基于LASSO回归构建肝移植术后急性肾损伤的预测模型  

Establishment of a prediction model for acute kidney injury after liver transplantation based on LASSO regression

作　　者：桑晓辉 赵晋明 白磊 王智鹏 李涛 何翼彪 SANG Xiaohui;ZHAO Jinming;BAI Lei;WANG Zhipeng;LI Tao;HE Yibiao(Department of Digestive and Vascular Surgery Center/Liver and Laparoscopic Surgery,the First Affiliated Hospital of Xinjiang Medical University,Urumqi,Xinjiang 830054,China)

机构地区：[1]新疆医科大学第一附属医院消化血管外科中心/肝脏·腹腔镜外科,新疆乌鲁木齐830054

出　　处：《肝胆胰外科杂志》2025年第4期227-235,242,共10页Journal of Hepatopancreatobiliary Surgery

摘　　要：目的 构建基于LASSO回归的肝移植术后急性肾损伤(AKI)的预测模型。方法 回顾性收集2015年6月至2024年6月新疆医科大学第一附属医院施行活体或公民逝世后器官捐献(DCD)供肝肝移植的83例患者基本信息和临床资料,根据改善全球肾脏疾病组织(KDIGO)2012年修订的AKI诊断标准,将83例患者分为AKI组(n=42)和非AKI组(n=41)。单因素差异性分析筛选与肝移植术后AKI相关的因素,并通过最小绝对收缩和选择算子(LASSO)回归优化模型的特征选择,利用多因素Logistic回归构建预测模型并绘制列线图,受试者工作特征(ROC)曲线、校准曲线和决策曲线分析(DCA)评估预测模型的区分度、校准度和临床适用性,使用Bootstrap法进行内部验证。结果 单因素分析结果显示,供体类型、性别、BMI、糖尿病、高血压、无肝期时间、术中出血量、术中尿量、术后AST峰值、术后ALT峰值与患者肝移植术后AKI有关(P<0.05)。利用LASSO回归筛选出4个预测变量:BMI、无肝期时间、术中尿量、术后AST峰值;多因素Logistic回归结果显示,BMI≥24.0 kg/m^(2)(OR=14.209,95%CI 3.716-76.010)、无肝期时间≥120.0 min(OR=4.746,95%CI 1.240-22.270)、术中尿量≤3 710.0 mL(OR=6.238,95%CI 1.694-27.840)、术后AST峰值≥1 476.0 U/L(OR=12.252,95%CI 3.260-63.560)为患者肝移植术后AKI的独立危险因素(P<0.05)。根据上述因素构建列线图预测模型,ROC曲线下面积(AUC)为0.893(95%CI 0.828-0.959)。校准曲线结果显示模型预测结果与实际结果拟合良好,Spiegelhalter Z检验结果显示P=0.873,DCA结果显示该模型预测肝移植术后AKI具有较高的净获益。结论 本研究构建的用于预测肝移植术后AKI的模型具有较好的可靠性和临床适用性。Objective To establish a prediction model for acute kidney injury(AKI)after liver transplantation based on LASSO regression.Methods The basic information and clinical data of 83 patients undergoing liver transplantation by living donor or donation after circulatory death(DCD)donor in the First Affiliated Hospital of Xinjiang Medical University from Jun.2015 to Jun.2024 were retrospectively collected.According to the AKI diagnostic criteria revised by Kidney Disease Improving Global Outcomes(KDIGO)in 2012,the 83 patients were divided into the AKI group(n=42)and the non-AKI group(n=41).Univariate analysis was used to screen factors associated with AKI after liver transplantation,and further selection was optimized with the least absolute shrinkage and selection operator(LASSO)regression.Additionally multivariate Logistic regression was employed to construct a prediction model and draw a nomogram.The model’s differentiation,calibration,and clinical applicability were measured by the receiver operating characteristic(ROC)curve,calibration curve,and decision curve analysis(DCA).The Bootstrap method was used for internal verification.Results Univariate analysis showed that,AKI after liver transplantation was significantly correlated with donor type,gender,body mass index(BMI),diabetes,hypertension,anhepatic phase,volume of intraoperative blood loss,intraoperative urine output,peak postoperative AST,and peak postoperative ALT(P<0.05).LASSO regression identified 4 independent predictors:BMI,anhepatic phase,intraoperative urine output,and peak postoperative AST.Multivariate Logistic regression showed that,BMI≥24.02 kg/m^(2)(OR=14.209,95%CI 3.716 to 76.010),anhepatic phase≥120.0 min(OR=4.746,95%CI 1.240 to 22.270),intraoperative urine output≤3710.0 mL(OR=6.238,95%CI 1.694 to 27.840)and peak postoperative AST≥1476.0 U/L(OR=12.252,95%CI 3.260 to 63.560)were independent risk factors for AKI after liver transplantation(P<0.05).According to the above 4 factors,a nomogram prediction model was constructed,and the area u

关 键 词：肝移植 急性肾损伤 危险因素 列线图 预测模型 

分 类 号：R657[医药卫生—外科学] R692[医药卫生—临床医学]