Prognostication in acute-on-chronic liver failure (ACLF) remains a challenge in modern hepatology, where victory has been tantalizingly close yet distant [
1]. The nuances of prognostication in ACLF are as complex as the disease itself, spanning multiple domains, including spontaneous improvement versus the need for liver support mechanisms, continued medical management versus liver transplant (LT), and, ultimately, continuation of care versus futility decisions [
1].
Ever since its formal recognition as a syndrome associated with systemic inflammation and multiorgan dysfunction, it has become evident that it follows distinct pathways of recovery or rapid deterioration towards short-term mortality [
2]. Multiple studies have explored the pathways of ACLF progression, including proteins, cytokines, and growth factors that mediate the development of organ failures (OF) and predict outcomes [
3]. However, a key challenge has remained the absence of a non-LT definitive therapeutic modality that can make decision-making objective [
2]. Extracorporeal liver support devices evoked great enthusiasm based on their robust theoretical basis; however, results from clinical trials were disappointing, and the same was reiterated with granulocyte colony-stimulating factor [
4,
5]. Currently, exchange therapies, including therapeutic plasma exchange, continuous renal replacement therapy, and double plasma molecular adsorption systems, are emerging as potential treatment strategies, although with mixed clinical results [
6]. LT, therefore, remains the only definitive modality of management. However, traditional perceptions of poor outcomes in ACLF, uncertainties regarding the optimal timing of transplantation, organ shortages, and a lack of access impede the utilization of LT.7 In the face of these challenges, accurately prognosticating and identifying patients who will benefit from a particular pathway remains a daily dilemma for clinicians.
At the heart of the problems clinicians face when interpreting available literature on ACLF lies the heterogeneity and variability in disease definition, acute precipitants, and approaches to management [
8]. Recently, there have been calls for harmonizing the definitions of ACLF with the definition proposed by the Asia Pacific Association for the Study of the Liver (APASL) representing “early” disease and the definitions proposed by the European Association of the Study of the Liver (EASL) and North American Consortia for the Study of End Stage Liver Disease (NACSELD) to represent more advanced stages, associated with higher mortality [
9]. Although this would homogenize clinical data and possibly eliminate regional bias, questions regarding accurate prognostication and identifying the optimal transplant candidate remain.
In this edition of the journal, Yeo et al. [
10] report their ACLF models (patients with ≥2 OF) developed from retrospective data obtained from a publicly available database spanning over 11 years (2008–2019). The authors divided the patients into two major groups, based on the EASL definition (n=1,692) and the NACSELD definition (n=1,511). The overall cohort showed a male predominance, with alcohol use being the leading etiology of liver disease. Respiratory and circulatory failures were the leading OF in both patient groups, and the 30-day mortality rates in both groups were similar (NACSELD: 47.8% vs. EASL: 49.2%). The authors used baseline variables at intensive care unit (ICU) admission to develop definition-specific models, the CatBoost model in NACSELD and the random forest ACLF model for EASL. The use of the Shapley model enabled the identification of the contribution of individual parameters (hierarchical relationships of variables) to the prediction of mortality. Interestingly, the top 12 factors identified in both models had significant overlap, including total bilirubin, albumin, international normalized ratio (INR), and systemic inflammatory response (SIRS) parameters, as well as heart rate, systolic blood pressure, respiratory rate, and body temperature. This effectively underscores our current understanding of ACLF—a state of systemic inflammation (elevated SIRS) provoked by an acute precipitant, which results in a potential hyperimmune response leading to multiorgan injury unless effectively controlled [
3]. The presence of markers of hepatocyte function—albumin, INR (synthesis), and bilirubin (excretion)—highlights the central role of the liver as a key determinant of outcomes in these patients [
11]. The authors subsequently validated the models in a cohort of 498 patients from the ICU of a tertiary care centre in Germany. Notably, both developed scores outperformed conventional scoring systems (model for end-stage liver disease [MELD], MELD with sodium [MELD-Na], MELD 3.0, and chronic liver failure consortium [CLIF-C] ACLF) in predicting 30-day mortality in these patients.
While the results of this study were interesting and provided valuable insights into the multiple variables that influence the outcome of ACLF patients, certain caveats need to be kept in mind. In order to represent a sicker cohort of patients, the authors chose to collect patient data from an ICU setting. However, considering that the admission criterion of ICUs shows significant regional and institutional variations, this carries the risk of selection bias (such as the higher incidence of circulatory and respiratory failure in this cohort) [
8]. While the database accessed would provide information about a large number of patients from multiple ethnicities, the results of this study may not be generalizable to areas such as Southeast or East Asia, where precipitants like viral hepatitis and drug-induced liver injury alter the dynamics [
12,
13]. In reality, this has been one of the key points of debate over differences in definition and prognostication between the East and the West, with an apparent divergence in prognostic score performances.
The long assessment period chosen by the authors allowed for the recruitment of a large number of patients to utilize the machine learning algorithms, and also reflects a time period during which the understanding and management of ACLF have changed considerably. Thus, the modality of treatment used, time of referral, institution of antibiotics (including timing, empirical versus culture-based decisions on escalation or de-escalation), and resuscitation practices may have changed over this duration, influencing the outcomes of the included patients and, consequently, the predictive model [
14-
16]. The retrospective design limits the ability to capture such additional data, which would have added further granularity to the reported outcomes.
ACLF is a dynamic condition, with clinical decisions being made based on changes in parameters such as vitals, hepatic functions, lactate, sensorium etc. While the reported model is effective in predicting mortality, it does not reflect the dynamism of the disease. This is akin to the conundrum faced by most clinicians who rely on “static” tests (King’s College criteria) for prognostication and transplant referral in acute liver failure, while dynamic tests (e.g., acute liver failure early dynamic score) have reported better comparative outcomes [
17,
18]. Therefore, identifying the impact of changes in the top 12 parameters (and consequently developing threshold values for such clinical improvement) on patient outcomes would have helped generate a model that could be used clinically to tailor the therapeutic modality to individual patients. This would help generate a robust treatment algorithm, assessing patients at multiple, definitive timepoints and providing guidance on the clinical decisions to be made at that time, based on the trends of the observed values. Inclusion of patients who had undergone LT for ACLF, along with controls who did not, would help in identifying optimal transplant candidates as well as help develop timelines for transplant evaluation and referral. Future studies may further investigate this aspect.
In conclusion, the current study represents a significant first step in streamlining ACLF prognostication and management using machine learning tools. The authors are to be commended for their considerable efforts in attempting to highlight the important clinical and biochemical factors that determine prognosis, regardless of the definition used. This calls for larger, collaborative global efforts to standardize the definitions, optimize management protocols, and finally, achieve our collective goal of improving ACLF outcomes.
FOOTNOTES
-
Authors’ contribution
SB, writing original draft. AR, conceptualization, editing, and supervision.
-
Conflicts of Interest
The authors have no conflicts to disclose.
Abbreviations
acute-on-chronic liver failure
Asia Pacific Association for the Study of the Liver
chronic liver failure consortium
European Association of the Study of the Liver
international normalized ratio
model for end-stage liver disease
model for end-stage liver disease with sodium
North American Consortia for the Study of End Stage Liver Disease
systemic inflammatory response
REFERENCES
- 1. Hernaez R, Solà E, Moreau R, Ginès P. Acute-on-chronic liver failure: an update. Gut 2017;66:541-553.
- 2. European Association for the Study of the Liver. EASL clinical practice guidelines on acute-on-chronic liver failure. J Hepatol 2023;79:461-491.
- 3. Engelmann C, Zhang IW, Clària J. Mechanisms of immunity in acutely decompensated cirrhosis and acute-on-chronic liver failure. Liver Int 2025;45:e15644.
- 4. Ballester MP, Elshabrawi A, Jalan R. Extracorporeal liver support and liver transplantation for acute-on-chronic liver failure. Liver Int 2025;45:e15647.
- 5. Engelmann C, Herber A, Franke A, Bruns T, Reuken P, Schiefke I, et al. Granulocyte-colony stimulating factor (G-CSF) to treat acute-on-chronic liver failure: a multicenter randomized trial (GRAFT study). J Hepatol 2021;75:1346-1354.
- 6. Beran A, Mohamed MFH, Shaear M, Nayfeh T, Mhanna M, Srour O, et al. Plasma exchange for acute and acute-on-chronic liver failure: a systematic review and meta-analysis. Liver Transpl 2024;30:127-141.
- 7. Belli LS, Duvoux C, Artzner T, Bernal W, Conti S, Cortesi PA, et al. Liver transplantation for patients with acute-on-chronic liver failure (ACLF) in Europe: results of the ELITA/EF-CLIF collaborative study (ECLIS). J Hepatol 2021;75:610-622.
- 8. Aggarwal A, Biswas S, Arora U, Vaishnav M, Shenoy A, Swaroop S, et al. Definitions, etiologies, and outcomes of acute on chronic liver failure: a systematic review and meta-analysis. Clin Gastroenterol Hepatol 2024;22:2199-2210.e25.
- 9. Kulkarni AV, Sarin SK. Acute-on-chronic liver failure - steps towards harmonization of the definition! J Hepatol 2024;81:360-366.
- 10. Yeo YH, Zhang M, McCoy MS, Zu J, He Y, Liu Y, et al. Predictive machine learning model in intensive care unit patients with acute-on-chronic liver failure and two or more organ failures. Clin Mol Hepatol 2025;31:1355-1371.
- 11. Zaccherini G, Weiss E, Moreau R. Acute-on-chronic liver failure: definitions, pathophysiology and principles of treatment. JHEP Rep 2021;3:100176.
- 12. Devarbhavi H, Joseph T, Sunil Kumar N, Rathi C, Thomas V, Prasad Singh S, et al. The Indian network of drug-induced liver injury: etiology, clinical features, outcome and prognostic markers in 1288 patients. J Clin Exp Hepatol 2021;11:288-298.
- 13. Devarbhavi H, Asrani SK, Arab JP, Nartey YA, Pose E, Kamath PS. Global burden of liver disease: 2023 update. J Hepatol 2023;79:516-537.
- 14. Singer M, Deutschman CS, Seymour CW, Shankar-Hari M, Annane D, Bauer M, et al. The third international consensus definitions for sepsis and septic shock (Sepsis-3). JAMA 2016;315:801-810.
- 15. Plevin R, Callcut R. Update in sepsis guidelines: what is really new? Trauma Surg Acute Care Open 2017;2:e000088.
- 16. Choudhury A, Kulkarni AV, Arora V, Soin AS, Dokmeci AK, Chowdhury A, et al. Acute-on-chronic liver failure (ACLF): the ‘Kyoto Consensus’-steps from Asia. Hepatol Int 2025;19:1-69.
- 17. Kumar R, Sharma H, Goyal R, Kumar A, Khanal S, et al. Prospective derivation and validation of early dynamic model for predicting outcome in patients with acute liver failure. Gut 2012;61:1068-1075.
- 18. Biswas S. Liver transplantation for acute liver failure-indication, prioritization, timing, and referral. J Clin Exp Hepatol 2023;13:820-834.
Citations
Citations to this article as recorded by
- Correspondence to editorial on “Predictive machine learning model in intensive care unit patients with acute-on-chronic liver failure and two or more organ failures”
Mengyi Zhang, Yee Hui Yeo, Jian Zu, Jonel Trebicka, Fanpu Ji
Clinical and Molecular Hepatology.2026; 32(3): e346. CrossRef
