肺移植后急性肾衰竭预测模型的开发与验证  

Development and validation of a prediction model for acute renal failure after lung transplantation

作　　者：陈胜 潘晨 李少翔 张丙正 矫文捷[1] CHEN Sheng;PAN Chen;LI Shaoxiang;ZHANG Bingzheng;JIAO Wenjie(Department of Thoracic Surgery,Affiliated Hospital of Qingdao University,Qingdao,266071,Shandong,P.R.China;Medical Department of Nantong University,Nantong,226000,Jiangsu,P.R.China)

机构地区：[1]青岛大学附属医院胸外科,山东青岛266071 [2]南通大学医学院,江苏南通226000

出　　处：《中国胸心血管外科临床杂志》2025年第4期473-481,共9页Chinese Journal of Clinical Thoracic and Cardiovascular Surgery

摘　　要：目的分析肺移植后急性肾衰竭(acute renal failure,ARF)的危险因素并建立预测模型。方法本研究的数据来源于美国器官资源共享网络(UNOS)数据库,纳入2015—2022年间接受单侧或双侧肺移植的患者,分析患者在术前和术后的多项临床特征。结合随机森林算法和最小绝对收缩和选择算子(least absolute shrinkage and selection operator,LASSO)回归,筛选出与肺移植术后ARF发生相关的关键因素,并据此建立列线图模型。在训练集和验证集中分别评估模型的预测性能,并使用受试者工作特征曲线和曲线下面积(area under the curve,AUC)等指标对模型的效果进行验证和比较。结果纳入15110例肺移植患者,其中男6041例、女9069例,中位年龄为62.00(54.00,67.00)岁。结果表明,术后肾透析与未透析患者术前肺部诊断、预计肾小球滤过率、机械通气、术前是否在ICU接受治疗、体外膜肺氧合支持、术前2周是否感染、Karnofsky功能状态评分、在等待名单上的时间、双肺移植、缺血时间差异均具有统计学意义(P<0.05)。通过随机森林和LASSO回归筛选出与肺移植后ARF相关的5个变量(受体预计肾小球滤过率、术前在ICU治疗、使用体外膜肺氧合、双肺移植和缺血时间),建立了列线图模型。模型评估结果显示,所构建的预测模型在训练集和验证集中均具有较高的准确性,且AUC值表现良好,验证了该模型的有效性和可靠性。结论根据肺移植后ARF常见的危险因素,开发了一个效能良好的预测模型,具有一定的临床应用价值。Objective To identify and analyze risk factors for acute renal failure(ARF)following lung transplantation and to develop a predictive model.Methods Data for this study were obtained from the United Network for Organ Sharing(UNOS)database,encompassing patients who underwent unilateral or bilateral lung transplantation between 2015 and 2022.We analyzed both preoperative and postoperative clinical characteristics of the patients.A combined approach utilizing random forest and least absolute shrinkage and selection operator(LASSO)regression was employed to identify key factors associated with the incidence of ARF post-transplantation,based on which a nomogram model was developed.The predictive performance of the constructed model was evaluated in both training and validation sets,using receiver operating characteristic(ROC)curves and area under the curve(AUC)metrics to verify and compare model effectiveness.Results A total of 15110 lung transplantation patients were included in the study,consisting of 6041 males and 9069 females,with a median age of 62.00 years(interquartile range:54.00 to 67.00).The analysis revealed statistically significant differences between postoperative renal dialysis and non-dialysis patients regarding preoperative lung diagnosis,estimated glomerular filtration rate(eGFR),mechanical ventilation,preoperative ICU treatment,extracorporeal membrane oxygenation(ECMO)support,infections occurring within two weeks prior to transplantation,Karnofsky Performance Status(KPS)score,waitlist duration,double-lung transplantation,and ischemia time(P<0.05).Five key variables associated with ARF after lung transplantation were identified through random forest and LASSO regression:recipients’eGFR,preoperative ICU treatment,ECMO support,bilateral lung transplantation,and ischemia time.A nomogram model was subsequently established.Model evaluation demonstrated that the constructed predictive model achieved high accuracy in both training and validation sets,with favorable AUC values,confirming its validity and r

关 键 词：肺移植 急性肾衰竭 危险因素 列线图 

分 类 号：R655.3[医药卫生—外科学]