S2780 ACLF Prognostication of 28- and 90-Day Outcomes: A Real-World Liver ICU Experience

Introduction: Early prognostication is vital in the management of critically ill patients with acute-on-chronic liver failure (ACLF). The Chronic Liver Failure Consortium (CLIF-C) and the North American Consortium for the Study of End-Stage Liver Disease (NACSELD) scores are well-validated tools for diagnosis and prognosis in this population. The global effort to standardize prognostic models for ACLF reflects an urgent need driven by clinical practice and research priorities. This study aims to assess the predictive performance of these models within a high-acuity liver Intensive Care Unit (ICU) cohort. Methods: We conducted a retrospective analysis of 463 first-time admissions to a dedicated liver ICU between August 2018 and August 2023. Included patients met criteria for ACLF on ICU Day 1, defined as CLIF-C ACLF grade ≥1 or the presence of ≥2 organ failures per NACSELD. Patients with ACLF grade 0 or missing 28- and 90-day outcome data were excluded. Day 1 CLIF-C and NACSELD scores were calculated. Based on CLIF-C ACLF grading, patients were stratified into low-grade (grade 1) and high-grade (grades 2–3) groups. The primary endpoints were 28- and 90-day mortality or liver transplant (LT). Prognostic performance was evaluated using receiver operating characteristic (ROC) curves. Results: Among 463 patients (56.4% man; median age 57 years [IQR 46–64]), the median ICU length of stay was 7.4 days (IQR 4.0–13.2). Alcohol-related liver disease was the predominant etiology (55.3% of cases). CLIF-C demonstrated better 28-day mortality prognostic performance compared to NACSELD (AUROC 0.682 vs 0.560; P < 0.0001) with similar trends at 90 days (AUROC 0.643 vs 0.540; P = 0.0026). Among patients with low-grade ACLF, CLIF-C yielded non-significantly higher AUROCs at both 28 days (0.641 vs 0.592; P = 0.34) and 90 days (0.604 vs 0.580; P = 0.580). In high-grade ACLF, CLIF-C performed with slightly better discriminatory AUROCS at 28 days (0.647 vs 0.530; P = 0.0006) and 90 days (0.629 vs 0.510; P = 0.0149). Conclusion: In this real-world liver ICU cohort, the CLIF-C and NACSELD scoring systems demonstrated only modest prognostic accuracy with AUROC values falling below thresholds typically considered clinically robust. These findings are consistent with international initiatives aimed at refining and harmonizing ACLF definitions and prognostic models. Our results highlight the limitations of current tools and emphasize the need for multicenter granular data to guide clinical decision-making and research priorities.