结石相关型胆囊上皮内瘤变的相关因素分析及预测模型建立

Objective To evaluate the related factors of gallstones related gallbladder intraepithelial neoplasia(GBIN) and establish the prediction models for gallstones related GBIN. Methods The clinicopathological data of 750 patients who underwent cholecystectomy for gallstones at Department of Hepatobiliary Surgery of the First Affiliated Hospital of Xi′an Jiaotong University from January 2013 to December 2018 and the postoperative pathological examination showed chronic cholecystitis or GBIN were analyzed retrospectively, including 150 cases of gallstones with GBIN and 600 cases of gallstones with chronic cholecystitis. There were 264 males and 486 females with age of (51.3±14.5) years (range: 18 to 90 years). The related factors for gallstones related GBIN were screened by χ² test and Logistic regression model, and the prediction models were established based on independent related factors and internal validation was conducted. The original data were randomly divided into a training cohort(526 cases) and a validation cohort(224 cases) at a ratio of 7∶ 3, and the nomogram and tree augmented naïve Bayes were conducted to establish the prediction model for gallstones related GBIN.The consistency index (C-index), calibration chart, area under the receiver operating characteristic curve(AUC) and confusion matrix were used to evaluate the prediction performance of the two models. Results Univariate analysis showed that age, gallstones history(years), gallbladder size, whether the gallbladder mucosa smooth or not, whether the gallbladder wall thickened or not, gallstones diameter, and number of gallstones were related factors for the occurrence of gallstones related GBIN (χ² = 19.957, 8.599, 9.724, 9.301, 8.341, 15.288, 9.169, all P<0.05).Multivariate analysis showed that age (OR= 2.23, 95%CI: 1.50-3.31, P<0.01), gallbladder size (OR=2.11, 95%CI: 1.17-3.80, P=0.013), whether the gallbladder mucosa smooth or not (OR=1.80, 95%CI: 1.13-2.88, P=0.014), gallstones diameter(OR= 2.98, 95%CI: 1.71-5.21, P<0.01), and number of gallstones (OR=2.14, 95%CI:1.34-3.42, P<0.01) were independent related factors for the occurrence of gallstones related GBIN; the C-index of the nomogram in training cohort and validation cohort were 0.708 and 0.696, respectively. The AUC of the two models in training cohort were 70.60% and 70.73%, and in validation cohort were 68.14% and 67.47%, respectively. The accuracy of the two models in training cohort were 69.96% and 70.72%, and in validation cohort were 66.96% and 67.41%, respectively. Conclusion Age, gallbladder size, whether the gallbladder mucosa smooth or not, gallstones diameter and number of gallstones are independent related factors for the occurrence of gallstones related GBIN, and the nomogram and tree augmented naïve Bayes prediction models based on the above factors can be used to predict the occurrence of GBIN.