Critical care hepatology: A narrative review of current concepts and management

Table 4 Prognostic models and liver transplantation listing criteria in acute liver failure

Score	Components	Outcome predicted	Use/threshold	Type
Established prognostic and LT-listing criteria
KCC[109-111]	Acetaminophen: PH < 7.3, OR all 3 of (INR > 6.5, creatinine > 300 μmol/L, HE grade III-IV); lactate added in 2002 update. Non- acetaminophen: INR > 6.5 alone, OR any 3 of 5 factors (age < 10/> 40, non-A/B/drug etiology, jaundice-to-HE > 7 days, INR > 3.5, bilirubin > 300 μmol/L)	Death without ELT; need for emergency liver transplantation; PPV 70%-100%, specificity ~89%-95% in APAP-ALF	Meeting KCC → list for ELT; endorsed by AASLD and EASL. Sensitivity limited (~59%-69%); sequential use improves specificity. Lactate > 3.5 mmol/L (4 hours) or > 3.0 (12 hours post-resuscitation) incorporated as APAP-KCC criterion	LT listing
MELD/MELD-Na[104,112]	Creatinine, bilirubin, INR (+ sodium). A continuous score; rising values help guide urgency, particularly in ALF	Short-term mortality; ELT urgency. MELD correlates with severity but was not designed for ALF	MELD > 30 → high-urgency ELT listing; dynamic worsening more predictive than single value. Better for non-APAP-ALF (higher sensitivity) than KCC; both used complementarily	Prognostic
ALFSG-PI[113]	HE coma grade, INR, bilirubin, phosphate (≥ 3.7 mg/dL vs < 3.7 mg/dL), log10M30 (caspase-cleaved cytokeratin-18 apoptosis marker) - derived in 500 United States ALF patients; validated in independent 250-patient cohort	21-day transplant-free survival; need for LT or death at study entry. AUROC = 0.822 - superior to KCC (0.654) and MELD (0.704)	Higher index = worse prognosis; covers all ALF etiologies; requires M30 ELISA (limits routine use); 85.6% sensitivity/64.7% specificity for LT/death	Prognostic
CVC[114]	HE grade III-IV + factor V < 20% (age < 30 years) or < 30% (age ≥ 30 years); originally derived in fulminant hepatitis B patients	Survival without LT; primarily validated in viral (HBV) and non-paracetamol ALF in French cohorts	Criteria met → consider ELT listing; particularly used in France and Germany. Factor V assay not universally available. Sensitivity 69%-75%, specificity 50%-56% in mixed etiology (Ichai et al[114], 2015)	LT listing
APACHE II	Acute physiology (12 variables), age, chronic health - 0-71 points; validated in general ICU; applied to ALF as a general severity tool	ICU mortality in ALF patients; adjunctive severity stratification	APACHE II > 15 in ALF correlates with poor outcome; used alongside KCC; AUROC comparable to SOFA and ALFSG-PI in single-center studies	Prognostic
SOFA[115]	PaO2/FiO2, platelets, bilirubin, MAP/vasopressors, GCS, creatinine - 0-4 per organ; total 0-24	Multiorgan failure trajectory; daily ICU reassessment in ALF	Serial SOFA ≥ 15 or rapidly rising score used in futility discussions; SOFA outperforms KCC in multiorgan failure setting; AUROC ~0.84-0.85 in ALF studies (comparable to ALFSG-PI)	Prognostic
Bernal/United Kingdom ALF dynamic model[116]	Day 1: Age, GCS, arterial pH, lactate, creatinine, INR, circulatory failure; day 2: Change in lactate + change in INR - two time-point model specific to APAP-ALF	Death without LT in APAP-induced ALF; dynamic 2-day model improves over single time-point criteria. AUROC day 1: 0.82, day 2: Higher	Improves sensitivity over KCC at early admission; particularly useful in first 48 hours to identify non-survivors who do not yet meet standard KCC; used adjunctively in United Kingdom centers	LT-listing
Serum phosphate[117]	Serum phosphate at 48-96 hours post-paracetamol ingestion (threshold > 1.2 mmol/L); reflects failure of hepatic regeneration (phosphate uptake by regenerating hepatocytes)	Hepatic regeneration failure in APAP-ALF; high specificity for death without LT (used as add-on to KCC)	Phosphate > 1.2 mmol/L at 48-96 hours → failure to regenerate → consider ELT; incorporated as optional KCC criterion; specificity ~89%, PPV ~89% for death	LT-listing
Serum lactate[110]	Arterial lactate at admission and post-resuscitation: > 3.5 mmol/L at 4 hours, or > 3.0 mmol/L at 12 hours after adequate fluid resuscitation, in APAP-ALF	Early mortality indicator in APAP-ALF; tissue hypoxia and mitochondrial dysfunction marker; predicts poor outcome before HE develops	Lactate > 3.5 mmol/L at 4 hours (pre-resuscitation) = KCC criterion. Widely available and rapidly measurable; used for early risk stratification to initiate LT referral before KCC fully met	Prognostic
Emerging and investigational scores in ALF
miRNA-based prognostic model[118,119]	Regeneration-linked miRNA signature (early model) + cell-death miRNA signature (late model), combined with MELD score and vasopressor use - machine-learning derived, ALFSG cohort	21-day transplant-free survival in APAP and non-APAP ALF. Early model AUROC 0.78; late model AUROC 0.83 - both outperformed ALFSG-PI and KCC	Not yet in clinical use; requires standardised miRNA profiling. Represents next generation of biomarker-enriched dynamic ALF models; promising for identifying regeneration potential. Further prospective validation needed	Prognostic
Factor V-based composite scores[120]	Factor V level + bilirubin + vasopressor use + HE coma grade - derived within ALFSG cohorts as a dynamic biomarker model; factor V half-life 12-15 hours makes it sensitive to rapid change	Death or LT at 21 days in ALF; outperformed KCC and MELD but not ALFSG-PI in head-to-head comparison within ALFSG dataset	Factor V assay not universally available; currently investigational. Potential to complement KCC particularly in non-APAP ALF where factor V correlates better with prognosis than INR alone	Prognostic
AI/ML models[120]	Variable depending on model: Multi-parameter clinical + laboratory data (INR, bilirubin, creatinine, HE grade, vasopressors, lactate, ammonia, imaging); some incorporate omics data	Transplant-free survival and LT/death at 21-28 days; aim to be dynamic (daily recalculation) and etiology-agnostic	No AI/ML model currently validated for routine clinical ALF decision-making; several retrospective cohort studies promising (AUROC: 0.85-0.92). Major barriers: Explainability, prospective validation, regulatory approval. Active area of research	Prognostic
ALFED model[120]	Dynamic tracking of 4 variables over 3 days: HE grade > II, serum bilirubin, INR, arterial ammonia - assesses whether each remains above threshold or worsens. Derived in 380 non-APAP ALF patients (India)	Death without LT in non-paracetamol ALF (predominantly hepatitis E and hepatitis B etiology); designed for resource-limited settings	Promising for non-APAP ALF in Asia/developing world where hepatitis E predominates; AUROC superior to KCC in derivation cohort. Requires external validation in diverse global cohorts before routine use	Prognostic
Novel biomarker composites (CPS1, FABP1, Gc-globulin)[118,121,122]	CPS1 (hepatocyte-specific mitochondrial enzyme); FABP1 (hepatocyte damage); Gc-globulin (actin-free; reflects hepatic regeneration capacity) - all added to existing models (ALFSG-PI or KCC)	Death or LT in ALF; each has shown incremental improvement in AUROC when added to ALFSG-PI or KCC in ALFSG biobank studies. FABP1 > 350 ng/mL strongly associated with non-survival	All investigational; none in routine clinical use. FABP1 and CPS1 may be commercially assayable in future. Gc-globulin reflects regeneration rather than injury - concept of “regeneration potential” increasingly recognized as key missing element in current models	Prognostic

This table summarizes established and emerging prognostic models used in acute liver failure, including criteria for liver transplantation listing and short-term outcome prediction. It outlines key variables, predicted outcomes, and clinically relevant thresholds to support early risk stratification and time-sensitive transplant decision-making. LT: Liver transplantation; KCC: King’s College Criteria; OR: Odds ratio; INR: International normalized ratio; HE: Hepatic encephalopathy; ELT: Emergency liver transplantation; PPV: Positive predictive value; APAP: Acetaminophen (paracetamol); ALF: Acute liver failure; MELD: Model for end-stage liver disease; MELD-Na: Model for end-stage liver disease including sodium; ALFSG: Acute Liver Failure Study Group; PI: Prognostic index; AUROC: Area under the receiver operating characteristic curve; ELISA: Enzyme-linked immunosorbent assay; CVC: Clichy-Villejuif criteria; HBV: Hepatitis B virus; APACHE: Acute Physiology and Chronic Health Evaluation; ICU: Intensive care unit; SOFA: Sequential Organ Failure Assessment; PaO₂/FiO₂: Ratio of arterial oxygen partial pressure to fraction of inspired oxygen; MAP: Mean arterial pressure; GCS: Glasgow Coma Scale; miRNA: MicroRNA; AI: Artificial intelligence; ML: Machine learning; CPS1: Carbamoyl phosphate synthetase 1; FABP1: Fatty acid binding protein 1.