Análise comparativa da acurácia de escores prognósticos na insuficiência hepática aguda: revisão de literatura e proposta de protocolo de encaminhamento em hospital não-transplantador

Abstract

Acute Liver Failure (ALF) is a rare condition associated with high mortality, characterized by rapid deterioration of liver function in patients without preexisting chronic liver disease, defined by coagulopathy (INR ≥1.5), hepatic encephalopathy, and symptom onset within 26 weeks. Its progression may lead to multiorgan failure and cerebral edema, making the decision between spontaneous recovery and emergency liver transplantation particularly challenging. Prognostic scores are therefore essential for risk stratification and referral decisions. This study aimed to comparatively analyze the accuracy (AUROC) of major prognostic scores used in ALF — King’s College Criteria (KCC), MELD, Clichy, ALFSG, and ALFED — and to support the development of a triage algorithm for non-transplant centers. An integrative literature review was conducted, searching PubMed/MEDLINE, LILACS, and SciELO (2010–2025). Classical scores (KCC, MELD, and Clichy) showed moderate accuracy (AUROC 0.72–0.78), with performance varying by etiology. KCC demonstrated better specificity in acetaminophen-induced ALF, whereas MELD performed better in non- acetaminophen etiologies. Dynamic models such as ALFSG and ALFED showed superior prognostic performance, particularly within the first 48–72 hours, as they incorporate clinical progression and serial laboratory markers. The addition of variables such as arterial lactate, creatinine, and phosphate improved the predictive value of static models. No single score demonstrated ideal performance. A multimodal approach prioritizing high sensitivity to avoid undertriage appears more appropriate in non-transplant hospitals. The proposed flowchart represents an evidence-based theoretical tool requiring prospective validation before systematic implementation.