Pre-operative prediction of pancreatic fistula: is it possible?

Background/objectives: Understanding a patient's risk of pancreatic fistula (PF) prior to pancreatoduodenectomy (PD) would permit an individualised approach to patient selection, consent and, potentially, treatment. Various intra and post operative factors including pancreatic duct width and steatosis are associated with PF. We sought to identify whether information available in the pre-operative phase can predict PF.

Methods: Associations between patient characteristics, pre-operative blood test results, data from pre-operative CT imaging and PF were explored. Pancreatic density (Hounsfield units, Hu), pancreatic duct size and gland thickness were measured using CT imaging.

Results: PF occurred in 42 of 155 cases (types A, B and C: 32, 8, 2 respectively). An inverse relationship between duct width and PF was observed. The odds ratio of PF, for each 1 mm increase in duct width, was 0.639 (95% CI = 0.531-0.769, p < 0.001). The gland thickness and density at the pancreatic resection margin were positively associated with PF (both p = 0.03). No patient variable was associated with PF.

Conclusions: Pancreatic duct width has previously been assessed at the time of operation and simply regarded as normal or wide. Consideration of duct width as a continuous variable using pre-operative CT imaging can be used to simply predict risk of PF. The association between pancreatic density and PF is a novel finding. Whether pancreatic density in Hu relates to steatosis, as it does for hepatic steatosis, merits further review given the association between pancreatic steatosis and PF.