OBJECTIVETo evaluate the clinical value of a preoperative predictive scoring system which was established by the National Cancer Center Hospital (NCCH) for the postoperative pancreatic fistula (POPF) after pancreaticoduodenectomy.

METHODSThe clinical data of 269 patients who underwent pancreaticoduodenectomy at the Affiliated Provincial Hospital of Anhui Medical University from February 2008 to February 2014 were studied retroprospectively. The five indexes which including gender, portal invasion, pancreatic cancer, main pancreatic duct index and intra abdominal fat thickness were calculated in the NCCH predictive score system. Patients with a score over 4 were defined as high risk of POPF, and those with score less than 4 were defined as low risk of POPF. Then the factors associated with POPF were analyzed by Logistic regression test. The enumeration data and measurement data were compared with χ2 test and t test. Risk factors for postoperative pancreatic fistula were analyzed through single factor and multiple factors Logistic regression analysis. The sensitivity and specificity of the predictive scoring system were determined by receiver operating characteristic (ROC) curve analysis.

RESULTSA total of 33 patients were diagnosed as POPF, including 15 in grade A, 11 in grade B and 7 in grade C. The univariate analysis showed that the factors associated with POPF are gender, total serum bilirubin level, pancreatic cancer, portal invasion, the pancreatic texture, main pancreatic duct diameter and the pancreaticojejunostomy. The multivariate analysis showed that gender, pancreatic texture, portal invasion and main pancreatic duct diameter were the independent risk factor of POPF. The rate of pancreatic fistula of high risk group was 53.8% (14/26), and the rate of pancreatic fistula of the low risk group was 7.8% (19/243). There were significant differences in the pancreatic fistula rate between the high risk and low risk of POPF (χ2=46.231, P<0.01). The results of ROC curve analysis showed that the sensitivity and specificity of the predictive scoring system were 87.9% and 94.1%, respectively. The area under the curve was 0.946 (95% CI: 0.895-0.997).

CONCLUSIONSThe NCCH preoperative predictive scoring system could accurately predict the occurrence of POPF. While large, multicenter prospective randomized controlled trials is still needed to further confirm it.