Predicting 28-day mortality in artificial liver support-treated HBV–ACLF: development and validation of a novel prognostic model

Le résumé fourni par la source

This study aimed to develop and internally validate the ANIT (Age, Neutrophil count, INR, Total bilirubin) score for predicting 28-day mortality in patients with hepatitis B virus-related acute-on-chronic liver failure (HBV–ACLF) undergoing artificial liver support systems (ALSS), and to compare its prognostic performance with MELD and COSSH–ACLF II. In this retrospective single-center cohort study, 380 HBV–ACLF patients treated with ALSS between January 2017 and July 2024 were included and randomly divided into a discovery cohort ( n = 265) and a validation cohort ( n = 115). Candidate predictors were selected through correlation-based redundancy reduction, collinearity assessment, and multivariable logistic regression with backward stepwise elimination. Model discrimination, calibration, and decision-curve performance were compared with MELD and COSSH–ACLF II. The final ANIT model included age, neutrophil count, INR, and total bilirubin. In the discovery cohort, the AUC of ANIT was 0.840 (95% CI 0.792–0.887), compared with 0.719 (95% CI 0.654–0.785) for MELD and 0.777 (95% CI 0.719–0.835) for COSSH–ACLF II. In the validation cohort, the corresponding AUCs were 0.919 (95% CI 0.866–0.971), 0.873 (95% CI 0.805–0.941), and 0.845 (95% CI 0.776–0.915), respectively. ANIT showed higher discrimination than both comparator models in the discovery cohort and higher discrimination than COSSH–ACLF II while showing comparable discrimination to MELD in the validation cohort. Calibration was acceptable across all three models in both cohorts, and ANIT had the lowest Brier score in both cohorts. In bootstrap internal validation, the ANIT model showed an optimism-corrected C-index of 0.8690 (95% CI 0.8354–0.9037), an optimism-corrected Brier score of 0.1452, a calibration intercept of − 0.0184, and a calibration slope of 0.9670, indicating limited overfitting and stable internal performance. The ANIT score demonstrated clinically competitive prognostic performance relative to MELD and COSSH–ACLF II in HBV–ACLF patients undergoing ALSS. Its simple four-variable structure based on routinely available measurements supports its potential utility as a practical bedside risk-stratification tool. Prospective multicenter external validation is warranted.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.