IL-1 receptor antagonist plus pentoxifylline and zinc for severe alcohol-associated hepatitis

Abstract

Background and Aims:

Patients with severe alcohol-associated hepatitis (AH) have high mortality. Corticosteroids improve survival only for 30 days. We targeted inflammation, cellular injury, and gut leakiness in a randomized clinical trial comparing combination therapy to corticosteroids on 180-day survival.

Approach and Results:

Subjects with a clinical diagnosis of severe AH (Model for End-Stage Liver Disease [MELD] >20, Maddrey discriminant function [MDF] >32) were randomized to receive methylprednisolone (PRED; 28 days) or a combination of anakinra (14 days) plus pentoxifylline (28 days) plus zinc (COMB; 180 days). The primary endpoint was survival at 180 days. The study was designed in 2013, initiated in October 2014, and completed in March 2018. Five hundred patients were screened to randomize 104 subjects with a clinical diagnosis of AH with a MELD score >20. Fifty-three patients were randomized into the COMB and 50 to the PRED treatment; 1 dropped out of the study before randomization. Mean age was 45.3 ± 10.4 years; 60.6% were males, 92.3% White, and mean MELD 25.7 ± 3.9. Kaplan-Meier survival estimate at 180 days was 67.9% in COMB and 56% in PRED (HR = 0.69; p = 0.3001). Survival curves separated by 90 days (COMB, 69.8%; PRED, 58.0%; HR = 0.69; p = 0.28). Survival at 28 days was similar between the COMB (83.4%) and PRED groups (81.2%; HR = 0.91; p = 0.85). There were no unexpected serious adverse events, and incidence of infection was comparable between groups. MELD 20– 25 and MELD >26 strata showed nonsignificant treatment effects in favor of COMB.

Conclusions:

A combination of anakinra, pentoxifylline plus zinc provides similar survival benefits compared to corticosteroid therapy in severe AH.

Graphical Abstract

INTRODUCTION

Alcohol-associated hepatitis (AH) is characterized by rapid onset of jaundice, abdominal pain, anorexia, and features of systemic inflammatory response syndrome (SIRS) with elevated serum aminotransferases.^[1,2] Both rates of hospitalization and health care costs for AH increased by >30% between 2002 and 2010.^[3,4] Presence of SIRS increases the risk of developing acute kidney injury (AKI) and multiorgan dysfunction syndrome (MODS), which predicts high mortality in AH.^[5] Patients with severe AH, defined by a Maddrey discriminant function (MDF) >32^[6] or a Model for End-Stage Liver Disease^[7] (MELD) score >20, have a 90-day mortality rate near 30%.^[8]

Corticosteroid therapy is currently recommended for treatment of severe AH based on randomized, controlled trials showing improved survival at 30 days^[6,8–10]; however, the survival benefit is not sustained beyond the first month.^[8] An increased risk of infection becomes apparent after 3 weeks of treatment with corticosteroids.^[11,12] Because of the complex pathogenesis of AH, combination therapy targeting multiple pathophysiological targets may improve long-term survival.

The trigger for developing severe AH remains uncertain, and studies have identified that key elements in the pathogenesis are alcohol-induced hepatocyte damage, and changes in the gut and systemic inflammatory cytokine activation (Figure S1). Dysbiosis of the gut microbiome^[13] and disruption of gut mucosal integrity contribute to endotoxin and other microbial products entering the portal circulation.^[14,15] These pathogen-associated molecular patterns (PAMPs) act together with damage-associated molecules (DAMPs) released by hepatocytes injured by alcohol and activate prionflammatory pathways causing release of inflammatory cytokines.^[16–18] Circulating IL-1β is increased and TNFα levels correlate with mortality in patients with severe AH.^[19–21] In a preclinical model of alcohol-associated liver disease (ALD), inhibition of IL-1 signaling with the IL-1 receptor antagonist (IL-1RA), anakinra, attenuated liver damage, prevented inflammation, and improved liver regeneration.^[22,23]

Given the complexity of the pathobiology of AH, the present study evaluated the safety and efficacy of a combination of three agents: anakinra (IL-1RA; to improve the acute inflammatory component); zinc sulfate (Zn; to increase gut mucosal integrity); and pentoxifylline (PTX; to protect against development of hepatorenal syndrome),^[23–25] in comparison to corticosteroids (Figure S1). Efficacy was based on survival at 180 days, and safety was based on the rate of serious adverse events (SAEs), infections, and AKI.

MATERIALS AND METHODS

Study design and oversight

The Defeat Alcoholic Steatohepatitis (DASH) study was a multicenter, randomized, double-blind trial with a permuted block design to evaluate efficacy in improving survival and safety of the combination (COMB) of anakinra (IL-1RA; 100 mg by s.c. injection daily for 14 days) plus PTX (400 mg orally three times daily for 28 days) plus zinc supplements (220 mg of Zn given orally for 6 months), in comparison to methylprednisolone (PRED; 32 mg daily),^[6] as described.^[26]

The trial was conducted and reported with fidelity to the protocol and to USA and international standards of Good Clinical Practice (U.S. Food and Drug Administration [FDA] regulations 21 CFR 312 for investigational new drug studies and FDA guidance E6) for all studies. All procedures were conducted in accordance with the respective institutional review board guideline approvals and Declaration of Helsinki. Written informed consent was obtained from each participant. Study data were collected and managed using REDCap (Research Electronic Data Capture).^[27]

Patients

Human subjects with a clinical diagnosis of severe AH were recruited from four USA centers between February 2014 and March 2018. Details of the clinical diagnosis of AH and inclusion and exclusion criteria are detailed in the supplementary methods of the Supporting Information and reported earlier.^[1,28] Enrollment for each study site is listed in Table 1.

TABLE 1.

Baseline characteristics of the patients

	IL-1ra/PTX/Zn (n = 53)	Prednisolone (n = 50)	Total (n = 103)
Age, years (mean ± SD)	45.7 ± 10.6	45.3 ± 10.2	45.3 ± 10.4
Sex, male (N, %)	35 (66.0)	28 (56.0)	63 (60.6)
Race, White (N, %)	48 (90.6)	47 (94.0)	96 (92.3)
Ethnicity, non-Hispanic (N, %)	38 (71.7)	38 (76.0)	76 (74.5)
Family history of ALD (N, %)	14 (26.4)	9 (18.0)	23 (22.1)
BMI, kg/m2 (mean ± SD)	30.0 ± 7.3	31.1 ± 8.9	30.6 ± 8.0
CONUT scorea (mean ± SD)	8.2 ± 2.2	7.9 ± 2.1	8.1 ± 2.2
Maximum hand-grip strength, pounds per square inch	49.3 ± 27.8	44.2 ± 22.0	46.9 ± 25.3
Clinical site randomization (N, %)
Cleveland Clinic	11 (20.8)	11 (22.0)	22 (21.8)
University of Louisville	10 (18.9)	8 (18.0)	18 (18.5)
University of Massachusetts	10 (18.9)	7 (14.0)	17 (16.5)
University of Texas Southwestern	22 (41.5)	23 (46.0)	45 (43.7)
Alcohol consumption: AUDIT scoresb
Women
Hazardous alcohol use (7–15)	3 (16.7)	2 (9.1)	5 (12.2)
Harmful alcohol use (16–19)	3 (16.7)	4 (18.2)	7 (17.1)
Possible dependence (20+)	8 (44.4)	13 (59.1)	21 (51.2)
Men
Hazardous alcohol use (8–15)	5 (14.3)	2 (7.1)	7 (11.1)
Harmful alcohol use (16–19)	4 (11.4)	2 (7.1)	6 (9.5)
Possible dependence (20+)	21 (60.0)	16 (57.1)	37 (58.7)
Encephalopathy (N, %)
None	39 (73.6)	39 (78.0)	78 (75.0)
Grades 1–2	12 (22.6)	10 (20.0)	22 (21.1)
Grades 3–4	1 (1.9)	1 (2.0)	2 (1.9)
Ascites (N, %)
None	12 (22.6)	11 (22.0)	23 (22.1)
Mild/moderate	33 (62.3)	36 (72.0)	69 (66.3)
Severe	7 (13.2)	3 (6.0)	10 (9.6)
Laboratory results (mean ± SD)
Albumin, g/dl	2.5 ± 0.4	2.5 ± 0.5	2.5 ± 0.5
Alkaline phosphatase, U/L	163.5 ± 62.6	168.6 ± 81.4	166.0 ± 72.0
ALT, U/L	45.2 ± 24.3	42.1 ± 23.5	43.7 ± 23.9
AST, U/L	134.8 ± 68.2	125.7 ± 57.0	130.4 ± 62.9
Bilirubin, mg/dl	18.9 ± 7.6	20.7 ± 7.6	19.7 ± 7.6
Creatinine, mg/dl	0.9 ± 0.5	0.9 ± 0.5	0.9 ± 0.5
Globulin, g/dl	3.8 ± 1.0	3.7 ± 0.9	3.7 ± 0.9
Total protein, g/dl	6.1 ± 0.9	6.0 ± 0.8	6.1 ± 0.8
WBC, × 103/mm3	11.3 ± 6.6	11.1 ± 5.1	11.2 ± 5.9
Hematocrit, %	29.1 ± 6.0	28.5 ± 5.2	28.8 ± 5.6
Absolute neutrophil count, cells/µl	2173.3 ± 4466.0	1382.9 ± 3724.4	1805.9 ± 4128.1
Hemoglobin, g/dl	10.6 ± 3.8	9.8 ± 2.0	10.2 ± 3.1
Mean platelet volume, × 103/mm3	12.3 ± 15.0	10.3 ± 1.5	11.3 ± 10.7
Absolute lymphocyte count, cells/µl	456.7 ± 853.7	309.5 ± 641.0	386.3 ± 757.6
Absolute eosinophil count, cells/µl	127.5 ± 631.5	32.2 ± 81.0	81.9 ± 458.9
Absolute monocyte count, cells/µl	287.2 ± 591.8	231.5 ± 468.7	260.6 ± 533.3
BUN, mg/dl	13.5 ± 9.5	14.1 ± 12.3	13.8 ± 10.9
Prothrombin time, s (N, %)
<1.7	11 (20.7)	15 (30.0)	26 (25.0)
1.7–2.3	36 (67.9)	24 (48.0)	60 (57.7)
>2.3	5 (9.4)	11 (22.0)	16 (15.4)
Prognostic scores
MDFc (mean ± SD)	58.2 ± 18.1	62.2 ± 30.9	59.9 ± 25.1
MELDd (mean ± SD)	25.6 ± 3.2	25.8 ± 4.5	25.7 ± 3.9
Child-Turcotte-Pugh scoree (mean ± SD)	10.7 ± 1.3	10.6 ± 1.3	10.7 ± 1.3
7–9 points (N, %)	8 (15.1)	10 (20.0)	18 (17.3)
10–15 points (N, %)	43 (81.1)	39 (78.0)	82 (78.8)
Lille score, day 7 (N, %)
Good prognosis: Lille <0.45	19 (35.9)	23 (46.0)	42 (40.4)
Poor prognosis: Lille >0.45	20 (37.7)	20 (40.0)	40 (38.5)

Note: Lille score was calculated at day 7. The number of subjects with a favorable score <0.45 (%) and the number with unfavorable scores (%) are shown.

Abbreviations: BUN, blood urea nitrogen; WBC, white blood cells.

^aCONUT Score is a measure of undernutrition degree where 0–1 = normal, 2–4 = light, 5–8 = moderate, and 9–12 = severe.

^bAUDIT Questionnaire = Alcohol Use Disorders Identification Test.

^cDiscriminant function is calculated as 4.6 × (patient’s prothrombin time in seconds − lab control’s prothrombin time in seconds) + patient’s serum bilirubin level in milligrams per deciliter; a value of ≥32 indicates severe AH that carries an adverse prognosis.

^dIn the MELD, scores range from 6 to 40, with higher scores indicating worse prognosis.

^eChild-Turcotte-Pugh Score classes for cirrhosis severity: A = 5–6 points; B = 7–9 points; and C = 10–15 points.

Subjects (n = 500) were screened to determine eligibility (Figure 1). A total of 104 subjects were randomized to participate in the trial, but 1 subject withdrew before receiving a treatment assignment, leaving 103 subjects for analysis. An equal number of subjects in each group discontinued treatment before completion primarily because of adverse events (AEs) or progression of disease despite therapy. To evaluate whether subjects with missing primary outcomes (vital status at 180 days) were similar to those with available primary outcomes, we performed a logistic regression with missingness as the outcome and 20 baseline variables (including treatment group assignment) as predictors. We defined missingness of the primary outcome as a subject last seen alive ≥60 days before 180 days from randomization. None of the 20 variables were significantly related to missingness. Ten subjects (19%) in the COMB group and 8 subjects (16%) in the PRED group were missing the primary outcome. Nausea and vomiting were the most common AEs that led to discontinuation of treatment (n = 5). There was no difference in relapse to alcohol use by treatment groups (p = 0.47; Table S5).

FIGURE 1.

CONSORT (Consolidated Standards of Reporting Trials) diagram. *n = 314 of 347 excluded (90.5%) enrolled in the moderate or observational studies. ^†One participant was randomized but withdrew from the study before receiving a treatment assignment or study drugs. Therefore, we report on 103 participants.

Randomization

The Pharmaceutical Randomization and Inventory Management portal for the Alcoholic Hepatitis Study was created at the University of Massachusetts Medical School. The randomization scheme was generated using a permuted block design with random block sizes of 2, 4, or 6, stratified by clinical site and MELD score (20–25, 26–31, and ≥32).

Endpoints

The primary endpoint was 180-day mortality. Secondary endpoints included 28-and 90-day mortality and MELD score at 30, 90, and 180 days.

Statistical analysis

Analysis strategy

For the primary efficacy endpoint (180-day mortality), the two treatment groups were compared using Kaplan-Meier product limit estimates with a log-rank test at a two-sided alpha level of 0.05 with survival censored at 180 days, as specified in the protocol. Kaplan-Meier survival analyses were prespecified for the 28-and 90-day endpoints as well as survival analyses stratified by MELD strata (20–25, 26–31, and ≥32). All analyses used intention-to-treat principles.

For the prespecified secondary analysis of the efficacy outcomes, analyses used Cox proportional hazards regression models for efficacy outcomes to adjust the treatment effect for other patient characteristics and predictive factors. Univariate models were run for prognostic scores (MDF, MELD, and Lille score),^[28] patient characteristics and clinical variables (e.g., age, sex, time since diagnosis), laboratory variables (e.g., alanine aminotransferase [ALT], aspartate aminotransferase [AST], and creatinine), and selected AEs, to examine their association with mortality at the prespecified time points. The final Cox model was fit to the data to adjust the treatment effect for the other predictors with a p ≤ 0.05 in the univariate models. AEs were coded using MedDRA and grouped by body system and preferred term for presentation. Incidence of AEs and SAEs was presented as frequencies and proportions of patients for each preferred term.

Data safety monitoring board

The data safety monitoring board (DSMB) reviewed recruitment, patient progress, protocol deviations, patient characteristics, descriptive presentation of study endpoints without inference, laboratory values, and safety (AEs and SAEs) at 6-month intervals from the start of the study. The DSMB viewed the unblinded data in a closed session. A total of five interim analyses and one final analysis were performed. East study planning software (Version 6.4, survival analysis module; Cytel, Inc., Cambridge, MA) was used to establish the futility monitoring bounds for severe patients. The study design parameters in East were the same as for the study with the additional specification of six looks at the data. East calculated O’Brien-Fleming-Harrington monitoring boundaries (both for efficacy and for futility) using Lan-Demets alpha-and beta-spending functions. All statistical analyses were conducted in SAS software (v9.4; SAS Institute Inc., Cary, NC).

Sample size

To estimate the sample size for a log-rank test, the study design parameters were: two-sided log-rank test at alpha = 0.05; two treatment groups (a control and an intervention groups); 80% power; an HR of 0.4447; the proportion surviving at 6 months in the control group is 0.64 and 0.82 in the intervention group^[28]; total study duration of 2.5 years with uniform patient accrual in the first 2 years; and the proportion of dropout in each group is 0.10 during the 6 months with no treatment crossovers. Using the approach of Lakatos,^[19] the estimated sample size for a log-rank test with these design parameters was 130 subjects (65 per group). Because several approaches to sample-size estimates for survival analysis are available, a sensitivity analysis of these estimates was performed, using the Lachin-Foulkes approach,^[29] with the same assumptions as above, which yielded a similar number of patients. Thus, the sample size was set as 130 patients in total or 65 patients in each treatment group. PASS II (NCSS, Kaysville, UT) was used for all sample-size calculations.^[30]

RESULTS

Patients

Baseline clinical and demographic data were comparable between the two treatment groups (Table 1). Laboratory parameters, MELD score (25.7 ± 3.9), and mean MDF (59.9 ± 25.1) were similar in both groups (Table 1). Lille score at day 7 could not be calculated in 21 patients (14 COMB, 7 PRED) because of missing bilirubin on day 7. The majority of subjects were overweight or obese with a mean body mass index (BMI) of 30.6 ± 8.0 kg/m². BMI was not corrected for an estimate of the volume of ascites because only 9.6% had severe ascites. Despite the high BMI, the average Controlling Nutritional Status (CONUT)^[29] score indicated moderate-to-severe malnutrition (Table 1). Compliance with study drugs was similar in both groups, with the exception of PTX (49.4% ± 40.2%) versus placebo (65.5% ± 35.3%; p = 0.06; Table S1). Nausea was the primary reason for noncompliance. Liver biopsies performed on 20 patients (9 COMB, 11 PRED) confirmed the diagnosis of AH.

Endpoints

Survival at 180 days, the primary endpoint, was 36 of 53 (67.9%) in COMB and 28 of 50 (56.0%) in PRED (p = 0.3001). At 28 days, 46 of 53 patients (86.8%) in COMB and 41 of 50 (82%) in PRED were alive. At 90 days, 37 of 53 (69.8%) in COMB and 29 of 50 (58.0%) in PRED were alive. Four patients in each group were taken off study medication prematurely by the investigators or at the request of the patient because of SAEs or deterioration in clinical condition (presumed futility). Survival based on Kaplan-Meier estimates was similar in both groups at 28 days (86.8% for COMB vs. 82.0% for PRED), 90 days (69.8% for COMB vs. 58.0% for PRED), and 180 days (67.9% for COMB vs. 56% for PRED; Figure 2). The trend toward better survival in the group treated with COMB was not statistically significant (Figure 2A; p = 0.30; critical p value adjusted for interim monitoring looks at the data described above, p ≤ 0.021). Mortality in both groups was directly related to the initial MELD score at the time of randomization (Figure S2; log rank, p = 0.0027). Of note, mortality in either group did not change significantly between 90 (35.9%) and 180 days (37.9%; Figure S2). Survival in the COMB group was not inferior to PRED at each time point after stratifying for baseline MELD score (Figure 2B,C). MELD scores in patients surviving 28, 90, and 180 days were similar in both groups (Table S2).

FIGURE 2.

Kaplan-Meier survival plots by treatment assignment of all patients enrolled (panel A), by treatment assignment and disease severity in MELD 20–25 group (B), by treatment assignment and disease severity in MELD greater or equal 26 group (C), and by treatment assignment and disease severity measured by day 7 Lille score (D).

Several baseline factors, including age, prothrombin time/international normalized ratio (INR), MDF, MELD, and Lille score at day 7, were significantly associated with mortality at 180 days as assessed by univariate analysis (Table 2). Lille score at 7 days was predictive of survival at 90 and 180 days for subjects treated with PRED or COMB (Figure 2D; p = 0.002). Using multivariate analysis, higher baseline age and prothrombin time/INR were predictive of mortality at 180 days in both groups (Table 2). Treatment group was not predictive of mortality at 180 days in univariate analysis.

TABLE 2.

Analysis of factors associated with mortality at 180 days

	Univariate analysisa		Multivariate analysisb
	HR (95% CI)	p value	HR (95% CI)	p value
Baseline factorsc
MDF	1.02 (1.01–1.03)	<0.001
MELD	1.12 (1.04–1.20)	0.002	1.10 (1.03–1.18)	0.004
Prothrombin ratio or INR	2.63 (1.42–4.86)	0.002
Bilirubin	1.04 (0.99–1.08)	0.10
Creatinine	1.56 (0.97–2.51)	0.07
Age	1.07 (1.03–1.10)	<0.001	1.06 (1.03–1.10)	<0.001
White blood cell count	1.02 (0.97–1.07)	0.47
Albumin	0.81 (0.41–1.62)	0.56
Alcohol intake
Harmful vs. hazardous	0.45 (0.11– 1.88)	0.27
Possible dependence vs. hazardous	1.10 (0.42– 2.88)	0.85
HE	1.94 (0.76–4.95)	0.17
IL-1ra/PTX/Zn vs. prednisolone	0.72 (0.38–1.35)	0.30	0.80 (0.42–1.50)	0.48

^aValues reported for all variables were obtained at baseline.

^bA separate univariate model was fitted for each baseline factor.

^cSignificant variables in the univariate analysis were added to a multivariate model. Individual components of the MELD were not added to the model because of multicollinearity. Because of the high correlation between the Maddrey and MELD scores, only the MELD was added.

Serum IL-1ra antagonist and endotoxin levels

Circulating levels of IL-1ra were elevated on enrollment in AH patients compared to normal controls (p < 0.0001). In the COMB group, IL-1ra levels on day 7 showed a significant (0.0001) increase compared to baseline or day 28 levels. This was consistent with administration of the COMB treatment in the days 1–14 period. In the PRED group, there was no significant change over time in IL-1ra levels compared to baseline (Figure S3).

Endotoxin levels were significantly elevated in AH patients compared to healthy controls at days 0 and 7 (Figure S4). We found that endotoxin levels significantly declined by day 28 in each treatment group compared to earlier time points.

AEs

Sixty-three of 103 patients developed SAEs during the 180-day study period, with 33 (62.3%) in the COMB and 30 (60%) in the PRED group (Table 3). Of the SAEs, 24.5% in the COMB and 32% in the PRED group led to death. SAEs that occurred in ≥5% of patients included infections, AKI, and upper gastrointestinal (GI) bleeding (Table 3). Fungal infections in PRED treatment (p < 0.02) was the only significantly different SAE between treatment groups.

TABLE 3.

SAEs

	No. of patients (%)
Event	IL-1ra/PTX/Zn (n = 53)	Prednisolone (n = 50)	Total (n = 103)	p value*
SAEs
Any	33 (62.3)	30 (60.0)	63 (61.2)	0.84
Any that led to death	13 (24.5)	16 (32.0)	29 (28.2)	0.51
SAEs occurring in ≥5% of patients, by preferred term
AKI	10 (18.9)	13 (26.0)	23 (22.3)	0.48
Upper GI hemorrhage	4 (7.5)	4 (8.0)	8 (7.8)	0.99
Fungal infections (aspergillus, histoplasmosis, norcardia, candida)	0 (0.0)	5 (10.0)	5 (4.9)	0.02
SAEs occurring in ≥5% of patients, by system organ class
Renal and urinary disorders	13 (24.5)	15 (30.0)	28 (27.1)	0.66
Vascular disorders	7 (17.0)	10 (22.0)	17 (19.4)	0.43
Infections and infestations	12 (22.6)	9 (18.0)	21 (20.4)	0.63
Respiratory, thoracic, and mediastinal disorders	4 (7.5)	8 (24.0)	12 (15.5)	0.23
General disorders and administration site conditions	7 (13.2)	5 (10.0)	12 (11.7)	0.76
Hepatobiliary disorders	9 (17.0)	3 (6.0)	12 (11.7)	0.12
Cardiac disorders	7 (13.2)	2 (4.0)	9 (8.7)	0.27
Metabolism and nutrition disorders	3 (5.7)	5 (10.0)	8 (7.8)	0.48

Note: “n” refers to the number of unique patients experiencing the given AE.

^*p values from Fisher’s exact tests.

AKI was significantly associated with mortality at 180 days, usually from MODS, and was similar in both treatment groups (Tables S3 and S4). The total number of SAEs related to infection (13 in COMB and 11 in PRED) was similar in the two groups. There were seven SAEs related to fungal infections (range of days on study, 21–47); three from Aspergillus, one from Nocardia, one from Histoplasmosis, and two from Candidia in subjects treated with PRED whereas there were no fungal infections in patients treated with COMB (Table S3). Although the incidence of bacterial infections was similar in both groups, PRED had a higher incidence of systemic fungal infections (10% vs. 0%) than those treated with COMB (Table S3). Frequency of other nonserious AEs was similar between the groups (Tables S3 and S4).

DISCUSSION

Poor clinical outcomes in severe AH correlate with the complex and multifactorial pathogenesis of this disease. Our study was designed to take a three-pronged approach to target key elements in the disease process: (1) short-term anti-inflammatory intervention with the IL-1ra antagonist, anakinra, to disrupt the vicious cycle of proinflammatory cascade activation; (2) mid-term treatment for 28 days with PTX to attenuate hepatorenal complications and provide additional anti-inflammatory effects; and (3) long-term nutrient supplementation with zinc to improve intestinal barrier function. The results of this randomized, double-blind, controlled trial show that survival at 180 days in subjects treated with a COMB was higher, but not statistically significantly better (p = 0.30), than those treated with PRED, the current standard of treatment for severe AH. The current trial was designed to detect a difference of 30% in survival between the two groups based on enrollment of 130 subjects with an expected survival rate of 50% at 180 days for subjects treated with corticosteroids. The trend toward better survival in the COMB group is notable, although incomplete enrollment (80%) because of lack of continued funding may have limited statistical outcomes. The COMB treatment was well tolerated, with a similar number of both AEs and SAEs compared to PRED.

There was little difference in survival at 180 days compared to 90 days in the present study, suggesting that 90 days might be a better endpoint for future phase II/III trials in AH as suggested recently.^[30] A 90-day endpoint would minimize confounding related to relapse to drinking that contributes to longer-term outcomes in AH.^[31] At the time of initiation of our study in 2014, the conventional endpoint was 30 days, but because the efficacy of steroids is limited, longer duration of trials was necessary. Recommendation on 90-day endpoint was only recently adopted, and this is also supported by our findings. Both MDF >32 and MELD >20 were required for inclusion in the current study to avoid including subjects with high MELD scores related primarily to renal insufficiency in the setting of chronic liver disease. Subjects were stratified by MELD and randomized by site to ensure that the groups were balanced by severity and that outcomes attributed to differences in supportive care would be minimized. The 28-, 90-, and 180-day survival rates for subjects treated with PRED were similar to other trials with similar baseline MELD scores.^{[8,10,25,32,33]} As expected, survival in both groups correlated inversely with baseline MELD score. Data from the STOPAH (Steroids or Pentoxifylline for Alcoholic Hepatitis) trial indicated that a MELD cutoff of 25 had a higher negative predictive value for survival.^[34] Although the number of patients with MELD >32 was small in our study, survival was slightly better, though not statistically different, in those treated with COMB therapy compared to PRED.

Lille score has been used to predict risk of mortality at 180 days after 7 days of treatment with corticosteroids^[28,35] and is commonly used to determine the futility of continuing corticosteroids beyond 7 days. In the present study, Lille score at 7 days was predictive of survival at 90 and 180 days, both in PRED and COMB groups (Figure 2D; p = 0.0017), indicating the predictive use of the Lille score in patients not treated with corticosteroids. In the present study, a Lille score >0.45 was not used as a “stopping rule” to avoid compromising the double-blind design.

Infections were a major cause of death and a trigger for development of MODS in both treatment groups. Increased number of deaths in the PRED group was related to serious infections developing between 30 and 60 days that was not observed in the COMB group. Fungal infections developed in the 3-to 6-week period after initiation of treatment. A higher rate of systemic fungal infections in the PRED group is similar to that in previous reports with corticosteroids.^[11,12,36] Intervention with anakinra to dampen innate immune activation, unlike the previously used anti-TNFα-based therapies, was safe and well tolerated in AH patients and resulted in no increased incidence of infections. Thus, IL-1 inhibitor treatment may offer benefits over PRED in patients with high risk for fungal infections.

PTX was included in the COMB therapy to reduce the incidence of AKI and hepatorenal syndrome.^[10,37] However, studies such as the STOPAH, published after the initiation of our trial, suggest that PTX does not improve outcomes in severe AH.^[8,10,38,39] Although the total number of patients who developed AKI was similar between the two groups, AKI was significantly lower in the group with MELD 20–25 in the COMB compared to the PRED group (Table S3). We also noted that compliance with PTX was poor because of GI side effects.

Limitations of our study include the lower-than-proposed enrollment of subjects. This highlights the enrollment challenges in treatment trials for severe AH even in multicenter studies. Despite this limitation, this is one of the largest contemporary treatment trials in a well-characterized population with severe AH. Another limitation is that a placebo group was not included. However, in a disease with a high acute and medium-term mortality, there are ethical issues with a placebo-controlled trial, and a superiority study design was chosen to overcome this concern. Overall mortality was lower than in previously reported studies for similar levels of severity of liver disease because of improved nutritional status and/or standard of care. Given the stringent inclusion and exclusion criteria, the generalizability of our observations may be limited.

In conclusion, our survival analyses indicate that 180-and 90-day survival of patients receiving COMB therapy was similar to those receiving PRED, the current standard of care for severe AH (p = 0.30). Survival analysis of 28-day survival of patients receiving COMB therapy was similar to those receiving PRED. COMB therapy was safe and well tolerated. Future studies in severe AH should examine treatments targeted to specific mechanisms in the pathogenesis of AH.

Abbreviations:

AEs

adverse events

AH

alcohol-associated hepatitis

ALT

alanine transaminase

AST

aspartate aminotransferase

AKI

acute kidney injury

BMI

body mass index

COMB

combination of anakinra, pentoxifylline, zinc

CONUT

Controlling Nutritional Status

DAMPs

damage-associated molecular patterns

FDA

U.S. Food and Drug Administration

GI

gastrointestinal

IL-1RA

interleukin-1 receptor antagonist, anakinra

INR

international normalized ratio

MELD

Model for End-Stage Liver Disease

MDF

Maddrey discriminant function

MODS

multiorgan dysfunction syndrome

PAMPs

pathogen-associated molecular patterns

PRED

methylprednisolone

PTX

pentoxifylline

SAEs

serious adverse events

SIRS

systemic inflammatory response syndrome

Zn

zinc sulfate

DATA AVAILABILITY STATEMENT

Anonymized individual patient data will be made available to interested investigators at the discretion of all authors. Data requests should be sent to the corresponding author.