Complicatons in gastro-intestinal surgery: From clinical factors to AI-driven management

Part I: Implications of Complications The first part of the study investigates the impact of neoadjuvant treatment and anastomotic leakage on recurrence rates after esophageal cancer surgery. Chapter 2 presents a retrospective study on esophageal cancer patients who received neoadjuvant chemoradiotherapy followed by surgery at Amsterdam UMC between 2010 and 2018. The study found that 45% of patients experienced cancer recurrence, primarily outside the initial radiotherapy field. Patients with loco-regional recurrence treated with curative intent had prolonged survival, indicating a need for further research to identify which patients might benefit most from such treatments. Chapter 3 explores the relationship between neoadjuvant treatment, anastomotic leakage, and cancer recurrence. Contrary to previous studies, it found that anastomotic leakage did not increase the risk of recurrence. However, patients who had both recurrence and anastomotic leakage experienced a shorter time to recurrence. This finding suggests the need for further research into the follow-up strategies for patients with anastomotic leakage. Part II: Prevention of Complications This section discusses the benefits of hospital mergers and surgical techniques to prevent complications. Chapter 4 assesses the impact of the merger between the Academisch Medisch Centrum and the Vrije Universiteit medisch centrum. The retrospective cohort study showed no decline in short-term quality metrics post-merger, suggesting that hospital mergers can improve specialized care without compromising service quality. Chapter 5 compares different techniques for robot-assisted intrathoracic anastomosis during Ivor Lewis esophagectomy. It found the circular stapled technique to be the most uniform and easiest to learn, recommending its use as the starting point for surgeons transitioning to robot-assisted surgery. Chapter 6 examines surgical approaches for high-risk esophageal cancer patients using data from the Dutch Upper GI Cancer Audit (DUCA). The study found that high-risk patients had more complications post-surgery. However, those who underwent the transhiatal approach had lower morbidity and mortality rates compared to the transthoracic approach, suggesting that the transhiatal method may be safer for high-risk patients. Part III: Prediction of Complications The final part focuses on using artificial intelligence (AI) to predict surgical complications. Chapter 7 reviews the literature on AI's application in predicting complications in major abdominal surgery, noting its potential for accurate predictions but emphasizing the need for rigorous testing and validation before clinical implementation. Chapter 8 delves into the challenges and recommendations for integrating AI models into clinical practice for predicting postoperative complications, highlighting the need for proper model development and implementation strategies. Chapter 9 compares AI with logistic regression in predicting postoperative complications using data from the Dutch Pancreatic Cancer Audit (DPCA). The study found that AI did not outperform logistic regression when using structured clinical data, suggesting that logistic regression remains a reliable method. Chapter 10 outlines the protocol for the A1Check study, which tests a machine learning model's ability to predict colorectal anastomotic leakage during surgery. This study aims to validate the model externally, with the hypothesis that AI can accurately predict leakage in real-time, potentially improving surgical decision-making and outcomes.