基于术前评分模型对腹腔镜胰十二指肠切除术术后胰瘘风险的预测分析

Prediction of postoperative pancreatic fistula risk after laparoscopic pancreaticoduodenectomy based on preoperative scoring model

目的探究腹腔镜胰十二指肠切除术(LPD)术后胰瘘的危险因素,构建“术前-术中”两位一体评分模型,在术前预测LPD术后胰瘘发生风险,并评估模型的临床应用价值。方法回顾性分析2014年1月至2019年1月湖南省人民医院及常德市第一人民医院176例LPD患者的临床资料,其中30例患者在术后发生B级及以上胰瘘(B级胰瘘28例,C级胰瘘2例)。通过对患者的临床特点、影像学表现、病理特征、治疗及术后恢复等相关数据进行处理,基于术前数据估计术中数据,进而构建“术前-术中”两位一体评分模型预测胰瘘发生风险。结果多因素Logistic回归结果显示,年龄(P<0.001)、血小板计数(P<0.001)、胰腺质地(P=0.003)、胰管直径(P<0.001)、术中出血量(P<0.001)为影响胰瘘发生的显著因素。基于年龄、血小板计数、胰腺质地和胰管直径这4个术前指标数据,采用MissForest算法预估术中出血量,预测术后胰瘘准确率达89.2%(157/176)。基于术前数据和估计出的术中数据,构建LPD术后胰瘘风险评分模型,最终预测效能较好,AUC值达0.792。结论本研究构建的“术前-术中”两位一体评分模型能够基于患者术前身体指标以及估计出的术中出血量,较为准确地预估出LPD术后胰瘘的发生风险。

Objective To investigate the risk factors of postoperative pancreatic fistula(POPF)after laparoscopic pancreaticoduodenectomy(LPD),to construct a preoperative-intraoperative integrated scoring model,which was used to predict the risk of POPF preoperatively and its clinical application value was evaluated.Methods This retrospective analysis was conducted on clinical data of 176 LPD patients from Jan.2014 to Jan.2019 at Hunan Provincial People’s Hospital and the First People’s Hospital of Changde City.Among them,30 patients experienced grade B or higher pancreatic fistula postoperatively(28 case of grade B and 2 case of grade C).Relevant data including clinical characteristics,imaging findings,pathological features,treatment,and postoperative recovery were processed.Based on preoperative data,intraoperative data were estimated,and a preoperative-intraoperative integrated scoring model was constructed to predict the risk of POPF.Results Multivariate Logistic regression results indicated that age(P<0.001),platelet count(P<0.001),pancreatic texture(P=0.003),pancreatic duct diameter(P<0.001),and intraoperative blood loss(P<0.001)were significant factors affecting the occurrence of POPF.Based on the preoperative data of these 4 indicators—age,platelet count,pancreatic texture,and pancreatic duct diameter—the MissForest algorithm was used to estimate the intraoperative blood loss,and the accuracy rate of predicting postoperative pancreatic fistula reached 89.2%(157 out of 176 cases).The constructed scoring model,based on preoperative and estimated intraoperative data,demonstrated good predictive performance with an AUC value of 0.792.Conclusion Based on patients’preoperative physical indicators and estimated intraoperative blood loss,the preoperative-intraoperative integrated scoring model constructed in this study can accurately predict the risk of POPF after LPD.