Establishment And Clinical Validation Of A Predictive Model For The Risk Assessment Of Pancreatic Fistula After Pancreaticoduodenectomy

Part One Exploration the Risk factors of pancreatic fistula after pancreaticoduodenectomyBackground:Pancreatoduodenectomy(PD)is the only effective treatment for the ampullary disease.But the high rate of complication of PD is an unavoidable problem,including(POPF),delayed gastric emptying(DGE)after pancreatectomy(PPH),bleeding,bile leakage(BL)etc.Despite improvements of surgical techniques and perioperative management,postoperative pancreatic fistula(POPF)is still a challenge after pancreaticoduodenectomy.Even in a large center,the rate of POPF is very high,about 5-45%.But the standard of pancreatic fistula in previous studies were made reference to the 2005 International pancreatic surgery group standards,it is the standard of any measurable drainage fluid amylase level more than 3 times the upper limit of normal serum for postoperative pancreatic fistula,and according to the degree of influence on clinical outcome for class A,B,C grade 3 grades,the standard of diagnosis of operation,there are some defects.2016 pancreatic surgery international research group(ISGPS)the definition of pancreatic fistula and its grading system was updated,divided into biochemical leakage,fistula,B grade C grade pancreatic fistula,conclusion of previous research may have some changes,the are past,pancreatic texture and body mass index the duct diameter and the nature of the disease,jaundice,nutritional status,anastomosis and the operative time,bleeding volume,postoperative pancreatic fistula so the pattern of change,risk factors or risk factors are not previously discussed,the purpose of this study is the clinical risk factors of pancreatic fistula surgery after 2016 ISGPS definition and classification system based on evaluation.Research methods:A total of 225 cases of periampullary lesions between 2009 and 2015 were retrospectively analyzed and the clinical data of PD were reviewed.The clinical data for this study include age,gender,smoking history,drinking history,history of diabetes,body mass index(BMI),serum CA19-9,CA-125,CEA,blood glucose,blood test,serum bilirubin,serum albumin,blood coagulation indexes(PT,INR),blood electrolytes.Data collected after operation included serum albumin,blood sugar,blood routine,serum bilirubin,serum albumin level and blood electrolyte.Other data included tumor size,location,pathological primary lesion,surgical technique type(PD and PPPD),anastomosis mode,intraoperative bleeding,bile duct and pancreas size and texture(hard or soft),diameter of main pancreatic duct and the scope of lymphatic dissection.The classification variables were analyzed with chi square test,and the continuous variables were analyzed by t test.Logistic regression analysis was used for the single factor analysis of all variables with P value less than 0.1 in multivariate analysis.The effect of the prediction model on the test data is evaluated by using the ROC and the area under the curve(AUC).The P value is less than 0.05 is considered statistically significant.A single factor and multiple factor Logistic regression analysis was conducted in the study group through the potential effects of preoperative and surgical factors on CR-POPF.Result:Among the 225 patients,the average age of the patients was 58.2+12.1 years,58.2%for men and 41.8%for women.191(84.9%)patients were treated with pancreatoduodenectomy for malignant disease and 34(15.1%)of benign disease.The average operation time was 510 mim,and the average bleeding amount was 300 ml during the operation.The median time of postoperative hospitalization was 29 days.40 patients(17.8%)had cr-popf after pancreatoduodenectomy.Univariate analysis revealed that CR-POPF patients had pancreatic texture(OR = 2.09,95%CI 1.10-3.98,P = 0.025),main pancreatic duct diameter(MPD)(OR = 2.72,95%CI 1.23-5.99P =0.013)and serum albumin(<25 g/L versus>30 g/L)on the first day after surgery.OR = 5.12,95%CI 1.82-14.41,P = 0.002;25g-30g/L versus>30g/L:OR = 1.43,95%CI 1.02-2.00,P = 0.037),and extended lymph node dissection(OR = 1.57,95%Cl 1.13-2.18,P = 0.007)were independent risk factors.Conclusion:This study found that the quality of the pancreas,serum albumin at 1 days after the operation and enlarged lymph node dissection were independent risk factors for CR-POPF.Soft texture and low postoperative protein levels are widely accepted risk factors for pancreatic fistula,but we firstly report that extended lymph dissection is associated with pancreatic fistula after pancreatoduodenectomy.Part two Establishment and clinical validation of a predictive model for the risk assessment of pancreatic fistula after pancreaticoduodenectomyBackgroundPancreatic fistula is arguably one of the most serious and troublesome complications after pancreaticoduodenectomy,.Because of the risk of postoperative pancreatic fistula,many studies are predicting the risk factors of pancreatic fistula.There are many predictive models,the postoperative pancreatic fistula risk scale(FRS)and the Japanese integral prediction system are important.NCCH prediction system,as well as the prediction model proposed by Chinese and French experts also be used,these prediction models are based on the 2005 version of the classification.whether these models are applicable the new classification,these models need to be verified.Our study which is based on the definition and classification of pancreatic fistula 2016,establish a predictive model and valid the model in clinical.Patients and methodsThe clinical records of 225 consecutive patients who underwent PD for periampullary tumor between 2009 and 2015 were reviewed retrospectively.According to the result of univariate analysis of the first part,pancreatic texture(P =0.025),main pancreatic duct diameter(MPD)(P = 0.013),serum albumin(<25g/L versus>30g/L,P = 0.002,25g-30g/L versus>30g/L,P = 0.037)and enlarged lymph node dissection(P = 0.007)were included in Multivariate logisitc regression analysis,get the beta value and constant,calculate Z = constant + beta 1X1 + beta 2X2+ beta 3X3+.......beta nXn,X is the risk factor and beta is the corresponding risk factor.The risk of pancreatic fistula was Y=1/(1 + e-z),e was the natural logarithm(?2.718281828459),and a concise scoring system was established according to the beta value.Finally,136 patients with pancreaticoduodenectomy admitted to our hospital from November 2015 to November 2017 were selected as the validation group to test the predictive value of the above models.ResultsRegression analysis showed that Z=-5.367+0.992 x pancreatic texture(1,soft pancreas;0,hard pancreas)+ 1.421 x MPD(1,<2.5 mm;0,>2.5 mm;+ 0.587 x extended lymph dissection(1,yes;0,no)+(3.87 x albumin(1,<25 g/L;0,30 g/L)or 0.419 x albumin(1,25-30 g/L;0,30 g/L).The risk value Y of each patient can be obtained by introducing the Z value into the pancreatic fistula risk formula Y = 1/(1 +e-z).The above formulas were substituted into the validation group according to the risk factors and assignment,and the risk value Y of pancreatic fistula was calculated.The ROC curve was drawn with whether have pancreatic fistula.The area under the curve was 0.777(95%CI:0.698-0.855),indicating that the above formulas were more effective in predicting the occurrence of pancreatic fistula.According to the different beta value of each risk factor,we approximate assigned the value of pancreatic fistula risk score:pancreatic texture(soft:15 hard:0,);MPD(<2.5mm:1.5,>2.5mm:0),enlarged lymph dissection(yes:0.5,no:0),serum albumin on the first day after surgery(<25g/L:3,25g-30g/L:0.5,>30g/L:0).According to the above-mentioned scoring principles,the training group and the verification group were scored,ROC curve was drawn according to the scoring results,and the predictive value of the scoring system for pancreatic fistula was evaluated.It was found that the area under the curve(AUC)was 0.813(95%Cl:0.737-0.889)in the training group and 0.806(95%Cl:0.735-0.878)in the verification group.It shows that the system has good predictive value.ConclusionsBy analyzing the clinical data of 225 patients undergoing pancreaticoduodenectomy in the training group,we concluded in the first part that pancreatic texture,serum albumin on the first day after operation,and enlarged lymph node dissection were risk factors.Multivariate Logisitc analysis was used to establish a risk prediction formula for pancreatic fistula and a corresponding concise scoring system.Through the validation of cases in the validation group,it was found that the formula had a good predictive value for predicting postoperative pancreatic fistula.