The effect of high intraoperative blood loss on pancreatic fistula development after pancreatoduodenectomy: An international, multi-institutional propensity score matched analysis

Abstract

Background: The association between intraoperative estimated blood loss and outcomes after pancreatoduodenectomy has, thus far, been rarely explored.

Methods: In total, 7,706 pancreatoduodenectomies performed at 18 international institutions composing the Pancreas Fistula Study Group were examined (2003-2020). High estimated blood loss (>700 mL) was defined as twice the median. Propensity score matching (1:1 exact-match) was employed to adjust for variables associated with high estimated blood loss and clinically relevant pancreatic fistula occurrence. The study was powered to detect a 33% clinically relevant pancreatic fistula increase in the high estimated blood loss group, with α = 0.05 and β = 0.2.

Results: The propensity score model included 966 patients with high estimated blood loss and 966 patients with lower estimated blood loss; all covariate imbalantces were solved. Patients with high estimated blood loss patients experienced higher clinically relevant pancreatic fistula rates (19.4 vs 12.6%, odds ratio 1.66; P < .001), as well as higher severe complication rates (27.8 vs 15.6%), transfusions (50.1 vs 14.3%), reoperations (9.2 vs 4.0%), intensive care unit transfers (9.9 vs 4.8%) and 90-day mortality (4.7 vs 2.0%, all P < .001). High estimated blood loss was an independent predictor for clinically relevant pancreatic fistula (odds ratio 1.78, 95% confidence interval 1.37-2.32), as were prophylactic Octreotide administration (odds ratio 1.95, 95% confidence interval 1.46-2.61) and soft pancreatic texture (odds ratio 5.32, 95% confidence interval 3.74-5.57; all P < .001). Moreover, a second model including 1,126 pancreatoduodenectomies was derived including vascular resections as additional confounder (14.0% vascular resections performed in each group). On multivariable regression, high estimated blood loss was confirmed an independent predictor for clinically relevant pancreatic fistula reduction (odds ratio 1.80, 95% confidence interval 1.32-2.44; P < .001), whereas vascular resection was not (odds ratio 0.64, 95% confidence interval 0.34-1.88; P = .156).

Conclusion: This study better establishes the relationship between estimated blood loss and outcomes after pancreatoduodenectomy. Despite inherent contributions to blood loss, its minimization is an actionable opportunity for clinically relevant pancreatic fistula reduction and performance optimization in pancreatoduodenectomy. Accordingly, practical insights are offered to achieve this goal.