胰十二指肠切除术后腹腔感染风险预测模型的构建及验证

Establishment and validation of risk prediction model for postoperative intraperitoneal infection in pancreaticoduodenectomy patients

目的构建一种新型的胰十二指肠切除术(PD)后腹腔感染(IAI)风险预测模型,为临床诊断、治疗提供理论依据。方法回顾性分析2016年6月-2022年6月在天津市第三中心医院行PD术的243例患者的临床相关资料,利用LASSO回归分析筛选PD术后IAI的危险因素,Logistic回归分析并建立PD术后IAI风险预测模型,C-index、受试者工作特征(ROC)曲线、校准图和决策曲线分析法(DCA)对其进行评估。结果243例行PD术患者,96例发生了术后IAI,感染率为39.5%;LASSO回归分析筛选出肿瘤大小、胰瘘、胆瘘、腹腔出血和血肌野(Scr)水平是PD术后发生IAI的危险因素(P<0.05);多因素Logistic回归分析显示,胰瘘、胆瘘、腹腔出血和Cr水平是PD术后IAI的独立危险因素(P<0.05);该模型C-index为0.779(95%CI:0.738~0.821),ROC曲线下面积(AUC)为0.778,显示出良好的区分度,校准曲线显示该模型的预测风险概率与实际发生概率基本吻合,DCA曲线显示该模型应用价值高。结论PD术后发生IAI的风险较高,LASSO结合Logistic回归分析构建的新模型可以更准确地用于PD术后IAI的预测。'

OBJECTIVE To establish a novel type of risk prediction model for postoperative intraperitoneal infection(IAI)in the patients undergoing pancreaticoduodenectomy(PD)so as to provide theoretical basis for clinical diag-nosis and treatment.METHODS The clinical data were collected from 243 patients who underwent PD in the Third Central Hospital of Tianjin from Jun 2016 to Jun 2022 and were retrospectively analyzed.The risk factors for postoperative IAI in the PD patients were screened out by means of LASSO regression analysis,logistic regression analysis was performed to establish the risk prediction model for the postoperative IAI.The performance of the model was evaluated by C-index,receiver operating characteristic(ROC)curve,calibration curve and decision curve analysis(DCA).RESULTS Of the 243 patients who underwent PD,96 had postoperative IAI,with the in-fection rate 39.5%.LASSO regression analysis showed that the size of tumor,pancreatic fistula,biliary fistula,intraabdominal hemorrhage and serum creatinine(Scr)were the risk factors for postoperative IAI in the PD pa-tients(P<0.05).C-index of the model was 0.779(95%CI:0.738-0.821),the area under ROC curve(AUC)was 0.778,showing excellent discrimination degree;the calibration curve indicated that there was a good consistency between the probability of prediction for risk and the actual probability of occurrence.DCA curve showed that the model had high application value.CONCLUSION The PD patients are at high risk of postoperative IAI.The novel model that is established based on LASSO combined with logistic regression analysis can predict the postoperative IAI more accurately.