K-CVD as the Best-Calibrated ASCVD Risk Equation Among Three Country-Specific Equations in 9,532 Chinese Adults: A 9-Year Real-World Validation From CHARLS With Decision Curve Analysis
- 1Department of Cardiology (Cardiology Unit I), Handan First Hospital, Handan 056000, China
- 2Handan Key Laboratory of Cardiac Precision Medicine, Handan First Hospital, Handan, China
Abstract
Background: Precise 10-year atherosclerotic cardiovascular disease (ASCVD) risk prediction matters for Chinese primary prevention, while domestic risk equations have not been validated in 9-year real-world cohorts, and the Korean K-CVD model has never been validated locally.
Methods: This study enrolled 9,532 ASCVD-free Chinese adults aged 45–79 years from the China Health and Retirement Longitudinal Study (CHARLS, 2011–2020) with a 9-year follow-up (114 ASCVD events, incidence 1.20%). Three risk tools (pooled cohort equation (PCE)-US, China-PAR, and K-CVD-Korea) were compared regarding discrimination (area under the receiver operating characteristic (ROC) curve (AUC)), calibration, and clinical utility via decision curve analysis (DCA); MICE was adopted to impute missing systolic blood pressure data.
Results: K-CVD presented optimal calibration (predicted/observed ratio = 0.97). PCE overestimated risk by 17%, and China-PAR underestimated risk by 10%. The three models had comparable AUCs (0.775–0.777, P > 0.05). Post logit recalibration yielded identical Brier score (0.0116) for all models. DCA showed negative net benefit at thresholds ≥ 5%; PCE and K-CVD gained slight superiority over China-PAR at 1–2% thresholds matching the low actual event rate.
Conclusions: K-CVD has the best native calibration in this Chinese cohort. All three equations need formal recalibration for clinical use, and low cut-offs (1–2%) are more suitable for low-risk Chinese adults instead of Western standard thresholds.
Related
- Development and temporal validation of a calibrated machine learning model for in-hospital blood culture positivity using routine clinical data
- Calibrated Reliance on Large Language Models for Personalized Surgical Care: A Critical Narrative Review of Confidence, Clinical Reasoning and Decision Support
- HistoMRIFuseNet: retrospective confidence-calibrated decision fusion of independent histopathology and DCE-MRI classifiers
- On real-time calibrated prediction for complex model-based decision support in pandemics: Part 1
- Do Calibrated-Probability Claims Hold Up in Clinical Diagnosis? A Pre-registered Evaluation of a Decision Model Against Frontier Language Models
BibTeX
@article{zhang2026kcvd,
title = {{K-CVD as the Best-Calibrated ASCVD Risk Equation Among Three Country-Specific Equations in 9,532 Chinese Adults: A 9-Year Real-World Validation From CHARLS With Decision Curve Analysis}},
author = {Ren Jie Zhang and Xiu Feng Zhao and Meng Dan Miao and Wen Jing Liu and Chao Chang},
journal = {Cardiology research},
year = {2026},
month = oct,
doi = {10.14740/cr2315},
url = {https://europepmc.org/article/PMC/PMC13719645},
note = {CC BY},
}
RIS
TY - JOUR
TI - K-CVD as the Best-Calibrated ASCVD Risk Equation Among Three Country-Specific Equations in 9,532 Chinese Adults: A 9-Year Real-World Validation From CHARLS With Decision Curve Analysis
AU - Ren Jie Zhang
AU - Xiu Feng Zhao
AU - Meng Dan Miao
AU - Wen Jing Liu
AU - Chao Chang
PY - 2026
DA - 2026/10/24
JO - Cardiology research
DO - 10.14740/cr2315
UR - https://europepmc.org/article/PMC/PMC13719645
AB - Background: Precise 10-year atherosclerotic cardiovascular disease (ASCVD) risk prediction matters for Chinese primary prevention, while domestic risk equations have not been validated in 9-year real-world cohorts, and the Korean K-CVD model has never been validated locally. Methods: This study enrolled 9,532 ASCVD-free Chinese adults aged 45–79 years from the China Health and Retirement Longitudinal Study (CHARLS, 2011–2020) with a 9-year follow-up (114 ASCVD events, incidence 1.20%). Three risk tools (pooled cohort equation (PCE)-US, China-PAR, and K-CVD-Korea) were compared regarding discrimination (area under the receiver operating characteristic (ROC) curve (AUC)), calibration, and clinical utility via decision curve analysis (DCA); MICE was adopted to impute missing systolic blood pressure data. Results: K-CVD presented optimal calibration (predicted/observed ratio = 0.97). PCE overestimated risk by 17%, and China-PAR underestimated risk by 10%. The three models had comparable AUCs (0.775–0.777, P > 0.05). Post logit recalibration yielded identical Brier score (0.0116) for all models. DCA showed negative net benefit at thresholds ≥ 5%; PCE and K-CVD gained slight superiority over China-PAR at 1–2% thresholds matching the low actual event rate. Conclusions: K-CVD has the best native calibration in this Chinese cohort. All three equations need formal recalibration for clinical use, and low cut-offs (1–2%) are more suitable for low-risk Chinese adults instead of Western standard thresholds.
SN - 1923-2829
SN - 1923-2837
ER -