High-risk adenoma predicts delayed post-polypectomy bleeding: a size-stratified retrospective cohort study and risk score development

Vollständiger Abstract

Worum geht es in dieser Arbeit?

Background Delayed post-polypectomy bleeding (DPPB) is a common adverse event after endoscopic mucosal resection (EMR) of colorectal polyps. However, the role of high-risk adenoma (HRA) in DPPB has been rarely studied, as prior work focused on polyp number, age, and comorbidities; its size-specific effect remains unclear. Methods Conducted at a single center, this retrospective cohort study enrolled 4,035 patients undergoing EMR for colorectal polyps (October 2016–March 2025) after excluding inflammatory bowel disease, familial polyposis, incomplete records, and non-EMR techniques. The HRA–DPPB association was evaluated with sequential multivariable logistic regression across three nested adjustment models, with interactions tested against high-risk serrated polyp (HRSP) and polyp size (≤1 cm vs. >1 cm). A simplified clinical risk score derived by the Sullivan method was internally validated through an optimism bootstrap that rebuilt the score on every resample, assessing discrimination (area under the curve, AUC) and calibration. Results Of 4,035 patients, 106 (2.63%) developed DPPB. HRA was independently associated with DPPB (adjusted odds ratio [OR] 2.90, 95% confidence interval [CI] 1.74–4.85), and the estimate strengthened rather than weakened with sequential adjustment (crude OR 2.49). Pathology-stratified analysis showed a dose–response pattern—HRA only (OR 2.90), HRSP only (OR 2.27), both (OR 5.54)—with an additive interaction that trended positive without reaching significance [relative excess risk due to interaction (RERI) 1.37, 95% CI −2.65 to 5.38]. The HRA × polyp size interaction ( P = 0.005) was driven by the overlap between the HRA size criterion (≥1 cm) and the size-based grouping: when HRA was redefined by its non-size criteria alone (high-grade dysplasia, villous component, or ≥3 adenomas), the interaction disappeared ( P = 0.95), leaving histology (OR ≈ 1.5–2.1) and size (OR 7.4 at 2 cm vs. 0.8 cm) as independent predictors of bleeding. The simplified score [age, sex, polyp size, count, body mass index (BMI)] achieved an optimism-corrected AUC of 0.729 with adequate calibration (Hosmer–Lemeshow P = 0.222; corrected slope 0.923), stratifying patients into low (0.73%), medium (2.03%), and high (8.84%) risk groups. Conclusion HRA is an independent risk factor for DPPB after EMR, and high-risk histology predicts bleeding across the size spectrum, complementing a strong, independent effect of polyp size. A simplified clinical score stratifies risk adequately, but as a single-center, internally validated instrument it awaits external confirmation before clinical adoption.