Skip to Main content Skip to Navigation
Master Thesis

Évaluation rétrospective des facteurs de risque de fistule pancréatique après pancréatectomie distale : établissement d'un score prédictif

Abstract : Postoperative Pancreatic Fistula (POPF) is a common and serious complication after Distal Pancreatectomy (DP). The number and the weight of the different risk factors reported in literature still remain controversial and an effective and accepted score to predict the occurrence of grade B and C (Clinically Relevant; CR) POPF does not exist. Methods: data regarding 103 consecutive patients undergoing DP were collected. A multivariate analysis of variables potentially predictive of fistula risk via multivariate logistic regression with backward selection was performed, in order to build a simplified score. The predictive performance of the score was then tested on the population studied. The accuracy in predicting a categorical outcome was evaluated using the Receiver Operating Characteristic (ROC) curves. Youden's J test was performed to evaluate the performance of a positive score on the POPF occurrence. Results: thirty-three patients developed a CR-POPF: 24 grade B fistulas and 9 grade C fistulas. Operation time >4 hours, amylase levels on drains’ fluid >500 UI on POD 3 and pancreatic thickness >10 mm were found to be predictive of CR-POPF on multivariate analysis. A 4 points score was created from the variables selected in the multivariate model, by assigning 1 point for each of the previous parameters and 1 if the patient had a BMI >30. The discriminating ability was tested on the ROC curve, showing an area under thecurve of 0.83 (CI 95% = 42 [0.75-0.92]). On the Hosmer-Lemeshow test, the difference between the prediction of the model and the observed value is not significant (p-value = 0.44). The score threshold was determined at 2 points/4, which has the highest value according to the Youden index (0.53). The sensitivity is calculated at 82% (CI 95% = [69-95]) and the specificity at 71 (CI 95% = [61-82]). A threshold at 3 points/4 allows to reach a specificity of 99% (CI 95% = [99-100]). Conclusion: a simple score based on operation time, obesity, amylase level on abdominal drains on POD3 and pancreatic thickness measured at the isthmus on preoperative CT can predict the postoperative risk of developing CR-POPF with good sensitivity using a cutoff of 2 or more point.
Document type :
Master Thesis
Complete list of metadata
Contributor : Scd Université de Montpellier <>
Submitted on : Friday, April 23, 2021 - 3:54:44 PM
Last modification on : Friday, June 4, 2021 - 11:50:03 AM


ROLLIN Nicolas_thèse.pdf
Files produced by the author(s)


Distributed under a Creative Commons Attribution - NonCommercial - ShareAlike 4.0 International License


  • HAL Id : dumas-03206833, version 1


Nicolas Rollin. Évaluation rétrospective des facteurs de risque de fistule pancréatique après pancréatectomie distale : établissement d'un score prédictif. Human health and pathology. 2021. ⟨dumas-03206833⟩



Record views


Files downloads