An ordinal model to predict the risk of symptomatic liver failure in patients with cirrhosis undergoing hepatectomy

[Display omitted] •Laparoscopy reduces the risk of liver failure after resection in a cirrhotic liver.•Remnant to total liver volume and platelets are other predictors of liver failure.•Intraoperative blood loss is a postoperative predictor of liver failure.•Predictive models are available at: https://prodeau.shinyapps.io/shiny/. Selection criteria for hepatectomy in patients with cirrhosis are controversial. In this study we aimed to build prognostic models of symptomatic post-hepatectomy liver failure (PHLF) in patients with cirrhosis. This was a cohort study of patients with histologically proven cirrhosis undergoing hepatectomy in 6 French tertiary care hepato-biliary-pancreatic centres. The primary endpoint was symptomatic (grade B or C) PHLF, according to the International Study Group of Liver Surgery’s definition. Twenty-six preoperative and 5 intraoperative variables were considered. An ordered ordinal logistic regression model with proportional odds ratio was used with 3 classes: O/A (No PHLF or grade A PHLF), B (grade B PHLF) and C (grade C PHLF). Of the 343 patients included, the main indication was hepatocellular carcinoma (88%). Laparoscopic liver resection was performed in 112 patients. Three-month mortality was 5.25%. The observed grades of PHLF were: 0/A: 61%, B: 28%, C: 11%. Based on the results of univariate analyses, 3 preoperative variables (platelet count, liver remnant volume ratio and intent-to-treat laparoscopy) were retained in a preoperative model and 2 intraoperative variables (per protocol laparoscopy and intraoperative blood loss) were added to the latter in a postoperative model. The preoperative model estimated the probabilities of PHLF grades with acceptable discrimination (area under the receiver-operating characteristic curve [AUC] 0.73, B/C vs. 0/A; AUC 0.75, C vs. 0/A/B) and the performance of the postoperative model was even better (AUC 0.77, B/C vs. 0/A; AUC 0.81, C vs. 0/A/B; p