基于LASSO回归的胰十二指肠切除术后胰瘘预测模型构建与评估  

Construction and Evaluation of Postoperative Pancreatic Fistula Prediction Model for Pancreaticoduodenectomy Based on LASSO Regression

作　　者：张文杰 魏安琪 偰光华 ZHANG Wenjie;WEI Anqi;XIE Guanghua(Yanbian University Hospital,Yanji City,Jilin Province 133000;不详)

机构地区：[1]延边大学附属医院肝胆胰外科,吉林省延吉市133000 [2]菏泽医学专科学校

出　　处：《医学理论与实践》2025年第5期735-739,743,共6页The Journal of Medical Theory and Practice

摘　　要：目的:运用LASSO回归及多变量回归分析筛选变量后构建预测模型,以预测术后胰瘘(POPF)发生的风险概率,及早发现并精准干预,从而减少POPF的发生率和死亡率。方法:回顾性分析2015年2月—2023年9月在延边大学附属医院肝胆胰外科行胰十二指肠切除术的102例患者临床资料,对纳入的21项术前术中特征采用LASSO回归及多变量logistic回归分析筛选出预测因素,从而建立预测模型,并绘制受试者工作曲线(ROC)及calibration校正曲线、计算C指数(C-index)、曲线下面积(AUC)、H-L拟合优度检验值评估模型的预测能力,运用内部K折交叉验证及Bootstrap评价模型泛化能力,最后绘制出临床决策曲线(DCA)以及临床影响曲线(CIC)评价临床实际效用。结果:POPF预测模型中包含的预测因素为BMI、胰腺质地、胰管直径以及术中出血量。该模型(AUC=0.898,95%CI:0.825~0.972)、H-L拟合优度检验(χ~2=7.5629,P=0.4773),具有较高的准确性和预测能力,校准曲线表明该模型具有良好的校准性。采用了K折交叉验证方法(准确率83.2%、Kappa值0.511、灵敏度0.910、特异度0.567)验证模型具有较高的准确性、稳定性,且在识别阳性样本和阴性样本方面表现良好,DCA及CIC曲线皆证明该预测模型具有一定的临床应用价值。结论:BMI、胰腺质地、胰管直径以及术中出血量是POPF的独立危险因素,构建的模型可以较好地预测POPF的发生风险,此模型可为临床医生提供一定的临床参考及指导,及早发现并精准干预POPF,从而减少POPF的发生率和死亡率。Objective:To construct a predictive model using LASSO regression and multivariable regression analysis to identify variables and predict the risk probability of postoperative pancreatic fistula(POPF),enabling early detection and precise intervention to reduce the incidence and mortality of POPF.Methods:A retrospective analysis was conducted on the clinical data of 102 patients who underwent pancreaticoduodenectomy at the hepatobiliary and pancreatic surgery department of Yanbian university hospital from February 2015 to September 2023.LASSO regression and multivariable logistic regression analyses were used to identify predictive factors from 21 preoperative and intraoperative characteristics,thereby establishing a predictive model.The model’s predictive ability was evaluated using the receiver operating characteristic(ROC)curve,calibration curve,C-index,area under the curve(AUC),and Hosmer-Lemeshow goodness-of-fit test.Internal K-fold cross-validation and Bootstrap methods were used to assess the model’s generalizability.Finally,the clinical decision curve analysis(DCA)and clinical impact curve(CIC)were plotted to evaluate the practical clinical utility of the model.Results:The predictive model for POPF included BMI,pancreatic texture,pancreatic duct diameter,and intraoperative blood loss as predictive factors.The model demonstrated high accuracy and predictive capability with an AUC of 0.898(95%CI:0.825~0.972)and a Hosmer-Lemeshow goodness-of-fit test value ofχ^(2)=7.5629,P=0.4773.The calibration curve indicated good calibration of the model.The K-fold cross-validation method validated the model’s high accuracy(83.2%),Kappa value(0.511),sensitivity(0.910),and specificity(0.567),showing good performance in identifying positive and negative samples.Both the DCA and CIC curves confirmed the clinical applicability of the predictive model.Conclusion:BMI,pancreatic texture,pancreatic duct diameter,and intraoperative blood loss are independent risk factors for POPF.The constructed model can effectively predict

关 键 词：胰十二指肠切除术 术后胰瘘 LASSO回归 危险因素 列线图 

分 类 号：R657.5[医药卫生—外科学]