Colorectal endoscopic submucosal dissection (ESD) is an effective curative technique for superficial colorectal lesions but remains technically demanding, with variable procedure times. This study aimed to develop and validate a preprocedure predictive model for colorectal ESD duration in the era of traction-assisted strategies.

We retrospectively analyzed prospectively collected data from patients who underwent colorectal ESD at Edouard Herriot Hospital (Lyon, France). The derivation cohort included 423 cases (January-December 2024), and an independent temporal validation cohort included 148 cases (January-May 2025). A generalized linear model with a gamma distribution and log-link function identified independent preprocedural predictors of ESD duration. Variables were selected on the basis of univariate significance and the Akaike information criterion. Model calibration and discrimination were assessed using R ² , mean absolute error, and classification accuracy at predefined thresholds (60, 90, and 120 minutes). A point-based score (Lyon ESD Dissection score, [LEDS]) and nomogram were derived from model coefficients.

Median procedure time in the derivation cohort was 50 minutes (IQR, 30-90). Independent predictors of ESD duration included dissection speed of endoscopists, lesion location (flexures, appendix, cecum, and ileocecal valve), predicted lesion size, and suspected fibrosis (all P < .05). The LEDS provides a direct estimation of the expected procedure time in minutes based on these variables, showing a strong correlation between predicted and observed durations ( R ² = 0.52). Classification accuracy was 76.1%, 82.2%, and 88.1% for thresholds of 60, 90, and 120 minutes, respectively. Validation confirmed good performance ( R ² = 0.48) with 91.8% accuracy for procedures ≥120 minutes.

This validated preprocedure scoring system accurately predicts colorectal ESD duration, especially for longer procedures. It may facilitate scheduling, optimize resource use, and improve workflow in endoscopy units.