Kunal Nandy, Aamir M Parray, Harsh Karun, Vinayaka S Bairannavar, Vikram Chaudhari, Shailesh V Shrikhande, Manish S Bhandare
In clinically relevant postoperative pancreatic fistula patients, bile leak and K pneumoniae infection independently predict severe postpancreatectomy hemorrhage. The pancreaticojejunal anastomosis was the most common bleeding site, whereas cryptogenic postpancreatectomy hemorrhage appears to represent a lower-mortality subgroup once arterial bleeding is excluded.
BACKGROUND: Postpancreatectomy hemorrhage complicates 3%-10% of pancreaticoduodenectomies and is associated with mortality rates up to 20%. Clinically relevant postoperative pancreatic fistula is a major risk factor. This study aimed to characterize bleeding sources and identify predictors of severe hemorrhage in patients with clinically relevant postoperative pancreatic fistula.
METHODS: We retrospectively analyzed 1,722 consecutive pancreaticoduodenectomies performed between January 2013 and December 2024 at a high-volume tertiary center. Bleeding sources were systematically classified, and microbiological profiles of peripancreatic collections were reviewed. Multivariable logistic regression identified independent predictors of grade B/C postpancreatectomy hemorrhage in patients with clinically relevant postoperative pancreatic fistula. Cryptogenic postpancreatectomy hemorrhage was defined as grade B/C hemorrhage without an identifiable source despite comprehensive evaluation.
RESULTS: Postpancreatectomy hemorrhage occurred in 135 patients (7.8%), including 89 cases (5.2%) of grade B/C hemorrhage. The pancreaticojejunal anastomosis was the most frequent bleeding source (15.6%), followed by branches of the superior mesenteric (9.6%) and hepatic (8.1%) arteries; the gastroduodenal artery accounted for only 3.7%. Independent predictors of severe hemorrhage were bile leak (odds ratio, 4.50; 95% confidence interval, 2.01-10.07; P < .001) and gram-negative infection, predominantly Klebsiella pneumoniae (odds ratio, 2.55; 95% confidence interval, 1.18-5.51; P = .017). A nomogram incorporating these factors showed acceptable discrimination (area under the curve, 0.701) with good internal validation (area under the curve, 0.791). Cryptogenic postpancreatectomy hemorrhage occurred in 24.7% of grade B/C cases and was associated with lower mortality than source-identified hemorrhage.
CONCLUSION: In clinically relevant postoperative pancreatic fistula patients, bile leak and K pneumoniae infection independently predict severe postpancreatectomy hemorrhage. The pancreaticojejunal anastomosis was the most common bleeding site, whereas cryptogenic postpancreatectomy hemorrhage appears to represent a lower-mortality subgroup once arterial bleeding is excluded.