Key result
XGBoost outperforms other models predicting VLST-induced type 4b AMI with an AUC of ~0.72.
Why the study?
Very late stent thrombosis (VLST) represents the most complex subset of stent thrombosis, but no studies have developed prediction models for VLST-induced type 4b AMI using machine learning techniques.
Can machine learning models accurately predict type 4b AMI caused by very late stent thrombosis in patients with prior stent implantation?
Observational
Can machine learning models accurately predict type 4b AMI caused by very late stent thrombosis in patients with prior stent implantation?
Effect estimate: AUC 0.724 (95% CI 0.640-0.809)
Machine learning models, particularly XGBoost, can predict very late stent thrombosis-induced type 4b AMI with moderate accuracy, potentially aiding in risk stratification.
Background Acute myocardial infarction (AMI) resulting from Stent thrombosis (ST) is classified as type 4b AMI. The risk factors for ST are time-dependent, with very late stent thrombosis (VLST) representing the most complex subset. No studies have developed prediction models for VLST-induced type 4b AMI using machine learning techniques. Purpose To analyze the risk factors of type 4b AMI caused by VLST using machine learning methods and construct its risk prediction model. Methods Patients with a history of coronary stent implantation ≥1 year prior, who developed AMI and underwent coronary angiography, were classified into two groups: the VLST-4b AMI group and the primary AMI group, based on the presence of VLST in the culprit vessel during angiography. Data from the initial coronary stent implantation, medication regimens, baseline characteristics during hospitalization, and current coronary angiography findings were collected and compared between the two groups. Machine learning techniques were employed to identify risk factors for VLST-4b AMI, and a predictive model for VLST-4b AMI was developed followed by an evaluation of its performance. Results A total of 19 statistically significant indicators were identified through univariate logistic regression analysis,These indicators were then further screened using lasso regression, resulting in the selection of 12 variables strongly associated with the occurrence of VLST-4bAMI. Machine learning models including eXtreme Gradient Boosting (XGBoost), support vector machine (SVM), random forest (RF), and k-nearest neighbor (KNN) were utilized to predict VLST-4bAMI. XGBoost provided the best predictive performance (AUC = 0.724, 95% CI: 0.640-0.809), followed by KNN (AUC = 0.698, 95% CI: 0.614-0.783), RF (AUC = 0.669, 95% CI: 0.583-0.755), and SVM (AUC = 0.641, 95% CI: 0.547-0.736). Conclusion Machine learning methods offer a robust foundation for predicting VLST-4bAMI, aiding in the stratification of high-risk populations. Among these, XGBoost may provide the most accurate predictive performance.
No takes yet. Share an insight, caveat, or question.
Gao et al. (2025) conducted an observational in Type 4b acute myocardial infarction caused by very late stent thrombosis. eXtreme Gradient Boosting (XGBoost) model vs. Other machine learning models (KNN, RF, SVM) was evaluated on Prediction of VLST-4b AMI (AUC 0.724, 95% CI 0.640-0.809). The eXtreme Gradient Boosting (XGBoost) machine learning model predicted type 4b acute myocardial infarction caused by very late stent thrombosis with an AUC of 0.724 (95% CI 0.640-0.809).
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: