Establishment of a risk prediction model for pancreatic fistula after pancreaticoduodenectomy： A study based on the 2016 edition of the definition and classification system of pancreatic fistula / 临床肝胆病杂志

ABSTRACT

ObjectiveTo investigate the differences in the risk factors for postoperative pancreatic fistula （POPF） after pancreaticoduodenectomy （PD） between the 2005 and 2016 editions of the definition and classification standards for pancreatic fistula， and to establish a risk prediction model for pancreatic fistula based on the 2016 edition. MethodsA retrospective analysis was performed for the clinical data of 303 patients who were admitted to Tianjin Third Central Hospital and underwent PD from January 2016 to May 2022， and the patients with POPF were identified based on the new and old editions. The independent-samples t test or the non-parametric Mann-Whitney U test was used for comparison of continuous data between groups， and the chi-square test was used for comparison of categorical data between groups. The univariate and multivariate logistic regression analyses were used to investigate the differences in the risk factors for pancreatic fistula after PD between the two editions； a risk prediction model was established for POPF based on the 2016 edition， and the receiver operating characteristic curve was used to invesitgate the accuracy of this model in predicting POPF and perform model validation. ResultsAccording to the 2005 edition， the univariate analysis showed that the diameter of the main pancreatic duct （χ2=31.641， P<0.001）， main pancreatic duct index （χ2=52.777， P<0.001）， portal vein invasion （χ2=6.259， P=0.012）， intra-abdominal fat thickness （χ2=7.665， P=0.006）， preoperative biliary drainage （χ2=5.999， P=0.014）， pancreatic cancer （χ2=5.544， P=0.019）， marginal pancreatic thickness （t=2.055， P=0.032）， pancreatic CT value （t=-3.224， P=0.002）， and preoperative blood amylase level （Z=-2.099， P=0.036） were closely associated with POPF， and the multivariate logistic regression analysis showed that main pancreatic duct index （odds ratio ［OR］=0.000， 95% confidence interval ［CI］： 0.000‍ ‍—‍ ‍0.011， P<0.05）， pancreatic cancer （OR=4.843， 95%CI： 1.285‍ ‍—‍ ‍18.254， P<0.05）， and pancreatic CT value （OR=0.869， 95%CI： 0.806‍ ‍—‍ ‍0.937， P<0.05） were independent risk factors； based on the 2016 edition， the univariate analysis showed the diameter of the main pancreatic duct （χ2=5.391， P=0.020）， main pancreatic duct index （χ2=11.394， P=0.001）， intra-abdominal fat thickness （χ2=8.899， P=0.003）， marginal pancreatic thickness （t=2.665， P=0.009）， pancreatic CT value （t=-2.835， P=0.004） were closely associated with POPF， and the multivariate logistic regression analysis showed that main pancreatic duct index （OR=0.001， 95%CI： 0.000‍ ‍—‍ ‍0.050， P<0.05） and pancreatic CT value （OR=0.943， 95%CI： 0.894‍ ‍—‍ ‍0.994， P<0.05） were independent risk factors. A risk prediction model was established for POPF after PD， and the ROC curve analysis showed that this model had an area under the ROC curve of 0.788 （95%CI： 0.707‍ ‍—‍ ‍0.870） in the modeling group and 0.804 （95%CI： 0.675‍ ‍—‍ ‍0.932） in the validation group. ConclusionMain pancreatic duct index and pancreatic CT value are closely associated with POPF after PD， and the risk prediction model for pancreatic fistula based on the 2016 edition has a good prediction accuracy.