Abstract
IMPORTANCE: Postpancreatectomy hemorrhage is an uncommon but highly morbid complication of pancreaticoduodenectomy. Clinical evidence often draws suspicion to the gastroduodenal artery stump, even without a clear source. OBJECTIVE: To determine the frequency of gastroduodenal artery bleeding compared to other sites and the results of mitigation strategies. DESIGN, SETTING, AND PARTICIPANTS: This cohort study involved a retrospective analysis of data for consecutive patients who had pancreaticoduodenectomy from 2011 to 2021 at Memorial Sloan Kettering Cancer Center (MSK) and Thomas Jefferson University Hospital (TJUH). EXPOSURES: Demographic, perioperative, and disease-related variables. MAIN OUTCOMES AND MEASURES: The incidence, location, treatment, and outcomes of primary (initial) and secondary (recurrent) hemorrhage requiring invasive intervention were analyzed. Imaging studies were re-reviewed by interventional radiologists to confirm sites. RESULTS: Inclusion criteria were met by 3040 patients (n = 1761 MSK, n = 1279 TJUH). Patients from both institutions were similar in age (median [IQR] age at MSK, 67 [59-74] years, and at TJUH, 68 [60-75] years) and sex (at MSK, 814 female [46.5%] and 947 male [53.8%], and at TJUH, 623 [48.7%] and 623 male [51.3%]). Primary hemorrhage occurred in 90 patients (3.0%), of which the gastroduodenal artery was the source in 15 (16.7%), unidentified sites in 24 (26.7%), and non-gastroduodenal artery sites in 51 (56.7%). Secondary hemorrhage occurred in 23 patients; in 4 (17.4%), the gastroduodenal artery was the source. Of all hemorrhage events (n = 117), the gastroduodenal artery was the source in 19 (16.2%, 0.63% incidence in all pancreaticoduodenectomies). Gastroduodenal artery hemorrhage was more often associated with soft gland texture (14 [93.3%] vs 41 [62.1%]; P = .02) and later presentation (median [IQR], 21 [15-26] vs 10 days [5-18]; P = .002). Twenty-three patients underwent empirical gastroduodenal artery embolization or stent placement, 7 (30.4%) of whom subsequently experienced secondary hemorrhage. Twenty percent of all gastroduodenal artery embolizations/stents (8/40 patients), including 13% (3/13 patients) of empirical treatments, were associated with significant morbidity (7 hepatic infarction, 4 biliary stricture), with a 90-day mortality rate of 38.5% (n = 5) for patients with these complications vs 7.8% without (n = 6; P = .008). Ninety-day mortality was 12.2% (n = 11) for patients with hemorrhage (3 patients [20%] with primary gastroduodenal vs 8 [10.7%] for all others; P = .38) compared with 2% (n = 59) for patients without hemorrhage. CONCLUSIONS AND RELEVANCE: In this study, postpancreatectomy hemorrhage was uncommon and the spectrum was broad, with the gastroduodenal artery responsible for a minority of bleeding events. Empirical gastroduodenal artery embolization/stent without obvious sequelae of recent hemorrhage was associated with significant morbidity and rebleeding and should not be routine practice. Successful treatment of postpancreatectomy hemorrhage requires careful assessment of all potential sources, even after gastroduodenal artery mitigation.