Controversies in Clinical Trials for Alcoholic Hepatitis

Abstract

Alcoholic hepatitis (AH) is the most severe form of alcoholic liver disease, contributing to significant morbidity and mortality. Yet, the only available therapies that improve survival are corticosteroids and liver transplantation with no new drugs successfully developed for decades. This article describes briefly the current state of affairs in AH therapy and examines the practical and ethical challenges to conducting controlled trials in patients with severe AH. While prednisolone is considered to be standard of care in severe AH, this recommendation remains controversial given the marginal benefits and questionable longterm safety of steroids. Placebo controlled trials without steroids may be necessary and ethically justified in certain populations of AH who have not been adequately investigated. Ultimately, we suggest the field will advance with the development of a plausible animal model of true AH, a consensus on a composite clinical end-point that does not rely solely on mortality for use in future RCTs, and the adoption of the recommendations of the NIAAA Alcoholic Hepatitis Consortia regarding standard definitions and when to require a liver biopsy prior to study entry.

I. Introduction

Alcoholic hepatitis (AH) represents one of the deadliest diseases in clinical hepatology contributing to significant morbidity and mortality among patients with alcoholic liver disease (ALD) [1]. Patients with severe AH often show signs of systemic inflammation and are prone to infections, kidney injury, and ultimately multiorgan dysfunction. Mortality rates range from 20 to 50% at 3 months [2,3]. Few advances have been made in the management of patients with AH [4,5]. To date, the only therapies that improve survival in AH are corticosteroids alone and liver transplantation [6–8]. The landscape for treatment of severe AH has remained relatively static since the early 1970s, at which time the use of corticosteroids was proposed to treat severe AH, which is in sharp contrast to the landmark developments in treatment of chronic viral hepatitis [9].

The lack of advances in the field of AH have been due to intrinsic difficulties in performing randomized clinical trials (RCTs) in patients with active addiction and advanced clinical illness as is the case in patients with severe AH. Furthermore, the lack of experimental models of advanced ALD has held back the discovery of treatments based on basic mechanisms, and hampered proof of concept therapeutic interventions [1,5,9,10]. Undoubtedly, there is an urgent need to develop novel therapies to treat AH. There is a burgeoning list of potential therapies, and an optimistic assessment would be that we are on the crest of a new wave of therapies for AH. The challenge remains, however, how to conduct phase 2 and phase 3 studies using new AH therapies wherein the studies are both practically and ethically sound.

In this review article, we will discuss RCTs to study novel therapies for AH. We will review the evidence surrounding the use of corticosteroids, the current “mainstay” or standard of care therapy. With this as our backdrop, we will focus on practical and ethical challenges to advancing therapy for patients with severe AH.

II. Current State of AH Treatment

All studies of AH treatment start with the notion that research subjects should abstain from all alcohol. Although this is a concept that could be studied, there have been no well-conducted studies aimed at enhancing sobriety in severe AH, either with motivational interventions or by pharmacologic interventions to reduce craving or enhancement after drinking has begun.

Corticosteroid therapy in AH has been studied in numerous RCTs spanning several decades (see Table 1). These studies have been limited by small sample sizes, heterogeneity in study design and patient population, variability in steroid dosing and duration, and a high risk of bias [11]. Furthermore, the definition of severe alcoholic hepatitis, and thus the population at risk, has changed over time encompassing the presence of hepatic encephalopathy and the widely adopted Maddrey’s discriminant function. Regardless, prednisolone is widely considered the first line therapy for severe AH. Both the American Association for the Study of Liver Disease (AASLD) and European Association for the Study of Liver (EASL) practice guidelines recommend the use of corticosteroids (i.e. prednisolone 40 mg daily for 4 weeks) for patients with severe AH, defined by Maddrey’s discriminant function (MDF) ≥ 32 [12], or the presence of hepatic encephalopathy [3,13,14].

Table 1.

Summary of placebo-controlled trials of corticosteroids in alcoholic hepatitis.

Author (Year)	Total subjects (N)	Therapeutic intervention	Control group	Biopsy	MDF used as threshold for steroid therapy	Endpoint	Mortality benefit of steroid therapy
Porter (1971)48	20	Prednisolone IV tapered over 15 days	Placebo	No	No	Mortality	No
Helman (1971)9	37	Prednisolone 40 mg daily × 4 weeks then tapered	Placebo	Yes	No	Mortality	Yes
Campra (1973)49	45	Prednisone 0.5 mg/kg/day for 3 weeks then 0.25 mg/kg/day for 3 weeks	Placebo	No	No	Mortality	No
Blitzer (1977)50	33	Prednisolone 10 mg QID × 14 days then tapered over 12 days	Placebo	No	No	Mortality	No
Shumaker (1978)51	27	Prednisolone 80 mg daily for 4–7 days then tapered over 4 weeks	Placebo	No	No	Mortality	No
Maddrey (1978)12	55	Prednisolone 40 mg/day × 30 days	Placebo	No	No	Mortality	Yes, for patients with high discriminant function
Depew (1980)52	28	Prednisolone 40 mg/day × 28 days then tapered over 2 weeks	Placebo	No	Yes	Mortality	No
Theodossi (1982)53	55	Methylprednisolone 1 g/day × 3 days	Placebo	No	No	Mortality	No
Mendenhall (1984)54	178	Prednisolone 60 mg/day × 4 days and tapered over 30 days	Placebo	No	No	30 day mortality and overall survival	No
Carithers (1989)55	66	Prednisolone 32 mg/day × 28 days then tapered	Placebo	No	Yes	28 day mortality	Yes
Ramond (1992)56	61	Prednisolone 40 mg × 28 days	Placebo	Yes	Yes	66 day mortality	Yes
Thursz (2015)8	1103	Prednisolone 40 mg/day × 28 days (with or without PTX)	Placebo and Placebo+PTX	No	Yes	28 day mortality	No*

PTX, pentoxifylline.

^*While the primary endpoint did not reach statistical significance, prednisolone was associated with a decreased odds ratio of 28 day mortality on a multivariable regression analysis adjusted for predictors of mortality.

The basis of these recommendations comes from the 2011 meta-analysis by Mathurin et al. of five previously published randomized controlled trials, with a combined recruitment of more than 400 patients with severe AH on the basis of the MDF ≥32, all of whom had biopsy-confirmed alcoholic steatohepatitis [15]. This reanalysis demonstrated a significant 28-day survival benefit of 79.97±2.8% in the corticosteroid-treated subjects compared to 65.7±3.4% (p=0.0005) in the control subjects. In addition, while a 2008 Cochrane meta-analysis of 15 trials (over 700 patients) by Rambaldi et al. showed no statistically significant reduction in mortality in the corticosteroid group compared with placebo or no intervention in the entire population, there was a significant reduction in mortality in patients with a discriminant function greater than 32 or hepatic encephalopathy [16]. Another recent French RCT in severe AH compared corticosteroids to corticosteroids plus intravenous n-acetyl cysteine (NAC) [17]. The primary end-point, which was survival at 6-months, was not improved by the combination of prednisolone and NAC. However, post hoc analysis showed a significant survival benefit for prednisolone and NAC at 28 days. Unfortunately, the majority of the trials were at high risk of bias owing to significant heterogeneity.

A more recent randomized trial published by Thursz et al. investigated steroids or pentoxyfilline for AH (given the acronym ‘STOPAH’) and concluded that corticosteroids had a beneficial survival effect in only the first 28 days [8]. This multicenter, double-blind, factorial (2×2) trial randomized 1103 AH patients with MDF ≥32 into one of four arms: placebo/placebo (i.e., the natural history of the disease), prednisolone/placebo, pentoxyfilline (PTX)/placebo, or prednisolone/PTX [8]. All diagnoses were made on clinical criteria and liver biopsies were not required. Furthermore, subjects were randomized based on a risk stratification, with high risk defined as recent gastrointestinal hemorrhage, renal insufficiency, or sepsis. Exclusionary criteria included severe renal failure (renal replacement therapy or a serum creatinine > 5.7 mg/dL or 500 μmoles/L), gastrointestinal hemorrhage, sepsis, or inotropic support after 7 days of therapy. There was a non-significant survival advantage during the first four weeks among recipients of prednisolone (OR 0.72; 95% CI 0.52–1.01, p=0.06). On cross-sectional analysis of 28-day survival, prednisolone was not associated with a survival benefit. However, on a post-hoc multivariable analysis, prednisolone was associated with improved 28-day survival, with an odds ratio of survival of 0.609 (p = 0.015), while having no effect on 90-day or 1-year mortality. Interestingly, the mortality rate from severe alcoholic hepatitis was lower in all treatment arms than was projected in the power analysis and lower than the mortality rate in the alcoholic hepatitis literature overall. Nevertheless, the STOPAH trial lends further evidence to the efficacy of corticosteroids in the treatment of severe alcoholic hepatitis.

On the basis of this short review, we conclude that the design of an RCT of AH treatment requires careful consideration of several design issues, and an awareness of the ethical challenges when studying this population. In the remainder of this paper we will discuss these challenges in further detail.

III. Controversies in Designing AH Trials

Inclusionary Criteria

Since AH is a clinical entity, diagnosis will be made with specified clinical parameters. The MDF has become the ubiquitous minimal measure of severity to ensure that patients with a good chance of spontaneous recovery are excluded. However, the range of mortality risk of patients with an MDF above 32 is very broad, and several authorities have recommended restricting entry into RTCs for a specific strata of patients within this general criterion of MDF >32. This could require a second prognostic score such as ABIC [18], MELD [19–21], Lille [22,23] or the Glasgow score [24,25]. A consequence of restricting recruitment in this way will be to limit recruitment.

A second area of controversy is whether or not to require liver biopsy confirmation of AH. The benefit of this policy is to avoid inclusion of subjects with conditions that mimic clinical AH while lacking alcoholic steatohepatitis. Examples are alcoholic cirrhosis with sepsis, or alcoholic foamy cell hepatitis [26,27]. Furthermore, immunohistochemistry data, in conjunction with laboratory and clinical data, can provide important diagnostic and prognostic information including risk for infection and mortality [28–30]. Once again the unwanted consequence is that of limiting recruitment. Indeed the unavailability of transjugular biopsy in many community hospitals would preclude their participation were a biopsy required. Even if biopsy is available, the histologic diagnosis of AH is not standardized across all pathologists and is subject to variable interpretation. Furthermore, requiring biopsy confirmation of histologic alcoholic steatohepatitis may limit the “real-world” application of clinical trial results in these same community hospitals where histologic confirmation of AH may not be possible. Recently, the NIAAA study consortia have proposed a partial way around this conundrum by characterizing severe AH as definite, probable and possible (Table 2) [31]. In clinical studies, biopsy would be required in order to include patients in the possible category, but not in probably or definite categories.

Table 2.

NIAAA Study Consortium Definitions of alcoholic hepatitis (AH)^*

CERTAINTY OF AH DIAGNOSIS	DEFINITION
DEFINITE	∎ Clinically diagnosed ∎ Biopsy proven
PROBABLE	∎ Clinically diagnosed without confounding factors (e.g. possible ischemic hepatitis from multiorgan failure, uncontrolled gastrointestinal bleeding, hypotension, cocaine use, etc.) ∎ Heavy alcohol use and typical liver tests ∎ Negative markers for immune and metabolic liver disease ∎ Absence of sepsis, shock, cocaine use, or recent drug use making DILI unlikely
POSSIBLE	∎ Clinically diagnosed ∎ Potential confounding factors (e.g. ischemic hepatitis from GI bleeding, septic shock, cocaine use, DILI, uncertain alcohol use) ∎ Atypical laboratory tests (e.g. AST < 50 IU/mL or > 400 IU/mL, AST/ALT ratio < 1.5, ANA Ab > 1:160, SMA > 1:80)

^*Permission granted to represent this adaptation of the definitions outlined by Crabb et al. Gastroenterology 2016; 150;4:785–90.

Abbreviations: ALT = alanine aminotransferase, ANA = antinuclear antibody, AST = aspartate aminotransferase, DILI = drug-induced liver injury, SMA = smooth muscle antibody.

Exclusionary Criteria

As we discussed above, the authors of future studies will have to decide the severity of liver injury needed to be included. Since AH is a dynamic condition, it makes sense, as in STOPAH, to avoid patients who are very likely to die irrespective of treatment. However, the converse is also true. Patients in the ICU, with severe renal failure, or with very high MDF and MELD scores are important populations to investigate. Determining whether these patients should be excluded or stratified into subgroups is important for the design of future studies. The principal benefit of using MDF > 32 as an entry requirement is to avoid including patients who would likely recover in any case. On foot of the salutary effect of alcohol withdrawal, some studies introduce a short waiting period before commencement of the RCT, in order to avoid diluting the cohorts with patients who would recover with simple medical management. However, as more restrictions to recruitment are applied, the harder it becomes to fulfil power requirements. The inclusion and exclusion criteria recently proposed by the NIAAA-funded consortia on alcoholic hepatitis are depicted in Table 3.

Table 3.

NIAAA Study Consortium Proposed Inclusion and Exclusion Criteria for Clinical Trials in alcoholic hepatitis (AH)^*

	RECOMMENDATIONS
INCLUSION CRITERIA	∎ Onset of jaundice within prior 8 weeks ∎ Alcohol Consumption:  ○ Females: > 40 g per day for at least 6 months  ○ Males: > 60 g per day for at least 6 months ∎ Less than 60 days of abstinence before onset of jaundice ∎ AST > 50 IU / L ∎ AST and ALT < 400 IU / L ∎ ALT / ALT > 1.5 ∎ Serum total bilirubin > 3.0 mg/dL ∎ Liver biopsy confirmation in patients with confounding factors (e.g. potential ischemic hepatitis, DILI, immune or metabolic liver disease, viral hepatitis, etc.)
EXCLUSION CRITERIA	∎ Patients with very severe disease (MDF > 60 or MELD > 30) ∎ Uncontrolled infections ∎ Multiorgan failure ∎ Uncontrolled upper gastrointestinal bleeding ∎ Preexisting kidney injury with severe creatinine > 2.5 mg/dL ∎ Other underlying liver diseases including HBV, autoimmune liver diseases, Wilson disease, suspected DILI ∎ HCC or other active malignancies except skin cancer ∎ Pregnancy ∎ Underlying diseases that might be exacerbated by proposed treatments (e.g. HCV, hemocromathosis, latent TB) ∎ Uncontrolled drug addiction
STRATIFICATION BASED ON SEVERITY	∎ MDF ≥ 32 assuming a control prothrombin of 12 seconds ∎ MELD > 20 ∎ Less ill patients may be appropriate for early phases or mechanistic studies

^*Permission granted to represent this adaptation of the definitions outlined by Crabb et al. Gastroenterology 2016; 150;4:785–90.

Abbreviations: ALT = alanine aminotransferase, AST = aspartate aminotransferase, DILI = drug-induced liver injury, HBV = hepatitis B virus, HCC = hepatocellular carcinoma, HCV = hepatitis C virus, MDF = Maddrey’s discriminant function test, MELD = Model for End-Stage Liver Disease, TB = tuberculosis.

Power Calculations and End Points

A common limitation to performing successful RCTs in the field of AH is the need for large subject numbers in each therapeutic arm. The issues surrounding power calculations are well demonstrated by STOPAH. The power analysis was based on a predicted 28-day survival in the placebo/placebo group similar to that reported by Mathurin et al. of 70%, with an expected improvement to 79% in the treated groups. In fact, the 28-day survival in the placebo/placebo group was 83% compared to 86% in subjects who received prednisolone. These findings pose several questions. Does the STOPAH outcome constitute a new normal? Should future studies be based on this estimated survival at 28 days? If this is the case, it will further complicate studies using mortality as the primary end-point and necessitate very large cohorts.

The selection of end-points is linked to the choice of duration of observation, and the power calculation. Even in very large studies, such as STOPAH, when power calculation expectations are not met, the end-point of mortality may not yield a clear result. The field would be advanced by the development of a consensus endpoint that included survival but also clinical markers of recovery or deterioration. In fact, there is a current effort to define new primary end points other than survival in order to reduce the number of patients needed for such studies [31].

Observation Interval and Monitoring Abstinence

We have already drawn attention to the studies of the AH treatment which have shown a survival benefit at 28 days, but not at 3 or 6 months. This is not surprising, since the events arising in the shorter time period are directly related to severity of liver injury, the capacity of the liver to recover, and the influence of related events such as infection and multiorgan failure. In this short interval, many patients are hospitalized or admitted to nursing home care, thereby restricting their access to alcohol. Furthermore, ill-health itself may act as a restraint on drinking. In contrast, as the interval from the onset of treatment of AH admission is extended out to 3 or 6 months, the contribution of alcohol use disorder tends to increase, as the patient recovers from the immediate effects of the AH episode. It seems reasonable to make two suggestions in light of the two separate pathways of AH and alcohol use disorder: first, that the interval of observation of AH treatment closely mirrored the expected natural history of the disease. Given that most of the clinical improvment from alcoholic hepatitis occurs within the first three months of alcohol abstinence, 90 days seems to be a reasonable duration of observation. Second, studies adopting end-points after 28 days should include some structured treatment of alcohol use disorder.

Since the greater the duration of study, the greater the risk of relapse, alcohol relapse is a potential powerful confounder in those studies that are continued beyond 28 days. Consequently these studies would be greatly enhanced by formal plans to monitor alcohol use. There is no consensus on how best to do this, but we would recommend a mixed approach involving asking questions about drinking and use of biomarkers [32].

IV. Ethics of Placebo Controlled Trials in AH

Placebo-controlled trials are scientifically necessary and important in advancing the therapy of deadly diseases such as AH. However, the question remains whether such studies are ethical when an established therapy exists. For a RTC to be ethical, it is essential that investigators have no preference for any particular intervention, a concept known as equipoise [33,34]. In circumstances where there is no established treatment, placebo-controlled trials provide the greatest methodologic rigor and robust data. However, it is usually unethical to include a placebo arm in an RCT when it deprives a subject from receiving an established therapy that is available and effective [35,36]. We discuss the ethical controversy of conducting placebo-controlled trials in the setting of AH and arguments for and against conducting such studies.

Argument For Placebo Controlled Trials in AH

There are compelling methodologic reasons to conduct placebo-controlled trials. Research in AH is limited by small sample sizes and placebo arms of RTCs ensure internal validity and provide stronger evidence of the effectiveness of novel therapies [37]. A well-designed study that shows superiority of a treatment to a control provides strong evidence of its effectiveness. On the contrary, a noninferiority trial that shows “equivalence” (i.e. little difference between a new drug and known active treatment) does not itself demonstrate that the new treatment is effective [35,38].

Additionally, proven therapies for AH (e.g. prednisolone) may not have long-term benefits and committing patients to potentially ineffective therapies just because it is “standard of care” in itself may be unethical and harmful. When the effectiveness of an available treatment is modest or inconsistent, placebo arms can shed light on the benefits of new treatments or different dosing of current therapies. In the setting of AH, the STOPAH trial demonstrated that prednisolone did not improve long-term health and potentially harmed individuals after one month given increased infectious risks. Based on this evidence, it is reasonable to believe that omitting steroids would pose no significantly greater harm to individuals than that experienced in the natural course of AH. That being said, the placebo arm should not be substantially more likely than those receiving active-treatment to die, to experience severe discomfort, or to suffer serious harm [37].

Furthermore, we acknowledge that placebo alone trials may be ethical in certain patient populations which have not been traditionally studied, and in which there are limited data regarding whether regional differences or patient characteristics might influence steroid efficacy. Examples of such special populations would be patients in countries in the developing world, or patients with severe renal disease, encephalopathy, or recent variceal hemorrhage that were excluded from the studies reviewed above.

In order for placebo-controlled trials in AH to be ethical, there must be robust informed consent. Individuals must be informed of the rationale for using a placebo arm, must be aware of the existence of any effective therapy, and understand the consequences of not receiving such therapy. If an individual believes that foregoing known therapy poses unacceptable risks and that receiving placebo is unreasonable, then he/she can choose not to participate in such a trial.

Argument Against Placebo Controlled Trials in AH

In the setting of a known therapy for AH, placebo-controlled trials violate the therapeutic obligation of clinicians to offer optimal medical care and pose unnecessary harm and risk to individuals [34,39]. According to the original Declaration of Helsinki: “In any medical study, every patient—including those of a control group, if any—should be assured of the best proven diagnostic and therapeutic method” [40]^,[41]. As discussed in the review above, prednisolone is currently considered the standard of care for severe AH with MDF >32 [11,15,42,43]. In high stakes situations in which morbidity and mortality are veritable risks, it is unethical to deprive individuals of a proven beneficial therapy (even if the benefit is marginal). If we consider steroids to have any benefit in terms of reducing morbidity and mortality, then it is unethical to use untreated controls or placebos in the setting of AH [39,44].

The benefits of prednisolone may have been marginalized given the divergent results between small clinical trials and the STOPAH study. First, these smaller studies were subject to Type II errors. Second, differences in mortality from AH varied significantly based on the study country and socioeconomic factors (i.e. in Mexico and India 3-month mortality averaged 50–60%) [45]. Lastly, these studies varied due to different exclusion criteria (i.e. presence of sepsis, acute kidney injury (AKI), severe encephalopathy as in STOPAH) and so did not have comparable study populations. Based on these considerations, it is conceivable that disease severity and local socioeconomic factors masked the efficacy of prednisolone. Although findings varied, controlling for these variables including locoregional differences may have revealed more consistent benefits of steroids among AH patients.

Lastly, individuals with AH comprise a critically ill patient population in which it may be difficult to obtain robust informed consent. For informed consent to be valid, however, comprehension and voluntariness are required. Individuals with substance abuse problems and hepatic encephalopathy may have impaired decision-making capacity and comprehension, limiting their ability to provide truly informed consent [46,47]. It is problematic to rely on these individuals to determine for themselves whether forgoing standard of care or known effective therapy is reasonable. Developing standardized methods for assessing and grading hepatic encephalopathy prior to research enrollment may help to ensure an ethical balance between protecting research participants with encephalopathy from undue influences and promoting their research participation.

Current state of clinical trials in alcoholic hepatitis

The current state of clinical trials in severe alcoholic hepatitis reflect the ethical dilemma outlined above. Frequently, novel investigational agents are used in combination with corticosteroids and outcomes compared to subjects treated with corticosteroids alone. Clinical trials investigating ASK-1 inhibitors, N-acetylcysteine, metadoxine, and granulocyte-colony stimulating factor (G-CSF) are examples of trials using novel agents in conjunction with corticosteroids. In a similar vein, true placebo controlled trials of investigational agents are frequently reserved for patients that fail corticosteroids. Current trials utilizing mycophenolate mofetil and G-CSF have inclusion criteria requiring corticosteroid failure prior to enrollment. Thus, true placebo controlled trials in severe alcoholic hepatitis are presently uncommon.

V. Conclusion

There are formidable barriers to developing high-quality, well-powered RCTs of therapy for severe AH. We have outlined several practical barriers and ethical challenges related to designing such trials. All call for a balance between clarity of purpose, and the inherent danger that greater stringency in regards to entry criteria, selection of end-points and the like, will damage recruitment. We specifically address the issue of placebo inclusion and ethical challenges to testing novel therapies using placebo control arms. We suggest the field will advanced by the development of a plausible animal model of AH, by the framing of a consensus on a composite clinical end-point that does not solely rely on mortality, and the adoption of the recommendations of the NIAAA Alcoholic Hepatitis Consortia regarding standard definitions, and when to require a liver biopsy prior to study entry. While the AASLD and EASL consider prednisolone to be standard of care in severe AH, this recommendation remains controversial given the marginal benefits and questionable longterm safety of steroids in severe AH. Placebo controlled trials without steroids may be necessary and ethically justified in certain populations of AH who have not been adequately investigated. Future trials will have to determine whether omitting corticosteroids is an ethical and pragmatic way to test novel therapies in AH.

Abbreviations

AH

alcoholic hepatitis

ALD

alcoholic liver disease

MDF

Maddrey Discriminant Function

MELD

Model for End-Stage Liver Disease