Prediction model and risk score for perforation in patients undergoing colorectal endoscopic submucosal dissection

Perforation is the adverse event of greatest concern during colorectal endoscopic submucosal dissection (ESD). Accurate risk prediction of perforation may enable prevention strategies and selection of the most efficient therapeutic option. This study aimed to develop and validate a risk prediction model for ESD-induced perforation.

A multicenter cross-sectional study was performed on 2046 patients who underwent colorectal ESD at 9 Korean ESD Study Group-affiliated hospitals. The enrolled patients were randomly divided into either a derivation set or a validation set. In the derivation set, a prediction score was constructed to assess the risk of perforation using preoperative and procedural-related predictors selected via logistic regression. Discrimination and calibration of the prediction model was assessed using the validation set.

An ESD-induced perforation occurred in 135 patients (6.6%). In the derivation set, multivariate logistic regression identified endoscopist experience (≥50 ESDs: odds ratio [OR] = 0.59; 95% confidence interval [CI], 0.35-1.00), tumor size (+1-cm increments: OR = 1.39; 95% CI, 1.19-1.62), colonic location (OR = 2.20; 95% CI, 1.24-3.89), and submucosal fibrosis (OR = 2.00; 95% CI, 1.04-3.87) as predictive factors (C-statistic = 0.678; 95% CI, 0.617-0.739). In the validation set, the model showed good discrimination (C-statistic = 0.675; 95% CI, 0.615-0.735) and calibration (P = .635). When a simplified weighted scoring system based on the OR was used, risk of perforation ranged from 4.1% (95% CI, 2.8%-5.9%) in the low-risk group (score ≤4) to 11.6% (95% CI, 8.5%-15.6%) in the high-risk group (score >4).

This study developed and internally validated a score consisting of simple clinical factors to estimate the risk of colorectal ESD-induced perforation. This score can be used to identify patients at high risk before colorectal ESD.