at https://professional.heart.org/en/guidelines-and-statements/prevent-calculator.

Michael H. Crawford, MD, is a Professor of Medicine and Consulting Cardiologist, University of California Health, San Francisco.