经颈静脉肝内门腔静脉分流术后早期肝性脑病列线图的建立和验证

Establishment and validation of nomogram in patients with early hepatic encephalopathy after TIPS

目的构建经颈静脉肝内门腔静脉分流术(transjugular intrahepatic portosystemic shunt,TIPS)后早期肝性脑病(hepatic encephalopathy,HE)的预测模型列线图并进行验证,以期为患者个体化风险预测与临床获益评估提供量化工具。方法收集南方医科大学南方医院262例患者临床资料进行回顾性分析,使用SPSS软件随机抽取70%病例纳入训练集构建临床预测模型(clinical prediction models,CPMs),剩余30%病例纳入验证集验证CPMs。通过Logsitic回归分析构建预测模型图,并计算ROC曲线下面积、Calibration plot、决策曲线分析法,对CPMs的区分度、校准度、临床实用性进行全面的临床效能评估与验证。结果(1)年龄(OR=1.975)、肝功能Child-Pugh分级(OR=1.110)是TIPS术后早期HE的独立危险因素,而血浆纤维蛋白原(OR=-1.724)是保护因素;(2)训练集:区分度:AUC=0.805;校准度:χ~2=5.98,P=0.817。验证集:区分度:AUC=0.817;校准度:χ~2=5.20,P=0.877,临床效能评估均具有较好的临床实用性。结论基于年龄、Child-Pugh分级、纤维蛋白原构建的列线图,可视化及可操作性较强,具有良好的准确性、稳定性和临床应用价值,为临床量化风险和获益评估提供了有效的参考工具。

Objective To establish and verify nomogram of early hepatic encephalopathy(HE)after transjugular intrahepatic portosystemic shunt(TIPS)and to provide quantitative tools for individual risk prediction and clinical benefit evaluation.Methods The clinical data of 262 patients were collected in Nanfang hospital affiliated to Southern medical university for further retrospective analysis,then 70%of the data collected from 262 patients(183 in total)is randomly chosen as the training set for establishing the clinical prediction model and the rest 30%(79 in total)was used for internal verification.Then,the data in the training set is used for single and multiple factors Logsitic regression analysis and the screened independent affecting factors were used for establishing the clinical prediction model.Afterwards,the assessment on clinical efficacy of nomogram in three aspects as discrimination ability,calibration and clinical practicability was carried out by adopting the area under the ROC curve,calibration plot,DCA decision curve respectively.Results It was demonstrated that age,Child-Pugh grade and ln(fibrinogen)were the independent affecting factors of the patients with HE in the early stage after TIPS.The discrimination of training set revealed AUC was 0.805 and calibration assessment wereχ~2=5.98,P=0.817.The verification of nomogram showed that AUC was 0.8170 and calibration assessment wereχ~2=5.20,P=0.877.Conclusions Age,Child-Pugh grade and ln(fibrinogen)are the independent affecting factors of the patients with HE in the early stage after TIPS.The clinical prediction model established in the subject is accurate,stable and of clinical application value.Moreover,it is also able to provide the effective reference tool for clinical quantitative risks and assessment on benefits.