Survival variability across hospitals after resection for pancreatic adenocarcinoma: A multilevel survival analysis on a high-resolution population-based study.

Introduction: Resection is the cornerstone of curative management for pancreatic ductal adenocarcinoma (PDAC). Hospital surgical volume influence post-operative mortality. Few is known about impact on survival.

Methods: Population included 763 patients resected for PDAC within the 4 French digestive tumor registries between 2000 and 2014. Spline method was used to determine annual surgical volume thresholds influencing survival. A multilevel survival regression model was used to study center effect.

Results: Population was divided into three groups: low-volume (LVC) (<41 hepatobiliary/pancreatic procedures/year), medium-volume (MVC) (41-233) and high-volume centers (HVC) (>233). Patients in LVC were older (p = 0.02), had a lower rate of disease-free margins (76.7% vs. 77.2% and 69.5%, p = 0.028) and a higher post-operative mortality than in MVC and HVC (12.5% and 7.5% vs. 2.2%; p = 0.004). Median survival was higher in HVC than in other centers (25 vs. 15.2 months, p < 0.0001). Survival variance attributable to center effect accounted for 3.7% of total variance. In multilevel survival analysis, surgical volume explained the inter-hospital survival heterogeneity (non-significant variance after adding the volume to the model p = 0.3). Patients resected in HVC had a better survival than in LVC (HR 0.64 [0.50-0.82], p < 0.0001). There was no difference between MVC and HVC.

Conclusion: Regarding center effect, individual characteristics had little impact on survival variability across hospitals. Hospital volume was a major contributor to the center effect. Given the difficulty of centralizing pancreatic surgery, it would be wise to determine which factors would indicate management in a HVC.