基于超声内镜下胃小间质瘤特征表现的诊断预测列线图模型建立  

A diagnostic prediction nomogram for small gastric stromal tumors based on features under endoscopicultrasonography

作　　者：章雁 陈晔[1] 孙会会[1] 陈莹[1] 熊杰 许树长[1] Zhang Yan;Chen Ye;Sun Huihui;Chen Ying;Xiong Jie;Xu Shuchang(Department of Gastroenterology,Tongji Hospital of Tongji University,Shanghai 200065,China)

机构地区：[1]同济大学附属同济医院消化内科,上海200065

出　　处：《中华消化内镜杂志》2023年第2期115-120,共6页Chinese Journal of Digestive Endoscopy

基　　金：上海市科学技术委员会资助项目(19411951604)。

摘　　要：目的建立基于超声内镜下特征表现的胃小间质瘤诊断预测列线图模型。方法回顾性收集2015年6月-2021年8月于同济大学附属同济医院消化内科行内镜下切除的长径<2 cm胃黏膜下肿瘤的189例患者临床病理资料。所有病例通过R软件随机函数按2∶1的比例分为建模组(n=126)和验证组(n=63)。在建模组中采用单因素和多因素Logistic回归分析筛选出超声内镜下诊断胃小间质瘤的独立影响因素, 构建列线图模型。在建模组和验证组中绘制受试者工作特征(receiver operator characteristic, ROC)曲线以评价模型的区分度, 采用Hosmer-Lemeshow检验和校准曲线以评价模型的校准度。结果患者年龄>60岁(OR=2.815, 95%CI:1.148~6.900, P=0.024)、病灶位于贲门/胃底(OR=5.210, 95%CI:1.225~22.165, P=0.025)、起源于固有肌层(OR=6.404, 95%CI:2.262~18.135, P<0.001)、呈腔外生长(OR=6.024, 95%CI:1.252~28.971, P=0.025)是在超声内镜下诊断胃小间质瘤的独立影响因素, 并以上述4个因素作为预测变量构建超声内镜下胃小间质瘤诊断预测列线图模型。该模型在建模组和验证组的ROC曲线下面积分别为0.834(95%CI:0.765~0.903)和0.780(95%CI:0.667~0.893)。Hosmer-Lemeshow拟合优度检验结果显示该模型具有较好的拟合度(建模组χ^(2)=10.23, P=0.176;验证组χ^(2)=2.62, P=0.918)。Bootstrap法绘制模型的校准图显示建模组和验证组的校准曲线与标准曲线均贴合良好。结论基于超声内镜下特征表现的胃小间质瘤诊断预测列线图模型具有良好的区分度和校准度, 为内镜医师在超声内镜下诊断胃小间质瘤提供了可视化参考工具。Objective To establish a nomogram based on features under endoscopic ultrasonography(EUS)for predicting the diagnosis of small gastric stromal tumors.Methods The clinicopathological data of 189 patients with gastric submucosal tumors(diameter less than 2 cm)who underwent endoscopic resection at the Department of Gastroenterology,Tongji Hospital of Tongji University from June 2015 to August 2021 were retrospectively collected.All patients were divided into the modeling group(n=126)and the validation group(n=63)at 2:1 by random function of software R.Independent influencing factors for the diagnosis of small gastric stromal tumors under EUS screened by univariable and multivariable logistic regression analysis were used to establish the diagnostic prediction nomogram.The receiver operator characteristic(ROC)curves were drawn to evaluate the discrimination of the model both in the modeling group and the validation group.Hosmer-Lemeshow test and calibration curve were used to evaluate the calibration of the model in both groups.ResultsThe age of patients>60 years(OR=2.815,95% CI:1.148-6.900,P=0.024),the lesions located in cardia/fundus(0R=5.210,95% CI:1.225-22.165,P=0.025),originated in muscularis propria(0R=6.404,95% CI:2.262-18.135,P<0.001)and of external growth(0R=6.024,95% CI:1.252-28.971,P=0.025)were independent influencing factors for the diagnosis of small gastric stromal tumors under EUS.The diagnostic prediction nomogram was established based on the four factors above.The areas under ROC curve of the modeling group and validation group were 0.834(95% CI:0.765-0.903)and 0.780(95%CI:0.667-0.893).Hosmer-Lemeshow test indicated that this model fit the data well(x^(2)=10.23,P=0.176 in the modeling group;x^(2)=2.62,P=0.918 in the validation group).Calibration charts of the model drawn by Bootstrap method showed that the calibration curves fit well with the standard curves in both groups.ConclusionThe nomogram based on features under EUS for predicting the diagnosis of small gastric stromal tumors provides a visual

关 键 词：超声检查 列线图 胃小间质瘤 诊断预测模型 

分 类 号：R735.2[医药卫生—肿瘤]