Establishment of a machine learning-based predictive model with dual-center external validation: investigating the role of robotic surgery in preventing delayed gastric emptying for right-sided colon cancer.

After colorectal surgery, delayed gastric emptying (DGE) is a clinically significant postoperative complication that significantly lowers patients' quality of life. The evolving application of robotic surgery in gastrointestinal oncology continues to prompt investigation into its dual therapeutic potential: achieving oncological efficacy while mitigating DGE complications. The primary objectives of this study were to identify high-risk factors for DGE, develop a predictive model, and conduct internal and external validation. In addition, we investigated the potential advantages of robotic surgery in preventing DGE through cohort analysis and predictive model. This study utilized data from two major clinical research centers: Cohort 1: 522 right hemicolectomy cases (Northern Jiangsu People's Hospital, 2019-2024); Cohort 2: 115 cases (Huai'an Cancer Hospital, 2019-2024). Machine learning algorithms and logistic regression were employed to construct predictive models. After comparing their performance, the logistic regression model was selected to predict DGE following radical resection of right-sided colon cancer to further screening of high-risk factors for DGE and evaluation of the advantages of robotic surgery. The predictive model demonstrated robust performance upon internal and external validation, incorporating seven variables including: age (OR = 3.08), obstruction (OR = 5.51), preoperative hyperglycemia (OR = 2.56), preoperative potassium (OR = 3.55), surgical type (OR = 4.65) and anastomotic leakage (OR = 14.56). These variables were consequently identified as significant risk factors for DGE. Notably, cohort analysis revealed a slight reduction in DGE incidence with robotic surgery compared to laparoscopic approaches without statistically significant (9.0% vs 11.2%). We have established a more reliable predictive model for DGE which can provide guidance for clinical practitioners and conclude that robotic surgery demonstrates comparable efficacy to laparoscopic surgery, with satisfactory clinical outcomes in preventing the incidence of DGE.