Introduction

Acute‑on‑chronic liver failure (ACLF) is a clinically distinct syndrome characterized by acute decompensation of cirrhosis, systemic inflammation, and organ failure, which is associated with high short‑term mortality.¹ Unlike traditional decompensated cirrhosis, ACLF involves extrahepatic organ failure and frequently requires intensive care unit (ICU) admission.² Among these failures, acute kidney injury (AKI) plays a pivotal role in worsening prognosis and guiding therapeutic decisions.^3,4

To capture the dynamic and systemic nature of ACLF, several prognostic scoring systems have been developed. The Chronic Liver Failure Consortium (CLIF‑C) ACLF score, developed by the European Foundation for the Study of Chronic Liver Failure, is one of the most accurate tools to assess short‑term mortality in patients with ACLF and to guide decisions regarding liver transplant and intensive care support.^1,2,5,6 This score incorporates organ failures based on the CLIF Sequential Organ Failure Assessment (CLIF‑SOFA), which in turn adapts the traditional SOFA framework to the context of liver disease. Both CLIF‑SOFA and SOFA remain essential in ICU‑based triage and have demonstrated strong predictive capacity for in‑hospital mortality.⁶

In parallel, liver‑specific scores, such as the Model for End‑Stage Liver Disease (MELD) and its derivatives—MELD with sodium correction (MELD‑Na) and integrated MELD (iMELD)—continue to be widely used for transplant prioritization. However, these indices primarily reflect hepatic and renal laboratory parameters and do not account for multiorgan dysfunction, which limits their predictive value in patients with ACLF.⁷ Other emerging tools, such as the MELD to serum Na ratio (MESO) and United Kingdom MELD (UKELD) score, integrate serum sodium and creatinine into mortality prediction models, but their utility in the ACLF population is still debated.

Given the systemic complexity of ACLF, biomarkers reflecting different pathophysiological components of the condition have been explored as potential adjuncts to clinical scores. For example, von Willebrand factor has been shown to reflect endothelial dysfunction and to correlate with both disease severity and short‑term prognosis.⁸ Neutrophil gelatinase–associated lipocalin (NGAL), a marker of tubular injury and systemic inflammation, has emerged as a candidate for early detection of AKI and mortality prediction in cirrhotic patients.^9,10 Prior studies have suggested its utility in differentiating hepatorenal syndrome from other AKI types and in predicting outcomes independently of serum creatinine.^4,9 However, data on the performance of NGAL in homogeneous ACLF cohorts remain limited. Ariza et al¹¹ evaluated urine NGAL in 716 cirrhotic patients (including 148 with ACLF), and found it independently predictive of ACLF and 28‑day transplant‑free mortality. Another recent study in an Indian ACLF cohort showed that fatty acid–binding protein 1 (FABP1) and cystatin C levels differed significantly between the patients who developed AKI and those who did not, suggesting their potential as early biomarkers.¹²

This study aimed to evaluate serum NGAL concentrations in patients with well‑characterized ACLF and end‑stage liver disease, and to assess its prognostic value in predicting survival and access to liver transplantation, as compared with established scoring systems, including CLIF‑C ACLF, SOFA, MELD, iMELD, UKELD, and MESO.

Patients and methods

This was a retrospective, observational, single‑center pilot study conducted at a tertiary liver transplant center. It included adult patients admitted with a diagnosis of ACLF. Their clinical, laboratory, and outcome data were extracted from electronic medical records. Serum NGAL concentrations were measured within 48 hours of ACLF development. For each patient, established prognostic scores were calculated using standard formulas, including MELD, MELD‑Na, iMELD, UKELD, MESO, SOFA, CLIF Organ Failure (CLIF‑OF), and CLIF‑C ACLF. Scoring was based on data collected during the first 48 hours of hospitalization to reflect the patient’s initial clinical status. Survival status and liver transplant data were recorded at discharge or death.

Written informed consent for the use of anonymized clinical data for research purposes was obtained from all patients upon hospital admission, in accordance with the institutional protocol. No prospective enrollment was performed. All patient data were anonymized prior to analysis to ensure confidentiality and integrity. The study protocol was approved by the Bioethics Committee of the Medical University of Warsaw (KB/55/2018), and conformed with the ethical guidelines of the 1975 Declaration of Helsinki (6th revision, 2008).

Results

A total of 43 adults with end‑stage liver disease who developed ACLF were included in the study. Serum NGAL concentrations were available for 41 patients. Baseline clinical and biochemical characteristics, as well as liver disease severity scores, are presented in Supplementary material, Tables S1 and S2. Among the cohort, 16 patients (37.2%) had ACLF grade 1, 11 (25.6%) had ACLF grade 2, and 16 (37.2%) had ACLF grade 3.

The median (IQR) serum NGAL concentration for the entire cohort was 586 (268–896) ng/ml. Serum NGAL concentrations did not differ significantly across ACLF severity grades. Median NGAL levels were 552 (338.5–981.5) ng/ml in ACLF grade 1, 507 (282.5–638) ng/ml in ACLF grade 2, and 614 (388–857) ng/ml in ACLF grade 3, with no significant differences observed (P = 0.77).

AKI was present in 33 of the 43 patients with ACLF; serum NGAL measurements were available in 31 of these patients. The remaining 10 patients did not develop AKI and all had available NGAL data. Median NGAL concentrations were nonsignifcantly higher in the AKI than the non‑AKI group (600 [334.5–938.5] vs 469 [263–603.5] ng/ml, respectively; P = 0.22).

When stratified by liver transplant status, the patients who underwent the procedure (n = 21) had lower NGAL concentrations (median, 430 [195–671] ng/ml) than those who did not undergo a transplant (median, 688.5 [535–983.25] ng/ml). Although this difference reached nominal significance (P = 0.03), it did not remain statistically robust given the sample size and multiple subgroup comparisons.

NGAL concentrations were also compared between the survivors (n = 24) and nonsurvivors (n = 19). Among the survivors, the median NGAL value was 497 (231–818) ng/ml, as compared with 625 (523–978) ng/ml in the nonsurvivors (P = 0.13).

NGAL did not significantly correlate with any liver‑specific prognostic scores, including MELD, MELD‑Na, iMELD, MESO, and UKELD, nor was it associated with CLIF‑derived scores, after adjustment for multiple testing (all FDR‑adjusted P >0.44). An association with SOFA (R = 0.32; P = 0.04) did not remain significant after the Benjamini–Hochberg correction (FDR‑adjusted P = 0.44). Full correlation data are provided in Table 1.

In contrast, several CLIF‑derived prognostic scores demonstrated clinically meaningful associations with mortality. The nonsurvivors had higher CLIF‑OF scores (median, 12 [12–14.5]) than the survivors (median, 10 [10–12]; P <⁠0.001). Similarly, CLIF‑C ACLF scores were markedly higher in the nonsurvivors, as compared with the survivors (median, 62 [54.5–64.5] vs 47 [43.75–55], respectively; P <⁠0.001). Conversely, traditional liver severity scores (MELD, iMELD, MELD‑Na, MESO, UKELD) did not differ significantly between the survivors and nonsurvivors (all P >0.24), indicating limited prognostic discrimination within this advanced ACLF cohort.

No significant differences in MELD‑based or CLIF‑based scores were observed between the transplanted and nontransplanted patients (all P >0.24), suggesting that decisions regarding eligibility for liver transplant in this cohort were based on broader clinical assessments rather than purely on numerical scoring.

Discussion

In this study, we evaluated the prognostic utility of serum NGAL concentration in a homogeneous cohort of patients with ACLF. Despite the high prevalence of organ dysfunction and short‑term mortality in our population, NGAL levels did not significantly differ across ACLF grades and were not independently associated with survival or access to orthotopic liver transplantation after correction for multiple comparisons. These findings diverge from earlier reports regarding the predictive value of NGAL. Previous studies suggested that NGAL levels, particularly urinary ones, could play an important role as a biomarker in 2 main areas: early identification and differentiation of AKI, as well as its response to treatment; and prediction of adverse events and mortality in patients with cirrhosis.^9,13,14 A significant correlation has been shown between serum NGAL levels and total bilirubin, international normalized ratio, and MELD score, as well as the risk of bacterial infections, hepatorenal syndrome, and 28‑day mortality. These findings were particularly significant when combined with dynamic changes in other parameters, such as creatinine or C‑reactive protein.^10,15,16 It has been suggested that NGAL, similarly to other emerging biomarkers, might be a general biomarker of multiorgan failure and inflammation rather than of any specific organ dysfunction.¹²

However, most of these studies involved heterogeneous cohorts including patients with decompensated cirrhosis but not necessarily fulfilling strict ACLF criteria. Many of them also included patients with an ACLF diagnosis based on the Asian Pacific Association for the Study of the Liver criteria, which differ significantly from the European Association for the Study of the Liver (EASL) CLIF‑C criteria.¹⁷ This could explain the differing findings. In addition, the timing and primary purpose of NGAL measurements varied substantially between previously published studies and our cohort. In several earlier reports, NGAL was measured either at the time of AKI diagnosis, on ICU admission, or perioperatively, most commonly to detect early renal injury or to predict AKI progression rather than mortality. Some studies incorporated repeated or dynamic measurements, whereas our design relied on a single baseline serum NGAL value obtained within 48 hours of ACLF onset. Moreover, only a minority of individuals included in previous cohorts met strict ACLF criteria, which limits the applicability of their prognostic results to this specific population. These methodological and temporal differences likely contribute to the discrepancy between the prognostic performance of NGAL reported in the literature and the limited predictive value observed in our ACLF cohort.

The systemic inflammatory response in ACLF is associated with complex, overlapping mechanisms of organ injury involving hypoperfusion, cytokine‑mediated damage, and mitochondrial dysfunction.¹ As a result, biomarkers such as NGAL, reflecting tubular stress, may lack specificity for mortality when isolated from other organ dysfunctions. Similar observations were recently reported in an Indian ACLF cohort, where NGAL and interleukin 18 did not predict 28‑day mortality or AKI, whereas composite or multiorgan‑based scores and other markers, such as cystatin C and FABP1, showed stronger prognostic performance.¹² This is further underscored by our finding that NGAL did not correlate with any of the CLIF‑derived mortality prediction scores.^6,7 In contrast, our analysis showed that scores reflecting multiorgan failure, particularly CLIF‑C ACLF, CLIF‑C ACLF 1- and 3‑month predictions, and SOFA, were strongly and significantly associated with patient survival. These findings are in line with large multicenter validations of the CANONIC (Chronic Liver Failure; Acute‑on‑Chronic Liver Failure in Cirrhosis) cohort and EASL guidelines, which emphasize the prognostic value of organ failure patterns over traditional liver‑specific indices, such as MELD or iMELD.⁷ Despite being widely used in transplant allocation, MELD‑based scores showed poor performance in our ACLF cohort, likely reflecting their limited capacity to capture systemic dysfunction.^18,19

Another important observation was the lack of significant differences in prognostic scores between the patients who underwent a transplant and those who did not. This may reflect nonclinical barriers to listing, including resource constraints, comorbidities, or perceived futility of transplantation in advanced ACLF.^5,20 While NGAL showed a nonsignificant trend toward lower levels in the patients who received a transplant, it was not predictive after multiple‑comparison correction, further limiting its utility in triage settings.

Taken together, our results highlight the challenge of applying single biomarkers in the context of a multifactorial syndrome such as ACLF, and support the primacy of CLIF‑derived composite scores in evaluating prognosis and guiding clinical decisions. Although NGAL retains potential utility in AKI subtyping and early detection of tubular injury,^9,10 its role in survival prediction and transplant prioritization remains uncertain in advanced ACLF. Future studies should explore biomarker panels, dynamic trends, and integration with clinical scores to enhance prognostication in this critically ill population.