Protein biomarkers for predicting incident carotid plaque: a nested case–control study in the ChinaHEART cohort with transportability assessment in the UK Biobank

Figures

• 

  Figure 1. Study design and participant selection. Flow chart of participant selection in the ChinaHEART cohort, illustrating the nested case–control design. Participants without carotid plaque at baseline were followed for 2 years, and incident cases were identified. Controls were selected from participants without carotid plaque during follow-up and matched to cases on age, sex, LDL-C, systolic blood pressure, body mass index, smoking status and diabetes history. Development of the prediction model in the ChinaHEART cohort, including biomarker measurement, variable selection and model construction. Assessment of the transportability and incremental predictive value of the selected biomarkers in the UK Biobank cohort for incident ASCVD. ASCVD, atherosclerotic cardiovascular disease; BMI, body mass index; BP, blood pressure; DM, diabetes mellitus; ChinaHEART, Health Evaluation And risk Reduction through nationwide Teamwork; CKD, chronic kidney disease; CVD, cardiovascular disease; LDL-C, low-density lipoprotein cholesterol; MI, myocardial infarction.

• 

  Figure 2. Model development and performance evaluation in the ChinaHEART cohort. (A) LASSO coefficient profiles of the candidate biomarkers across a range of penalty parameters (log λ). (B) The LASSO model used tenfold cross-validation to determine the optimal lambda. (C) Calibration curve of the logistic regression model for predicting incident carotid plaque, comparing predicted and observed probabilities. (D) Original ROC curve of the logistic model for predicting the risk of incident carotid plaque, demonstrating discriminative performance. The bootstrap-corrected AUC obtained from internal validation is reported separately in the Results section. (A, B) The variable selection process using LASSO regression, which informed the construction of the final logistic regression model shown in (C, D). AUC, area under the curve; ChinaHEART, Health Evaluation And risk Reduction through nationwide Teamwork; LASSO, least absolute shrinkage and selection operator; ROC, receiver operator characteristics.

• 

  Figure 3. ROC curve (A) and calibration plot of the model (B) after the addition of biomarkers in the UK Biobank. (A) Comparison of ROC curves between the traditional factor-building model and the model after the addition of biomarkers. (B) Calibration curve of the model after the addition of biomarker. AUC, area under the curve; LASSO, least absolute shrinkage and selection operator; ROC, receiver operator characteristics.