Case Control Study Open Access
Copyright ©The Author(s) 2025. Published by Baishideng Publishing Group Inc. All rights reserved.
World J Gastroenterol. Aug 14, 2025; 31(30): 110401
Published online Aug 14, 2025. doi: 10.3748/wjg.v31.i30.110401
Serum metabolomic characteristics and their predictive value for ninety-day prognosis in patients with acute-on-chronic liver failure
Yan Liu, Jing-Jing Zhang, Jian-Dong Zhang, Department of Clinical Laboratory, The Hebei Medical University Third Hospital, Shijiazhuang 050051, Hebei Province, China
Ying Xiao, Ze-Qiang Qi, Ya-Dong Wang, Department of Infectious Diseases, The Hebei Medical University Third Hospital, Shijiazhuang 050000, Hebei Province, China
Lian-Feng Ai, Technology Center of Shijiazhuang Customs, Shijiazhuang 050000, Hebei Province, China
Lei Dong, Key Laboratory of Molecular Medicine and Biological Diagnosis and Treatment, Aerospace Center Hospital, School of Life Science, Beijing Institute of Technology, Beijing 100000, China
ORCID number: Ya-Dong Wang (0000-0003-0140-0674).
Co-first authors: Yan Liu and Ying Xiao.
Author contributions: Liu Y and Xiao Y conceived the idea for the study and were responsible for metabolomics testing, data extraction, statistical analysis, and manuscript drafting, they contributed equally to this article, they are the co-first authors of this manuscript; Ai LF supervised the study and revised the manuscript for important intellectual content; Zhang JJ and Zhang JD assisted in collecting clinical data and serum specimens; Qi ZQ assisted with data analysis and statistical application; Dong L and Wang YD performed methodological assessment and revised the manuscript for important intellectual content; and all authors have read and approve the final manuscript.
Supported by the Hebei Natural Science Foundation, No. H2023206042; and Medical Science Research Project of Hebei, No. 20230670.
Institutional review board statement: This study was approved by the Medical Ethics Committee of the Hebei Medical University Third Hospital, approval No. W2023-043-1.
Informed consent statement: All procedures complied with the Declaration of Helsinki. Informed consent was obtained and documented from all participants before enrollment.
Conflict-of-interest statement: All the authors report no relevant conflicts of interest for this article.
STROBE statement: The authors have read the STROBE Statement-checklist of items, and the manuscript was prepared and revised according to the STROBE Statement-checklist of items.
Data sharing statement: The datasets used and/or analyzed during the current study are available from the first author on reasonable request (Yan Liu, Department of Clinical Laboratory, The Hebei Medical University Third Hospital, Shijiazhuang 050051, China. E-mail: 396758927@qq.com).
Open Access: This article is an open-access article that was selected by an in-house editor and fully peer-reviewed by external reviewers. It is distributed in accordance with the Creative Commons Attribution NonCommercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited and the use is non-commercial. See: https://creativecommons.org/Licenses/by-nc/4.0/
Corresponding author: Ya-Dong Wang, Department of Infectious Diseases, The Hebei Medical University Third Hospital, No. 275 Zhongshan West Road, Qiaoxi District, Shijiazhuang 050000, Hebei Province, China. wangyadong@hebmu.edu.cn
Received: June 7, 2025
Revised: June 26, 2025
Accepted: July 21, 2025
Published online: August 14, 2025
Processing time: 62 Days and 21.5 Hours

Abstract
BACKGROUND

Acute-on-chronic liver failure (ACLF) is characterized by severe metabolic disturbances; however, the specific metabolomic features and their predictive value on 90-day prognosis remain unclear.

AIM

To identify serum metabolomic changes in patients with ACLF with different prognoses to support clinical prediction of outcomes and treatment decisions.

METHODS

This non-interventional, observational case-control study enrolled 58 patients with ACLF. Fasting venous blood samples were analyzed using targeted metabolomics. Univariate and multivariate statistical analyses identified differential metabolites among 18 amino acids, 11 fatty acids, 5 gut microbiota-related metabolites, and 4 bile acid metabolites. Binary logistic regression identified independent mortality risk factors, visualized via forest plots and receiver operating characteristic curves.

RESULTS

Significant differences (P < 0.05) were observed between the death and survival groups in baseline age, model for end-stage liver disease score, model for end-stage liver disease with sodium, neutrophil-to-lymphocyte ratio (NLR), total bilirubin, serum creatinine, blood urea nitrogen, and platelet count. Metabolites, including L-carnitine, creatinine, alanine, arginine (Arg), proline, choline, and oleic acid, also showed statistically significant differences between the groups. Multivariate analysis identified age, NLR, and Arg as independent risk factors for 90-day mortality in patients with ACLF. The predictive model, age-NLR-Arg = -15.481 + 0.135 × age + 0.156 × NLR + 0.203 × Arg, with a cutoff of 0.759, achieved an area under the receiver operating characteristic curve of 0.945 with sensitivity of 84.0% and specificity of 87.9%.

CONCLUSION

The age-NLR-Arg model demonstrates a strong predictive value for 90-day mortality risk in patients with ACLF.

Key Words: Acute-on-chronic liver failure; Metabolomics; Artificial liver blood purification system; Modeling; Prognosis

Core Tip: Acute-on-chronic liver failure (ACLF) is a rapidly progressing condition with high mortality and limited treatment options. Traditional prognostic models fail to capture its dynamic metabolic disturbances. This study identifies seven key metabolites linked to 90-day ACLF prognosis, with Arginine as an independent risk factor. Age- neutrophil-to-lymphocyte ratio-arginine model expressed perfect predictive efficiency for 90-day prognosis of patients with ACLF. In addition, artificial liver blood purification system treatment modulated alanine and L-carnitine, reducing inflammation and promoting liver regeneration. These findings highlight the potential of metabolomics to enhance ACLF prognosis, offering a more precise approach for clinical assessment and management.



INTRODUCTION

Acute-on-chronic liver failure (ACLF) is a critical clinical syndrome characterized by rapid deterioration in patients with relatively stable chronic liver disease due to acute insults, manifesting as jaundice, coagulopathy, hepatorenal syndrome, hepatic encephalopathy, and ascites[1]. ACLF is a severe, rapidly progressing condition with no effective pharmacological treatments, resulting in a 28/90-day mortality rate of 20%-80%[2,3], considerably impacting patients’ quality of life. Early and accurate prediction of ACLF prognosis is crucial for developing evidence-based, individualized treatments, optimizing resource allocation, and improving clinical outcomes. Metabolomics, which uses advanced analytical platforms to identify and quantify small-molecule metabolites in biological samples, facilitates early biomarker discovery and guides clinical management, making it the omics discipline closest to the biological phenotype[4]. The liver regulates the metabolism of carbohydrates, lipids, proteins, and bile acids, and severe metabolic disruption is a key mechanism in the pathogenesis of ACLF[5]. Our previous study found that as Child-Turcotte-Pugh grades increase, 3-phosphopyruvate levels rise, peaking at Child-Turcotte-Pugh-B before declining. This indicates a shift from glucose or pyruvate oxidation to lipid oxidation with disease progression driving cellular growth reliance from glycolysis to lipid and amino acid metabolism[6]. Therefore, metabolomics is valuable for understanding liver metabolic functions and supporting early diagnosis and prognosis of liver diseases.

Current clinical practice relies on traditional indicators and scoring systems, such as the model for end-stage liver disease (MELD), chronic liver failure-sequential organ failure assessment, and Chinese group on the study of severe hepatitis B models to assess ACLF prognosis. Given the severe immune, inflammatory, and metabolic disturbances in ACLF, integrating metabolomic characteristics can more precisely reflect dynamic disease changes, enhancing the sensitivity and specificity of prognostic predictions. This study aimed to develop a more scientific and accurate prognostic prediction model to guide the clinical management of patients with ACLF by assessing baseline-serum metabolite differences in patients with ACLF with varying 90-day outcomes and their correlation with the efficacy of artificial liver blood purification system (ALBPS) treatment using tandem mass spectrometry.

MATERIALS AND METHODS
Case selection and inclusion/exclusion criteria

This single-center, prospective, non-interventional, observational case-control study included 58 patients with ACLF hospitalized at the Hebei Medical University Third Hospital between January 2023 and December 2024. ACLF diagnosis adhered to the 2019 Asia-Pacific Association for the Study of the Liver criteria[1]. All patients received standardized medical treatment (SMT), including etiology control, hepatoprotective therapy, and complication prevention. Treatment was individualized based on etiology, recovery status, tolerance, and adverse reactions, with adjustments per guideline. Of the 58 patients, 23 received ALBPS in addition to SMT, with treatment modalities and efficacy assessed per the 2022 Chinese Medical Association Hepatology Branch consensus[7]. Exclusion criteria included: (1) Presence of liver or other extrahepatic solid organ malignancies; (2) Coexisting metabolic disorders (e.g., metabolic syndrome, hemochromatosis, alpha-1 antitrypsin deficiency, glycogen storage disease, tyrosinemia, gout, and phenylketonuria); and (3) Pregnant or lactating women. This study was approved by the Medical Ethics Committee of The Hebei Medical University Third Hospital, approval No. W2023-043-1, and all procedures complied with the Declaration of Helsinki. Informed consent was obtained and documented from all participants before enrollment.

Clinical data collection

Clinical data included age, sex, underlying liver disease etiology, treatment regimen, and 90-day prognostic outcomes. Laboratory indices at initial ACLF diagnosis were collected, including: (1) Hematology: White blood cell count, neutrophils, lymphocytes, hemoglobin, platelets (PLT); (2) Serum biochemistry: Albumin, alanine transaminase, aspartate transaminase, alkaline phosphatase, γ-glutamyl transpeptidase, total bilirubin (TBil), serum creatinine (Scr), serum sodium (Na+); and (3) Coagulation: Activated partial thromboplastin time, international normalized ratio (INR).

MELD scores, MELD with Na (MELD-Na) scores, and neutrophil-to-lymphocyte ratio (NLR) were calculated based on these indices. The MELD score formula was: 9.6 × ln [Scr (mg/dL)] + 3.8 × ln [TBil (mg/dL)] + 11.2 × ln (INR) + 6.4 × etiology (0 for cholestatic or alcoholic cirrhosis, 1 for other causes). The MELD-Na score formula was: MELD score + 1.59 × [135 - Na+ (mmol/L)], with Na+ capped at 135 mmol/L if > 135 mmol/L, floored at 120 mmol/L if < 120 mmol/L, and otherwise calculated as the actual value. The NLR was: Neutrophil count/Lymphocyte count.

Metabolomics detection methods

Sample collection: To eliminate dietary interference, 2 mL of peripheral venous blood was collected after 12 hours fast via elbow venipuncture into tubes with inert separation gel and coagulant. Samples were centrifuged at 4000 rpm for 5 minutes at 20-25 °C within 8 hours, and the supernatant was collected and stored at -80 °C until analysis. Samples were prepared through extraction and reconstitution for targeted metabolomics analysis of serum fatty acids, bile acids, amino acids, and trimethylamine N-oxide (TMAO) levels.

Instrumentation and testing conditions: Fatty and bile acids were analyzed using the Xevo TQ-S Cronos LC-MS (Waters Corporation, United States); amino acids and TMAO were analyzed using the ACQUITY UPLC-Xevo TQ-S tandem quadrupole mass spectrometer (Waters Corporation, United States). Sample pretreatment and analysis conditions are detailed in Supplementary material 1.

Data analysis

Metabolomics data analysis: Metabolomics data were analyzed using SPSS 26.0 (International Business Machines Corporation, Chicago, IL, United States) for t tests and P value calculations. Fold change (FC) was used to assess differential metabolite expression between the groups, and volcano plots were generated using the OmicStudio online tool (https://www.omicstudio.cn/tool). Multivariate statistical analysis was performed using the SIMCA 14.1 software. Principal component (PC) analysis was used to assess overall trends and outliers, visualized via PC score plots, with each point representing an independent sample. Orthogonal partial least squares-discriminant analysis was applied to identify group-specific differences, visualized via variation coefficient plots. Variable importance in projection (VIP) scores were calculated, with VIP > 1 defining differential metabolites. Model reliability was evaluated using R2X, R2Y, and Q2Y values.

Statistical analysis

Data were analyzed using SPSS 26.0. Normally distributed continuous data were expressed as the mean ± SD deviation and compared using two-sample t tests; non-normally distributed data were presented as the median (quartiles) and compared using the Mann–Whitney U test. Categorical data were reported as counts (percentages), n (%) and analyzed with χ2 tests or Fisher’s exact test. Binary logistic regression identified independent risk factors for 90-day prognosis in patients with ACLF. Receiver operating characteristic curves and area under the curve (AUC) assessed model predictive value, with forest plots visualized using the R 4.4.0 forest plot package. All tests were two-sided, and statistical significance was set at P < 0.05.

RESULTS
Baseline demographic characteristics

This study enrolled 58 patients with ACLF and the data collected were complete. The mean age of the included population was 47 ± 11 years (range: 28-73), 44 males (75.86%) and 14 females (24.14%). Collected data is complete. Hepatitis B virus (HBV) infection was the predominant etiology (28 cases, 48.28%). Based on 90-day outcomes, patients were categorized into survival (n = 33; male:female = 25:8; mean age: 44 ± 10 years) and death (n = 25; male:female = 19:6; mean age: 51 ± 11 years) groups. The death group exhibited significantly higher age, MELD, MELD-Na, NLR, TBil, Scr, and blood urea nitrogen, and lower PLT than did the survival group (P < 0.05). Other indices showed no significant differences (P > 0.05) (Table 1).

Table 1 Baseline characteristics and laboratory indices by clinical outcome, mean ± SD.
Characteristic
Survival group (n = 33)
Death group (n = 25)
Statistic
P value
Age (years)44 ± 1051 ± 11t = -2.5200.015
Male/female (n)25/819/6Z = -0.0210.983
Liver disease etiology
HBV159NANA
Alcohol1310NANA
Other56NANA
MELD18.75 ± 6.6522.89 ± 6.03t = -2.4400.018
MELD-Na21.98 ± 8.0128.36 ± 8.41t = -2.9400.005
ALBPS sessions1.36 ± 2.091.61 ± 2.14t = -0.4940.623
Laboratory tests
WBC6.26 (4.36-9.85)8.23 (5.23-13.40)Z = -1.3820.167
NLR5.98 ± 6.5611.46 ± 9.78t = -2.5480.014
Hb104.47 ± 25.3195.92 ± 28.91t = 1.1980.236
PLT132.96 ± 84.8983.96 ± 58.93t = 2.4680.017
ALB31.18 ± 6.2030.35 ± 4.24t = 0.5750.567
ALT55.00 (28.00-212.50)47.00 (25.50-76.00)Z = -0.8640.388
AST107.00 (61.00-275.00)105.00 (55.00-150.00)Z = -1.0830.279
ALP132.00 (105.50-186.50)137.00 (119.00-213.00)Z = -0.9580.338
GGT68.00 (49.00-145.50)76.00 (40.50-164.50)Z = -0.1960.844
TBil235.10 (198.40-310.15)297.50 (235.45-378.50)Z = -2.2060.027
Scr53.00 (47.50-72.50)68.00 (56.00-89.50)Z = -2.2230.026
BUN5.25 ± 3.8410.16 ± 8.38t = -2.9780.004
Na+134.33 ± 4.14132.24 ± 6.19t = 1.5050.138
APTT42.61 ± 8.4945.97 ± 10.70t = -1.3340.188
INR2.35 ± 0.592.53 ± 0.87t = -0.9770.333

Based on treatment strategies, 23 patients received ALBPS. Compared with the SMT-only group, no significant differences were observed in age or sex ratio (P > 0.05); however, the SMT group had lower hemoglobin, Albumin, alanine transaminase, aspartate transaminase, TBil, and MELD scores (P < 0.05). Other indices showed no significant differences (P > 0.05) (Table 2).

Table 2 Clinical baseline characteristics and laboratory indices between the artificial liver blood purification system and standardized medical treatment groups.
Characteristic
Total (n = 35)
SMT group
Total (n = 23)
ALBPS group
P valuea
Survival (n = 22)
Death (n = 13)
Survival (n = 11)
Death (n = 12)
Age (years)45 ± 1142 ± 1050 ± 1150 ± 1147 ± 952 ± 120.128
Male/female (n)27/817/510/317/68/39/30.780
Liver disease etiology
HBV10821777NA
Alcohol18108532NA
Other743113NA
MELD18.83 ± 6.1317.42 ± 5.8621.21 ± 6.0323.13 ± 6.7421.41 ± 7.6024.70 ± 5.730.015
MELD-Na24.92 ± 10.0021.61 ± 8.1530.51 ± 10.6524.44 ± 6.4722.71 ± 8.0526.02 ± 4.370.841
ALBPS sessions---3.74 ± 1.604.09 ± 1.303.42 ± 1.83NA
Laboratory tests
WBC6.75 (4.57-11.05)6.51 (4.31-9.47)7.96 (4.32-13.47)7.32 (5.15-12.81)6.03 (4.17-12.81)9.38 (5.76-13.68)0.465
NLR5.88 (2.68-9.15)5.18 (2.03-7.80)9.15 (4.40-25.11)7.09 ± 6.315.55 ± 5.828.50 ± 6.660.541
Hb94.64 ± 28.3797.83 ± 26.7787.54 ± 30.64110.13 ± 22.28115.73 ± 18.33105.00 ± 25.040.031
PLT106.24 ± 81.53127.02 ± 91.5671.08 ± 45.07120.35 ± 73.66144.82 ± 72.2897.92 ± 70.380.506
ALB28.73 ± 5.3129.34 ± 6.0227.71 ± 3.8234.00 ± 3.8734.86 ± 4.9733.21 ± 2.460.000
ALT40.00 (20.00-70.00)36.50 (19.50-124.50)44.00 (22.50-70.00)79.00 (47.00-203.00)124.00 (79.00-222.00)64.00 (30.25-81.50)0.004
AST75.00 (43.00-161.00)77.50 (44.75-384.25)75.00 (30.50-141.50)148.00 (71.00-176.00)167.00 (95.00-179.00)126.50 (65.00-152.00)0.048
ALP149.69 ± 55.04146.19 ± 54.91155.62 ± 56.98133.00 (115.00-203.00)127.00 (107.00-203.00)133.00 (117.00-210.75)0.899
GGT68.00 (36.00-165.00)89.50 (36.75-171.25)53.00 (30.50-164.50)76.00 (49.00-108.00)65.00 (50.00-108.00)88.00 (44.25-172.25)0.844
TBil249.98 ± 80.28237.03 ± 79.44271.89 ± 79.93287.00 (223.30-353.70)225.00 (208.00-317.20)312.25 (279.38-392.23)0.046
Scr70.00 ± 37.3758.14 ± 20.5790.08 ± 50.2262.00 (52.00-79.00)53.00 (51.00-78.00)66.50 (60.25-83.50)0.426
BUN5.01 (3.75-9.00)4.43 (3.46-5.71)6.73 (4.50-17.04)6.64 ± 5.574.54 ± 2.228.57 ± 7.160.830
Na+132.34 ± 6.29134.23 ± 5.31129.15 ± 6.72135.09 ± 2.66134.55 ± 1.75135.58 ± 3.290.053
APTT43.22 ± 8.9141.93 ± 8.7045.39 ± 9.1745.33 ± 10.5843.96 ± 8.3146.59 ± 12.550.415
INR2.33 ± 0.592.26 ± 0.522.46 ± 0.692.57 ± 0.882.53 ± 0.692.61 ± 1.050.216
Serum metabolomic profile characteristics and candidate metabolite selection

Targeted metabolomics detected 38 metabolites in serum samples, including 18 amino acids, 11 fatty acids, 5 gut microbiota-related metabolites (TMAO), and 4 bile acids (Supplementary material 2). All data were collected completely.

Univariate analysis using FC and t-tests filtered metabolites with |log2FC| > 0.58 and P < 0.1, yielding 10 differentially expressed metabolites between the death and survival groups, visualized in a volcano plot (Figure 1A). Owing to the multidimensional and highly correlated nature of metabolomics data, traditional univariate analysis struggled to capture latent information effectively. Thus, multivariate analysis was conducted using the SIMCA 14.1 software for dimensionality reduction and maximization of intergroup differences. PC analysis revealed similar yet overlapping metabolic phenotypes in patients with ACLF with different 90-day outcomes (Figure 1B), with PC1 and PC2 explaining 30.9% and 12.9% of the variance, respectively. To enhance the visualization of high-dimensional metabolomics data, Orthogonal partial least squares-discriminant analysis was applied, producing a score plot with predictive and orthogonal components. This achieved basic group separation (R2Xcum = 0.383, R2Ycum = 0.397, Q2Ycum = 0.068, Figure 1C). Differential metabolites were selected based on model contribution significance, with VIP > 1 as the threshold (Figure 1D). Combining univariate and multivariate analyses, seven differentially expressed metabolites (|log2FC| > 0.58, P < 0.1, and VIP > 1) were identified: L-carnitine, creatinine, alanine (Ala), arginine (Arg), proline (Pro), choline, and oleic acid (C18:1) (Figure 1E, Table 3).

Figure 1
Figure 1 Principal component analysis, orthogonal partial least squares-discriminant analysis, and variable importance in projection distribution between the groups. A: Volcano plot of metabolite distribution, where each point represents one metabolite; B: Principal component analysis of metabolite distribution, where each point represents one sample; C: Orthogonal partial least squares-discriminant analysis of metabolite distribution, where each point represents one sample; D: Variable importance in projection diagram from orthogonal partial least squares-discriminant analysis (variable importance in projection > 1 indicates significant variables); E: Venn diagram of differential metabolites, showing seven shared metabolites from univariate and multivariate analyses (death vs survival groups). Ala: Alanine; Arg: Arginine; Pro: Proline; C18:1: Oleic acid; IIe: Isoleucine; Leu: Leucine; Val: Valine; Gly: Glycine; CDCA: Chenodeoxycholic acid; Thr: Threonine; Asp: Aspartate; Ser: Serine; Lys: Lysine; Tyr: Tyrosine; Glu: Glutamate; Gln: Glutamine.
Table 3 Candidate metabolites for assessing ninety-day prognosis in patients with acute-on-chronic liver failure.
Metabolite
Survival
Death
|Log2FC|
P value
VIP
Carnitine50.55 ± 27.0678.89 ± 35.861.5101.94
Creatinine61.75 ± 25.5592.07 ± 49.321.7401.62
Ala405.67 ± 110.92465.51 ± 211.055.040.081.49
Arg14.56 ± 8.4122.81 ± 11.931.5501.43
Pro266.13 ± 78.89322.25 ± 133.303.620.031.21
Choline57.26 ± 27.0476.70 ± 37.782.370.011.17
C18:1182.34 ± 221.42279.69 ± 322.041.620.091.05
Multivariate analysis of 90-day prognosis prediction

Baseline variables differing between the survival and death groups (age, MELD, NLR, blood urea nitrogen, and PLT) and differential metabolites (L-carnitine, creatinine, Ala, Arg, Pro, choline, and C18:1) were included in a binary logistic regression model to identify independent predictors of 90-day prognosis (survival = 0, death = 1). The results showed that high age [odds ratio (OR) = 1.145, 95% confidence interval (CI): 0.001-1.31, P < 0.05], high NLR (OR = 1.169, 95%CI: 1.007-1.356, P < 0.05), and high Arg (OR = 1.225, 95%CI: 1.05-1.429, P < 0.05) were independent risk factors for 90-day mortality in patients with ACLF (Table 4).

Table 4 Logistic regression analysis of factors influencing ninety-day mortality risk in patients with acute-on-chronic liver failure.
Variable
Coefficient
SE
Wald χ2
P value
OR value
95%CI
Age0.1350.0693.8930.0481.1451.001-1.31
NLR0.1560.0764.220.041.1691.007-1.356
MELD0.189−0.1401.8240.1771.2080.918-1.589
PLT-0.0170.0093.4920.0620.9830.965-1.001
BUN0.2440.1373.2010.0741.2770.977-1.669
Choline-0.0120.0190.3840.5360.9880.951-1.026
Creatinine-0.0170.0230.5360.4640.9830.940-1.029
L-Carnitine0.0220.0220.9730.3241.0220.979-1.068
Pro0.0040.0060.50.4791.0040.993-1.016
Arg0.2030.0796.6480.011.2251.05-1.429
Ala-0.0030.0040.7230.3950.9970.990-1.004
C18:10.0010.0020.180.6721.0010.997-1.004
Constant-15.4816.0446.5610.010-
Visualization and efficacy analysis of the predictive model

Logistic regression forest plots visually depicted risk factors potentially influencing 90-day prognosis in patients with ACLF (Figure 2A). The predictive values for 90-day mortality were AUC_age = 0.682, AUC_NLR = 0.704, and AUC_Arg = 0.721. MELD and MELD-Na scores yielded AUC_MELD = 0.663 and AUC_MELD-Na = 0.714. A combined predictive ANA model comprising age, NLR, and Arg was developed, with the formula ANA = -15.481 + 0.135 × age + 0.156 × NLR + 0.203 × Arg. The cutoff was 0.759, with AUC_ANA = 0.945, a sensitivity and positive predictive value of 84.0%, and specificity and negative predictive value of 87.9%, outperforming MELD and MELD-Na scores (Figure 2B, Table 5). Stratified analysis of the etiology revealed that the ANA model predicted a 90-day prognostic AUC of 0.830 for patients with HBV-ACLF and 0.838 for those with alcohol-related ACLF (Figure 2C).

Figure 2
Figure 2 Logistic regression forest plot and receiver operating characteristic curve analysis. A: Logistic regression forest plot, where the reference line (odds ratio = 1) indicates no statistical significance. B: Receiver operating characteristic curve areas under the curve for the model for end-stage liver disease, model for end-stage liver disease with sodium, and age-neutrophil-to-lymphocyte ratio-Arg models predicting 90-day mortality in 58 patients with acute-on-chronic liver failure; C: Receiver operating characteristic curve areas under the curve of the age-neutrophil-to-lymphocyte ratio-arginine for predicting 90-day mortality in hepatitis B virus-acute-on-chronic liver failure and alcohol-related acute-on-chronic liver failure patients. NLR: Neutrophil-to-lymphocyte ratio; PLT: Platelet; BUN: Blood urea nitrogen; ANA: Age-neutrophil-to-lymphocyte ratio-arginine; HBV: Hepatitis B virus; AUC: Area under curve; Ala: Alanine; Arg: Arginine; Pro: Proline; C18:1: Oleic acid; ROC: Receiver operating characteristic; MELD: Model for end-stage liver disease; MELD-Na: Model for end-stage liver disease with sodium; ACLF: Acute-on-chronic liver failure.
Table 5 Predictive value of models and parameters for ninety-day mortality in patients with acute-on-chronic liver failure.
Parameter
AUC
Sensitivity (%)
Specificity (%)
Optimal cut-off
95%CI
P value
Age0.6826866.70.3470.541-0.8230.019
NLR0.7046466.30.3370.569-0.8400.008
Arg0.7217669.70.4570.584-0.8580.004
MELD0.6636075.80.3580.521-0.8050.035
MELD-Na0.7148060.60.4060.580-0.8470.006
ANA0.9458887.90.7590.893-0.9970
Impact of ALBPS treatment on differential metabolite changes

To further validate ALBPS effects on the seven metabolites, VIP values were compared before (within 2 hours) and after (within 2 hours) the first and last ALBPS sessions. Only Ala and L-carnitine showed statistically significant changes (VIP > 1) post-treatment (Table 6).

Table 6 Changes in differential metabolites before and after artificial liver blood purification system treatment.
Metabolite
ALBPS
P value
VIP
BeforeAfter
Ala370.58 ± 143.66548.53 ± 339.560.0251.88
L-carnitine69.79 ± 40.90101.68 ± 70.240.0671.36
Arg22.48 ± 9.5622.95 ± 9.570.870.91
Creatinine90.21 ± 44.44112.29 ± 69.860.2080.88
Pro303.03 ± 106.05324.23 ± 171.560.6170.79
C18:1459.80 ± 304.72350.58 ± 243.920.1860.7
Choline81.41 ± 39.8889.30 ± 61.740.6090.41
DISCUSSION

Advancements in informatics and analytical technologies, coupled with integrative biological approaches, have empowered metabolomics to detect subtle early changes in biological pathways, providing insights into disease mechanisms[4]. ACLF is marked by heightened systemic catabolism, driving extensive lipolysis, glycogenolysis, and proteolysis. These processes release fatty acids, glucose, and amino acids into the bloodstream, fueling immune responses and peripheral organ dysfunction, leading to a cascade of immune–inflammatory–metabolic dysregulation[8]. Accurately predicting survival in patients with ACLF is paramount, as liver transplantation markedly improves outcomes for those unresponsive to SMT or ALBPS. Patients with poorer predicted prognoses are prioritized for transplantation. This study employed Xevo TQ-S Cronos LC-MS and ACQUITY UPLC-Xevo TQ-S tandem quadrupole mass spectrometry to analyze serum levels of fatty acids, bile acids, amino acids, and TMAO in patients with ACLF experiencing divergent 90-day outcomes. Through combined univariate and multivariate analyses, we identified seven differential metabolites-L-carnitine, creatinine, Ala, Arg, Pro, choline, and C18:1 - distinguishing the survival from death groups. Logistic regression confirmed Arg as an independent predictor of 90-day mortality risk. In addition, ALBPS treatment modulated Ala and L-carnitine levels. Although clinical evidence underscores ALBPS’s role as an effective adjunct to SMT and a bridge to transplantation, this study did not establish a definitive link between ALBPS-mediated metabolite change and improved prognosis, possibly owing to the limited sample size. Larger, prospective studies are needed to clarify this relationship.

The liver orchestrates amino acid metabolism through specialized transporters, converting portal vein-derived amino acids into glucose and fatty acids to meet energy demands and maintain amino acid homeostasis[9]. Ala, a critical gluconeogenic amino acid, is exclusively metabolized in the liver. Its heightened catabolism in ACLF contributes to hyperglycemia and skeletal muscle wasting[10]. Disrupted cellular energy metabolism is intricately tied to disease progression, with alternative metabolic pathways becoming essential for cell survival and proliferation. In hepatocellular carcinoma models, both human and animal, enzymes driving Pro synthesis are upregulated, while those involved in its breakdown are suppressed[11], offering a plausible explanation for Pro’s association with adverse ACLF outcomes. Arg, a pivotal intermediate in the urea cycle, modulates macrophage polarization and inflammation as a key metabolic regulator. Zhang et al[12] demonstrated through transcriptomic and metabolomic profiling that M1 macrophages preferentially convert Arg into ornithine and urea, whereas M2 macrophages favor citrulline and nitric oxide production. However, the role of these amino acids as metabolic checkpoints in the inflammatory storm conditions of ACLF remains underexplored. Prior research also suggests that liver sinusoidal endothelial cells exhibit lower metabolic activity than do hepatocytes, with excess circulating fatty acids triggering endothelial dysfunction[13]. Zhou et al[14] reported elevated L-carnitine levels in hepatectomized mice, associated with reduced lipid accumulation, enhanced hepatocyte proliferation (Ki67-positive cells), diminished inflammation (MPO-positive cells), and fewer apoptotic cells (TUNEL-positive), highlighting L-carnitine’s role in promoting lipid metabolism, hepatocyte regeneration, and suppressing sterile inflammation caused by lipid overload. C18:1, a monounsaturated fatty acid, plays a key role in altering plasma lipid and lipoprotein profiles, attenuating inflammation, oxidative stress, and coagulopathy[15]. Stearoyl-coenzyme A desaturase, the rate-limiting enzyme in C18:1 synthesis, regulates inflammation and stress responses, even though its specific role in ACLF progression remains unclear within the complexity of systemic metabolism. Creatinine, identified via metabolomics, serves as a marker of broader metabolic shifts. The urea/creatinine ratio, as noted in the REducing Deaths due to OXidative Stress study, reflects catabolic states and mortality risk[16]. Choline, involved in oxidation, phosphorylation, and acetylation pathways, becomes dysregulated in ACLF, triggering endoplasmic reticulum stress and mitochondrial dysfunction. This amplifies oxidative stress, hepatocyte injury, macrophage activation, and inflammation, exacerbating disease progression[17]. In the occurrence and development process of ACLF, severe hepatocyte damage disrupts normal metabolism, while metabolic imbalances impair hepatocyte function, survival, and regeneration, driving disease onset and severity. Although current treatments emphasize anti-inflammatory, antioxidant, and supportive care, targeting choline, carbohydrate, and lipid metabolism to restore inflammatory balance holds vital promise for preventing ACLF onset, mitigating severity, and enhancing prognosis.

Univariate and multivariate logistic regression analyses pinpointed age, NLR, and Arg as significant predictors of 90-day mortality in patients with ACLF. These findings align with prior studies identifying age and INR as independent risk factors for poor 28- and 90-day outcomes in patients with HBV-ACLF meeting the Chinese group on the study of severe hepatitis B-ACLF criteria[18]. Lashen et al[19] similarly validated NLR as a predictor of 30-day and 6-month mortality in ACLF. Notably, integrating these clinical markers with metabolites into the ANA model yielded a robust AUC of 0.945, with superior sensitivity, specificity, and predictive accuracy compared with those of the MELD and MELD-Na scores, offering a powerful tool for early detection of adverse ACLF outcomes.

Furthermore, this study compared clinical parameters between the ALBPS and SMT groups, reflecting real-world practice: ALBPS-treated patients exhibited more severe baseline conditions (for example, higher MELD scores). Although ALBPS frequency or application did not independently predict 90-day mortality, its effectiveness in improving outcomes in critically ill patients with ACLF underscores its value as an adjunctive therapy. Treatment modality significantly influences ACLF prognosis. Yadav et al[20] conducted plasma metabolomics in 45 patients with ACLF categorized into SMT, hemoperfusion adsorption (HA), and therapeutic plasma exchange (TPE) groups using high-resolution mass spectrometry. HA reduced Arg-Pro and TMAO metabolites after 7 days, while TPE transiently elevated phenylalanine and serine levels. Compared with TPE and SMT, HA more effectively curbed inflammation and secondary energy metabolism pathways, improving the plasma environment (about 80% probability). In our study, ALBPS increased Ala and L-carnitine post-treatment, which suggesting its capacity to clear inflammatory mediators, ameliorate metabolic dysregulation, and support hepatocyte regeneration. However, differing blood purification techniques may influence metabolite profiles, potentially explaining ALBPS’s limited prognostic impact. This limitation, combined with a small sample size, variable treatment modalities, and patient-specific metabolic differences, underscores the need for further investigation to elucidate ALBPS’s role in modulating metabolites and improving clinical outcomes. In the future, untargeted metabolomics and spatial metabolomics need to be used to further elucidate the all-round and dynamic effects of ALBPS therapy on metabolites in patients with ACLF, and to further assess the impact on patient prognosis.

CONCLUSION

In conclusion, using high-performance liquid chromatography-mass spectrometry, we identified seven metabolites - L-carnitine, creatinine, Ala, Arg, Pro, choline, and C18:1 - linked to 90-day ACLF prognosis, with Arg emerging as an independent mortality risk factor. ALBPS treatment modulated Ala and L-carnitine, reducing inflammation and promoting hepatocyte regeneration. However, this study had limitations: (1) A single-center case-control design with a small sample size; (2) A restricted scope of metabolite detection, potentially skewing results; and (3) Variability in baseline metabolism, complications, and disease severity despite standardized treatment, influencing indices and outcomes. In future, multicenter, large-scale studies should be carried out to refine metabolite profiling based on stratified ACLF severity and treatment regimens, enabling more precise diagnosis and management.

Footnotes

Provenance and peer review: Unsolicited article; Externally peer reviewed.

Peer-review model: Single blind

Specialty type: Gastroenterology and hepatology

Country of origin: China

Peer-review report’s classification

Scientific Quality: Grade B, Grade B, Grade B

Novelty: Grade B, Grade C, Grade C

Creativity or Innovation: Grade B, Grade C, Grade C

Scientific Significance: Grade B, Grade B, Grade C

P-Reviewer: Fan XC; Qiu WS S-Editor: Bai Y L-Editor: A P-Editor: Zheng XM

References
1.  Sarin SK, Choudhury A, Sharma MK, Maiwall R, Al Mahtab M, Rahman S, Saigal S, Saraf N, Soin AS, Devarbhavi H, Kim DJ, Dhiman RK, Duseja A, Taneja S, Eapen CE, Goel A, Ning Q, Chen T, Ma K, Duan Z, Yu C, Treeprasertsuk S, Hamid SS, Butt AS, Jafri W, Shukla A, Saraswat V, Tan SS, Sood A, Midha V, Goyal O, Ghazinyan H, Arora A, Hu J, Sahu M, Rao PN, Lee GH, Lim SG, Lesmana LA, Lesmana CR, Shah S, Prasad VGM, Payawal DA, Abbas Z, Dokmeci AK, Sollano JD, Carpio G, Shresta A, Lau GK, Fazal Karim M, Shiha G, Gani R, Kalista KF, Yuen MF, Alam S, Khanna R, Sood V, Lal BB, Pamecha V, Jindal A, Rajan V, Arora V, Yokosuka O, Niriella MA, Li H, Qi X, Tanaka A, Mochida S, Chaudhuri DR, Gane E, Win KM, Chen WT, Rela M, Kapoor D, Rastogi A, Kale P, Rastogi A, Sharma CB, Bajpai M, Singh V, Premkumar M, Maharashi S, Olithselvan A, Philips CA, Srivastava A, Yachha SK, Wani ZA, Thapa BR, Saraya A, Shalimar, Kumar A, Wadhawan M, Gupta S, Madan K, Sakhuja P, Vij V, Sharma BC, Garg H, Garg V, Kalal C, Anand L, Vyas T, Mathur RP, Kumar G, Jain P, Pasupuleti SSR, Chawla YK, Chowdhury A, Alam S, Song DS, Yang JM, Yoon EL; APASL ACLF Research Consortium (AARC) for APASL ACLF working Party. Acute-on-chronic liver failure: consensus recommendations of the Asian Pacific association for the study of the liver (APASL): an update. Hepatol Int. 2019;13:353-390.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Full Text (PDF)]  [Cited by in Crossref: 614]  [Cited by in RCA: 585]  [Article Influence: 97.5]  [Reference Citation Analysis (0)]
2.  Trebicka J, Hernaez R, Shawcross DL, Gerbes AL. Recent advances in the prevention and treatment of decompensated cirrhosis and acute-on-chronic liver failure (ACLF) and the role of biomarkers. Gut. 2024;73:1015-1024.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in Crossref: 5]  [Cited by in RCA: 13]  [Article Influence: 13.0]  [Reference Citation Analysis (0)]
3.  Mezzano G, Juanola A, Cardenas A, Mezey E, Hamilton JP, Pose E, Graupera I, Ginès P, Solà E, Hernaez R. Global burden of disease: acute-on-chronic liver failure, a systematic review and meta-analysis. Gut. 2022;71:148-155.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in Crossref: 37]  [Cited by in RCA: 141]  [Article Influence: 47.0]  [Reference Citation Analysis (0)]
4.  Marchev AS, Vasileva LV, Amirova KM, Savova MS, Balcheva-Sivenova ZP, Georgiev MI. Metabolomics and health: from nutritional crops and plant-based pharmaceuticals to profiling of human biofluids. Cell Mol Life Sci. 2021;78:6487-6503.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Full Text (PDF)]  [Cited by in Crossref: 7]  [Cited by in RCA: 62]  [Article Influence: 15.5]  [Reference Citation Analysis (0)]
5.  Li J, Liang X, Jiang J, Yang L, Xin J, Shi D, Lu Y, Li J, Ren K, Hassan HM, Zhang J, Chen P, Yao H, Li J, Wu T, Jin L, Ye P, Li T, Zhang H, Sun S, Guo B, Zhou X, Cai Q, Chen J, Xu X, Huang J, Hao S, He J, Xin S, Wang D, Trebicka J, Chen X, Li J; Chinese Group on the Study of Severe Hepatitis B (COSSH). PBMC transcriptomics identifies immune-metabolism disorder during the development of HBV-ACLF. Gut. 2022;71:163-175.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Full Text (PDF)]  [Cited by in Crossref: 22]  [Cited by in RCA: 82]  [Article Influence: 27.3]  [Reference Citation Analysis (0)]
6.  Xiao Y, Lu J, Xu S, Wu Z, Wang W, Ji R, Guo T, Qi Z, Tong H, Wang Y, Zhao C. Metabolic Differences among Patients with Cirrhosis Using Q Exactive Hybrid Quadrupole Orbitrap Mass Spectrometry Technology. J Proteome Res. 2024;23:5352-5359.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in RCA: 1]  [Reference Citation Analysis (0)]
7.  Chen Y, Han T, Duan Z; Severe Liver Disease and Artificial Liver Group, Chinese Society of Hepatology, Chinese Medical Association. Clinical application of artificial liver and blood purification: expert consensus recommendations. Hepatol Int. 2023;17:4-17.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in Crossref: 9]  [Cited by in RCA: 11]  [Article Influence: 5.5]  [Reference Citation Analysis (0)]
8.  Valainathan SR, Xie Q, Arroyo V, Rautou PE. Prognosis algorithms for acute decompensation of cirrhosis and ACLF. Liver Int. 2025;45:e15927.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in Crossref: 3]  [Cited by in RCA: 4]  [Article Influence: 4.0]  [Reference Citation Analysis (0)]
9.  Paulusma CC, Lamers WH, Broer S, van de Graaf SFJ. Amino acid metabolism, transport and signalling in the liver revisited. Biochem Pharmacol. 2022;201:115074.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in Crossref: 2]  [Cited by in RCA: 77]  [Article Influence: 25.7]  [Reference Citation Analysis (0)]
10.  Okun JG, Rusu PM, Chan AY, Wu Y, Yap YW, Sharkie T, Schumacher J, Schmidt KV, Roberts-Thomson KM, Russell RD, Zota A, Hille S, Jungmann A, Maggi L, Lee Y, Blüher M, Herzig S, Keske MA, Heikenwalder M, Müller OJ, Rose AJ. Liver alanine catabolism promotes skeletal muscle atrophy and hyperglycaemia in type 2 diabetes. Nat Metab. 2021;3:394-409.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in Crossref: 30]  [Cited by in RCA: 68]  [Article Influence: 17.0]  [Reference Citation Analysis (0)]
11.  Ding Z, Ericksen RE, Lee QY, Han W. Reprogramming of mitochondrial proline metabolism promotes liver tumorigenesis. Amino Acids. 2021;53:1807-1815.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in Crossref: 6]  [Cited by in RCA: 25]  [Article Influence: 6.3]  [Reference Citation Analysis (0)]
12.  Zhang Q, Song Q, Liu S, Xu Y, Gao D, Lu P, Liu Y, Zhao G, Wu L, Zhao C, Yang J. Integrated transcriptomic and metabolomic analysis reveals the metabolic programming of GM-CSF- and M-CSF- differentiated mouse macrophages. Front Immunol. 2023;14:1230772.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in RCA: 6]  [Reference Citation Analysis (0)]
13.  Matuszyk E, Sierka E, Rodewald M, Bae H, Meyer T, Kus E, Chlopicki S, Schmitt M, Popp J, Baranska M. Differential response of liver sinusoidal endothelial cells and hepatocytes to oleic and palmitic acid revealed by Raman and CARS imaging. Biochim Biophys Acta Mol Basis Dis. 2020;1866:165763.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in Crossref: 5]  [Cited by in RCA: 9]  [Article Influence: 1.8]  [Reference Citation Analysis (0)]
14.  Zhou X, Huang G, Wang L, Zhao Y, Li J, Chen D, Wei L, Chen Z, Yang B. L-carnitine promotes liver regeneration after hepatectomy by enhancing lipid metabolism. J Transl Med. 2023;21:487.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Full Text (PDF)]  [Cited by in Crossref: 8]  [Cited by in RCA: 9]  [Article Influence: 4.5]  [Reference Citation Analysis (0)]
15.  Piccinin E, Cariello M, De Santis S, Ducheix S, Sabbà C, Ntambi JM, Moschetta A. Role of Oleic Acid in the Gut-Liver Axis: From Diet to the Regulation of Its Synthesis via Stearoyl-CoA Desaturase 1 (SCD1). Nutrients. 2019;11:2283.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Full Text (PDF)]  [Cited by in Crossref: 38]  [Cited by in RCA: 88]  [Article Influence: 14.7]  [Reference Citation Analysis (0)]
16.  Haines RW, Fowler AJ, Wan YI, Flower L, Heyland DK, Day A, Pearse RM, Prowle JR, Puthucheary Z. Catabolism in Critical Illness: A Reanalysis of the REducing Deaths due to OXidative Stress (REDOXS) Trial. Crit Care Med. 2022;50:1072-1082.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in Crossref: 1]  [Cited by in RCA: 33]  [Article Influence: 11.0]  [Reference Citation Analysis (0)]
17.  Chen X, Qiu W, Ma X, Ren L, Feng M, Hu S, Xue C, Chen R. Roles and Mechanisms of Choline Metabolism in Nonalcoholic Fatty Liver Disease and Cancers. Front Biosci (Landmark Ed). 2024;29:182.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in RCA: 8]  [Reference Citation Analysis (0)]
18.  Ma S, Xie Z, Zhang L, Yang Y, Jiang H, Ouyang X, Zhao Y, Liu Q, Xu X, Li L. Identification of a Potential miRNA-mRNA Regulatory Network Associated With the Prognosis of HBV-ACLF. Front Mol Biosci. 2021;8:657631.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Full Text (PDF)]  [Cited by in Crossref: 2]  [Cited by in RCA: 10]  [Article Influence: 2.5]  [Reference Citation Analysis (0)]
19.  Lashen SA, Salem P, Ibrahim E, Abd Elmoaty D, Yousif WI. Hematological ratios in patients with acute decompensation and acute-on-chronic liver failure: prognostic factors. Eur J Gastroenterol Hepatol. 2024;36:952-960.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in Crossref: 1]  [Cited by in RCA: 2]  [Article Influence: 2.0]  [Reference Citation Analysis (0)]
20.  Yadav M, Maiwal R, Kumar Br V, Tripathi G, Sharma N, Sharma N, Bindal V, Mathew B, Pandey S, Singh SP, Tevathia HV, Maras JS, Sarin SK. Comparative metabolome analysis reveals higher potential of haemoperfusion adsorption in providing favourable outcome in ACLF patients. Liver Int. 2024;44:1189-1201.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in RCA: 6]  [Reference Citation Analysis (0)]