Validation of the pooled cohort equations in a Hong Kong Chinese cohort

Abstract

OBJECTIVE: The 2013 American College of Cardiology and the American Heart Association guidelines recommended the Pooled Cohort equations for evaluation of atherosclerotic cardiovascular disease risk of individuals. We investigated the usefulness of the Pooled Cohort equations in Chinese by validating this risk prediction model using the Hong Kong Cardiovascular Risk Factor Prevalence Study (CRISPS) cohort. METHODS: The Hong Kong CRISPS is a population-based prospective cohort study of cardiovascular risk factors among 2895 Chinese men and women (aged 25-74years) initiated in 1994. Cardiovascular (CV) events were ascertained until December 2013. The discrimination and calibration of the Pooled Cohort equations was evaluated and compared with the Framingham risk equation. A Hosmer-Lemeshow Chi-square statistic (X2) <20 indicated good calibration. RESULTS: The discrimination power of the 2 models in both men and women was moderate (C-statistic >0.7). However, the calibration score of both models was unacceptable in men (Pooled Cohort X2=24.1, Framingham X2=20.1). Since the Framingham model systematically over-estimated CV risk [average predicted risk 18.3% (95% CI 15.5-21.0) versus average observed risk 13.4% (95% CI 11.0-15.8)], this can be corrected by recalibration of the model using the CRISPS data [average predicted risk 11.5% (95% CI 9.2-13.7) versus average observed risk 11.8% (95% CI 9.6-14.1)]. Recalibration cannot be applied to the Pooled Cohort model because the degree of miscalibration varied across the different risk categories. In women, although calibration of Pooled Cohort (X2=10.1) and Framingham model (X2 =12.1) appeared similar, the accuracy of the Framingham model was better [average predicted risk 6.2% (95% CI 4.6-7.7) versus average observed risk 6.3% (95% CI 4.8-7.9)] than the Pooled Cohort model [average predicted risk 4.8% (95% CI 3.3-6.3) vs average observed risk 6.9% (95% CI 5.1-8.7)]. CONCLUSIONS: Risk prediction models should be able to discriminate between individuals with and without disease, and also well-calibrated so that predicted risk estimates matches as closely as possible the observed risk in the population. The Pooled Cohort equations provide poor calibration and moderate discrimination in Hong Kong Chinese, especially in men. The Framingham risk equation can be applied to the Hong Kong population but requires recalibration in men.