Objective:To assess the predictive performance of the risk of cardiovascular diseases (CVD) derived from the China Kadoorie Biobank (CKB-CVD) model, prediction for atherosclerotic cardiovascular disease (ASCVD) risk in China (China-PAR) model, and the risk of fatal and nonfatal ischemic cardiovascular diseases derived from the USA-People′s Republic of China Collaborative Study (USA-PRC) model in Chinese Multi-provincial Cohort Study (CMCS).Methods:In this prospective cohort study, a total of 21 948 individuals aged ≥35 years without CVD were selected from 8 provinces and cities in China during the CMCS survey from 1992 to 2005 for 10-year follow-up. The occurrence of CVD or ASCVD events during the follow-up period was used as the gold standard. The CKB-CVD and China-PAR models were used to calculate the predicted risk of CVD events, while the USA-PRC model was used to calculate the predicted risk of ASCVD events. The discrimination of the models was evaluated using the C-statistic, and the calibration was assessed using the Hosmer-Lemeshow χ2 test and decile plot. Results:During the 10-year follow-up, a total of 955 (4.4%) CVD events, including 791 (3.6%) ASCVD events, were recorded among the study participants. The C-index for the CKB-CVD, China-PAR, and USA-PRC models were 0.775 (95% CI: 0.757-0.793), 0.781 (95% CI: 0.763-0.798), and 0.769 (95% CI: 0.750-0.789) for men, and 0.762 (95% CI: 0.737-0.788), 0.769 (95% CI: 0.745-0.794), and 0.767 (95% CI: 0.741-0.794) for women, respectively. China-PAR model showed good calibration for men ( χ2=2.20), however, both CKB-CVD and USA-PRC models demonstrated poor calibration in both men and women ( χ2>20). The results indicated that the CKB-CVD model overestimated the risk of CVD events in both males and females, while the China-PAR model underestimated the risk in females. Furthermore, the USA-PRC model underestimated the risk of ASCVD in both males and females in most decile groups, but overestimated the risk in the highest decile group. Conclusion:The CKB-CVD, China-PAR, and USA-PRC risk assessment models show some degree of deviation from the actual risk of events in the CMCS cohort, but all exhibit good discrimination.