Related Experiment Video
Updated: May 16, 2025
Imaging In-Stent Restenosis: An Inexpensive, Reliable, and Rapid Preclinical Model
Published on: September 14, 2009
Systemic coagulation-inflammation index in the prediction of ISR in patients undergoing drug-eluting stents implant:
Luo Yinhua1, Yingying Hu1, Zeng Ziyue1
1Department of Cardiology, Zhongnan Hospital, Wuhan University, Wuhan, China; Institute of Myocardial Injury and Repair, Wuhan University, Wuhan, China.
Insights
The Systemic Coagulation-Inflammation index (SCI) can predict in-stent restenosis (ISR) in coronary artery disease (CAD) patients. A low SCI, male sex, older age, and smoking increase ISR risk, identified by an Xgboost model.
Area of Science:
- Cardiology
- Hematology
- Machine Learning
Background:
- The Systemic Coagulation-Inflammation index (SCI) is a novel metric reflecting coagulopathic and inflammatory states.
- Coronary artery disease (CAD) patients undergoing percutaneous coronary intervention (PCI) with drug-eluting stents (DES) are at risk for in-stent restenosis (ISR).
Purpose of the Study:
- To evaluate the prognostic impact of the SCI on ISR in patients with CAD.
- To develop and validate a machine learning model for predicting ISR.
Main Methods:
- A retrospective analysis of 724 CAD patients who underwent PCI with DES.
- Development of eight machine learning models to predict ISR, with the optimal model selected based on performance and clinical net benefit.
- External validation of the optimal model using data from Zhongnan Hospital of Wuhan University.
Main Results:
- The Xgboost model demonstrated superior performance (AUC = 0.971) for ISR prediction.
- Key predictors of ISR included male sex, increasing age, low SCI, high Gensini score, and smoking.
- External validation confirmed the model's predictive power (AUC = 0.92).
Conclusions:
- The Xgboost model is optimal for predicting ISR in CAD patients.
- SCI, sex, age, Gensini score, and smoking habits are significant predictors of ISR.
Background:
The Systemic Coagulation-Inflammation index (SCI) is an innovative hematological metric that accurately reflects both coagulopathic and inflammatory dynamics. In this paper, the objective of this paper is to explain the prognostic impact of SCI on ISR in patients with coronary heart disease (CAD).
Methods:
This retrospective study analyzed clinical data from 724 CAD patients who underwent PCI with DES between September 2017 and June 2023. The study compared preoperative clinical data between patients who developed ISR and non-ISR groups. To avoid overfitting, we implemented 5 folds of resampling prior to model development. We then divided the dataset into training and validation sets in a 7:3 ratio. We constructed eight different machine-learning models to predict the occurrence of ISR. We selected the optimal model based on its performance metrics and clinical net benefit, and further validated its predictive power using an independent external dataset. To interpret the results, we applied the Shapley Additive Explanation (SHAP) method to the final model, XgBoost, enabling a visual analysis of key predictors influencing ISR. Additionally, this paper externally validated the optimal model using data from Zhongnan Hospital of Wuhan University and plotted the corresponding working curves of the subjects. This approach offered robust and interpretable insights into factors related to ISR.
Results:
Significant differences in clinical characteristics were observed between the non-ISR and ISR groups. The internal validation results identified the Xgboost model as the optimal model due to its best performance (AUC = 0.971, 95 % CI 0.9569-0.9851) and favorable goodness-of-fit. Among the 10 predictive variables, sex, age, and SCI were strong predictors of ISR. Specifically, being male, increasing age, having a low SCI, a high Gensini score, and smoking habits were positively correlated with an increased risk of ISR. We validated the optimal model using external data from the Zhongnan Hospital of Wuhan University, which produced an excellent subject-worker curve with an area under the curve of 0.92.
Conclusion:
In this paper, we have identified the Xgboost model as the optimal model for predicting ISR. In addition, sex, age, Gensini score, and smoking habits were strong predictors of ISR.
