肝损伤相关指标构建的logistic预测模型在胆源性胰腺炎中的预测价值  

The predictive value of logistic model constructed by liver injury related index in biliary pancreatitis

作　　者：孙嘉龙 吴铁龙 薛育政 郁淯晟 任怡琳 刘天浩 戴圆圆 樊子君 盛颖玥 Sun Jialong;Wu Tielong;Xue Yuzheng;Yu Yusheng;Ren Yilin;Liu Tianhao;Dai Yuanyuan;Fan Zijun;Sheng Yingyue(Department of Gastroenterology,Affiliated Hospital of Jiangnan University,Wuxi 214000,China)

机构地区：[1]江南大学附属医院消化内科,无锡214000

出　　处：《中华肝胆外科杂志》2025年第3期167-171,共5页Chinese Journal of Hepatobiliary Surgery

基　　金：国家自然科学基金(82405210、32101964、32372302);无锡市第二届“双百”中青年医疗卫生人才资助(HB2023043、BJ2023046)。

摘　　要：目的基于肝损伤指标构建预测急性胆源性胰腺炎(ABP)的logistic回归预测模型并分析其预测效果。方法回顾性分析2020年10月至2022年12月在江南大学附属医院诊断为急性胰腺炎(AP)的210例患者资料,其中男性113例,女性97例,年龄52.0(43.0,58.0)岁。210例AP患者中的ABP患者纳入ABP组(n=88),急性非胆源性胰腺炎(ANBP)患者纳入ANBP组(n=122)。同时回顾性收集和分析2023年1月至12月在江南大学附属医院诊断为AP的101例患者资料作为测试集,其中男性60例,女性41例,年龄53.0(43.0,63.0)岁。基于前一数据集采用单因素和多因素logistic回归分析ABP的影响因素。同时基于多因素结果建立ABP预测概率Pre的公式即logistic预测模型,并利用受试者工作特征(ROC)曲线评估各指标预测ABP的效果。ROC曲线分析得出Pre预测ABP的最佳界值,以此数值诊断测试集中的ABP与ANBP。结果多因素logistic回归分析,直接胆红素、丙氨酸转氨酶(ALT)、天冬氨酸转氨酶(AST)、胆碱酯酶、血浆纤维蛋白原为ABP的影响因素。基于多因素结果建立ABP预测概率Pre公式,Pre=1/{1+exp[-(4.807+0.134×DBIL-1.859×AST/ALT-0.0003×CHE-0.387×FIB)]},DBIL为直接胆红素,CHE为胆碱酯酶,FIB为纤维蛋白原。ROC曲线分析Pre值预测ABP的曲线下面积为0.858,最佳界值为0.56,灵敏度为69.3%,特异度为91.0%。依据预测概率Pre的最佳界值0.56,Pre值≥0.56判断为ABP,Pre值<0.56判断为ANBP,以此标准预测测试集患者。Pre值诊断ABP的灵敏度为86.1%,特异度为92.3%。结论由肝损伤相关指标建立的ABP logistic预测模型有助于临床评估ABP。Objective To establish and evaluated a logistic regression model for predicting the acute biliary pancreatitis(ABP)based on liver-injury related indexes.Methods Clinical data of 210 patients diagnosed with acute pancreatitis(AP)at the Affiliated Hospital of Jiangnan University from October 2020 to December 2022 were retrospectively analyzed,including 113 males and 97 females,with a median age of 52 years(range,43 to 58).Among these,88 were diagnosed with ABP and 122 with acute non-biliary pancreatitis(ANBP).Additionally,a test cohort was created using data from 101 AP patients diagnosed between January and December 2023,including 60 males and 41 females,with a median age of 53 years(range,43 to 63).Based on the original dataset,univariate and multivariate logistic regression analyses were conducted to identify the factors influencing ABP.A prediction probability formula(Pre)was then established based on the multivariate results.The effectiveness of each indicator in predicting ABP was evaluated using the receiver operating characteristic(ROC)curve.The ROC curve analysis determined the optimal cutoff value of Pre,which was subsequently used to diagnose ABP and ANBP in the test cohort.Results Multivariate logistic regression analysis showed the factors influencing ABP include direct bilirubin(DBIL),alanine aminotransferase(ALT),aspartate aminotransferase(AST),cholinesterase(CHE),and fibrinogen(FIB).Based on the multivariate analysis results,the prediction probability formula(Pre)for ABP was established as follows:P=1/{1+exp[-(4.807+0.134×DBIL-1.859×AST/ALT-0.0003×CHE-0.387×FIB)]}.ROC curve analysis revealed that the area under the curve(AUC)for Pre in predicting ABP was 0.858,with an optimal cutoff value of 0.56,at which the sensitivity was 69.3%and the specificity was 91.0%.Using the cutoff value of 0.56 for Pre,ABP was diagnosed when Pre≥0.56 and ANBP was diagnosed when Pre<0.56.This criterion was applied to diagnose patients in the test cohort,where the sensitivity and specificity of Pre for diagnosing ABP

关 键 词：胰腺炎 肝损伤 Logistic预测模型 

分 类 号：R576[医药卫生—消化系统]