• Langue : Français
• Discipline : Médecine. Chirurgie générale
• Identifiant : 2017LIL2M221
• Type de thèse : Doctorat de médecine
• Date de soutenance : 15/06/2017

Résumé en langue originale

.

Résumé traduit

Objective : To determine predictors of symptomatic post-hepatectomy liver failure (PHLF) and construct prognostic models of PHLF in patients with cirrhosis. Background data Selection criteria for hepatectomy in patients with cirrhosis are debatable and none of the reported predictive models of PHLF has been widely accepted. Methods : A prospective cohort of patients with histologically-proven cirrhosis undergoing hepatectomy in 6 French HBP-centres from October 2012 to June 2016 was used. Primaryendpoint was symptomatic-PHLF, defined using the International Study Group of Liver Surgery criteria and classified as grade B or C. Twenty-seven preoperative and 4 intraoperative variables were considered. Multiple imputation method was used to deal with missing data (12%). An ordered logistic-regression with proportional odds-ratios method was used to predict three classes : O-A (No PHLF or PHLF-grade A), B (grade B) and C (grade C). The model's stability was controlled with a bootstrap method. Results : Of the 343 patients included, the main indication was hepatocellular carcinoma in 301 (88%). Hepatectomy was anatomical in 199 (58%) patients and a laparoscopic approach was used in 88 (26%). Three-month mortality was 5.25%. Classification of patients regarding observed PHLF was : 0-A : 61%, B : 28%, C : 11%. Three preoperative^ known variables (platelet-count, liver-remnant-volume-ratio and laparoscopic approach) and one intraoperative variable (blood-loss) were retained in the final model. Laparoscopy was associated with a lower risk of PHLF (OR 0.25; 95%CI 0.12-0.51). The ordinal model including only the three preoperative variables estimated the probabilities of being classified into each of the classes with acceptable discrimination (B/C vs 0-A : AUC 0.72; C vs 0-A/B : AUC 0.73). Adding intraoperative blood-loss to the model increased the prediction performance (B/C vs 0-A : AUC 0.77; C vs 0-A/B : AUC 0.81). Conclusions : By accurately predicting the risk of PHLF, the 3-variables model should be useful at patient's selection before surgery. Prediction can be adjusted at end-of-surgery by considering blood-loss in the 4-variables model and might influence postoperative management.