Development and validation of a novel prognostic model to predict 1-year post-transplant mortality for acute-on-chronic hepatitis B liver failure: a nationwide, multicentre, cohort study.
Journal Articles
Overview
Research
Identity
Additional Document Info
View All
Overview
abstract
BACKGROUND: Liver transplantation (LT) provides a potential cure for hepatitis B virus-related acute-on-chronic liver failure (HBV-ACLF). We aimed to develop and externally validate a prognostic model to predict 1-year post-LT mortality in patients with HBV-ACLF. METHODS: This retrospective, nationwide, observational cohort study was conducted at ten high-volume LT centres in China. 4378 adult patients who underwent primary LT between January 2015 and December 2021 were screened, and those with HBV-ACLF according to the COSSH-ACLF criteria (separated into three ACLF grades based on the number of organ failures) were included. The HBV-ACLF LT (HALT) model was developed in the derivation cohort and validated in the external testing cohort. The derivation cohort were derived from two LT centres in one province (Zhejiang). The external testing cohort were derived from eight LT centres in two provinces. For model development, univariable Cox regression analysis was used to identify risk factors associated with 1-year post-LT mortality. Variables with univariable p < 0.05 were entered into the least absolute shrinkage and selection operator (Lasso) analysis for further feature selection. 10-fold cross validation was used to choose the optimal lambda (penalty for the number of features) of the Lasso model. Multivariable Cox regression was applied to construct the HALT model based on the risk factors selected by Lasso analysis. Primary outcome was survival rate at 1-year after LT. Secondary outcomes were short-term (28- and 90-day) and long-term survival after LT (3- and 5-year). Model performance was compared with eight other models (COSSH-ACLF II, COSSH-ACLF, CLIF-C ACLF, AARC, MELD, MELD-Na, SALT-M and TAM scores), using receiver operating characteristic curve and C-index values. A nomogram was developed to analyse the probability of the primary outcome in different graft-recipient combinations based on recipient factors (age, number of organ failures [OF], lactate) and graft factors (donation after circulatory death [DCD] and cold ischaemia time [CIT]). FINDINGS: Between Jan 1, 2015, and Dec 1, 2021, 668 patients were included (derivation cohort, n = 418; external testing cohort, n = 250), with survival rates of 88.0%, 81.1%, 77.5%, 75.6% and 72.1% at 28-day, 90-day, 1-year, 3-year and 5-year post-LT, respectively. Three recipient's factors (age, number of OF and arterial lactate concentration) as well as two graft's parameters (DCD and CIT) were independently associated with 1-year post-LT mortality in the derivation cohort (all p < 0.05). The HALT model was established accordingly, showing better discriminative performance (C-index, 0.791) than eight current models in the external testing cohort (C-index, 0.529-0.627; all p < 0.001). If the sickest patients (age >55 years, OFs ≥3 and lactate ≥2.5 mmol/L) received high-risk grafts (DCD and CIT >10 h), the estimated 1-year post-LT mortality was 85.6%. INTERPRETATION: The HALT model showed superior predictive ability over eight current models and may help for LT candidate selection and optimal organ allocation. Though the findings need to be verified in prospective studies and among different patient populations. FUNDING: This work was supported by grants from the National Natural Science Foundation of China, Natural Science Foundation of Zhejiang Province, and the Research Project of Jinan Microecological Biomedicine Shandong Laboratory.
