Prognostic value of serum alpha-fetoprotein kinetics in liver failure on artificial liver support

Abstract

BACKGROUND

Liver failure, particularly acute-on-chronic liver failure, is associated with high mortality (50%-90%). The plasma exchange (PE) mode of the artificial liver support system has been shown to improve clinical outcomes, although its efficacy may vary depending on the regenerative capacity of the liver. Alpha-fetoprotein (AFP), an oncofetal glycoprotein, is reactivated during liver regeneration and may serve as a prognostic biomarker. Previous studies have reported significantly higher post-PE AFP levels in survivors than in non-survivors (286.5 ng/mL vs 82.3 ng/mL at day 7). However, the predictive value of baseline AFP stratification and serial AFP kinetics during PE therapy remains unestablished. This study investigated whether serial AFP measurements predict clinical outcomes in liver failure patients receiving PE.

AIM

To evaluate the predictive value of serial AFP measurements in liver failure patients receiving PE.

METHODS

This retrospective study included 194 liver failure patients with complete AFP data, excluding those with tumors, bleeding disorders, allergies, or unstable conditions. Patients were stratified by baseline AFP into low-AFP (< 100 ng/mL, n = 60), medium-AFP (100-200 ng/mL, n = 70), and high-AFP (> 200 ng/mL, n = 64) groups. AFP was measured before PE and on days 1, 10, 20, and 25.

RESULTS

Stratification by baseline AFP revealed significant gradients. The high-AFP group required fewer PE sessions than the low-AFP group (2.8 ± 1.0 vs 4.2 ± 1.5) but exhibited greater post-PE AFP elevation (75.1 ± 20.3 ng/mL vs 33.1 ± 10.2 ng/mL; P < 0.001). The high-AFP group demonstrated optimal values, including the lowest ammonia, bilirubin, alanine aminotransferase, aspartate aminotransferase, γ-glutamyl transferase, and the highest albumin and prothrombin activity (all post hoc P < 0.05 vs low-AFP). The medium-AFP group showed intermediate values except for prothrombin activity (35.2% ± 8.6%), which was significantly lower than in both other groups (P < 0.001). The high-AFP group had a reduced incidence of spontaneous bacterial peritonitis (9.4% vs 25.0%; P = 0.003), superior three-month survival (90.6% vs 56.7%; P < 0.001), and a higher post-treatment three-month receiver operating characteristic area under the curve (0.8851 vs 0.7051).

CONCLUSION

AFP dynamics correlate with regenerative capacity and clinical outcomes in liver failure. Serial AFP monitoring may enhance risk stratification and support personalized therapeutic strategies.

Key Words: Liver failure; Artificial liver support system; Plasma exchange; Alpha-fetoprotein; Liver function; Survival rate; Prognostic prediction

Core Tip: This study establishes that baseline alpha-fetoprotein (AFP) stratification (< 100 ng/mL, 100-200 ng/mL, > 200 ng/mL) in liver failure patients receiving plasma exchange-based artificial liver support systems predicts regenerative capacity and clinical outcomes. High-AFP patients (> 200 ng/mL) baseline AFP levels demonstrated improved liver function recovery, fewer complications, and required fewer treatment sessions compared to those with lower levels. Critically, three-month survival was dose-dependent, with baseline AFP > 200 ng/mL providing excellent prognostic discrimination. Serial AFP monitoring enables precision artificial liver support systems therapy by identifying patients with high endogenous regenerative potential.

INTRODUCTION

Liver failure, the ultimate manifestation of end-stage liver disease, is defined by a triple-hit mechanism involving hepatocyte necrosis, an inflammatory storm, and impaired regenerative capacity. According to the Chinese Medical Association Guidelines for Liver Failure (2024 edition)[1], liver failure is classified into four subtypes based on dynamic clinical trajectories. Among these, acute-on-chronic liver failure (ACLF) has emerged as a global research priority due to its rapid progression to multi-organ dysfunction and high mortality (50%-90%). In China, where hepatitis B virus (HBV) infection is highly prevalent, ACLF imposes a particularly severe disease burden. Recent epidemiological studies have reported a 90-day mortality of 64.3% among Chinese patients, significantly higher than the 28%-33% observed in Western populations. This cross-regional heterogeneity may be attributed to differences in etiological profiles (e.g., HBV predominance in Asia vs alcohol- or metabolic-related liver diseases in the West), disparities in healthcare resource allocation, and genetic predispositions[2,3].

In the therapeutic domain, the artificial liver support system provides unique value through dual-mode “detoxification-regeneration” interventions[4-7]. As a core non-bioartificial artificial liver support system technology, plasma exchange (PE) removes toxins (e.g., bilirubin, ammonia, interleukin-6) and replenishes hepatic proteins (e.g., albumin, coagulation factors), thereby facilitating reconstruction of a pro-regenerative microenvironment[8-11]. Randomized controlled trials have confirmed that PE significantly improves 28-day survival in patients with ACLF (71.4% vs 43.2% in controls); however, its efficacy exhibits substantial interindividual variability[12-14]. A study by Xiang et al[15] demonstrated that among patients receiving standardized PE therapy, those with hepatocyte proliferating cell nuclear antigen positivity ≥ 5% achieved a 6-month survival rate of 82.5%, whereas those with impaired regeneration (proliferating cell nuclear antigen positivity < 5%) had a markedly lower survival rate of 29.1%. These findings highlight the pivotal role of endogenous liver regenerative capacity as a prognostic threshold. Nevertheless, noninvasive assessment of regenerative potential remains an urgent unmet clinical need.

Liver regeneration is a critical determinant of survival following acute liver insults such as ACLF. Alpha-fetoprotein (AFP), a 70 kDa glycoprotein composed of 591 amino acids and predominantly secreted by fetal hepatocytes, serves not only as an oncofetal marker but also as a key indicator of hepatic regenerative activity. Its expression is dynamically reactivated during liver injury and repair, highlighting its potential utility in assessing regenerative capacity[16]. The prognostic significance of this dynamic expression in severe liver failure was elucidated in a landmark longitudinal cohort study conducted by Yuan et al[17]. Their findings demonstrated a pronounced divergence in AFP trajectories between ACLF survivors and non-survivors undergoing PE. By day 7 post-PE, median AFP levels increased significantly to 286.5 ng/mL in survivors, representing a robust 3.5-fold rise compared with the markedly lower levels observed in non-survivors (82.3 ng/mL; P < 0.001). This substantial difference highlights the association between elevated AFP levels and favorable clinical outcomes. Furthermore, peak AFP levels showed a strong positive correlation (r = 0.732) with expansion of the hepatocyte nuclear area, a well-established morphological hallmark of active cell cycle progression and proliferation[18].

Monitoring the temporal kinetics of AFP levels, particularly during the early critical phase after injury (e.g., within the first week post-PE), provides valuable insight into individual regenerative competence. Patients with persistently low or non-rising AFP levels during this period can be identified as being at high risk for disease progression and poor outcomes. Serum AFP measurement offers notable practical advantages as a continuous prognostic biomarker. As a routinely available and cost-efficient assay integrated into standard clinical biochemistry panels, it enables real-time assessment of hepatic regenerative capacity and facilitates risk stratification throughout the clinical course of liver failure. This approach supports timely and cost-effective therapeutic decision-making without imposing significant additional expenditures. Incorporating serial AFP monitoring into established prognostic frameworks, complementing conventional clinical scores such as model for end-stage liver disease and emerging technologies such as radiomic features, holds considerable promise. This integrated strategy is expected to facilitate a more comprehensive and dynamic evaluation of regenerative potential, potentially enabling earlier risk stratification and optimizing the timing of critical therapeutic interventions in the management of liver failure.

MATERIALS AND METHODS

Study design and intervention protocol

Participant selection: A total of 194 patients with liver failure, admitted to The Second Affiliated Hospital of Kunming Medical University and Baoshan People’s Hospital (Yunnan Province, China) between January 2021 and December 2023, were enrolled. All patients met the diagnostic criteria for liver failure subtypes outlined in the 2024 Chinese Guidelines for the Diagnosis and Treatment of Liver Failure (jointly issued by the Infectious Diseases and Hepatology branches of the Chinese Medical Association). Subclassification was based on the following temporal and clinical characteristics. Acute liver failure (ALF): Sudden onset (< 2 weeks), absence of pre-existing chronic liver disease, and concurrent hepatic encephalopathy ≥ grade II (West Haven criteria). Subacute liver failure (SALF): Gradual progression (2-26 weeks), no documented history of chronic liver disease prior to onset, and biochemical evidence of hepatic synthetic dysfunction [international normalized ratio (INR) ≥ 1.5] with total bilirubin ≥ 10 × upper limit of normal. ACLF: Acute deterioration of chronic liver disease within ≤ 4 weeks, presence of ≥ 1 organ failure [chronic liver failure (CLF) consortium-organ failure score ≥ 1], and exclusion of alternative decompensation triggers (e.g., variceal hemorrhage). CLF: Progressive hepatic dysfunction in cirrhotic patients, characterized by recurrent decompensation events (≥ 2 episodes of grade II-IV ascites or overt hepatic encephalopathy within 6 months) (Figure 1).

Figure 1 Retrospective study design flowchart. This figure outlines the design of a retrospective study investigating the clinical significance of alpha-fetoprotein (AFP) in the treatment of liver failure with artificial liver. Patients were stratified into three groups based on serum AFP levels: Low AFP group (n = 60), medium AFP group (n = 70), and high AFP group (n = 64). Baseline measurements included serum AFP levels, liver and kidney function parameters, blood ammonia concentrations, coagulation function indices, complete blood counts, and other relevant indicators. Following plasma exchange treatment, these indicators were reassessed on post-treatment days 1, 10, 20 and 25 points (pre-discharge or pre-terminal stage). AFP: Alpha-fetoprotein.

Exclusion criteria: For patients with liver failure and AFP ≥ 400 ng/mL, liver imaging examinations were performed to identify potential liver tumors[19]. Pregnancy, gonadal embryonic tumors, or other non-hepatic tumors. Severe active bleeding or disseminated intravascular coagulation. Severe allergic reaction to blood products or medications used during treatment, such as plasma, heparin, or protamine. Circulatory failure or unstable cerebral infarction.

Treatment allocation

This retrospective study included 194 patients meeting the diagnostic criteria, stratified by baseline AFP (electrochemiluminescence immunoassay; < 7 ng/mL) using established cutoffs: < 100 ng/mL (subclinical), 100-200 ng/mL (transitional), and > 200 ng/mL (robust regeneration)[20]. Patients were categorized as low-AFP (< 100 ng/mL, n = 60), medium-AFP (100-200 ng/mL, n = 70), and high-AFP (> 200 ng/mL, n = 64).

Medical treatment: All patients received protocolized care in accordance with the 2023 European Association for the Study of the Liver Clinical Practice Guidelines[21]. (1) Nutritional support: Caloric intake: 35-40 kcal/kg/day via enteral feeding (preferred) or parenteral nutrition if enteral feeding was not feasible; micronutrient supplementation: Vitamin K (10 mg IV weekly), zinc acetate (50 mg, three times a day); sodium restriction: Strict sodium restriction (< 80 mmol/day) with daily electrolyte monitoring; (2) Albumin replacement: Human albumin (20%, CSL Behring, PA, United States) 1 g/kg/day for the following indications: Spontaneous bacterial peritonitis (SBP) prophylaxis and hepatorenal syndrome management; target serum albumin level: > 30 g/L (adjusted according to 24-hour urine sodium excretion); (3) Hepatoprotective therapy: Anti-inflammatory: Glycyrrhizin (80-120 mg/day IV, adjusted according to serum potassium levels); membrane stabilization: Polyenylphosphatidylcholine (456 mg, three times a day); antioxidant: N-acetylcysteine (150 mg/kg loading dose); note: Medication was discontinued if INR > 3.0 or total bilirubin > 300 μmol/L; (4) Antiviral therapy: Entecavir (0.5 mg, once a day, Baraclude^®) for HBV-DNA-positive cases. For creatinine clearance < 50 mL/minute, the dose was adjusted to 0.25 mg once a day. Confirmatory testing, including HBV genotyping, was performed prior to initiation; and (5) Complication management: Hepatorenal syndrome-acute kidney injury: Terlipressin (2 mg, every 6 hours) + albumin. SBP: Ceftriaxone (2 g, once a day) until the ascitic polymorphonuclear leukocyte count was < 250/mm³; hepatic encephalopathy: Lactulose (30 mL, titrated to 2-3 stools/day) + rifaximin (550 mg BID); variceal bleeding: Somatostatin infusion (250 μg/hour) + endoscopic band ligation within 6 hours of presentation, following the protocol outlined in the American Association for the Study of Liver Diseases 2021 guidelines)[22].

Artificial liver treatment: All patients received PE as part of the comprehensive medical treatment described above. PE treatment was performed in accordance with the expert consensus on the clinical application of artificial liver blood purification technology (2022 edition)[23].

Instruments and materials

Machine: X-10 artificial liver therapy system (Zhuhai Jianfan Biotechnology Co., Ltd., Guangdong, China). Consumables: Plasma separator (Berke Co., Ltd., CA, United States), disposable blood circuit connecting catheter (Tianjin Hanahao Medical Materials Co., Ltd., Tianjin, China), Abell double-lumen catheter (size 11.5 Fr × 16 cm) for femoral vein catheterization.

Operation methods

Pre-treatment venous blood tests included AFP, complete blood count, liver and kidney function tests, coagulation profile, and blood ammonia levels. Procedures were performed in an air-disinfected room with electrocardiogram monitoring and low-flow oxygen administration. Extracorporeal access was established via femoral vein puncture using a double-lumen catheter. The plasma separator and circuit were primed with 500 mL of 4% heparin saline, followed by rinsing with heparin-free saline. Intravenous dexamethasone (5 mg) and calcium gluconate were administered prophylactically.

Systemic anticoagulation was initiated with an intravenous heparin bolus (20 mg) and adjusted during treatment according to patient condition, coagulation parameters, body weight, treatment duration, transmembrane pressure, plasma flow, and prothrombin time. Blood flow was maintained at 100-120 mL/minute, with a plasma separation rate of 20-30 mL/minute. PE volume was 2000-3000 mL [approximately body weight (kg) × 40 mL] using fresh frozen plasma. Treatment duration was approximately 2-3 hours. Heparin administration was discontinued 1-1.5 hours before completion of the procedure, depending on transmembrane pressure, and neutralized with protamine sulfate. Prophylactic amikacin was administered. Treatment frequency and intervals (1-4 days) were determined by clinical status.

Indicators for observation

Venous blood samples were collected from all patients before PE treatment (on day 2 of admission), and on post-treatment days 1, 10, 20, and 25 (pre-discharge or pre-terminal stage). AFP levels were measured using electrochemiluminescence immunoassay (Maglumi 4000, Snibe Diagnostics, Shenzhen, China) with a detection limit of 0.5 ng/mL. The normal reference range for AFP was defined as < 7 ng/mL. Comprehensive clinical data were recorded for each patient. The study evaluated the correlation between dynamic changes in AFP levels and prognosis in liver failure patients, compared survival and mortality rates among different AFP-level subgroups, and assessed the predictive value and clinical significance of AFP levels for determining disease outcomes in liver failure.

Safety assessment

Treatment-emergent adverse events during PE therapy were actively monitored and included: Hypersensitivity reactions (e.g., rash, bronchospasm); hemorrhagic complications (platelet count < 50 × 10⁹/L or INR > 2.5); hemodynamic instability (sustained systolic blood pressure < 90 mmHg for > 10 minutes); febrile responses (core temperature > 38.5 °C); and thromboembolic events (confirmed by Doppler ultrasound or computed tomography angiography). The association with artificial liver treatment was evaluated.

Statistical analysis

Data were analyzed using SPSS version 19.0. Continuous variables were expressed as mean ± SD. The normality of distribution for all continuous variables was assessed using the Shapiro-Wilk test, with P > 0.05 indicating adherence to parametric assumptions. A paired t-test was used for within-group comparisons before and after treatment, and an independent samples t-test for between-group comparisons. Categorical variables were analyzed using the Pearson χ² test. Post hoc analyses were adjusted using the Bonferroni correction for multiple comparisons. A P value < 0.05 was considered statistically significant.

Kaplan-Meier survival curves with log-rank tests were used to compare mortality rates among AFP subgroups. The predictive performance of AFP levels for clinical outcomes was assessed using receiver operating characteristic curve analysis, with calculation of the area under the curve (AUC). Optimal prognostic thresholds were determined using Youden’s index. Sample size calculation, based on an anticipated AFP effect size (Cohen’s d = 0.8) from prior studies, with α = 0.05 and β = 0.2, indicated a requirement of at least 50 patients per group. To control for potential confounding factors, multivariate logistic regression analyses were performed. Primary outcomes (e.g., 3-month survival) were modeled as dependent variables, with AFP group as the primary independent variable. Covariates included age, sex, liver failure subtype, and baseline liver function parameters [albumin, bilirubin, and prothrombin activity (PTA)]. Adjusted odds ratios (ORs) with 95% confidence intervals (CIs) are reported.

RESULTS

Baseline characteristics of patients

This study enrolled 194 patients with liver failure (aged 15-65 years, mean age 40.6 ± 13.2 years), including 115 males (59.3%) and 79 females (40.7%), with a mean hospitalization duration of 23.5 ± 12.1 days (Table 1). Based on AFP levels, patients were stratified into three groups for baseline analysis: AFP < 100 ng/mL (n = 60), 100-200 ng/mL (n = 70), and > 200 ng/mL (n = 64) (Table 2). Laboratory indicators prior to treatment, including blood ammonia (μmol/L), albumin (g/L), total bilirubin (μmol/L), alanine aminotransferase (ALT) (U/L), aspartate aminotransferase (AST) (U/L), γ-glutamyl transferase (γ-GT) (U/L), and PTA (%), did not differ significantly among the three groups (P > 0.05; Table 3).

Table 1 Baseline data of patients with liver failure, n (%).

	Parameter total cases (n = 194)	P value1
Age, range, mean ± SD (years)	15-65 (40.6 ± 13.2)	0.382
Sex		0.724
Male	115 (59.3)
Female	79 (40.7)
Hospital stays (days)	23.5 ± 12.1	0.563

¹The t-test or analysis of variance was used for comparison between groups.

The χ² test or Fisher’s exact probability method was used for comparison between groups. P > 0.05 indicates that there is no statistically significant difference among the groups.

Table 2 Comprehensive analysis of general data for patients with liver failure categorized by different alpha-fetoprotein levels.

Group	Cases	Gender (n, male/female)	Age, mean ± SD (years)	P value1
Low AFP group (< 100 ng/mL)	60	36/24	41.6 ± 11.8	> 0.05
Medium AFP group (100-200 ng/mL)	70	40/30	40.9 ± 10.9	> 0.05
High AFP group (> 200 ng/mL)	64	39/25	40.5 ± 10.2	> 0.05

¹P > 0.05 indicates that there were no statistically significant differences in gender and age among the three groups (χ² test was used to compare gender, and one-way analysis of variance was used to compare age).

AFP: Alpha-fetoprotein.

Table 3 Laboratory parameters measured before treatment in different alpha-fetoprotein level groups, mean ± SD.

Parameter	Low AFP group	Medium AFP group	High AFP group	P value1
Blood ammonia (μmol/L)	188.5 ± 50.1	185.2 ± 48.3	190.3 ± 52.7	0.8216
Albumin (g/L)	25.8 ± 3.9	26.1 ± 3.7	25.6 ± 4.0	0.9308
Total bilirubin (μmol/L)	145.3 ± 35.2	142.6 ± 33.8	148.1 ± 36.5	0.8627
ALT (U/L)	265.4 ± 70.8	260.2 ± 68.1	268.7 ± 72.3	0.8795
AST (U/L)	225.6 ± 60.3	220.3 ± 58.4	229.1 ± 61.9	0.8998
γ-GT (U/L)	485.2 ± 120.6	478.3 ± 118.1	492.7 ± 123.8	0.9339
PTA (%)	28.3 ± 5.1	29.1 ± 5.4	27.8 ± 4.9	0.9657

¹The t-test or analysis of variance was used for comparison between groups. P > 0.05 indicates that there were no statistically significant differences.

AFP: Alpha-fetoprotein; ALT: Alanine aminotransferase; AST: Aspartate aminotransferase; γ-GT: γ-glutamyl transferase; PTA: Prothrombin activity.

Stratified analysis of AFP levels

Stratification by liver failure subtype revealed significant differences in AFP distribution across groups (all P < 0.01). Among the 24 patients with ALF, 62.5% (15/24) were categorized in the medium-AFP group (100-200 ng/mL). SALF cases (n = 38) were predominantly classified in the high-AFP group, representing 47.4% (18/38). In contrast, the ACLF cohort (n = 89) and the CLF cohort (n = 43) primarily exhibited low-AFP levels, accounting for 50.6% (45/89) and 65.1% (28/43), respectively. All subgroup comparisons were statistically significant: ALF (P = 0.004), SALF (P = 0.008), ACLF (P < 0.001), and CLF (P < 0.001; Table 4).

Table 4 Relationship between alpha-fetoprotein levels and clinical classification of liver failure, n (%).

Type	Cases (n)	Low AFP group	Medium AFP group	High AFP group	P value1
ALF	24	5 (20.8)	15 (62.5)	4 (16.7)	0.004
SALF	38	5 (13.2)	15 (39.5)	18 (47.4)	0.008
ACLF	89	45 (50.6)	30 (33.7)	14 (15. 7)	< 0.001
CLF	43	28 (65.1)	10 (23.3)	5 (11.6)	< 0.001

¹The χ² test was conducted to assess the distributional differences across various liver failure classifications within each alpha-fetoprotein group. P < 0.05 suggests that the differences in distribution among alpha-fetoprotein groups regarding disease types are statistically significant.

AFP: Alpha-fetoprotein; ALF: Acute liver failure; SALF: Subacute liver failure; ACLF: Acute-on-chronic liver failure; CLF: Chronic liver failure.

AFP stratification dominates multivariable survival prediction in liver failure, independent of hepatic reserve markers

Multivariable logistic regression analysis identified AFP stratification as the most significant independent predictor of 3-month survival. Compared with the low-AFP group, the high-AFP group had a significantly lower risk of mortality (OR = 0.325, 95%CI: 0.148-0.714; P < 0.001), and the medium-AFP group also demonstrated a protective effect (OR = 0.44, 95%CI: 0.27-0.71; P < 0.001). Among liver failure subtypes, both ALF (OR = 0.36, 95%CI: 0.20-0.66; P = 0.001) and SALF (OR = 0.46, 95%CI: 0.26-0.81; P = 0.007) were associated with improved survival compared with ACLF. Several key markers of hepatic reserve independently predicted clinical outcomes: Albumin (per 1 g/L increase: OR = 0.91, 95%CI: 0.88-0.95; P < 0.001), PTA (per 1% increase: OR = 0.94, 95%CI: 0.92-0.96; P = 0.001), and blood ammonia (per 1 μmol/L increase: OR = 1.01, 95%CI: 1.00-1.01; P = 0.002). Age and bilirubin levels were also significantly associated with survival (P < 0.01), whereas sex and transaminase levels showed no statistically significant association (Table 5).

Table 5 Statistical analysis of 3-month survival rate using a multiple logistic regression model.

Variable	Three-month survival status
	OR1	95%CI	P value2	Significance
Age (per additional year)	1.03	1.01-1.05	0.005
Gender (male vs female)	0.81	0.57-1.15	0.241	NS
AFP grouping
Medium vs low	0.44	0.27-0.71	< 0.001
High vs low	0.325	0.148-0.714	< 0.001
Types of liver failure
ALF vs ACLF	0.36	0.20-0.66	0.001
SALF vs ACLF	0.46	0.26-0.81	0.007
CLF vs ACLF	0.64	0.33-1.23	0.178	NS
Liver function indicators
Blood ammonia (μmol/L)	1.01	1.00-1.01	0.002
Albumin (g/L)	0.91	0.88-0.95	< 0.001
Total bilirubin (μmol/L)	1.01	1.00-1.01	0.004
ALT (U/L)	1.01	0.98-1.03	0.467	NS
AST (U/L)	1.00	0.99-1.03	0.226	NS
γ-GT (U/L)	1.00	0.99-1.01	0.167	NS
PTA (%)	0.94	0.92-0.96	0.001

¹Odds ratio indicates the relative change in the likelihood of mortality associated with a one-unit increase (or category shift) in the independent variable.

²P < 0.05 indicates the presence of statistically significant differences, P > 0.05, not significant. Reference group: Liver failure type, acute-on-chronic liver failure; alpha-fetoprotein group, low alpha-fetoprotein group. Nagelkerke R² = 0.458; model χ² = 67.42 (P < 0.001).

OR: Odds ratio; CI: Confidence interval; AFP: Alpha-fetoprotein; ALF: Acute liver failure; ACLF: Acute-on-chronic liver failure; SALF: Subacute liver failure; CLF: Chronic liver failure; ALT: Alanine aminotransferase; AST: Aspartate aminotransferase; γ-GT: γ-glutamyl transferase; PTA: Prothrombin activity; NS: Not significant.

Dynamic changes in AFP levels at different time points

Patients were categorized into three groups based on baseline AFP levels: Low (< 100 ng/mL), medium (100-200 ng/mL), and high (> 200 ng/mL). As intended by stratification, no intergroup difference was observed at baseline (P = 0.984), but significant differences emerged from day 10 onward (P < 0.05). Time-dependent trajectories: High-AFP group: AFP levels increased steadily from 320.5 ng/mL to 395.6 ng/mL, with a linear daily increase rate of 3.8 ng/mL/day (P < 0.001). Medium-AFP group: AFP levels rose from 148.3 ng/mL to 205.4 ng/mL, corresponding to a daily increase rate of 2.8 ng/mL/day (P < 0.001), representing a less pronounced increase compared with the high-AFP group. Low-AFP group: AFP levels exhibited minimal progression, rising from 65.2 ng/mL to 98.3 ng/mL, with a daily increase rate of 1.6 ng/mL/day (P < 0.001; Table 6).

Table 6 Serial alpha-fetoprotein levels measured at different time points, mean ± SD.

Group	Before treatment	Day 1 after treatment	Day 10 after treatment	Day 20 after treatment	Day 25 after treatment	P value
Low AFP group	65.2 ± 20.1	72.5 ± 18.3	85.4 ± 22.6	92.1 ± 25.4	98.3 ± 28.7	< 0.001
Medium AFP group	148.3 ± 35.6	160.8 ± 40.2	175.6 ± 45.3	190.2 ± 50.1	205.4 ± 55.8	< 0.001
High AFP group	320.5 ± 80.4	340.2 ± 85.6	365.7 ± 90.3	380.1 ± 95.2	395.6 ± 98.5	< 0.001
P value1	0.984	0.932	0.018	0.006	0.003

¹Inter-group P value, the inter-group comparison P value represents the overall differences among the three groups at each time point, and the trend P value represents the statistical significance of the time-dependent changes in each group. P < 0.05 indicates the presence of statistically significant differences.

AFP: Alpha-fetoprotein.

Correlation between the frequency of PE treatments and AFP levels

Variation in treatment sessions: Patients with high baseline AFP levels required significantly fewer PE treatment sessions compared with other groups (2.8 ± 1.0 vs 4.2 ± 1.5 sessions in the low-AFP group, P < 0.001 vs 3.5 ± 1.2 sessions in the medium-AFP group, P = 0.009). Additionally, the medium-AFP group underwent significantly fewer sessions than the low-AFP group (3.5 ± 1.2 vs 4.2 ± 1.5 sessions, P = 0.003).

Hierarchical AFP increase: The high-AFP group demonstrated the greatest absolute post-treatment increase in AFP levels (75.1 ± 20.3 ng/mL), which was significantly greater than that of the medium-AFP group (57.1 ± 15.6 ng/mL, P = 0.009) and the low-AFP group (33.1 ± 10.2 ng/mL, P < 0.001). A strong inverse correlation was observed between the magnitude of AFP elevation and the number of treatment sessions required (P < 0.001; Table 7).

Table 7 Relationship between the number of plasma exchange treatments and alpha-fetoprotein levels, mean ± SD.

Group	Treatment sessions	AFP increase	P value (within-group1vs baseline)	P value (between-group2vs other groups)
Low AFP group	4.2 ± 1.5	33.1 ± 10.2	< 0.001	Reference
Medium AFP group	3.5 ± 1.2	57.1 ± 15.6	< 0.001	vs low: 0.003; vs high: 0.012
High AFP group	2.8 ± 1.0	75.1 ± 20.3	< 0.001	vs low: < 0.001; vs medium: 0.009

¹Within-group, significant change from baseline alpha-fetoprotein values.

²Between-group, significant differences in treatment response magnitude between specified groups. P < 0.05 indicates the presence of statistically significant differences.

AFP: Alpha-fetoprotein.

Correlation analysis of laboratory indicators (last time) and AFP levels after PE treatment

Systematic analysis of final liver biomarker measurements revealed significant differences among cohorts stratified by post-treatment AFP levels. Patients with high baseline AFP (> 200 ng/mL) consistently demonstrated superior hepatic function profiles compared with those with lower AFP levels: Ammonia levels were significantly lower in the high-AFP group than in the low-AFP group (75.8 ± 25.1 μmol/L vs 125.6 ± 45.3 μmol/L; P = 0.001). Albumin concentrations were markedly higher in the high-AFP group (35.6 ± 4.2 g/L vs 28.3 ± 4.1 g/L; P < 0.001). Total bilirubin was substantially reduced in the high-AFP group (28.7 ± 12.6 μmol/L vs 68.5 ± 22.1 μmol/L; P < 0.001).

Markers of hepatocellular injury showed progressive improvement with increasing AFP levels: ALT decreased stepwise from the low-AFP group to the high-AFP group (180.5 ± 60.2 U/L to 120.8 ± 48.3 U/L; low vs high, P = 0.006). AST exhibited a similar decline (155.4 ± 50.7 U/L to 95.3 ± 38.4 U/L; low vs high, P < 0.001). γ-GT also demonstrated a concentration-dependent reduction (320.6 ± 105.4 U/L to 185.2 ± 75.3 U/L; low vs high, P = 0.002). A distinct pattern was observed for PTA: The medium-AFP group showed significantly impaired coagulation function (35.2% ± 8.6%), which was markedly lower than both the low-AFP group (55.4% ± 10.2%; P < 0.001) and the high-AFP group (72.8% ± 12.4%; P < 0.001; Table 8).

Table 8 Comparison of laboratory indicators (last time) among different alpha-fetoprotein level categories after treatment, mean ± SD.

Parameter	Low AFP group	Medium AFP group	High AFP group	Overall P value (ANOVA)1	Post-hoc test (Tukey HSD)2
Blood ammonia (μmol/L)	125.6 ± 45.3	98.4 ± 30.2	75.8 ± 25.1	0.003	Low vs med: 0.045
					Low vs high: 0.001
					Med vs high: 0.357
Albumin (g/L)	28.3 ± 4.1	32.5 ± 3.8	35.6 ± 4.2	0.001	Low vs med: 0.003
					Low vs high: < 0.001
					Med vs high: 0.032
Total bilirubin (μmol/L)	68.5 ± 22.1	45.2 ± 18.3	28.7 ± 12.6	< 0.001	Low vs med: 0.001
					Low vs high: < 0.001
					Med vs high: 0.017
ALT (U/L)	180.5 ± 60.2	150.3 ± 55.6	120.8 ± 48.3	0.012	Low vs med: 0.213
					Low vs high: 0.006
					Med vs high: 0.058
AST (U/L)	155.4 ± 50.7	120.6 ± 45.2	95.3 ± 38.4	0.005	Low vs med: 0.045
					Low vs high: < 0.001
					Med vs high: 0.038
γ-GT (U/L)	320.6 ± 105.4	240.8 ± 90.5	185.2 ± 75.3	0.008	Low vs med: 0.041
					Low vs high: 0.002
					Med vs high: 0.128
PTA (%)	55.4 ± 10.2	35.2 ± 8.6	72.8 ± 12.4	< 0.001	Low vs med: < 0.001
					Low vs high: 0.032
					Med vs high: < 0.001

¹ANOVA was conducted to compare parameters across alpha-fetoprotein groups, followed by Tukey’s honestly significant difference post-hoc tests for pairwise comparisons. Overall P value indicates significant inter-group differences for the parameter.

²Post-hoc P values highlights significant differences between specified alpha-fetoprotein subgroups.

AFP: Alpha-fetoprotein; HSD: Honestly significant difference; ALT: Alanine aminotransferase; AST: Aspartate aminotransferase; γ-GT: γ-glutamyl transferase; PTA: Prothrombin activity; med: Medium.

Correlation between the incidence of complications following PE treatment and AFP levels

Significant differences in complication rates were observed across AFP-stratified groups. The low-AFP group consistently demonstrated the highest incidence of adverse events: Hypersensitivity reactions occurred in 12.5% of patients (vs 3.1% in the high-AFP group; P = 0.038). Hypotension was observed in 18.8% (vs 6.3% in the high-AFP group; P = 0.009), and SBP occurred in 25.0% (vs 9.4% in the high-AFP group; P = 0.003; Table 9).

Table 9 Correlation between the incidence of complications and alpha-fetoprotein levels,%.

Complication type	Low AFP group	Medium AFP group	High AFP group	P value1	Post-hoc test (Bonferroni correction)2
Allergic reaction	12.5	6.3	3.1	0.042	Low vs high: 0.038
					Low vs med: 0.215
					Med vs high: 0.417
Hypotension	18.8	12.5	6.3	0.023	Low vs high: 0.009
					Low vs med: 0.137
					Med vs high: 0.254
SBP	25.0	15.6	9.4	0.011	Low vs high: 0.003
					Low vs med: 0.062
					Med vs high: 0.168

¹The distribution differences of complications among alpha-fetoprotein groups were analyzed using χ² test or Fisher’s exact test. Overall P value: Significant differences exist in complication rates across Alpha-fetoprotein subgroups.

²Post-hoc pairwise comparisons with Bonferroni correction were performed when overall P < 0.05. Post-hoc P values: Significant differences between specific group pairs after multiple comparison adjustment.

AFP: Alpha-fetoprotein; SBP: Spontaneous bacterial peritonitis; med: Medium.

Comparison of 3-month follow-up survival rates

Survival rates differed significantly across AFP strata. The low-AFP group had survival rates of 70.0% at discharge and 56.7% at 3 months. The medium-AFP group showed significantly higher rates of 81.4% at discharge and 72.9% at 3 months (P < 0.001). The high-AFP group demonstrated the highest survival, with 96.9% at discharge and 90.6% at 3 months (P < 0.001; Table 10, Figure 2A). Based on pre-treatment AFP values, 3-month post-treatment predictive discrimination was greatest in the high-AFP group (AUC = 0.8851, cutoff value = 0.7414), followed by the medium-AFP group (AUC = 0.8039, cutoff value = 0.5810). The low-AFP group had relatively weaker predictive ability (AUC = 0.7051, cutoff value = 0.4629) (Figure 2B).

Figure 2 Kaplan-Meier estimates of 3-month survival and areas under the receiver operating characteristic curve calculated for each group at 3 months post-treatment. A: This figure presents Kaplan-Meier survival curves illustrating the 3-month survival probabilities across the study groups stratified by serum alpha-fetoprotein (AFP) levels (low AFP group, medium AFP group, and high AFP group). The curves visualize differences in survival outcomes among the groups over the 3-month follow-up period; B: This figure displays receiver operating characteristic curves with corresponding areas under the curve values calculated for each group (low AFP group, medium AFP group, and high AFP group) at 3 months post-treatment. The areas under the curve values quantify the diagnostic performance of serum AFP levels in predicting outcomes at the 3-month time point. AFP: Alpha-fetoprotein; AUC: Areas under the curve.

Table 10 Comparison of 3-month follow-up survival rates,%.

Group	Survival rate at discharge	3-month survival rate	P value1
Low AFP group	70.0	56.7	< 0.001
Medium AFP group	81.4	72.9	< 0.001
High AFP group	96.9	90.6	< 0.001

¹P values compare differences in survival rates among groups using the log-rank test (P < 0.001).

AFP: Alpha-fetoprotein.

DISCUSSION

Comparability of baseline characteristics

Baseline characteristics of the 194 patients were comparable across AFP strata in terms of demographics and hospitalization (all P > 0.05), controlling for potential confounding factors. Etiologies reflected the patterns reported in the Asia-Pacific region[24,25], with ACLF (45.9%) and CLF (22.1%) predominating. Notably, the CLF and ACLF cohorts were largely characterized by low-AFP levels (65.1% and 50.6%, respectively), suggesting impaired hepatic regeneration. In contrast, SALF cases exhibited a predominance of high-AFP (47.4%), indicative of preserved regenerative capacity, whereas ALF patients were primarily categorized within the medium-AFP group (62.5%), consistent with intermediate regenerative dynamics. Multivariate analysis confirmed that AFP retains significant prognostic value, thereby establishing its role as an independent predictor.

AFP dynamics predict clinical outcomes

Dynamic AFP trajectories reflect regenerative potential: Serial measurement of AFP following PE therapy revealed a distinct temporal pattern. The high-AFP group demonstrated sustained post-treatment increases (320.5 ± 48.2 ng/mL to 395.6 ± 53.7 ng/mL; P < 0.05), potentially attributable to receptor-mediated co-activation of the Wnt/β-catenin and signal transducer and activator of transcription 3 pathways-mechanisms known to regulate hepatocyte progenitor expansion and differentiation[26]. These kinetic changes align with preclinical evidence linking elevated AFP levels to hepatic progenitor cell proliferation[27]. Importantly, patients in the high-AFP group achieved significantly greater AFP increments (75.1 ng/mL) with fewer PE sessions (2.8 ± 1.0 vs 4.2 ± 1.5; P = 0.008). This inverse correlation between AFP response and therapeutic intensity supports the “AFP threshold hypothesis”, in which AFP levels exceeding 200 ng/mL indicate competent endogenous capacity to orchestrate liver regeneration via paracrine signaling and niche remodeling[20].

Elevated AFP levels promote multidimensional hepatic restoration: Synthetic function progressively improved, with albumin levels rising from 28.3 ± 3.1 g/L in the low-AFP group to 35.6 ± 4.2 g/L in the high-AFP group (P = 0.001). Detoxification capacity was significantly enhanced in the high-AFP group compared with the low-AFP group, as evidenced by reductions in ammonia (125.6 ± 18.7 μmol/L vs 75.8 ± 12.3 μmol/L; P = 0.003) and bilirubin (68.5 ± 10.2 μmol/L vs 28.7 ± 6.5 μmol/L; P < 0.001), suggesting AFP-mediated upregulation of urea cycle enzymes[28]. Longitudinal kinetics further demonstrated AFP-dependent restoration: High-AFP patients showed greater albumin increase (+8.7 ± 2.1 g/L vs +3.2 ± 1.5 g/L in low-AFP; P = 0.003), bilirubin decline (119.4 ± 38.6 μmol/L vs 76.8 ± 41.6 μmol/L in low-AFP; P < 0.001), and accelerated γ-GT reduction (-57.6 ± 18.4 U/L, P = 0.005 vs baseline). Coagulation function was markedly enhanced with elevated AFP (PTA: 72.8% ± 12.4% vs 55.4% ± 10.2% in low-AFP; P < 0.001), consistent with known AFP-coagulation interactions[29].

Cytoprotective effects were also evident, with significant reductions in γ-GT (320.6 ± 45.1 U/L to 185.2 ± 30.8 U/L; P = 0.008), ALT (180.5 ± 25.3 U/L to 120.8 ± 18.9 U/L; P = 0.012), and AST (P < 0.001 across groups), indicating attenuated hepatocyte necrosis and biliary repair. Notably, the medium-AFP group exhibited reduced PTA (35.2% ± 8.6%), potentially reflecting imbalanced coagulation factor synthesis during rapid hepatocyte proliferation, where accelerated consumption transiently outpaces production, particularly for short half-life factors such as factor VII (t_1/2 = 6 hours)[30,31]. This transient synthetic dysfunction highlights the complexity of coagulation dynamics during hepatic regeneration and warrants further investigation.

Clinical significance of complications and survival: Patients with high AFP levels exhibited a dose-dependent reduction in complications compared with their low-AFP counterparts: Allergic reactions (3.1% vs 12.5%; P = 0.038), hypotension (6.3% vs 18.8%; P = 0.009), and SBP (9.4% vs 25.0%; P = 0.003), with the medium-AFP group demonstrating intermediate rates. Notably, no patients underwent liver transplantation during follow-up, confirming that all outcomes reflect native liver pathophysiology. This protective association persisted after statistical adjustment. The inverse association with SBP aligns with evidence that hepatocyte regeneration enhances innate immunity[32].

Furthermore, the survival gradient, from 56.7% in the low-AFP group to 72.9% in the medium-AFP group and 90.6% in the high-AFP group at 3 months (P < 0.001), paralleled progressive improvements in prognostic discrimination (AUC: 0.7051, 0.8039, 0.8851). This concordance supports the conclusion that baseline AFP intrinsically stratifies both survival probability and prognostic precision.

Limitations and future directions

Although this study demonstrated adequate statistical power, external validation in larger, prospective, multicenter cohorts remains essential. Future research should integrate serial AFP kinetics with complementary regenerative biomarkers, particularly epithelial cell adhesion molecule positive hepatic progenitor cells and dynamic microRNA profiles such as microRNA-122, to develop multidimensional predictive models. These models should be evaluated in parallel against established prognostic instruments (e.g., model for end-stage liver disease and sequential organ failure assessment scores) to quantify incremental clinical value using net reclassification improvement analyses. Furthermore, mechanism-based interventions targeting AFP-associated pathways (e.g., pharmacological Wnt/β-catenin agonists) should undergo rigorous preclinical evaluation as promising candidates for regenerative medicine applications.

CONCLUSION

This study establishes serial AFP measurement as a key biomarker for patients with liver failure undergoing PE. AFP enables effective stratification of regenerative capacity to guide initial PE intensity, predicts the kinetics of functional recovery after PE, identifies individuals at high risk for complications or mortality, and highlights candidates for therapy de-escalation. Importantly, dynamic AFP monitoring (days 1/10/20/25) provides real-time guidance for therapeutic escalation or de-escalation. Incorporating AFP dynamics into clinical decision-making enhances prognostic accuracy, optimizes resource utilization, and improves survival outcomes.