This study was a collaboration between ESCP's Cohort Studies team and the University Basel (Department of biomedical engineering) and Cantonal Hospital Baselland. It provides insights about anastomotic leak in right hemicolectomy, taking a different approach to EAGLE 1 and EAGLE 2.
The study asked 'Can a predictive model that integrates multiple base models, predictions and patient-specific features accurately predict the incidence of anastomotic leakage?
In this retrospective prognostic study involving 9120 patients, the predictive performance of a meta-model achieved an overall F1 score of 87% and 70% using the cross-validation and external validation test sets, respectively.
These results suggest that a predictive model integrating multisource information can predict anastomotic leakage more accurately than individual base models, demonstrating the potential of advanced machine learning techniques to enhance clinical decision-making.
This prognostic study evaluates a meta-model that integrates logits from different existing predictive models combined with patient-specific features to predict anastomotic leakage risk after colorectal surgery.
The study was published on 20 October 2025 in JAMA Network Open:
Development of a Clinical Prediction Model for Anastomotic Leakage in Colorectal Surgery
Vincent Ochs, Stephanie Taha-Mehlitz, Joël L Lavanchy et al for the 2015 European Society of Coloproctology Collaborating Group. 2025;8;(10):e2538267. doi:10.1001/jamanetworkopen.2025.38267