Validation and Recalibration of PCE, China-PAR, and PREVENT Models for Estimating ASCVD Risk in China

Figures

• 

  Figure 1. Selection and Characteristics of the Study Population Graphical overview of cohort selection. We first validated the Pooled Cohort Equations (PCEs), Prediction of atherosclerotic cardiovascular disease risk in China (China-PAR), and Predicting Risk of Cardiovascular Disease EVENTs (PREVENT) model in the whole cohort. To avoid overfitting, we randomly split our cohort into the derivation set (60%) and the validation set (40%). Original PCE, China-PAR, and PREVENT model were recalibrated in the derivation set and validated in the validation set.

• 

  Figure 2. Calibration Plots of Predicted vs Observed Rate in Whole Cohort The observed rate of atherosclerotic cardiovascular disease (ASCVD) events at 10 years was estimated based on the Kaplan-Meier method. Predicted 10-year risk of ASCVD was calculated using original PCE, China-PAR, and PREVENT models. Data points represent mean and error bars are 95% CI. Abbreviations as in Figure 1.

• 

  Central Illustration. Validation of the Original and Recalibrated PCE, China-PAR, and PREVENT Calibration plots depict the mean observed and predicted 10-year risk of atherosclerotic cardiovascular disease (ASCVD) events across quintiles of predicted risk in the validation set. China-PAR = Prediction of ASCVD risk in China; PCE = Pooled Cohort Equation; PREVENT = Predicting Risk of Cardiovascular Disease EVENTs.

• 

  Figure 3. Decision Curves for Original and Recalibrated Models in Validation Set The 10-year treatment probability threshold was calculated using the original and recalibrated the PCE, China-PAR, PREVENT models. Abbreviations as in Figures 1 and 2.

• 

  Figure 4. Performance for Original vs Updating Models in Whole Cohort As presented in the forest plots of Harrell’s C-index (A), calibration curves (B), and decision curves (C), the original PCE, China-PAR, and PREVENT models were compared with their updated counterparts. The latter were constructed by integrating 2 key biomarkers—triglyceride-glucose (TyG) index (linked to insulin resistance) and high-sensitivity C-reactive protein (hsCRP) (associated with inflammation)—into the original model. Abbreviations as in Figures 1 and 2.