Machine-Learning Prediction of Bleeding After Endoscopic Submucosal Dissection for Early Gastric Cancer: A...

Area of Science:

• Gastroenterology
• Oncology
• Medical Informatics

Background:

• Endoscopic submucosal dissection (ESD) is a key treatment for early gastric cancer (GC).
• Post-ESD bleeding is a significant and unpredictable complication.
• Accurate prediction of post-ESD bleeding is crucial for patient management.

Purpose of the Study:

• To develop and validate machine learning (ML) models for predicting post-ESD bleeding.
• To identify significant risk factors associated with post-ESD bleeding.
• To compare the performance of ML models against traditional logistic regression.

Main Methods:

• Retrospective analysis of a multicenter clinical database of patients undergoing ESD for early GC.
• Development of an ML model using patient characteristics and perioperative data.
• Comparison of ML model performance with a logistic regression model, including feature importance analysis.

Main Results:

• The ML model demonstrated superior predictive performance (AUC 0.80) compared to the non-ML model (AUC 0.71, p=0.03).
• ML-identified risk stratification showed increasing bleeding rates with higher predicted probabilities (2.3% low, 8.8% intermediate, 28.6% high).
• Anticoagulant use in atrial fibrillation was identified as a key predictor of post-ESD bleeding.

Conclusions:

• ML models effectively predict and help rule out post-ESD bleeding.
• The developed ML model aids in identifying clinically relevant risk factors.
• These findings support the use of ML for personalized prophylactic strategies in early GC patients undergoing ESD.