Procedure duration is a key determinant of safety, resource utilization, and scheduling in colorectal endoscopic submucosal dissection (ESD). Although several lesion- and patientrelated factors have been associated with procedural difficulty, a validated model for predicting prolonged duration is lacking. This study aimed to identify independent predictors of procedure duration in colorectal ESD and quantify their relative contribution through categorical regression (CATREG), and develop and validate a clinically applicable risk score to predict prolonged procedures.

We retrospectively analyzed 114 consecutive colorectal ESDs recorded in a prospectively maintained database. Procedure duration was the primary outcome. Normality testing demonstrated significant deviation from normality, so non-parametric statistics were employed. Variables with p<0.05 were entered into a multivariable CATREG model with optimal scaling and ridge regularization (tolerance ≥0.40). Independent predictors were used to construct a weighted risk score based on normalized CATREG F-statistics. The score was validated against prolonged duration (≥1.5 hours) using receiver operating characteristic (ROC) analysis. 

Duration correlated positively with age, lesion size, maximal axis and circumferential involvement (all p<0.001). No significant differences were observed for sex, location, type of sedation, antithrombotic use, bleeding, perforation, en bloc resection, clip and band traction or underwater technique, even if the last two substantially contributed to the procedure effectiveness. Significantly longer procedures occurred in cases with complications (p=0.001), fibrosis (p=0.023), tunnel technique (p=0.001), LST-G morphology (p=0.028), previous interventions (p=0.007), intermuscular involvement (p<0.001), and R1 resection (p=0.014). Multivariable CATREG identified seven independent predictors which were used to construct a weighted risk score (range 0–22 points) for predicting prolonged procedures with excellent performance. The predictors included: circumferential involvement (p<0.001, 9 points) lesion size (p<0.001, 4 points), age (p=0.001, 3 points), complications (p=0.011, 3 points), intermuscular involvement (p=0.010, 2 points) LST-G morphology (p=0.037, 1 point) and previous interventions (p=0.046, 1 point). CATREG-derived optimal cutoffs included: size ≥12 mm, age ≥70 years, circumference ≥180°. 

Colorectal ESD duration is driven primarily by lesion complexity—particularly circumferential extent and lesion size—with additional contributions from age, intermuscular dissection, complications, LST-G morphology, and prior interventions. An ESD risk score incorporating these factors reliably predicts prolonged procedures and offers a practical tool for preprocedural planning. Prospective validation is warranted.