Establishment and evaluation of a predictive model for rebleeding after endoscopic treatment of esophageal and gastric varices

Objective To establish a predictive model for rebleeding after endoscopic injection of lauromacrogol combined with tissue adhesive in patients with esophageal and gastric varices, and to investigate its accuracy by cross validation. Methods A total of 180 patients with esophagogastric variceal bleeding who were admitted to Beijing Ditan Hospital from January 2014 to December 2016 were enrolled, and a retrospective analysis was performed for the clinical data of 126 patients, including age, sex, laboratory markers, Child-Pugh score, and degree of esophageal and gastric varices. A logistic regression analysis was used to screen out independent predictive factors for rebleeding and establish a predictive model, and the clinical data of the other 54 patients were used for cross validation of the accuracy of this model.The t-test was used for comparison of continuous data between two groups, and the chi-square test was used for comparison of categorical data between two groups. The logistic regression method was used to perform multivariate analysis and establish the predictive model. The receiver operating characteristic ( ROC) curve was used for evaluation of this model and cross validation of its accuracy. Results The 1-year rebleeding rate was 46. 83% in 126 patients, and there were significant differences between the rebleeding group and the non-rebleeding group in platelet count ( t =-7. 488, P < 0. 001) , international normalized ratio ( t = 3. 145, P = 0. 002) , and degree of esophageal varices ( χ2= 8. 841, P = 0. 031) . The multivariate logistic regression analysis showed that sex ( odds ratio [OR]= 3. 366, 95% confidence interval [CI]: 1. 015-11. 166, P = 0. 047) , platelet count ( OR = 0. 922, 95% CI: 0. 893-0. 951, P < 0. 001) , and degree of esophageal varices ( OR = 2. 422, 95% CI: 1. 179-4. 977, P = 0. 016) were independent predictive factors for rebleeding. The predictive model based on the combination of these three factors had an area under the ROC curve of 0. 876, a sensitivity of 86. 0%, and a specificity of 83. 1%.Cross validation based on the clinical data of the other 54 patients showed that this predictive model had an accuracy of 92. 6%. Conclusion Male sex and severe varices are high-risk factors for rebleeding, and platelet count is a protective factor. The predictive model based on thecombination of these three factors has a certain value in predicting rebleeding, with an accuracy of 92.6%.