A risk-prediction model for en bloc resection failure or perforation during endoscopic submucosal dissection of colorectal neoplasms.

Abstract

Technical difficulties in colorectal endoscopic submucosal dissections (ESD) result in en bloc resection failure or perforation. This study aimed to develop and validate a risk score for predicting en bloc resection failure or perforation in ESD of colorectal neoplasms. This single-center observational study included 1133 colorectal neoplasms treated with ESD in a Japanese tertiary cancer center. With a derivation set (n = 716), we performed multiple logistic regression to identify significant risk factors for en bloc resection failure or perforation. Based on odds ratios, we developed a risk score, ranging from 0 to 10: 0-1 'low risk' (LR); 2-4 'moderate risk' (MR); and 5-10 'high risk' (HR). An independent validation set comprised prospectively enrolled subjects (n = 417) that underwent ESDs from January 2014 to August 2016. The performance of the risk score for predicting en bloc resection failure or perforation for each risk tier was evaluated. The baseline incidences of en bloc resection failure or perforation were 14.5% and 5.5% in the derivation and validation sets, respectively. We identified the following significant risk factors: endoscopist experience, tumor location, morphology, scope operability, underlying fold, and fold convergence. In the validation set, the incidences of en bloc resection failure or perforation were 0% in the LR tier (n = 62; 14.8%), 2.3% in the MR tier (n = 293; 70.4%), and 25.8% in the HR tier (n = 62; 14.8%) (P < 0.001, Cochran-Armitage trend test). A risk scoring system, which was developed and prospectively validated, can successfully estimate the incidence of en bloc resection failure or perforation.