Pruvot, François René, Sauvanet, Alain, Delpero, Jean Robert, Truant, Stéphanie, Bendersky, Noelle, Farges, Olivier
Objective: Measure the caseload of pancreatectomies that influences their short-term outcome, at a national level, and assess the applicability of a centralization policy.
Background: There is agreement that pancreatectomies should be centralized. However, previous studies have failed to accurately define a “high-volume” center.
Methods: French healthcare databases were screened to identify all adult patients who had elective pancreatectomies between 2007 and 2012. The patients’ age, comorbidities, indication, and extent of surgery, and also the hospital administrative-type and location were retrieved. The annual-caseload of pancreatectomy was calculated for each hospital facility. The primary endpoint was 90-day mortality. Spline modeling was used to identify the different annual-caseload that influenced mortality. Logistic regressions were performed to assess if their influence was independent of confounders, and the accuracy of the model calculated.
Results: Overall, 22,366 patients underwent a pancreatectomy and the mortality was 8.1%. Two cut-offs were identified (25 and 65 per year): compared with centers performing >65 resections per year, the adjusted OR of mortality was 1.865 (1.529–2.276) in centers performing ≤25 resections per year and 1.234 (1.031–1.478) in those performing 26 to 65 resections per year. The average number of facilities performing ≤25, 26 to 65, and >65 pancreatectomies per year was 456, 20, and 9, respectively. The percentage of patients operated in these facilities was 56.6%, 19.9%, and 23.3%, respectively.
For pancreaticoduodenectomies (12,670 patients; mortality 9.2%), there were 2 cut-offs (16 and 40 pancreaticoduodenectomies per year), and both were independent predictors of mortality (adjusted OR of 1.979 and 1.333). For distal pancreatectomies (7085 patients; 6.2% mortality), there were 2 cut-offs (13 and 25 distal pancreatectomies per year), but neither was an independent predictor of outcome (area under the receiver-operating characteristic curve of the model = 0.778).
Conclusions: Centralization of pancreatic surgery is theoretically justified, but currently unrealizable. As the incidence of pancreatic cancer increases, there is an urgent need to improve the training of surgeons and develop both intermediate and high-volume centers.