Serum Hepatocyte Growth Factor Predicts Outcomes in Non‐Acetaminophen Drug‐Induced Acute Liver Failure

ABSTRACT

Background and Aims

Regenerative capacity and inflammatory responses are key determinants of outcome in acute liver failure (ALF). However, their prognostic significance in non‐acetaminophen drug‐induced liver injury with acute liver injury or ALF (non‐APAP‐DILI‐ALI/ALF) remains poorly defined. We aimed to identify biomarkers of regeneration and inflammation associated with clinical outcomes in this population.

Methods

We evaluated clinical laboratory parameters and circulating biomarkers in 79 patients with non‐APAP‐DILI‐ALI/ALF (20 with ALI, 59 with ALF) and 19 healthy donors. We used multivariable logistic regression to identify independent predictors of 21‐day transplant‐free survival among these patients. These predictors were used to create a composite model that was compared with the Model for End‐Stage Liver Disease (MELD) score.

Results

Markers reflecting hepatic synthetic dysfunction (TBIL, INR, MELD) were strongly associated with the outcomes, whereas aminotransferases were not. HGF levels were markedly elevated in non‐APAP‐DILI‐ALI/ALF patients versus healthy donors (5070 ± 4001 vs. 276 ± 581 pg/mL; p < 0.001), increased with injury severity, and were highest in patients who died or underwent transplantation (6421 ± 4359 vs. 4102 ± 3456 pg/mL; p = 0.009). Among inflammatory mediators, only interleukin‐6 showed a modest association with outcome. We developed an H‐Score based on the three variables independently associated with survival on multivariable analysis: HGF (odds ratio, 1.0002; p = 0.048), TBIL and INR. The H‐Score outperformed MELD overall (area under the curve, 0.819 vs. 0.713) and in ALF only patients (0.811 vs. 0.643; p = 0.040).

Conclusions

HGF provides independent prognostic value beyond standard measures in non‐APAP‐DILI‐ALF. The H‐Score may improve risk stratification over the MELD score and warrants validation in a prospective cohort.

Highlight

• Hepatocyte growth factor (HGF) plays a central role in liver regeneration, but its prognostic significance in non‐acetaminophen drug‐induced acute liver injury and acute liver failure (non‐APAP‐DILI‐ALI/ALF) remains poorly defined. We found that serum HGF levels increased with disease severity, were associated with poor clinical outcomes, and provided independent prognostic value. An HGF‐based score developed by combining HGF, total bilirubin and INR improved prediction of transplant‐free survival, with significantly better discriminative performance observed in patients with drug‐induced acute liver failure as compared with the MELD score.

• Patients with non‐APAP‐DILI‐ALI/ALF exhibited broad dysregulation of inflammatory cytokines and chemokines, reflecting substantial immune perturbation during disease progression. However, only IL‐6 showed a modest association with clinical outcome, and this association did not persist after multivariable adjustment, indicating limited prognostic value of circulating inflammatory mediators compared with HGF in non‐APAP‐DILI‐ALF.

Lay Summary

Better biomarkers are needed to accurately predict patient outcomes in non‐acetaminophen drug‐induced liver injury beyond current clinical measures. By analysing circulating growth factors and inflammatory signals in non‐APAP‐DILI‐ALI/ALF, we identified hepatocyte growth factor (HGF) as a key marker of disease severity and prognosis, with higher levels associated with poorer outcomes. Integrating HGF with standard laboratory measures improved prediction of outcomes compared to the MELD score with significant difference in the ALF subgroup which may help refine risk assessment and guide clinical decision‐making in acute liver failure.

Abbreviations

ALF

acute liver failure

ALI

acute liver injury

APAP

acetaminophen

CCL

C‐C Motif Chemokine Ligand 2

DILI

drug‐induced liver injury

EGF

epidermal growth factor

HGF

hepatocyte growth factor

IL

interleukin

IL‐2R

soluble interleukin‐2 receptor

IL1RA

interleukin‐1 Receptor Antagonist

IP10

interferon gamma‐induced protein 10

MCP1

monocyte chemoattractant protein‐1

MIG

monokine induced by gamma

OR

odds ratio

RANTES

Regulated on Activation, Normal T Cell Expressed and Secreted

1. Introduction

Although uncommon, non‐acetaminophen drug‐induced acute liver failure (non‐APAP‐DILI‐ALF) is a major cause of ALF. It is a life‐threatening syndrome characterized by severe hepatocyte injury, hepatic encephalopathy and coagulopathy [1, 2, 3, 4]. Unlike APAP‐ALF, which has predictable dose‐dependent mechanisms and an approved antidote, non‐APAP‐DILI‐ALF exhibits unpredictable, idiosyncratic toxicity, presenting significant challenges in clinical risk assessment, management and drug development. Moreover, non‐APAP‐DILI‐ALF carries a poorer prognosis than APAP‐ALF [5]. Besides ALF, some cases of non‐APAP‐DILI manifest as acute liver injury (DILI‐ALI), characterized by hepatocellular injury and coagulopathy but no hepatic encephalopathy. While most DILI‐ALI cases recover, a subset progresses to ALF, some of whom require liver transplantation for survival [2, 3, 6, 7, 8, 9, 10, 11]. Thus non‐APAP‐DILI‐ALF and ALI present significant challenges in diagnosis, clinical management and drug development compared to liver injury due to APAP.

Established biomarkers including serum alanine aminotransferase (ALT), aspartate aminotransferase (AST), total bilirubin (TBIL) and INR remain critical for diagnosis and prognosis [1, 2, 12]. While Model for End‐Stage Liver Disease (MELD)‐based scores provide valuable outcome prediction, their prediction accuracy could be improved [9, 13, 14]. Biomarkers like soluble CD163 and microRNA‐122 show potential for predicting clinical outcomes but lack sufficient validation, and no studies have specifically evaluated their prognostic value in non‐APAP‐DILI‐ALF [12, 15].

The liver's regenerative capacity is crucial for recovery from acute injury. Hepatocyte growth factor (HGF) is primarily produced by platelets and various mesenchymal cells across multiple tissues. Within the liver, HGF is produced by stellate cells, Kupffer cells and sinusoidal endothelial cells. In turn, a high proportion of HGF is normally rapidly cleared by the liver through c‐Met‐mediated endocytosis and cell‐surface binding to heparan sulfate proteoglycan molecules on liver cells, followed by intracellular degradation [16]. HGF has emerged as a promising biomarker of liver regeneration [17, 18]. In animal models, HGF and its receptor c‐Met have been identified as potent mitogenic drivers for liver regeneration and repair [19]. HGF functions as a primary mitogen for hepatocytes and endothelial cells, promoting proliferation, inhibiting apoptosis and exerting anti‐inflammatory effects [17, 20, 21, 22]. Serum HGF levels rise following liver injury and may be associated with outcomes after hepatectomy, DILI, and ALF [18, 23]. In a small cohort of 31 ALF patients, higher HGF levels were observed in non‐survivors compared with survivors [23]. Other mitogens for hepatocytes including epidermal growth factor (EGF), vascular endothelial growth factor (VEGF), and fibroblast growth factor 2 (FGF2) also contribute to liver regeneration [22, 24]. It has been shown that the EGF receptor is required for liver regeneration in a mouse partial hepatectomy model, and serum EGF levels could serve as a potential prognostic marker for ALF [25, 26].

Dysregulated inflammation is a hallmark of severe DILI and ALF. Patients may develop a systemic inflammatory response syndrome (SIRS) characterized by a surge of cytokines (e.g., interleukin‐6 [IL‐6]) and chemokines (e.g., CCL2) released by activated immune cells, including Kupffer cells and T cells [1, 27, 28]. These inflammatory mediators not only contribute to clinical manifestations of hepatic and extrahepatic injury but also influence the regenerative process. For example, IL‐6 has been shown to modulate liver regeneration in ALF, an effect partly due to its influence on HGF expression [29].

Although the roles of HGF and inflammatory mediators have been studied in APAP‐ALF and mixed‐aetiology ALF, their prognostic potential in non‐APAP‐DILI‐ALF remains largely unknown [18, 25, 26, 28, 29, 30]. The prognostic utility of these regenerative and inflammatory markers, particularly in relation to established clinical scores, has not been systematically evaluated in a dedicated non‐APAP DILI cohort.

In this study, we systematically assessed clinical laboratory biomarkers and their association with outcomes in non‐APAP‐DILI‐ALI and/or non‐APAP‐DILI‐ALF. We analysed serum levels of HGF, EGF and cytokines/chemokines in healthy donors and patients, examining their associations with clinical outcomes and their independent prognostic value relative to MELD to evaluate their potential as predictive markers to enhance clinical decision‐making [31, 32].

2. Patients and Methods

2.1. Non‐APAP‐DILI‐ALF Patients and Controls

Patient serum samples and corresponding clinical data were obtained from the deidentified Acute Liver Failure Study Group (ALFSG) biorepository and database [33]. The serum samples were collected at the time of the study enrollment following the ALFSG protocol. Enrollment in the ALFSG registry required acute hepatic illness of less than 26 weeks with an international normalized ratio (INR) ≥ 1.5 and hepatic encephalopathy of any grade.

Clinical outcomes at 21 days following enrollment were categorized as spontaneous transplant‐free survival, liver transplantation, or death. The study cohort included 79 patients (59 with acute liver failure [ALF] and 20 with ALI) attributed to 57 distinct drug categories, excluding acetaminophen‐induced cases [33]. Drug‐induced liver injury (DILI) causality was assessed according to U.S. DILI Network criteria, and only cases classified as probable, highly likely, or definite were included. Written informed consent was obtained from all participants, and the study protocol received approval from the Institutional Review Board of each participating center for the ALFSG registry. Detailed demographic and clinical characteristics are summarized in Table 1. Human serum samples from 19 healthy individuals were purchased from BioIVT.

TABLE 1.

Outcome of non‐APAP‐DILI‐ALI/ALF: spontaneous survival versus liver transplant or death.

Characteristic	Overall N = 79 a	Spontaneous survival N = 46 a	Liver transplant or death N = 33 a	p b
Age (years)	46.2 ± 17.7	45.7 ± 17.3	47.0 ± 18.5	0.7
Sex				0.6
Female	53 (67%)	32 (70%)	21 (64%)
Male	26 (33%)	14 (30%)	12 (36%)
Ethnicity				0.8
African American	17 (22%)	11 (24%)	6 (18%)
Asian	4 (5.1%)	3 (6.5%)	1 (3.0%)
Caucasian	52 (66%)	29 (63%)	23 (70%)
Other	6 (7.6%)	3 (6.5%)	3 (9.1%)
Biochemical types				0.4
Cholestatic	3 (3.8%)	1 (2.2%)	2 (6.1%)
Hepatocellular	60 (76%)	37 (80%)	23 (70%)
Mixed	12 (15%)	5 (11%)	7 (21%)
Unknown	4 (5.1%)	3 (6.5%)	1 (3.0%)
ALT(IU/L)	2492.7 ± 2616.9	2797.0 ± 2907.3	2041.5 ± 2081.6	0.23
AST(IU/L)	2744.7 ± 3120.3	3308.2 ± 3721.2	1928.6 ± 1702.2	0.067
INR	2.9 ± 1.7	2.6 ± 1.3	3.4 ± 2.1	0.061
Total Bilirubin (mg/dL)	17.4 ± 11.9	12.9 ± 9.2	24.2 ± 12.4	3.71E‐5
Creatinine (mg/dL)	1.8 ± 1.7	1.7 ± 1.8	1.8 ± 1.6	0.743
MELD score	28.6 ± 7.0	26.7 ± 6.5	31.3 ± 7.0	0.0055
ALI/ALF				0.005
non‐APAP‐DILI‐ALF	59 (75%)	29 (63%)	30 (91%)
non‐APAP‐DILI‐ALI	20 (25%)	17 (37%)	3 (9.1%)

^a Mean ± SD; n (%).

^b Continuous variables analysed by Mann–Whitney U test; Categorical variables analysed by Pearson's Chi‐squared test or Fisher's exact test.

2.2. Luminex

Concentration of growth factors, cytokines, and chemokines in serum samples from patients and healthy donors were measured in duplicate using a human 30‐plex kit (Thermo Fisher Scientific, Catalogue number: LHC6003M) as per the manufacturer's instructions. The 30‐plex panel include the following: growth factors: EGF, FGF‐basic, HGF, VEGF; Cytokines: G‐CSF, GM‐CSF, IFN‐α, IFN‐γ, IL‐1β, IL‐1RA, IL‐2, IL‐2R, IL‐4, IL‐5, IL‐6, IL‐7, IL‐8, IL‐10, IL‐12 (p40/p70), IL‐13, IL‐15, IL‐17, TNF‐α; Chemokines: Eotaxin(CCL11), IP‐10(CXCL10), MCP‐1(CCL2), MIG(CXCL9), MIP‐1α (CCL3), MIP‐1β(CCL4), RANTES(CCL5). Fluorescence was measured using a BioPlex 200 analyser (Bio‐Rad Laboratories, Hercules, CA, USA), and analyte concentrations were calculated using Bio‐Plex Manager software (version 6.2) with a five‐parameter logistic regression model.

2.3. Statistical Analysis

Thirty cytokines and chemokines were quantified in 79 patients and 19 healthy controls. Analytes with fewer than 5% undetectable levels were retained for analysis, resulting in 14 cytokines and chemokines meeting inclusion criteria. Values below the lower limit of detection were imputed using the minimum observed value for each analyte, while values exceeding the upper limit of detection were imputed using the maximum observed value.

Associations between clinical laboratory parameters, growth factors, cytokines and spontaneous transplant‐free survival were evaluated using univariate and multivariable logistic regression analyses. Statistical analyses of continuous and categorical variables among different groups were applied by Mann–Whitney U test or chi‐squared test performed using R software (version 4.5.1). A composite prognostic model (H‐Score) was developed by selecting variables from the multivariable analysis with p < 0.05 and weighting them according to their regression coefficients to estimate the probability of spontaneous survival.

3. Results

3.1. Clinical Laboratory Biomarkers and Clinical Outcomes and Severity in Non‐APAP‐DILI‐ALI/ALF

Among the entire non‐APAP‐DILI‐ALI/ALF cohort, 46 patients (58%) achieved spontaneous survival, while 33 patients (42%) either required liver transplantation or died (Table 1). The cohort had a mean age of 46.2 ± 17.7 years with female predominance (67%). The majority of patients were Caucasian (66%), followed by African American (22%), with smaller proportions being Asian (5.1%) and other ethnicities (7.6%). The most common biochemical pattern observed was hepatocellular (N = 60, 76%) (Table 1). Age, sex, ethnicity and biochemical injury pattern categorized by R‐values were not significantly associated with clinical outcomes or disease severity (Table 1 and Table S1).

Based on disease severity classification, 59 patients (75%) were classified as non‐APAP‐DILI‐ALF, and 20 patients (25%) were classified as the less severe non‐APAP‐DILI‐ALI (Table S1). The distribution of non‐APAP‐DILI‐ALI versus non‐APAP‐DILI‐ALF differed significantly between clinical outcome groups (p = 0.005). Among spontaneous survivors, 38% were classified as non‐APAP‐DILI‐ALI while 62% met criteria for ALF. Conversely, in the transplant/death group, only 9% had non‐APAP‐DILI‐ALI while 91% presented with non‐APAP‐DILI‐ALF, indicating that initial disease severity at the time of enrollment was strongly associated with clinical outcome (Table S1).

Additionally, creatinine levels were significantly higher in the non‐APAP‐DILI‐ALF group compared to the non‐APAP‐DILI‐ALI group (2.0 ± 1.9 vs. 1.0 ± 0.6 mg/dL, p = 0.005), as were bilirubin levels (19.2 ± 12.0 vs. 12.2 ± 10.2 mg/dL, p = 0.036). The MELD score was significantly higher in non‐APAP‐DILI‐ALF compared to non‐APAP‐DILI‐ALI patients (30.5 ± 6.7 vs. 23.2 ± 4.7, p < 0.001) (Table S1).

Aminotransferase levels were numerically higher in the spontaneous survival group compared to the death/transplant group (ALT: 2797.0 ± 2907.3 vs. 2041.5 ± 2081.6 IU/L, p = 0.23; AST: 3308.2 ± 3721.2 vs. 1928.6 ± 1702.2 IU/L, p = 0.067), although these differences did not reach statistical significance (Table 1). Similarly, no significant differences in ALT or AST were observed between non‐APAP‐DILI‐ALI and non‐APAP‐DILI‐ALF groups (Table S1).

In contrast, markers of synthetic dysfunction and disease severity differed significantly between outcome groups. Spontaneous survivors had substantially lower serum TBIL levels (12.9 ± 9.2 vs. 24.2 ± 12.4 mg/dL, p < 0.001) and lower INR values (2.6 ± 1.3 vs. 3.4 ± 2.1, p = 0.061) than those who died or underwent liver transplantation. Correspondingly, the MELD score was significantly lower in the spontaneous survival group (26.7 ± 6.5 vs. 31.3 ± 7.0, p = 0.0055). Notably, creatinine levels, a key component of the MELD score, did not differ between outcome groups (Table 1). Similarly, MELD score and TBIL were significantly higher in non‐APAP‐DILI‐ALF than in non‐APAP‐DILI‐ALI patients. However, creatinine levels were also significantly higher in non‐APAP‐DILI‐ALF compared with non‐APAP‐DILI‐ALI (Table S1).

In summary, lower TBIL along with lower MELD scores, were associated with better outcomes in non‐APAP‐DILI‐ALF. Lower TBIL and creatinine levels, as well as lower MELD scores, were associated with less severe disease in non‐APAP‐DILI‐ALI compared to non‐APAP‐DILI‐ALF. The majority of patients with non‐APAP‐DILI‐ALI achieved 21‐day spontaneous survival, whereas spontaneous recovery was less common among those with non‐APAP‐DILI‐ALF.

3.2. HGF levels, but Not EGF Levels, Increase With Disease Severity and Poor Clinical Outcomes

Given the central role of hepatocyte growth factor (HGF) in liver regeneration following ALF and previous reports of elevated HGF levels in ALF patients compared to healthy donors, we first examined serum HGF concentrations across the spectrum of non‐APAP‐DILI‐ALI/ALF [26]. HGF levels were markedly elevated in all patients compared with healthy controls (non‐APAP‐DILI‐ALI/ALF: 5070.1 ± 4001.14 vs. Healthy: 275.7 ± 581.4 pg/mL; p < 0.001, Figure 1A).

FIGURE 1.

Serum hepatocyte growth factor (HGF), but not epidermal growth factor (EGF), is associated with disease severity and clinical outcome in non‐APAP‐DILI‐ALI/ALF. (A) Serum HGF concentrations in patients with combined non–acetaminophen drug–induced acute liver injury and acute liver failure (non‐APAP‐DILI‐ALI/ALF) compared with healthy controls. (B) Serum HGF concentrations in healthy controls, non‐APAP‐DILI‐ALI and non‐APAP‐DILI‐ALF patients. (C) Serum HGF concentrations according to clinical outcome (spontaneous survival vs. death or liver transplantation) in combined ALI/ALF patients. (D) Serum EGF concentrations in combined non‐APAP‐DILI‐ALI/ALF patients compared with healthy controls. (E) Serum EGF concentrations in healthy controls, non‐APAP‐DILI‐ALI and non‐APAP‐DILI‐ALF patients. (F) Serum EGF concentrations according to clinical outcome in combined ALI/ALF patients. Data are presented as mean ± SD. Statistical comparisons were performed using the Mann–Whitney U test. p < 0.05; p < 0.01; p < 0.001.

Disease stage stratification revealed a progressive elevation pattern. Non‐APAP‐DILI‐ALI patients demonstrated significantly higher HGF concentrations than healthy donors (non‐APAP‐DILI‐ALI: 3750.6 ± 3805.1 pg/mL; p < 0.001, Figure 1B), while non‐APAP‐DILI‐ALF patients showed further significant increases relative to non‐APAP‐DILI‐ALI patients (non‐APAP‐DILI‐ALF: 5517.76 ± 3997.7 pg/mL; p = 0.0317, Figure 1B). Importantly, HGF levels were higher in poor outcome (death or transplant) patients compared with spontaneous survivors (death/transplant: 6420.8 ± 4358.7 pg/mL vs. survivors: 4101.6 ± 3455.8 pg/mL; p = 0.009, Figure 1C).

In addition to HGF, other growth factor pathways including those mediated by epidermal growth factor (EGF), vascular endothelial growth factor (VEGF) and basic fibroblast growth factor (FGF2) are known to contribute to liver regeneration [34, 35, 36]. In contrast to HGF, however, it has been reported that EGF is decreased in patients with APAP‐ALF and mixed‐aetiology ALF compared with healthy donors [26]. Consistent with these findings, our data showed that serum EGF levels were significantly reduced in non‐APAP‐DILI‐ALI/ALF patients compared with healthy controls (non‐APAP‐DILI‐ALI/ALF: 54.62 ± 86.54 pg/mL vs. healthy controls: 150.71 ± 416.2 pg/mL; p < 0.001, Figure 1D). However, disease stage stratification revealed no significant differences in EGF levels between non‐APAP‐DILI‐ALI and non‐APAP‐DILI‐ALF patients (ALF: 59.75 ± 97.92 pg/mL vs. ALI: 39.7 ± 35.6 pg/mL; p = 0.378, Figure 1E), and EGF concentrations did not correlate with clinical outcomes within the non‐APAP‐DILI‐ALI/ALF cohort (death/transplant: 48.38 ± 45.27 pg/mL vs. survivors: 59.20 ± 107.54 pg/mL; p = 0.705, Figure 1F). Serum VEGF and FGF2 levels were below detection limits in most samples and were excluded from further analysis (Table S2).

Our data demonstrate that HGF levels progressively increase with disease severity, from healthy donors to non‐APAP‐DILI‐ALI to non‐APAP‐DILI‐ALF and reach the highest concentrations in patients with the poorest outcomes, indicating that elevated HGF levels are associated with worse prognosis. By contrast, while mean EGF levels are significantly reduced in non‐APAP‐DILI‐ALI/ALF patients compared with healthy donors, they do not vary significantly across disease stages or correlate with outcomes.

3.3. Dysregulation of Levels of Cytokines and Chemokines in Non‐APAP‐DILI‐ALI and/or Non‐APAP‐DILI‐ALI/ALF Patients

Patients with non‐APAP‐DILI‐ALI/ALF demonstrated broad elevations in both proinflammatory and anti‐inflammatory mediators, similar to findings in APAP‐ALF and mixed‐aetiology ALF populations [27, 28, 29, 37]. Analysis of cytokine profiles revealed significant elevations in multiple inflammatory markers compared to healthy controls (Figure 2A). Levels of the proinflammatory cytokines IL‐6 and soluble IL‐2R, as well as the chemokines CCL2, CXCL9 and CXCL10, were significantly higher in non‐APAP‐DILI‐ALI/ALF patients compared with healthy controls (all p < 0.001, Figure 2A,D). The anti‐inflammatory cytokine IL‐10 was also significantly elevated in non‐APAP‐DILI‐ALI/ALF patients (2847.03 ± 24400.92 pg/mL vs. healthy controls: 10.58 ± 116.28 pg/mL; p < 0.001, Figure 2A), suggesting concurrent activation of compensatory anti‐inflammatory pathways. In contrast, IL‐12, IL‐1RA, IL‐8, CCL4 and CCL11 showed no significant differences between patients and healthy donors, while CCL5 levels were reduced in non‐APAP‐DILI‐ALI/ALF patients compared to controls (Figure 2A,D).

FIGURE 2.

Cytokine and chemokine profiles in non‐APAP‐DILI‐ALI/ALF. (A) Serum cytokine levels (IL‐6, IL‐10, IL‐12, IL‐1RA, IL‐2R and IL‐8) in healthy controls (n = 19) compared with the combined non‐APAP‐DILI‐ALI/ALF (n = 79). (B) Cytokine levels stratified by disease severity: healthy controls, non‐APAP‐DILI‐ALI and non‐APAP‐DILI‐ALF patients. (C) Cytokine levels according to clinical outcome (spontaneous survival vs. death or liver transplantation) in the combined ALI/ALF cohort. (D) Serum chemokine levels (CCL2, CCL4, CCL5, CCL11, CXCL9 and CXCL10) in healthy controls compared with the combined non‐APAP‐DILI‐ALI/ALF cohort. (E) Chemokine levels stratified by disease severity: healthy controls, non‐APAP‐DILI‐ALI and non‐APAP‐DILI‐ALF patients. (F) Chemokine levels according to clinical outcome in the combined ALI/ALF cohort. Data are shown as individual values with median indicated. Statistical significance is denoted by asterisks: p < 0.05; p < 0.01; p < 0.001.

Stratification by disease stage revealed distinct cytokine expression patterns (Figure 2B,E). Non‐APAP‐DILI‐ALI patients demonstrated significantly higher levels of IL‐6, IL‐2R, CXCL9 and CXCL10 compared with healthy donors. CCL5 levels were consistently decreased in both ALI and ALF patients compared to healthy controls (Figure 2E). Non‐APAP‐DILI‐ALF patients exhibited further elevations in IL‐6, IL‐10, IL‐2R, IL‐8 and CCL4 relative to non‐APAP‐DILI‐ALI patients. However, CCL5 and CXCL9 levels were further decreased in non‐APAP‐DILI‐ALF patients compared to non‐APAP‐DILI‐ALI patients. These data suggest that progression from non‐APAP‐DILI‐ALI to non‐APAP‐DILI‐ALF is associated with broad dysregulation across multiple inflammatory pathways, with levels of some cytokines and chemokines increasing while others decrease.

Analysis of clinical outcomes within the non‐APAP‐DILI‐ALI/ALF cohort revealed limited associations between inflammatory markers and patient survival (Figure 2C,F). Among all measured cytokines and chemokines, only IL‐6 showed a significant association with clinical outcome. Patients who died or required liver transplantation had significantly higher IL‐6 levels compared with spontaneous survivors (360.4 ± 697.4 pg/mL vs. 310.9 ± 708.25 pg/mL; p = 0.023, Figure 2C). Other inflammatory markers, including IL‐10, IL‐2R, IL‐8 and the chemokines CCL2, CCL4, CCL5, CXCL9 and CXCL10, showed no significant differences between survivors and non‐survivors despite their elevation in disease stages (Figure 2C,F).

The paradoxical reduction in certain chemokines (CCL5, CXCL9) with disease severity alongside elevation of others suggests dysregulated rather than simply amplified inflammatory responses. While multiple cytokines and chemokines show altered levels during disease progression, only IL‐6 levels showed a signficant association with clinical outcomes, reinforcing its unique dual role as a hepatocyte mitogen and immune modulator in non‐APAP‐DILI‐ALF; although the magnitude of increase is marginal.

3.4. HGF, but Not Cytokines and Chemokines, Demonstrated Independent Prognostic Value for Outcome of Non‐APAP‐DILI‐ALI/ALF Patients

Given that levels of HGF and IL‐6 were associated with patient outcomes, we further investigated the prognostic association of HGF and IL‐6 with clinical outcome in the non‐APAP‐DILI‐ALI/ALF cohort using logistic regression analysis. In univariate analysis, TBIL (odds ratio [OR] 1.019, 95% confidence interval [CI] 1.010–1.028, p = 3.71 × 10⁻⁵), HGF (OR 1.0003, 95% CI 1.0001–1.0006, p = 0.010), INR (OR 1.068, 95% CI 0.998–1.141, p = 0.061) and AST (OR 0.9996, 95% CI 0.9999–1.0003, p = 0.067) showed associations with outcome, reaching at least borderline significance (p < 0.1). ALT (p = 0.230), creatinine (p = 0.743) and IL‐6 (p = 0.773) were not significantly associated with outcomes (Table 2).

TABLE 2.

Logistic regression analysis for the prognostic association of traditional liver biomarkers, HGF and IL6 with clinical outcome in the non‐APAP‐ALI/ALF cohort.

	Univariate analysis		Multivariate analysis
	Odds ratio (95% CI)	p	Odds ratio (95% CI)	p
ALT	1.0 (1.0–1.0)	0.230
AST	0.9996 (0.999–1.0003)	0.067	1.0 (1.0–1.0)	0.826
INR	1.068 (0.998–1.141)	0.0610	1.079 (1.019–1.142)	0.011
Total bilirubin	1.019 (1.010–1.028)	3.71E‐5	1.020 (1.010–1.029)	9.57E‐5
Creatinine	1.011 (0.945–1.083)	0.743
HGF	1.0003 (1.0001–1.0006)	0.0104	1.0002 (1.0001–1.0004)	0.0477
IL6	1.0 (0.9998–1.0002)	0.773

Note: Only the factors that reached a borderline significant in univariate analysis (p < 0.1) were included for multivariate analysis.

The subsequent multivariate analysis confirmed that only TBIL (OR 1.020, 95% CI 1.010–1.029, p = 9.57 × 10⁻⁵), INR (OR 1.079, 95% CI 1.019–1.142, p = 0.011) and HGF (OR 1.0002, 95% CI 1.0001–1.0004, p = 0.048) remained significantly associated with clinical outcome, while AST lost statistical significance (p = 0.826) (Table 2). This demonstrates that HGF acts as an independent predictor of survival outcomes, separate from TBIL and INR.

Given the independent prognostic value of these three biomarkers, we developed a novel composite score, the H‐Score, incorporating HGF, TBIL and INR to predict the likelihood of spontaneous transplant‐free survival. The H‐Score was calculated using the following formula: H‐Score = (0.01945 × TBIL) + (0.07591 × INR) + (0.00002494 × HGF). To evaluate the potential clinical utility of this HGF‐based model, we performed receiver operating characteristic (ROC) curve analysis comparing the H‐Score's discriminative performance with the established MELD score (Figure 3).

FIGURE 3.

Receiver operating characteristic (ROC) curve analysis comparing the H‐Score and MELD score for prediction of spontaneous transplant‐free survival. The H‐Score incorporates hepatocyte growth factor (HGF), total bilirubin (TBIL) and international normalized ratio (INR) using the formula: H‐Score = (0.01945 × TBIL) + (0.07591 × INR) + (0.00002494 × HGF). (A) ROC curves comparing discriminative performance of the H‐Score and MELD score in the combined non‐APAP‐DILI‐ALI/ALF cohort (n = 79). (B) ROC curves comparing H‐Score and MELD score performance in the non‐APAP‐DILI‐ALF subgroup (n = 59). Area under the curve (AUC) values and statistical comparisons between models are shown in each panel.

In the combined non‐APAP‐DILI‐ALI/ALF cohort, the H‐Score demonstrated superior performance with an AUC of 0.819 compared with the MELD score's AUC of 0.713, though this difference did not reach statistical significance (p = 0.082) (Figure 3A). However, in the non‐APAP‐DILI‐ALF subgroup (n = 59), the H‐Score had a significant higher AUC over the MELD score (AUC 0.811 vs. 0.643, p = 0.040) (Figure 3B), representing an improvement of 0.168 in discriminative ability. These results demonstrate that the HGF‐based H‐Score model offers significantly enhanced prognostic capability compared to the standard MELD Score in non‐APAP‐DILI‐ALF patients.

4. Discussion

In this cohort of patients with non‐acetaminophen drug‐induced acute liver injury and acute liver failure (non‐APAP‐DILI‐ALI/ALF), markers of hepatic synthetic dysfunction emerged as the dominant determinants of disease severity and clinical outcome. Total bilirubin and MELD score robustly differentiated spontaneous transplant‐free survival from death or liver transplantation, whereas aminotransferase levels were not predictive. These findings are consistent with previous studies indicating that hepatic synthetic and clearance functions, rather than the extent of aminotransferase elevation, are predictors of poor outcome in acute liver failure [1, 2].

Among nontraditional biomarkers, hepatocyte growth factor (HGF) was the most clinically informative indicator of prognosis. Circulating HGF levels were markedly elevated compared with healthy controls, increased progressively from ALI to ALF, and were highest in patients who died or required liver transplantation. Our finding is consistent with reports that elevated HGF levels are also seen in poor‐outcome patients with acetaminophen‐induced and mixed‐aetiology ALF, suggesting that this circulating factor has prognostic value across different forms of acute liver failure [18, 23, 38]. Although recombinant HGF has been shown to enhance liver regeneration in experimental models and early clinical studies, our findings suggest that augmenting HGF levels alone is unlikely to confer clinical benefit in advanced disease [17, 19, 30].

Importantly, lower HGF remained independently associated with spontaneous transplant‐free survival after adjustment for bilirubin and INR, indicating that assessment of regenerative stress provides prognostic information beyond conventional measures of hepatic synthetic function. Therefore, we developed a prognostic model using a composite H‐Score that incorporates HGF, total bilirubin and INR. The H‐Score demonstrated a higher discriminative performance compared with MELD, particularly in patients with established ALF, where accurate prognostication is most critical. If the H‐Score is validated as a predictive tool for non‐APAP DILI outcomes in further studies, this improvement may have a practical prognostic role, given the known limitations of MELD in acute liver failure and DILI‐specific populations [9, 13, 14].

Inflammatory profiling demonstrated broad and complex immune dysregulation, with alterations in multiple cytokines and chemokines. However, only interleukin‐6 (IL‐6) showed a modest association with clinical outcome, and this association did not persist after multivariable adjustment. Previous experimental studies have shown that IL‐6 can augment HGF‐mediated liver regeneration. However, IL‐6 was not independently associated with outcome in our cohort. One possible explanation is that the increases in circulating IL‐6 levels were merely less pronounced than that observed for HGF, potentially contributing to their limited prognostic value. Other explanations include differences between experimental models and human disease, patient heterogeneity and variability in the timing of sample collection [27, 29, 39].

Our study has several limitations. First, biomarker measurements were obtained at a single timepoint at enrollment. HGF has a very short circulating half‐life (approximately 3–5 min) and is rapidly cleared by the healthy liver [16]. Despite this, we observed sustained HGF elevation in non‐APAP‐DILI‐ALI/ALF patients (5070 ± 4001 pg/mL vs. 276 ± 581 pg/mL in controls), likely reflecting massive endogenous release coupled with profoundly impaired hepatic clearance secondary to severe necrosis. The paradoxical association between higher HGF levels and worse outcomes suggests that circulating HGF may serve as a biomarker of failed hepatic clearance rather than simply reflecting growth factor availability. Nevertheless, single‐timepoint measurements limit our ability to assess whether dynamic profiles of rising and falling HGF levels provide superior prognostic information. Future studies with serial sampling are needed to define the temporal kinetics of HGF and other regenerative biomarkers.

Second, the ALFSG database does not systematically capture plasma product administration relative to sample collection, precluding adjustment for this potential confounder. However, several observations suggest our findings reflect disease biology rather than transfusion artefact: the magnitude of HGF elevation (> 18‐fold) substantially exceeds expected dilutional effects; HGF demonstrated a clear biological gradient across disease stages and outcomes; HGF retained independent prognostic value after adjusting for INR, which is directly affected by plasma infusion; and given HGF's short half‐life, transient elevations from exogenous plasma would rapidly dissipate. Nevertheless, prospective validation studies should document transfusion data to enable sensitivity analyses, with biomarker sampling ideally occurring prior to plasma infusion when feasible.

Third, the relatively modest sample size may limit power for subgroup analyses, and use of a specialized ALF registry may limit generalizability to broader DILI populations [33]. Despite these limitations, our findings identify HGF as a robust independent prognostic biomarker and demonstrate that the H‐Score improves outcome prediction beyond the MELD score alone. The H‐Score should be validated in a prospective, independent cohort before being considered for routine clinical application.

In conclusion, our study identify HGF as an independent prognostic biomarker in non‐APAP‐DILI‐ALF. By integrating the regenerative biomarker with established measures of hepatic synthetic function, use of the H‐Score improves outcome prediction beyond MELD alone and represents a promising tool for risk stratification and transplant decision‐making. Nonetheless, our findings must be prospectively validated in independent cohorts. Future studies should also explore longitudinal biomarker assessment and therapeutic strategies for restoring hepatic responsiveness to circulating regenerative factors in this high‐risk population [12].

Author Contributions

Tao Wang and Minjun Chen conceived the concept, designed the study and drafted the manuscript; Silvia M. Bacot and Tao Wang conducted the Luminex experiment and compiled raw data for statistical analysis; Proceedings of the National Academy of Sciences. All authors discussed and provided comprehensive editorial input in the writing of this manuscript.

Funding

Funding for this work was provided by the U.S. Food and Drug Administration.

Disclosure

Disclaimer: This article reflects the views of the authors and does not necessarily reflect those of the U.S. Food and Drug Administration. Any mention of commercial products is for clarification only and is not intended as approval, endorsement, or recommendation.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Table S1: Clinical and demographic factors for non‐APAP‐DILI‐ALI/ALF: ALF versus ALI.

Table S2: The assay measurements of the growth factors, cytokine assays and chemokines on serum samples from patients and healthy donors.

Wang T., Bacot S. M., Patterson K., et al., “Serum Hepatocyte Growth Factor Predicts Outcomes in Non‐Acetaminophen Drug‐Induced Acute Liver Failure,” Liver International 46, no. 11 (2026): e70895, 10.1111/liv.70895.

Data Availability Statement

The data that support the findings of this study are available in the Supporting Information of this article.