Validation and update of a clinical score model to predict technical difficulty of colorectal endoscopic submucosal dissection: a multicenter prospective cohort study

Background and Aims The Zhongshan colorectal endoscopic submucosal dissection (CR-ESD) score model was proposed to grade the technical difficulty of CR-ESD. The objective of this study was to prospectively validate and update the score model. Methods A multicenter prospective cohort analysis of CR-ESD was conducted. Individual data on patients, lesions, and outcomes of CR-ESD were used to validate the original model and further refine the difficulty of the prediction model. Data were randomly divided into discovery and internal validation cohorts. A multivariate Cox regression analysis was conducted on the discovery cohort to develop an updated risk-scoring system, which was then validated. Results Five hundred forty-eight patients with 565 colorectal lesions treated by ESD from 4 hospitals were included. In the prospective validation cohort, the area under the receiver-operating characteristic (ROC) curve for the original model was .707. Six risk factors were identified and assigned point values: tumor size (2 points for 30-50 mm, 3 points for ≥50 mm), at least two-thirds circumference of the lesion (3 points), tumor location in the cecum (2 points) or flexure (2 points), laterally spreading tumor-nongranular lesions (1 point), preceding biopsy sampling (1 point), and NBI International Colorectal Endoscopic type 3 (3 points). The updated model had an area under the ROC curve of .738 in the discovery cohort and of .782 in the validation cohort. Cases were categorized into easy (score = 0-1), intermediate (score = 2-3), difficult (score = 4-6), and very difficult (score ≥7) groups. Satisfactory discrimination and calibration were observed. Conclusions The original model achieved an acceptable level of prediction in the prospective cohort. The updated model exhibited superior performance and can be used in place of the previous version. (Clinical trial registration number: ChiCTR2100047087.) The Zhongshan colorectal endoscopic submucosal dissection (CR-ESD) score model was proposed to grade the technical difficulty of CR-ESD. The objective of this study was to prospectively validate and update the score model. A multicenter prospective cohort analysis of CR-ESD was conducted. Individual data on patients, lesions, and outcomes of CR-ESD were used to validate the original model and further refine the difficulty of the prediction model. Data were randomly divided into discovery and internal validation cohorts. A multivariate Cox regression analysis was conducted on the discovery cohort to develop an updated risk-scoring system, which was then validated. Five hundred forty-eight patients with 565 colorectal lesions treated by ESD from 4 hospitals were included. In the prospective validation cohort, the area under the receiver-operating characteristic (ROC) curve for the original model was .707. Six risk factors were identified and assigned point values: tumor size (2 points for 30-50 mm, 3 points for ≥50 mm), at least two-thirds circumference of the lesion (3 points), tumor location in the cecum (2 points) or flexure (2 points), laterally spreading tumor-nongranular lesions (1 point), preceding biopsy sampling (1 point), and NBI International Colorectal Endoscopic type 3 (3 points). The updated model had an area under the ROC curve of .738 in the discovery cohort and of .782 in the validation cohort. Cases were categorized into easy (score = 0-1), intermediate (score = 2-3), difficult (score = 4-6), and very difficult (score ≥7) groups. Satisfactory discrimination and calibration were observed. The original model achieved an acceptable level of prediction in the prospective cohort. The updated model exhibited superior performance and can be used in place of the previous version. (Clinical trial registration number: ChiCTR2100047087.)