Serum Fibroblast Growth Factor-21 Discriminates Between Decompensated Alcohol-Associated Cirrhosis and Severe Alcohol-Associated Hepatitis

Abstract

INTRODUCTION: 

We hypothesized that fibroblast growth factor-21 (FGF-21) would be highly expressed in patients with alcohol-associated hepatitis (AH) and could be a novel and biologically relevant predictive biomarker to reliably distinguish severe AH and decompensated alcohol-associated cirrhosis (AC).

METHODS: 

We identified a discovery cohort of 88 subjects with alcohol-associated liver disease (ALD) of varying disease severity from our ALD repository. Our validation cohort consisted of 37 patients with a biopsy-proven diagnosis of AH, AC, or absence of ALD with Model for End-Stage Liver Disease scores ≥10. Serum from both groups during index hospitalization was assayed for FGF-21 by ELISA. We performed receiver operating characteristic analysis and prediction modeling in both cohorts to discriminate between AH and AC in high Model for End-Stage Liver Disease (≥20) patients.

RESULTS: 

In both cohorts, FGF-21 concentrations were highest in subjects with moderate to severe AH compared with those having alcohol use disorder or AC (mean: 2,609 pg/mL, P < 0.0001). The discovery cohort area under the curve of FGF-21 between AH and AC was 0.81 (95% confidence interval: 0.65–0.98, P < 0.01). In the validation cohort, FGF-21 levels were higher in severe AH compared with AC (3,052 vs 1,235 pg/mL, P = 0.03), and the area under the curve was 0.76 (95% confidence interval: 0.56–0.96, P < 0.03). A survival analysis showed that patients with FGF-21 serum levels in the second interquartile had the highest survival compared with all other quartiles.

DISCUSSION: 

FGF-21 performs well as a predictive biomarker to distinguish severe AH from AC and may be helpful in the management and clinical investigation of patients with severe alcohol-associated liver diseases.

INTRODUCTION

Since 2000, cirrhosis-related mortality in the United States has increased by 31%, despite the decline in hepatitis B-related and C-related liver disease (¹). Alcohol-associated liver disease (ALD) and nonalcoholic fatty liver disease (NAFLD) have emerged as the primary indications for liver transplantation in the United States in an increasingly young population (²). Of particular concern is the increased incidence of alcohol-associated hepatitis-related hospitalizations since the early 2000s, as reported in several single-center studies (³). Severe alcohol-associated hepatitis (AH) represents a fulminant phenotype of ALD known to have a high 90-day mortality rate with no highly effective therapeutics to date (⁴). Two significant limitations in the effective management of AH include challenges in accurately diagnosing AH vs decompensated alcohol-associated cirrhosis (AC) and few identified effective therapeutic targets (⁵). Clinically, both AC and AH can present with high Model for End-Stage Liver Disease (MELD) and portal hypertension; further, AH can be superimposed on established cirrhosis. Therefore, in the absence of a biopsy, clinical features are often insufficient to diagnose AH accurately, and available biopsy data suggest that approximately 15% of patients are incorrectly diagnosed (⁶). It is, therefore, urgently necessary to identify noninvasive biomarkers that can aid the diagnosis and prognostication of patients with ALD and presumed AH.

Mechanisms of ethanol-induced liver injury have been studied in animal and human models, and the pathobiology of steatosis, inflammation, and eventual fibrosis involves numerous processes. A subclass of fibroblast growth factors (FGFs-19, 21, and 23) lacks the traditional heparin-binding domain (⁷), allowing for endocrine function (⁸). FGF-21, highly expressed in the liver, shows promise as a potential therapeutic target in metabolic diseases. FGF-21 transgenic mice on a ketogenic diet showed improved insulin sensitivity, glucose clearance, reduced body weight, and leaner hepatic tissue. By contrast, murine knockout FGF-21 models conversely demonstrated fatty liver, hyperlipidemia, and reduced ketone levels on a ketogenic diet (⁹).

In humans, there likewise seems to be an important relationship between hepatic steatosis and FGF-21, although the regulation of FGF-21 in humans remains incompletely defined. Several cross-sectional studies have consistently observed a positive correlation between obesity, impaired glucose tolerance, hyperlipidemia, biochemical liver injury, and increased circulatory FGF-21 (¹⁰). Despite animal models showing FGF-21 expression increases in response to a fasted or a ketogenic state, FGF-21 does not seem to be nutritionally regulated in human subjects (^9,10). In a small prospective cohort study, patients with histologically confirmed NAFLD and nonalcoholic steatohepatitis measured higher levels of FGF-21 by body mass index (BMI) and degree of steatosis. However, this effect was less pronounced in subjects with nonalcoholic steatohepatitis, suggesting advanced fibrosis may impair response (¹⁰). Thus, although regulatory mechanisms remain incompletely elucidated, FGF-21 has been tightly linked to NAFLD.

Emerging evidence also suggests a critical intersection between FGF-21 and ALD. Recently, a study of human subjects highlighted a 40-fold elevation of circulating FGF-21 in response to binge alcohol consumption without playing a role in ethanol clearance or hepatic metabolism (¹¹). Several animal FGF-21 knockout models demonstrated severe hepatic steatosis, inflammation, and injury in response to chronic alcohol exposure compared with wild-type mice (¹²). Finally, prospective cohort studies have observed that exogenous administration of FGF-21 leads to reduced alcohol consumption in murine models and can limit the development of alcohol-induced liver injury. Notwithstanding, its effect in humans with alcohol-induced liver injury is less clearly elucidated (^11–13).

We hypothesized that FGF-21 expression becomes dysregulated in the setting of AH and could contribute to AH pathogenesis. In this article, we investigate whether circulating FGF-21 levels can predict ALD disease state and severity. In this study, we used a clinically characterized discovery cohort and a pathology-defined validation cohort to determine the risk factors and outcomes associated with FGF-21 expression. We further sought to determine whether this potential biologically relevant biomarker would yield a highly accurate diagnostic tool to distinguish between AH and decompensated AC in patients with similar disease severity by the MELD score.

METHODS

The Massachusetts General Hospital Alcohol Liver Disease Repository is a prospectively enrolled cohort of patients with ALD. The cohort enrolled 500 hospitalized patients admitted to a tertiary hospital center from 2019 to 2021 with a history of alcohol use disorder (AUD) and clinically defined ALD. Patients were included if they were older than 18 and could provide consent. All subjects for this study were selected from this prospective cohort. Subjects were excluded if serum testing or histopathology (validation cohort only) was unavailable within 6 months of enrollment or if a secondary etiology for the chronic liver disease had previously been diagnosed. The hospital institutional review board (protocol number 2010P000220) approved this study, and all patients provided informed consent.

Documentation supporting a diagnosis of AUD by self-report and physician assessment was paired with evidence of liver injury to diagnose ALD clinically in patients enrolled. To this end, serum analysis, ultrasound or cross-sectional imaging, and biopsies were collected from the ALD biorepository. To determine the effect of current alcohol consumption, data from a chart review regarding the subjects' last alcohol use within 2 weeks were collected in the discovery cohort and corroborated with serum ethanol levels or urine ethyl glucuronide. The MELD score stratified the severity of the liver disease. The subgroups consecutively enrolled in the discovery cohort included controls without liver disease; patients with AUD without clinically apparent liver injury (normal aminotransferases); mild AH as evidence of an alcohol-associated liver injury with abnormal aminotransferases and MELD scores less than or equal to 10; moderate AH was diagnosed using the National Institute on Alcohol Abuse and Alcoholism (NIAAA) criteria of AH (onset of jaundice within the prior 8 weeks with preexisting heavy use of alcohol (>60 g/d in men and >40 g/d in women)), hyperbilirubinemia greater than 3 mg/dL, aspartate transaminase (AST): alanine transaminase (ALT) >1.5, and MELD score <20; severe AH was defined by NIAAA diagnostic inclusion criteria for clinical trials with MELD scores of greater or equal to 20; and AC was defined by known history of cirrhosis, imaging features of cirrhosis with portal hypertension, and/or with a decompensating event, in the absence of clinical features of AH (¹⁴). Age-matched and sex-matched patients diagnosed by liver biopsy with NAFLD were identified from the [hospital] NAFLD tissue repository for comparison with the ALD cohort. Clinical diagnoses were made based on a careful review of the medical record by 2 independent physicians.

Discovery cohort

Descriptive analyses of a pilot cohort were performed to ensure an adequate sample size and the establishment of minimum thresholds for FGF-21 values between ALD phenotypes. Fasting sera were collected in the morning in all patients. All samples were stored at −80°C. We then used the serum for quantitative measurement of the hepatokine FGF-21 using enzyme-linked immunosorbent assays (EMP Millipore FGF21 ELISA, Billerica, MA) as per manufacturer instructions. The minimum detectable concentration was 15 pg/mL. All measurements were performed in duplicate, with the results averaged. The processors were blind to the clinical diagnoses of the respective patients.

Validation cohort

To validate our findings, we selected 37 patients from the [hospital] ALD repository with complete clinical data and liver biopsies performed within 6 months of enrollment and initial serum collection. All subjects in this cohort underwent biopsy according to routine clinical care, and all subjects had MELD ≥10. Because of timing, these patients were enrolled and had sera accessed after the data collection of the discovery cohort. Based on the observed difference between FGF-21 values between controls, patients with severe AH, and patients with AC in the serum cohort, the sample size of the biopsy cohort was deemed adequate to complete the analysis. Two independent physicians performed chart reviews to categorize the diagnosis of interest, characteristics of biopsies, and outcomes. A third independent physician reviewed and adjudicated each classification based on histopathology alone. The histopathologic diagnosis of AH was defined by the NIAAA diagnostic criteria and the presence of hepatocyte ballooning with or without Mallory-Denk bodies, neutrophil-predominant parenchymal infiltration on a background of steatosis ± fibrosis. If established cirrhosis was observed on biopsy without evidence of steatohepatitis, these subjects were classified as having AC (^15,16). Patients with evidence of AH and advanced fibrosis or cirrhosis were pooled within the AH cohort. In this cohort, some subjects had undergone biopsy, which then demonstrated no evidence of alcohol-associated liver disease (no ALD). All samples were initially processed and stored at −80°C. The sera were subsequently used for quantitative measurement of the hepatokine FGF-21 by the EMD Millipore ELISA kit. All our assays were duplicated, and the absorbance was determined using a microplate reader.

Statistical analyses

Descriptive statistics were calculated for both cohorts, with interquartile ranges reported for the expression of FGF-21. Categorical variables were reported as frequencies and percentages. Our discovery cohort used ANOVA (Kruskal-Wallis for nonparametric data) to determine significant differences in FGF-21 expression between our clinical groups. t Tests and Wilcoxon rank-sum tests were used to determine significant differences between clinical phenotypes for the demographic data in Table 1. We used similar parametric and nonparametric tests for FGF-21 expression in the validation cohort. Where applicable, missing data were addressed by inputting the median value within each clinical phenotype (¹⁷). Spearman rank correlations were used to evaluate relationships between FGF-21 and prospective modifiable risk factors, including age, BMI, AST levels, and MELD scores.

Table 1. - Demographics of the discovery cohort

	Control (n = 4)	AUD (n = 19)	Mild AH (n = 20)	ModSev AH (n = 29)	AC (n = 16)	Overall (n = 88)	P < 0.05
Age (yr), mean ± SD	45.5 ± 10.8	46.3 ± 11.8	51.4 ± 12.3	46.1 ± 10	56.3 ± 9.06	51.9 ± 12	0.007
Sex (% female) (SD)	0 (0)	21.1 (0.42)	20.0 (0.41)	41.1 (0.50)	25.0 (0.45)	27.3 (0.45)	0.97
Race, n (%)
White	4 (100)	16 (84.2)	15 (75)	24 (82.8)	14 (87.5)	73 (83)
Other	0 (0)	3 (15.8)	2 (10)	3 (10.3)	2 (12.5)	10 (11.4)
Asian	0 (0)	0 (0)	2 (10)	1 (3.4)	0 (0)	3 (3.4)
Black	0 (0)	0 (0)	1 (5)	1 (3.4)	0 (0)	2 (2.3)
Ethnicity, n (%)
Hispanic	1 (25)	1 (5.3)	2 (10)	2 (6.9)	2 (12.5)	8 (9.1)
Non-Hispanic	3 (75)	18 (94.7)	18 (90)	26 (89.7)	14 (87.5)	79 (89.8)
Unknown	0	0 (0)	0 (0)	1 (3.4)	0 (0)	1 (1.1)
BMI (kg/m2), mean ± SD	29.3 ± 3.8	25.8 ± 5.9	24.9 ± 5.4	30 ± 9.0	25.7 ± 3.8	27.1 ± 6.9	0.065
ALT (U/L), mean ± SD	32.8 ± 8.1	31.3 ± 16.2	89.6 ± 91.8	62.7 ± 98.2	44 ± 37.1	57.3 ± 75.5	0.017
AST (U/L), mean ± SD	26.5 ± 5.0	40.3 ± 24.6	115 ± 79.8	127 ± 64	81.8 ± 48.5	92.8 ± 68	<0.001
MELD score, median (range)	6.5 (6–8)	7.1 (6–11)	6.9 (6–9.4)	22.4 (11–45)	11.1 (7–21)	9.5 (6–45)	<0.001

AC, alcohol-associated cirrhosis; AH, alcohol-associated hepatitis; ALT, alanine transaminase; AST, aspartate transaminase; AUD, alcohol use disorder; BMI, body mass index; MELD, Model for End-Stage Liver Disease.

We next characterized the subjects by FGF-21 concentration by interquartile rank at least above or below the 75% percentile. We performed prediction modeling to assess the discriminatory ability of FGF-21 to determine clinical diagnosis (AH vs AC) in both cohorts. We then plotted a Kaplan-Meier curve to complete a survival analysis. Multivariate logistic regression analysis was performed to adjust for MELD score, age, sex, BMI, and race/ethnicity. For outcomes, the primary end points were death or liver transplantation by the end of the study. The secondary end points were readmission at 3 or 6 months.

RESULTS

Discovery cohort

Ninety-six patients were identified from the [Massachusetts General Hospital] Alcohol Liver Disease Biorepository for the discovery cohort, and 88 of these participants were eligible for final analysis. Eight subjects were removed for the following reasons: meeting exclusion criteria, follow-up loss, or liver transplantation before enrollment and serum collection. The patients' characteristics are shown in Table 1. The median follow-up duration for the discovery cohort was 733 days, ranging from 2 to 935 days. The mean age of participants was 51.9 years (SD 12.0). Of the 96 patients, 64 (72.7%) were men. Female subjects (27.3%) were more frequently observed to have moderate to severe AH (41.4% of subjects), compared with AUD (12.5%) or AC (25%), but this variation was not statistically significant (P = 0.97, ANOVA). 17.2% of the cohort died or required liver transplantation within the study's follow-up period, and 40.9% of subjects were readmitted within 6 months.

The average age of patients in the discovery cohort varied by clinical phenotype (Table 1), and patients with AH were younger compared with patients who had AC (46.1 vs 56.3 years, P < 0.001). The BMI of subjects in the discovery cohort was not statistically different between groups (P = 0.065), although 51.7% of the cohort were considered overweight by their observed BMI. Differences within the cohort by clinical phenotype between AST, ALT, and MELD scores were statistically significant (Table 1).

FGF-21 concentrations significantly correlated with both MELD scores (P < 0.001) and AST values (P < 0.001) but not with ALT, age, or BMI (Figure 1a–d). We further explored for a relationship between BMI and FGF-21 expression by stratifying by clinical diagnosis in both cohorts, and it remained without association (see Supplementary Figure 1A and B, Supplementary Digital Content 1, https://links.lww.com/CTG/A928). Although FGF-21 concentrations trended higher in patients with 2 weeks of documented sobriety in both the AH and AC subgroups, this was however not statistically significant (P = 0.39). Similarly, in the validation cohort, no difference was observed (see Supplementary Figure 2A and B, Supplementary Digital Content 1, https://links.lww.com/CTG/A928).

Figure 1.: FGF-21 expression correlated with MELD scores (a) and AST (I/U) (b) but did not correlate with age (years) (c) or body mass index (d) in subjects of the discovery cohort (Kruskal-Wallis ANOVA P < 0.001, P = 0.429, and P = 0.960; used due to nonparametric FGF-21 expression). AST, aspartate transaminase; FGF-21, fibroblast growth factor-21.

Substantial differences in FGF-21 levels were observed between clinical phenotypes. FGF-21 expression was significantly higher in subjects with severe AH than in all other clinical phenotypes, with a median value of 2,124 pg/mL (P < 0.0001, Figure 2). In the discovery cohort, 5 of 31 subjects (16.1%) with clinically diagnosed moderate to severe alcoholic hepatitis had FGF-21 levels in the lowest 2 quartiles (<786.8 pg/mL); the remainder had FGF-21 equal to or greater than 787 pg/mL. In the biopsy-proven validation cohort, all the patients in the highest quartile (>2,124 pg/mL) had moderate to severe AH (see Supplementary Figure 3, Supplementary Digital Content 1, https://links.lww.com/CTG/A928). We also compared FGF-21 expression between our ALD cohort with patients with NAFLD and found that the expression in patients with NAFLD was lower than all ALD phenotypes (see Supplementary Figure 4, Supplementary Digital Content 1, https://links.lww.com/CTG/A928).

Figure 2.: FGF-21 expression is highest in patients with severe alcohol-associated hepatitis in the discovery cohort (ANOVA Р < 0.001). Here, AUD is alcohol use disorder; MildAH is alcohol-associated hepatitis (MELD <11); ModSev AH is alcohol-associated hepatitis (MELD ≥11); and AC is alcohol-related cirrhosis. Intergroup comparisons using the Tukey test for post hoc analysis showed a meaningful difference between control and AH, but not AUD or AC (P = 0.87), and a meaningful difference between severe AH, mild AH, AUD, and AC (difference P < 0.001). FGF-21, fibroblast growth factor-21; MELD, Model for End-Stage Liver Disease.

Validation cohort

To determine the ability of FGF-21 to discriminate between AH and AC, we identified subjects in our ALD repository who had undergone liver biopsy as part of indicated clinical care within 6 months of enrollment. Thirty-seven subjects met the inclusion criteria. Of these, 16 had pathology-defined AH, 15 had established AC, and 6 had no evidence of alcohol-associated liver disease (no ALD) (the clinical diagnoses are summarized in Supplementary Table 1, see Supplementary Digital Content 1, https://links.lww.com/CTG/A928).

The clinical characteristics of the validation cohort are shown in Table 2. The median follow-up duration was 215 days (3–464 days). Women comprised 35% of the total cohort and were frequently observed to have AH (50% women), compared with AC (28%) or no ALD (17%). Sixteen percent of the validation cohort died or required liver transplantation during the follow-up period, and 72.9% of subjects were readmitted to the hospital within 6 months of enrollment.

Table 2. - Demographics of the validation cohort

	AH	AC	No ALD	Overall	P
Age (yr), mean ± SD	45.4 ± 10.8	46.3 ± 11.8	50.3 ± 12.3	46.6 ± 11.1	0.66
Sex (% female) (SD)	50 (0.52)	28 (0.46)	17 (0.41)	35.1 (0.48)
Race, n (%)
Other	1 (6.3)	3 (23.1)	0 (0)	4 (10.8)
White	15 (93.8)	11 (73.3)	4 (66.7)	30 (81.1)
Asian	0 (0)	1 (6.7)	1 (16.7)	2 (5.4)
Black	0 (0)	0 (0)	1 (16.7)	1 (2.7)
Ethnicity, n (%)
Hispanic	1 (6.3)	2 (13.3)	0 (0)	3 (8.1)
Non-Hispanic	15 (93.8)	13 (86.7)	6 (100)	34 (91.9)
BMI (kg/m2), mean ± SD	29.8 ± 7.0	28.5 ± 5.7	23.4 ± 4.3	28.2 ± 6.4	0.1
ALT (U/L), mean ± SD	43.8 ± 22.3	29.1 ± 19.2	309 ± 572	80.8 ± 237	0.03
AST (U/L), mean ± SD	112 ± 54.7	70.1 ± 31.3	299 ± 557	125 ± 226	0.04
MELD score, median (range)	24 (9–43)	27 (10–40)	9 (6–16)	24 (6–43)	0.003

AC, alcohol-associated cirrhosis; AH, alcohol-associated hepatitis; ALD, alcohol-associated liver disease; ALT, alanine transaminase; AST, aspartate transaminase; BMI, body mass index; MELD, Model for End-Stage Liver Disease

The average age and BMI of patients in the biopsy cohort did not vary significantly by clinical phenotype; the mean age was 45 and 46 years in the AH and AC subgroups, respectively, and 50 years in the no ALD group. The overall mean BMI was 28.2 kg/m². Differences by clinical phenotype between AST, ALT, and MELD scores were statistically significant (Table 2).

Differences in FGF-21 concentrations were highly significant when compared among the 3 clinical phenotypes, with the highest levels again observed in AH (3,052 pg/mL) when compared to AC (1,235 pg/mL, P = 0.03) and no ALD (325.5 pg/mL, P = 0.02) (Figure 3). We additionally examined the differences in expression stratified by MELD score. FGF-21 levels were higher in AH than AC for MELD scores ranging from 20 to 30 (P = 0.01, Figure 4b), and FGF-21 scores were higher in patients with AH with a MELD between 10 and 19 compared with patients without ALD (P = 0.02, Figure 4a). However, in our small subset of patients with a MELD score >30, FGF-21 levels were only slightly higher in patients with AH and the finding did not reach significance (Figure 4c).

Figure 3.: FGF-21 expression is highest in patients with severe alcohol-associated hepatitis in the validation cohort (ANOVA P < 0.01). Here, “no ALD” is used for samples whose liver biopsy demonstrated no evidence of alcohol-associated liver disease; AH is severe alcohol-associated hepatitis as defined by the NIAAA criteria; and AC is alcohol-related cirrhosis. Intergroup comparisons using the t test for post hoc analysis showed meaningful differences between AC and AH (P = 0.03), and between AC and no ALD (P = 0.02). No significant difference was seen between AC and no ALD (P = 0.25). FGF-21, fibroblast growth factor-21; NIAAA, National Institute on Alcohol Abuse and Alcoholism.

Figure 4.: Differences in FGF-21 expression were significant in the biopsy-proven validation cohort for MELD ranges 10–19 and 20–30 when compared among the 3 clinical phenotypes. Notably, there were no patients with MELD >20 without alcohol-associated liver disease (no ALD). Where MELD scores were greater than 30, there was no significant difference. AC, alcohol-related cirrhosis; AH, alcohol-associated hepatitis (all MELD); FGF-21, fibroblast growth factor-21; MELD, Model for End-Stage Liver Disease.

Notably, in subjects with AC and MELD >30, FGF-21 levels were significantly higher than those with MELD scores between 20 and 30 (see Supplementary Figure 5A and B, Supplementary Digital Content 1, https://links.lww.com/CTG/A928).

Receiver operating characteristic modeling and survival analysis

A receiver operating characteristic curve for the ability of FGF-21 to distinguish between moderate to severe AH from AC in the discovery cohort found the area under the curve (AUC) to be 0.81 (95% confidence interval [CI]: 0.65–0.98, P < 0.01). Examination of the histology-proven validation cohort yielded similar results: FGF-21 levels were again higher in AH compared with AC (3,444 vs 1,279 pg/mL, P < 0.01), and the AUC was 0.76 (95% CI: 0.56–0.96, P = 0.03). Finally, in the combined groups for subjects with MELD ≥20, the AUC to discriminate between AH and AC was 0.80 (95% CI: 0.68–0.92, P < 0.0001, Figure 5a).

Figure 5.: In the combined cohorts for subjects with severe alcohol-associated liver disease, FGF-21 expression reliably discriminated between decompensated cirrhosis and severe AH when MELD scores were greater than 20 (a) (AUC 0.80 P < 0.0001, 95% CI: 0.68–0.91). Survival analysis for the combined cohort showed that patients with FGF-21 expression within the second quartile range had the best survival (P < 0.05), even when age-matched and sex-matched (b). AUC, area under the curve; CI, confidence interval; FGF-21, fibroblast growth factor-21; MELD, Model for End-Stage Liver Disease.

A survival analysis performed on the combined cohorts demonstrated that patients with serum FGF-21 levels in the second quartile (267–844 pg/mL) had significantly improved survival compared with all other quartiles (P < 0.05, Figure 5b). When only subjects with moderate to severe AH were analyzed, there was a survival benefit in both the first and second quartiles (P = 0.016, see Supplementary 6A, Supplementary Digital Content 1, https://links.lww.com/CTG/A928). However, when the analysis was further restricted to 28- or 90-day survival in patients with moderate to severe AH, no significant difference was seen between quartiles, although the same trend was observed (P = 0.44, see Supplementary Figure 6B, Supplementary Digital Content 1, https://links.lww.com/CTG/A928, P = 0.33, see Supplementary Figure 6C, Supplementary Digital Content 1, https://links.lww.com/CTG/A928).

DISCUSSION

In this cohort of subjects with a broad clinical spectrum of ALD, the most striking findings were the exceptionally elevated circulating FGF-21 levels in patients with severe AH. Previous studies have suggested a similar trend (¹⁸), but we now confirm these findings in a large and clinically diverse cohort. Importantly, we found that the expression of FGF-21 in this population was independent of BMI or age. These findings suggest that FGF-21 may accurately distinguish AH from other disease states.

The clinical distinction between severe AH (with or without advanced fibrosis) and decompensated AC without active steatohepatitis can be challenging, and previous studies suggest that nearly 20% of patients with severe AH are clinically miscategorized (⁶). This has important clinical implications for managing these critically ill patients, clinical trial design, and drug development. Importantly, our data indicate that FGF-21 is a highly reliable biomarker, with an AUC of 0.80 to discriminate between AH and AC in patients with a high MELD score (≥20), precisely the group in whom the need is the greatest for an accurate diagnosis.

We did not confirm a clear link between the proximity of alcohol use and FGF-21 levels. In our discovery cohort, recent alcohol exposure showed a trend toward lower FGF-21 concentrations, but we did not observe the same in our validation cohort. In a previously published study, significantly higher FGF-21 levels were seen in Austrian patients with positive ethyl glucuronide tests compared with patients with a negative ethyl glucuronide test for those with both alcohol-related and nonalcoholic cirrhosis (¹⁸). Our conflicting findings may reflect the relatively small sample size or recall bias in patient reporting of alcohol use.

Interestingly, in the survival analysis, patients across all diagnoses with FGF-21 concentrations within the second quartile (267–844 pg/mL) had significantly improved survival compared with all other quartiles. This effect was preserved when age, sex, and ethnicity were accounted for. The lower survival observed in the third and fourth quartiles is expected because these high levels are associated with more severe diagnoses. Because some patients with decompensated AC did not develop elevated FGF-21 concentrations, we posited that patients with AC influenced the poor survival of the first quartile of the combined cohorts. Indeed, when looking only at moderate to severe AH, we still found that higher FGF-21 expression had worse long-term survival. However, for 28- and 90-day survival, although there was a trend toward worse survival in the highest quartile, this was not significant. Mechanistic questions remain in determining the possible protective effect of FGF-21 overexpression and how chronic alcohol-mediated liver injury may perturb the physiologic response.

Investigations to date have demonstrated that alcohol consumption increases circulating levels of FGF-21 in humans and mice, and there may be feedback responses that curtail alcohol consumption (^19,20). When mice are administered FGF-21 (to supraphysiologic levels), they have decreased sweet and alcohol preferences. Other published work has shown that FGF-21 may inhibit macrophage-mediated inflammation by suppressing the nuclear factor κB signaling pathway, thereby reducing the expression of tumor necrosis factor α, interleukin 1β, interleukin 6, and interferon γ (²¹). In murine models, FGF-21 knockout mice have increased hepatic lipogenesis and inflammatory injury in the presence of alcohol exposure. Our data suggest that this hepatoprotective response may become markedly dysregulated in severe ALD, and there may be a degree of FGF “resistance” in these patients.

Several limitations are noted within this study, including the relatively small size and potential selection bias that could arise from reporting out of a tertiary center. This study design may have confounding variables that influence FGF-21 expression here not accounted for, particularly at the extremes of ALD clinical severity, including superimposed infection or systemic inflammatory responses.

Based on our findings, FGF-21 concentrations may represent a novel predictive biomarker with a strong ability to discriminate between AH and decompensated cirrhosis. This may serve as a valuable biomarker for routine management and clinical investigation in patients with severe alcohol-associated liver disease.

CONFLICTS OF INTEREST

Guarantor of the article: Paige McLean Diaz, MD, and Esperance A.K. Schaefer, MD, MPH.

Specific author contributions: Study design: P.M.D., A.V., R.P.G., E.A.K.S., and R.T.C. Statistical analysis: A.V., P.M.D., E.A.K.S., and A.J. Data collection: A.V., P.M.D., E.A.K.S., R.P.G., J.L., R.E.M., and K.C. Interpretation of results: P.M.D., A.V., E.A.K.S., E.M.P., and J.L. All authors read and approved the final version of the manuscript.

Financial support: P.M.D. and E.M.P. are supported by NIH award T32DK007191 and this manuscript was further supported by K08DK115881 (RPG) and a Burroughs Wellcome Career Award for Medical Scientists (R.P.G.).

Potential competing interests: P.M.D., A.V., E.M.P., R.T.C., J.L., R.P.G., and E.A.K.S. have no relevant disclosures to declare. A.J. is employed by Regeneron Pharmaceuticals. The work presented in this paper was conducted before joining Regeneron.

Study Highlights

WHAT IS KNOWN

• ✓ Alcohol-associated hepatitis is a clinical diagnosis, and some patients are misclassified even by expert hepatologists.

WHAT IS NEW HERE

• ✓ Circulating FGF-21 is markedly elevated in AH and can accurately distinguish AH from decompensated cirrhosis.