Analysis Of Risk Factors And Model Construction Of Esophageal Stenosis After Endoscopic Submucosal Dissection

Background:Endoscopic submucosal dissection(ESD)has become the main treatment for early esophageal cancer and precancerous lesions.Due to its high complete resection rate,low postoperative recurrence rate,definite pathological evaluation,less trauma,and rapid recovery.However,esophageal stenosis is still a common complication after ESD,and the occurrence of esophageal stenosis will seriously reduce the quality of life of patients.At present,the primary treatment for esophageal stenosis is endoscopic dilatation.Some patients with severe esophageal stenosis may require repeated endoscopic dilatation treatments.At present,the methods for judging the occurrence of esophageal stricture and the number of endoscopic dilatation of postoperative are subjective,so the purpose of this study is to establish a predictive model to more objectively and accurately identify the high-risk patients with esophageal stricture after ESD and predict the number of esophageal stricture dilatation.Aims:1.To investigate the risk factors of esophageal stricture and refractory esophageal stricture after endoscopic submucosal dissection of early esophageal cancer and precancerous lesions;2.To establish a model for predicting the risk of esophageal stricture and the number of endoscopic esophageal dilatation after ESD.Methods:1.Retrospectively analyzed the patients with early esophageal cancer and precancerous lesions underwent ESD in the First affiliated Hospital of Zhengzhou University between January 2015 and April 2020.Collected demographic characteristics,clinical data,and developed a nomogram model.Risk factors for stenosis were assessed by Multivariate Logistic regression analysis.The performance of the model was evaluated by the consistency index(C-index)and the calibration curve.Bootstrap was applied to complete internal verification to avoid over-fitting of the model.2.Among the above-collected data of esophageal stricture,eligible patients were screened out by inclusion and exclusion criteria,risk factors of refractory esophageal stricture were analyzed by SPSS26.0,the best parameters of the model were screened out by python software using cross-validation,the prediction model of the number of esophageal stricture dilation was constructed,and the prediction effect of the model was evaluated by mean square error(mean squared error,MSE).Results:1.A total of 421 patients with early esophageal cancer and precancerous lesions followed by ESD were enrolled in the model of the risk of esophageal stenosis,of which 89 patients developed esophageal stenosis.Significant differences include postoperative pathology,depth of infiltration,longitudinal resection length,circumferential specimen size,circumferential resection range,and muscular injury between the stenosis and non-stenosis group(P<0.05).Logistic regression analysis showed that the circumferential resection range(odds ratio(OR)14.748,7.057 to30.820;P<0.05)and muscular injury(odds ratio(OR)3.450 1.519 to7.836;P<0.05)were independent risk factors for esophageal stenosis after ESD.The Nomogram model was constructed through the lasso method by taking the longitudinal resection length,circumferential resection range,and muscle injury as predictive factors.The concordance index(C-index)of the apparent nomogram was 0.9339(95%confidence interval:0.909-0.959),which was confirmed to be 0.9308 after 100 times of Bootstrap internal sampling.The prediction probability of the model in the calibration curve was in good agreement with the actual observation probability.2.A total of 55 patients were included in the model of esophageal dilatation times,of which 8(14.5%)patients developed refractory esophageal stricture.There was a significant difference in the circumferential resection range and the muscular injury between the refractory esophageal stricture group and the non-refractory esophageal stricture group(P<0.05).Based on the XGBoost algorithm to build a regression model,the first three important features in the analysis are the muscular injury,the circumferential resection range,and the longitudinal resection length,indicating that these three characteristics play an important role in the number of esophageal stricture dilatation.The mean square error of this model is 0.0754.Conclusions:1.The extent of the circumferential resection range and muscular injury are independent risk factors for postoperative esophageal stenosis.The Nomogram model is validated and helps for setting up standard guidelines of early intervene in patients with a high risk of stenosis after ESD.2.The extent of the circumferential resection range and the muscular injury are the risk factors of refractory esophageal stenosis.In this study,a regression tree model is constructed based on the XGBoost algorithm.The muscular injury,circumferential resection range,and the longitudinal resection length were important factors affecting the number of esophageal stenosis dilatation.The prediction effect of this model is good and has a certain application value in clinical practice.