The Establishment And Validation Of Predictive Model And Scoring System For Delayed Bleeding After Endoscopic Submucosal Dissection Of Colorectal Lesions

Objective:With the development of endoscopic techniques and the implementation of screening programs worldwide,a growing number of colorectal lesions are diagnosed and treated endoscopically.Endoscopic Submucosal Dissection（ESD）has become one of the indispensable treatments for gastrointestinal lesions.Meanwhile,its complication,such as delayed bleeding,has received widespread attention.The study aimed to investigate the risk factors for delayed bleeding after colorectal ESD as well as to develop and validate a predictive model and scoring system,assisting clinicians in making reasonable diagnosis and treatment decisions.Methods:The data of patients who underwent colorectal ESD treatment at the First Affiliated Hospital of Nanchang University Gastrointestinal Endoscopy Center from June 2008to January 2021 were retrospectively collected,including general data,lesion-related data,and ESD-related data.The enrolled patients were randomly divided into a modeling group and a validation group at a ratio of 3:1.The former is used to establish the model and scoring system and the latter is used for validation.Univariate and multivariate analysis were used to select independent risk factors associated with delayed bleeding after colorectal ESD and to establish a logistic regression model.A scoring system was established based on the regression coefficientβvalue of each variable in the model.The risk stratification was performed according to the optimal cut-off value and divided into low-risk and high-risk groups.Hosmer-Lemeshow（H-L）test was used to evaluate the consistency of the model.Area under the curve（AUC）of receiver operating characteristic（ROC）curve was used to evaluate the discriminating ability of regression model and scoring system.Finally,the validation group data were substituted into the model and scoring system for validation.Results:1.A total of 911 patients included in this study were randomly divided into a modeling group（n=684）and a validation group（n=227）at the ratio of 3:1.The rate of delayed bleeding in modeling group and validation group were 3.4%（23/684）and 2.2%（5/227）respectively.2.In the modeling group,multivariate analysis showed that lesion location in the rectum（OR:4.621;95%CI:1.296-16.470;P=0.018）,lesion size≥2mm（OR:4.077;95%CI:1.474-11.275;P=0.007）,using argon plasma coagulation（APC）for wound management（OR:2.502;95%CI:1.016-6.161;P=0.046）,and partially closure with clips（OR:3.465;95%CI:1.051-11.424;P=0.041）were independent risk factors for delayed bleeding after colorectal ESD.As the number of risk factors increased,the incidence of delayed bleeding after colorectal ESD also increased.3.The predictive model:Logit P=-5.776+1.531X₁+1.405X₂+0.917X₃+1.243X₄(X₁=lesion location,X₂=lesion size,X₃=using APC for wound management,X₄=clip closure.The H-L test showed the model fit well（P=0.920>0.05）;the AUC of the model is 0.782（95%CI:0.692-0.871）,which indicates that the model has a good discrimination.4.A scoring system was established based on the scores calculated from the regression coefficientβvalues of the above 4 variables,with a total score of 0 to 6points.The AUC of the scoring system was 0.769,95%CI（0.670-0.867）,indicating that the scoring system had a good prediction.According to the optimal cut-off value,the total score>2points was classified as high-risk group,while score≥2points was classified as low-risk group.The diagnostic sensitivity,specificity and accuracy were78.3%,64.4%and 64.9%.The incidence of delayed bleeding between the two groups were 1.2%and 7.1%respectively,which has significant differences（P<0.001）.5.In the validation group,model showed good discrimination,which the H-L test’s P value was more than 0.05（P=0.951）and the AUC was 0.697,95%CI（0.478-0.917）.As for the scoring system,the AUC of scoring system was 0.603,95%CI（0.325-0.881）.With the increase of the score,the incidence of bleeding also tended to increase.With the same risk categories as defined in the modeling group,the incidence of delayed bleeding in the low-risk and high-risk groups were 2.1%and 2.4%respectively（P=0.856）.Conclusions:1.Lesion location in the rectum,lesion size≥20 mm,using APC for wound management,and partially closure with clips were independent risk factors for delayed bleeding after colorectal ESD.2.The predictive model and scoring system developed in this study are simple and practical,with good differentiation and predictive ability.It has good clinical application value to predict the risk of delayed bleeding after colorectal ESD,and provides a reference basis for clinicians to identify high-risk individuals and make optimal clinical decisions.