JoVE - Journal of Visualized Experiments
JoVE Visualize
Contact Us

Prospective multicenter validation of a machine learning model for predicting anastomotic leakage in patients with gastric adenocarcinoma undergoing total or proximal gastrectomy

  • 0Department of Surgery, Tongji Hospital, Tongji Medical College, Huazhong University of Science and Technology, Wuhan, Hubei, China.
International Journal of Surgery (london, England) +

|

July 23, 2025

|

July 23, 2025

View abstract on PubMed

Summary

This summary is machine-generated.

A new machine learning model effectively predicts anastomotic leakage (AL) risk in gastric cancer surgery. This tool helps identify high-risk patients and allows low-risk patients to avoid intensive monitoring, improving surgical outcomes.

Area Of Science

  • Surgical Oncology
  • Artificial Intelligence in Medicine
  • Gastrointestinal Surgery

Background

  • Anastomotic leakage (AL) is a severe complication after esophagogastric and esophagojejunal surgeries.
  • Traditional risk assessment methods for AL are often inadequate due to complex patient and operative variables.

Purpose Of The Study

  • To evaluate the clinical utility of a real-time machine learning (ML) model for predicting anastomotic leakage (AL).
  • To assess the ML model's ability to stratify patients based on AL risk in esophagogastric and esophagojejunal reconstructions.

Main Methods

  • Prospective enrollment of 512 gastric adenocarcinoma patients undergoing gastrectomy across four centers (January 2022 - January 2024).
  • Real-time application of a developed ML model during surgery to assess AL risk.
  • Primary outcome measured was the occurrence of AL.

Main Results

  • The ML model demonstrated an Area Under the Curve (AUC) of 0.780, with a sensitivity of 0.769 and a negative predictive value of 0.990.
  • AL occurred in 2.54% of patients (13/512).
  • The high-risk group (221 patients) had a significantly higher AL rate (4.5%) compared to the low-risk group (1.0%) (P=0.027), with ~35% potentially avoiding intensive monitoring.

Conclusions

  • The developed ML model offers effective risk stratification for AL in gastric cancer patients undergoing esophagogastrostomy or esophagojejunostomy.
  • The model achieves high sensitivity and can identify a significant proportion of patients at low risk for AL, potentially optimizing postoperative care.

Keywords: