AUTHOR=Wang Yun-Feng , Guo Zi-Qi , Han Jing-Xiang , Gao Lin-Na , Liu Yu-Ming , Jia Kai , Chen Hao , Yao Tian , Huang He 

TITLE=Analysis of risk factors for esophagojejunal anastomotic leakage after total gastrectomy based on Bayesian network model

JOURNAL=Frontiers in Medicine

VOLUME=Volume 12 - 2025

YEAR=2025

URL=https://www.frontiersin.org/journals/medicine/articles/10.3389/fmed.2025.1632214

DOI=10.3389/fmed.2025.1632214

ISSN=2296-858X

ABSTRACT=ObjectivesThis research aims to develop a nomogram for predicting esophagojejunal anastomotic leakage (EJAL) after total gastrectomy and analyze the relationship between individual risk factors through the Bayesian network model.Materials and methodsThe research enrolled 238 patients who underwent total gastrectomy and esophagojejunal Roux-en-Y anastomosis for gastric cancer between January 2017 and June 2022 in the Department of Gastrointestinal Surgery of the First Hospital of Shanxi Medical University and retrospectively collected clinical data of the patients. Multivariable logistic regression was used to explore the risk factors of EJAL and a nomogram based on the results was constructed. The predictive ability of the model was assessed by receiver operating characteristic (ROC) curve and calibration curve. In addition, the clinical benefit was indicated by decision curve analysis (DCA). Ultimately, a Bayesian network model was developed to analyze the interrelationship between the risk factors.ResultsEsophagojejunal anastomotic leakage occurred in 13 of 238 patients (5.4%). End-to-side anastomosis, diabetes mellitus (DM), preoperative albumin (ALB) ≤ 33.6 g/L, drinking history and systemic inflammation response index (SIRI) > 1.18 were identified as independent risk factors for EJAL based on multivariable logistic regression. A nomogram containing the aforementioned factors was constructed, with an area under the receiver operating characteristic curve (AUROC) of 0.880. Likewise, the model showed good predictive ability and clinical application in the calibration curve and DCA. Ultimately, the Bayesian network model demonstrates that type of anastomosis (ToA), DM, and ALB were directly associated with EJAL development, while gender, age, drinking history, smoking history, hypertension, and SIRI were conditionally dependent on EJAL given the presence of mediator variables.ConclusionSurgeons should be alert to the occurrence of EJAL, especially in patients with end-to-side anastomosis, DM, drinking history, preoperative lower ALB, and higher SIRI. Also, males, advanced age, smoking history, and hypertension can affect the development of EJAL.