Preoperative CT Scan Helps to Predict the Occurrence of Severe Pancreatic Fistula After Pancreaticoduodenectomy

Abstract

Objective: To assess the influence of body fat distribution, estimated by a preoperative computed tomographic (CT) scan, on pancreatic fistula (PF) risk after pancreaticoduodenectomy (PD). Background: Pancreatic fatty infiltration is a predictive factor of PF, but accurate preoperative assessment is challenging. We hypothesized that it could be associated with an increased visceral obesity and could be assessed preoperatively. Methods: Over 18 months, 103 consecutive patients with PD and pancreaticogastrostomy were studied. Demographic, radiologic, and pathologic data were correlated to PF occurrence. Radiologic data included on a nonenhanced CT acquisition: pancreas, spleen, and liver density measures (Dpancreas, Dspleen, and Dliver [densities of the pancreas, spleen, and liver in hounsfield units], respectively), retro-renal fat thickness, and at the level of the umbilicus, total, visceral, and subcutaneous fat area (TFA [total fat area], VFA [visceral fat area], and SFA [subcutaneous fat area], respectively). Pancreatic fatty infiltration was graded histologically. Logistic regression analysis was used to identify independent predictors of PF-graded B and C according to the International Study Group on the Pancreatic Fistula. Results: Among the 103 patients, 37% (n = 38) developed a PF (47.4% grade A, 39.5% grade B, and 13.1% grade C). PF risk was correlated with pancreatic fatty infiltration (P = 0.017). In univariate analysis, male gender (P = 0.023), body mass index (BMI) over 25 kg/m² (P = 0.02), retro-renal fat thickness over 15 mm (P = 0.006), TFA over the median (>233 cm²; P = 0.023), and VFA over the median (>84 cm²; P < 0.0001) were significantly associated with an increased risk of symptomatic PF (grade B and C). In multivariate analysis, VFA greater than 84 cm² (OR = 8.16, P = 0.002) was the only independent predictive factor of grade B or C PF. Using the same model, a VFA greater than 84 cm² was the only independent factor associated with the presence of fatty pancreas on pathologic examination. Conclusions: Preoperative assessment of body fat distribution by a CT scan, as a surrogate for fatty pancreas infiltration, can help to predict the occurrence of clinically significant PF after PD.