Nomogram模型用于胆道闭锁早期诊断的效能验证  

Efficacy verification of Nomogram tool for an early diagnosis of biliary atresia

作　　者：刘坤慧 龚一鸣[1] 蔡威[1] Liu Kunhui;Gong Yiming;Cai Wei(Department of Pediatric General Surgery,Affiliated Xinhua Hospital,Shanghai Jiao Tong University School of Medicine,Shanghai 200092,China)

机构地区：[1]上海交通大学医学院附属新华医院儿普外科,上海200092

出　　处：《临床小儿外科杂志》2023年第3期211-217,共7页Journal of Clinical Pediatric Surgery

基　　金：上海市卫健委上海市临床重点专科建设项目-出生缺陷(shslczdzk05702)

摘　　要：目的对当前热门的胆道闭锁(biliary atresia,BA)Nomogram预测模型进行单中心数据验证,并评估其诊断效能。方法回顾性分析上海交通大学医学院附属新华医院2009—2020年因婴儿梗阻性黄疸行手术探查的631例患儿临床资料,根据术中胆道探查及造影结果分为胆道闭锁组(500例)和非胆道闭锁组(131例)。在Nomogram模型中输入同一患儿的性别、体重、碱性磷酸酶(alkaline phosphatase,ALP)、γ-谷氨酰转肽酶(gamma-glutamyltransferase,GGT)和直接胆红素(direct bilirubin,DB)数值,得到该患儿预测为BA的风险概率。分别按手术年份和手术时日龄对患儿进行分组。将患儿BA预测概率值与其他肝功能数值在同一坐标系内绘制受试者操作特征(receiver operating characteristic,ROC)曲线,得到诊断BA的曲线下面积(area under the curve,AUC)、灵敏度、特异度,评估该Nomogram预测模型对BA患儿早期诊断的准确性。结果按手术年份分组,性别、术前体重、手术时日龄和GGT在每个年份亚组内BA患儿和非BA患儿中的差异均有统计学意义(P<0.05),ALP和DB的差异均无统计学意义(P>0.05)。在前期组(2009—2014年)和后期组(2015—2020年)的对比中,前期组Nomogram模型和GGT单独诊断的AUC值、灵敏度均明显优于后期组;前期组Nomogram模型的AUC=0.865、灵敏度为84.1%,GGT的AUC=0.851、灵敏度为85.8%;后期组Nomogram模型的AUC=0.787、灵敏度为68.2%,GGT的AUC=0.813、灵敏度为65.2%。按手术时日龄分层,日龄≤50 d和81~90 d的患儿中,Nomogram模型的AUC分别为0.806和0.866,均低于GGT的0.807和0.896;而日龄51~60 d、61~70 d、71~80 d和>90 d的患儿中Nomogram模型的AUC值均高于GGT。结论Nomogram预测模型诊断BA较单一指标诊断有较高的AUC值和灵敏度,但在手术时日龄≤50 d或>80 d的患儿中诊断BA的AUC值较GGT单独诊断要低,因此采用该模型对所有日龄的患儿进行BA预测可能不够准确。Objective To evaluate the efficacy of currently popular Nomogram prediction model for biliary atresia(BA)through a single-center data model.Methods Retrospective review was preformed for 631 children undergoing surgical explorations for infantile obstructive jaundice from 2009 to 2020.According to the findings of intraoperative cholangiography,they were assigned into two groups of BA(n=500)and non-BA(n=131).Gender,body weight,ALP(alkaline phosphatase),GGT(gamma glutamyltransferase)and DB(direct bilirubin)were input into the model for estimating the predicted risk probability of BA in the same child.They were grouped according to operative year and day age at surgery.The predicted probability of BA and other liver function parameters were plotted as receiver operating characteristic curve(ROC)curve in the same coordinate system.And area under the curve(AUC),sensitivity and specificity for diagnosing BA were obtained.Also the accuracy of model for an early diagnosis of BA was evaluated.Results Nomogram predictive model had higher AUC value and sensitivity for diagnosing BA than single parameter diagnosis.However,AUC value for diagnosing BA in children aged≤50 day or>80 day was lower than that for GGT alone.Thus the prediction of pediatric BA using this model might not be accurate enough.Conclusion Nomogram predictive model has higher AUC value and sensitivity for diagnosing BA than single parameter diagnosis.However,AUC value for diagnosing BA in children aged≤50 day or>80 day is lower than that for GGT alone.

关 键 词：胆道闭锁 列线图 Γ-谷氨酰转肽酶 碱性磷酸酶 早期诊断 

分 类 号：R726.5[医药卫生—儿科]