Validation of a Model for Predicting Anastomotic Leakage

Validation of a Machine Learning Model for Predicting Anastomotic Leakage of Esophagogastrostomy and Esophagojejunostomy: A Multicenter Prospective Study

This study will validate a machine learning model for predicting anastomotic leakage of esophagogastrostomy and esophagojejunostomy.

Anastomotic leakage is a fatal complication after total and proximal gastrectomy in gastric cancer patients. Identifying patients with high-risk of AL is important for guiding the surgeons' decision making, such as a more rigorous anastomotic operation， placing a jejunal feeding tube and dual-lumen flushable drainage catheter. We have developed a high-performance machine learning model based on 1660 gastric cancer patients, which showed good discrimination of anastomotic leakage. Hence, this multi-center prospective study will validiate the usability of the model for predicting anastomotic leakage in gastric cancer patients who receive total and proximal gastrectomy.

Inclusion Criteria:

Inclusion Criteria:

1. Aged older than 18 years and younger than 85 years.
2. Primary gastric adenocarcinoma confirmed by preoperative pathology.
3. Expected curative resection via total or proximal gastrectomy.
4. American Society of Anesthesiologists (ASA) class I, II, or III.
5. Written informed consent.

Exclusion Criteria:

1. Pregnant or breastfeeding women.
2. Severe mental disorder or language communication disorder.
3. Other surgical procedures of gastrectomy is performed.
4. Interrupted of surgery for more than 30 minutes due to any cause.
5. Malignant tumors with other organs