External validation and optimization of the Limoges score to improve the delayed bleeding risk prediction after...

The Limoges Bleeding Score estimates an individual's risk of clinically significant delayed bleeding (CSDB) after colorectal endoscopic submucosal dissection (ESD). We aimed to validate and update this model in a Western setting.

Procedural data and complications were prospectively recorded in French-Belgian (FECCO) and Spanish cohorts. The Limoges score was externally validated. A revised Western score was derived. Score performance was determined by discrimination and calibration. Internal validation was performed using bootstrapping and leave-one-out cross-validation. The score's performance was independently assessed in both cohorts.

4767 ESDs were included: 33.7 % rectal and 37.7 % proximal; median lesion size 50 mm; mean patient age 68.1 years; American Society of Anesthesiologists (ASA) score I-II 72.9 %; anticoagulants 10.9 %, and antiplatelets 17.0 %. CSDB prevalence was 6.8 %. The performance of the Limoges score was modest. A new score called DEBE (Delayed Bleeding ESD) was developed: age ≥ 75 years (2 points), lesion size ≥ 50 mm (5 points), ASA classification III-IV (4 points), location in the rectum (2 points) or proximal colon (1 point), anticoagulants (7 points) and antiplatelets (3 points). The DEBE score ranged from 0 to 23 points and categorized the patients into two groups (low risk 3.9 %; medium-high risk 14.2 %). The score showed acceptable discrimination (area under the curve 0.712), adequate calibration, and consistent performance after internal validation.

The DEBE score, based on seven preprocedural variables, allowed a personalized assessment of bleeding risk. It determined the individual CSDB risk, identified patients who would benefit from prophylactic treatment, and defined those who require monitoring after ESD.