Skip to main content

ORIGINAL RESEARCH article

Front. Oncol.
Sec. Surgical Oncology
Volume 14 - 2024 | doi: 10.3389/fonc.2024.1471137
This article is part of the Research Topic Surgical Management and Outcomes for Gastric Cancer View all articles

Identification of Intraoperative Hypoxemia and Hypoproteinemia as Prognostic Indicators in Anastomotic Leakage Post-Radical Gastrectomy: An 8-Year Multicenter Study Utilizing Machine Learning Techniques

Provisionally accepted
Yuan  Liu Yuan Liu 1Songyun  Zhao Songyun Zhao 2*Xingchen  Shang Xingchen Shang 1*Wei  Shen Wei Shen 1Wenyi  Du Wenyi Du 1*Ning  Zhou Ning Zhou 1*
  • 1 Department of General Surgery, Wuxi People's Hospital Affiliated to Nanjing Medical University, Wuxi, China
  • 2 Department of Neurosurgery, Wuxi People’s Hospital Affiliated to Nanjing Medical University, Wuxi, Jiangsu Province, China

The final, formatted version of the article will be published soon.

    Background: Complications and mortality rates following gastrectomy for gastric cancer have improved over recent years; however, complications such as anastomotic leakage (AL) continue to significantly impact both immediate and long-term prognoses. This study aimed to develop a machine learning model to identify preoperative and intraoperative high-risk factors and predict mortality in patients with AL after radical gastrectomy. Methods: For this investigation, 906 patients diagnosed with gastric cancer were enrolled and evaluated, with a comprehensive set of 36 feature variables collected. We employed three distinct machine learning algorithms-extreme gradient boosting (XGBoost), random forest (RF), and k-nearest neighbor (KNN)-to develop our models. To ensure model robustness, we applied k-fold cross-validation for internal validation of the four models and subsequently validated them using independent datasets.Results: In contrast to the other machine learning models employed in this study, the XGBoost algorithm exhibited superior predictive performance in identifying mortality risk factors for patients with AL across one, three, and five-year intervals. The analysis identified several common risk factors affecting mortality rates at these intervals, including advanced age, hypoproteinemia, a history of anemia and hypertension, prolonged operative time, increased intraoperative bleeding, low intraoperative percutaneous arterial oxygen saturation (SPO2) levels, T3 and T4 tumors, tumor lymph node invasion, and tumor peripheral nerve invasion (PNI). Conclusion: Among the three machine learning models examined in this study, the XGBoost algorithm exhibited superior predictive capabilities concerning the prognosis of patients with AL following gastrectomy. Additionally, the use of machine learning models offers valuable assistance to clinicians in identifying crucial prognostic factors, thereby enhancing personalized patient monitoring and management.

    Keywords: Gastric tumor, Gastrectomy, Anastomotic leakage, prognosis, risk factor, machine learning

    Received: 26 Jul 2024; Accepted: 12 Nov 2024.

    Copyright: © 2024 Liu, Zhao, Shang, Shen, Du and Zhou. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

    * Correspondence:
    Songyun Zhao, Department of Neurosurgery, Wuxi People’s Hospital Affiliated to Nanjing Medical University, Wuxi, 214023, Jiangsu Province, China
    Xingchen Shang, Department of General Surgery, Wuxi People's Hospital Affiliated to Nanjing Medical University, Wuxi, China
    Wenyi Du, Department of General Surgery, Wuxi People's Hospital Affiliated to Nanjing Medical University, Wuxi, China
    Ning Zhou, Department of General Surgery, Wuxi People's Hospital Affiliated to Nanjing Medical University, Wuxi, China

    Disclaimer: All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article or claim that may be made by its manufacturer is not guaranteed or endorsed by the publisher.