Screening Predictive Factors And The Establishment Of Prediction Model Of Clinically Relevant Postoperative Pancreatic Fistula After Pancreaticoduodenectomy

Objective:Analyze the predictive factor of clinically relevant postoperative pancreatic fistula after pancreaticoduodenectomy,then establish the prediction model and verify the model’s prediction effectiveness,feasibility and accuracy.The purpose is to use the model to screen out the CR-POPF high-risk patients after PD.It can provide the objective basis for clinical surgeons to take active and effective preventive measures and reduce the incidence of CR-POPF after PD.And it can improve the prognosis of patients.Methods:Collected the clinical indicators of PD patients in General Surgery Department of the Hospital from July,2016 to October,2019.Conformed the inclusion and exclusion criteria to select the study samples.All study samples were randomly divided into the model group and the test group with a ratio of 3:1.With the occurrence of CR-POPF as the dependent variable,to analyze the 22 clinical indicators of modeling group to screen out the independent risk factors of CR-POPF.And the risk prediction model is established with these factors as the predictor,and the accuracy of the prediction model will be tested.Results:The number of the sample included in this study was 296(222 in the model group and 74 in the test group).Among them,CR-POPF occurred after PD in 45cases(32 cases in the model group and 13 cases in the test group),with an incidence of 15.2%.Multivariate analysis results in model group showed that Age≥65 years(p=0.001;OR=5.609,95%CI:1.998~15.746),Preoperative Albumin<30g/L(p=0.002;OR=7.266,95%CI:2.093~25.223),Bleeding volume≥600ml(p<0.001;OR=9.233,95%CI:3.218~26.483),Soft Pancreas(p=0.004;OR=4.133,95%CI:1.579~10.820)and Pancreatic Duct Diameter<0.3cm(p=0.003;OR=4.947,95%CI:1.725~14.186)were independent risk factors for POPF.Age,preoperative albumin,bleeding volume,pancreatic texture,and pancreatic duct diameter were used as predictors to establish a mathematical model for risk prediction.The cut-off point of the ROC curve was 15.6%,the sensitivity was 78.1%and the specificity was 81.6%,and AUC was 0.875(95%CI:0.820-0.930).Test results of model calibration:the H-L test results show that X~2=2.837,p=0.725,so it showed that the mathematical model has a good calibration.The clinical data of the test group were substituted into the prediction model,and fitted the ROC curve.The AUC=0.862(95%CI:0.779~0.945),and the comparison with the AUC line Z test of the model group showed no difference between the two groups(p=0.49).The values of AUC of the two groups were between 0.7 and 09.The prediction performance of the mathematical model was good.Conclusions:Age≥65years,Preoperative Albumin<30g/L,Bleeding Volume≥600ml,Soft Pancreas and Pancreatic Duct Diameter<0.3cm are independent risk factors for CR-POPF after PD.The accuracy of the CR-POPF risk prediction model after PD is good.It will be helpful to screen out the patients with high risk of CR-POPF and take targeted preventive measures.