External validation of American heart association predicting risk of cardiovascular disease EVENTs (PREVENT) equations in a Chinese population

Developed by the American Heart Association (AHA), the Predicting Risk of Cardiovascular Disease Events (PREVENT) exhibits good performance in CVD risk assessment among the American population. However, their applicability to the Chinese population, particularly within specific age cohorts, remains unevaluated. Thus, this study aims to externally validate the performance of the PREVENT equations in a nationwide middle-aged and elderly Chinese population. This retrospective study analyzed 2011-2018 data from the China Health and Retirement Longitudinal Study (CHARLS), including 10,068 participants aged ≥ 45 years (4854 males and 5214 females). The primary outcome was CVD (including heart disease and stroke). The association between PREVENT scores and CVD risk was evaluated using univariate and multivariate logistic regression models integrated with restricted cubic splines. Receiver operating characteristic (ROC) curves, calibration analysis, clinical decision curve analysis (DCA), subgroup analysis and sensitivity analysis were used to assess the PREVENT equations' performance to predict CVD risk in this Chinese population. Additionally, we used the ROC curve and the Delong test to compare the performance of PREVENT scores and China-PAR in CVD risk prediction. Univariate logistic regression showed that each 1% increase in PREVENT scores was associated with a significantly higher risk of CVD in both males (OR = 1.05, 95% CI 1.04-1.07) and females (OR = 1.06, 95% CI 1.05-1.07). However, ROC analysis demonstrated an AUC of 0.61 (95% CI 0.59-0.64) for CVD prediction in males and 0.62 (95% CI 0.60-0.64) in females, substantially lower than US validation performance (0.757-0.794). The calibration slopes were 0.51 (95% CI 0.44-0.58) for males and 0.47 (95% CI 0.41-0.53) for females, with intercepts of -0.79 and -0.54 and Brier scores of 0.10-0.11, respectively. While PREVENT exhibited significantly better discrimination than China-PAR (AUC: 0.52, Delong test: Z = -6.983, P < 0.001), its clinical net benefit remained marginal. Although the PREVENT scores were significantly associated with increased CVD risk, the equations exhibit reduced discrimination compared to US validation in CVD prediction among middle-aged and elderly Chinese individuals. Consequently, clinicians should proceed with caution when directly applying these equations to this population without extensive statistical recalibration and further validation.