SSAT - PREOPERATIVE PREDICTION OF THE PANCREATIC FISTULA GRADE B/C AFTER PANCREATODUODENECTOMY USING RADIOLOGICAL AND CLINICAL PARAMETERS

Back to 2017 Posters

PREOPERATIVE PREDICTION OF THE PANCREATIC FISTULA GRADE B/C AFTER PANCREATODUODENECTOMY USING RADIOLOGICAL AND CLINICAL PARAMETERS
Hryhoriy Lapshyn^*, Michael Thomaschewski, Ekaterina Petrova, Kim C. Honselmann, Franck Billmann, Dirk Bausch, Tobias Keck, Ulrich F. Wellner
Clinic for Surgery, University Clinic Schleswig-Holstein, Campus Lübeck, Lübeck, Germany

Introduction: Postoperative pancreatic fistula (POPF) and further complications remain the Achilles heel of pancreatic surgery. Almost all models for risk prediction refer to parameters which are known only postoperatively. The aim of the current study was the prediction of clinical relevant POPF after pancreatoduodenectomy by preoperatively available parameters.

Methods: 89 patients with PD operated from December 2012 to June 2015 were included. Pancreas morphology was characterized by radiological parameters. The statistical analysis was carried out using a Classification and Regression Tree (CART) algorithm of the R software. Nine radiological-morphological parameters and seven clinical routine parameters were included in the prediction model.

Results: The incidence of POPF grade B/C according to ISGPS classification was 15%. A simple CART model on the basis of two parameters (BMI and radiological parenchyma thickness) predicted 91% of the cases correctly (positive/negative predictive value 66% and 96%). There was a significant correlation between prediction and pancreatic texture as well as Clavien-Dindo classification.

Conclusion: Clinical and radiological parameters combined in a CART model can correctly predict a clinical relevant POPF after pancreatoduodenectomy already before the operation. This can be used for clinical decision and stratification in randomized trials.

Back to 2017 Posters