Novel Prediction of Pancreatic Anastomotic Failure After Pancreatoduodenectomy Using Preoperative CT Imaging With the Evaluation of Remnant Pancreatic Volume and Body Composition

INTRODUCTION: An increased body mass index (BMI) and pancreatic duct size are known predictors of pancreatic anastomotic failure (PAF) after pancreatoduodenectomy (PD). However, the impact of anthropomorphic measurements (remnant pancreatic (parenchymal) volume (RPV) and body composition on PAF) are unknown. The aim was to determine if pancreatic remnant volume, subcutaneous/visceral adipose tissue (SAT/VAT) area, and skeletal muscle (SM) area calculated from the preoperative computed tomography (CT) predict PAF after PD. METHODS: In173 patients undergoing preoperative CT and PD at a single institution between 2004 and 2009, SM area and SAT/VAT cross-sectional area at the 3rd lumbar vertebra were quantitated using the preoperative CT. Muscle and adipose tissue were identified semi-automatically using the CT Hounsfield threshold method and remnant pancreatic volume as a volumetric sum of pancreatic parenchymal area to the left of the surgical margin (left border of SMA) over multiple cuts. Pancreatic duct size and parenchymal hardness were assessed by surgeon. The definition of PAF was the International Study Group of Pancreatic Fistula (ISGPF) Classification System; Grades B and C PAF were considered clinically-relevant PAF. Patient demographics of those with clinically relevant PAF were compared to those without PAF. Associations with PAF by univariate logistic regression models were summarized with odds ratios and 95% confidence intervals (CI). The predictive ability for several models was described using a concordance index (c-index). RESULTS: PAF occurred in 22 patients (13%); Grades B and C were present in 15 (9%) and 7 (4%) patients resp. In univariate logistic regression analysis, RPV, VAT, SM, BMI, SAT, pancreatic duct size, and pancreatic texture (soft or hard) were all predictors of PAF with P values of <0.001, <0.001, 0.001, 0.001, 0.02, 0.03, and 0.04, resp. A multivariate model with the known predictors using BMI and duct size had a c-index of 0.75 (BMI; Odds ratio (OR): 1.13, 95%CI: 1.04−1.24, P =0.005, duct size; OR: 0.71, 95%CI: 0.51−0.95, P =0.036). A better multivariate model included preoperative CT factors VAT and SM with a c-index = 0.96 (VAT; OR: 1.24, 95%CI: 1.16−1.37, P <.001, SM; OR: 0.02, 95%CI: 0.01−0.08, P <0.001). CONCLUSIONS: Compared to established risk factors for PAF, RPV and VAT were better predictors of PAF after PD. Prediction of the risk of PAF after PD may be best estimated by including these anthropomorphic measures from the preoperative CT as well as using intraoperative findings.