Developing a novel diagnostic model for identifying high-risk plaques in new onset unstable angina pectoris using coronary CT angiography.

BACKGROUND

Limited evidence supports the use of electronic health records for developing prediction models to identify high-risk plaques in patients with unstable angina pectoris (UAP). This study aimed to develop and validate a practical high-risk plaque prediction model in patients with new onset UAP.

METHODS

We prospectively enrolled consecutive patients presenting with new-onset UAP who underwent both coronary angiography and coronary computed tomography angiography (CCTA) at our center from January 2021 to December 2021. Based on the CCTA findings, the patients were categorized into two distinct groups: a high-risk plaque group (n=57) and a low-risk plaque group (n=26). We utilized LASSO regression and the Boruta algorithm for feature selection and performed multivariate logistic regression analyses to identify variables associated with high-risk plaque. Internal validity of the predictive model was assessed using bootstrapping (500 replications).

RESULTS

We developed a nomogram to predict high-risk plaque likelihood using LASSO regression, the Boruta algorithm, and multivariate logistic regression analyses. This approach identified four clinical features as significant predictors: diabetes mellitus, current smoking, total cholesterol, and lipoprotein(a). The area-under-the-curve (AUC) values, calculated using the bootstrap method with 500 replicates, for evaluating high-risk plaque in both the development and validation cohorts, were 0.851, accompanied by a 95% Confidence Interval (CI) ranging from 0.768 to 0.935. The nomogram exhibited satisfactory calibration when assessed with the bootstrap method (500 replicates), indicating a strong correlation with high-risk plaque as determined by CCTA. Furthermore, decision curve analysis indicated the clinical utility of this nomogram in accurately predicting high-risk plaque. And a web-based dynamic nomogram was further built to facilitate the prediction procedure.

CONCLUSIONS

Our prediction nomogram, developed using electronic health records, demonstrated robust capability in accurately identifying high-risk plaque among new onset patients with UAP. The implementation of this predictive tool holds great potential for tailoring individualized treatment strategies.