Abnormal cytokine metabolism, especially of tumor necrosis factor (TNF)-α, is a major feature of alcoholic liver disease. In 1989, we reported dysregulated TNF metabolism in patients with alcoholic hepatitis, noting that cultured monocytes, which produce the majority of systemic circulating TNF and are a surrogate marker for Kupffer cells, spontaneously produced TNF and exhibited an increased TNF response to an stimulus in contrast to control monocytes.1 Several groups then reported increased serum TNF concentrations in alcoholic hepatitis, and values correlated with disease severity and mortality.2–4 A TNF promoter polymorphism was linked with an increased susceptibility to alcoholic hepatitis, suggesting that a subset of people who drink alcohol may be genetically predisposed to the development of alcoholic liver disease.5 Concomitant with these human studies, complementary studies in rats, mice, and tissue culture evaluated the role of TNF in experimental models of liver disease.6–8 More than a decade ago it was shown that rats chronically fed alcohol were more sensitive to hepatotoxic effects of injected LPS (endotoxin).9 Subsequently, investigations showed that rats chronically fed alcohol had much higher stimulated plasma levels of TNF than control rats, and liver injury could be attenuated by agents such as a prostaglandin analogue that downregulated TNF production.10 Studies using the intragastric alcohol-feeding model demonstrated that the development of liver injury coincided with an increase in TNF messenger RNA in the liver and in isolated Kupffer cells.11, 12 Rats fed ethanol intragastrically also had high blood endotoxin levels and induction of cytochrome P450 2E1 which could cause generation of reactive oxygen species.13 Indeed, markers for oxidative stress and lipid peroxidation were noted in these animals.13 Several strategies were then devised to decrease either oxidative stress or cytokine production/activity in an attempt to block or attenuate liver injury; all of these strategies were successful in attenuating alcohol-induced liver injury in rats.14–16 Perhaps the most compelling data relating TNF to alcohol-induced liver injury come from the late Dr. Ron Thurman's laboratory, which successfully used anti-TNF antibody to prevent liver injury in alcohol-fed rats.17 Similarly, mice lacking the TNF type I receptor also did not develop alcoholic liver injury.18 Thus, there is compelling evidence that TNF plays an etiological role in the development of alcoholic liver disease.
In the previous issue of HEPATOLOGY, a large double-blind randomized control trial is described using either prednisolone treatment alone or prednisolone treatment plus infliximab (anti-TNF antibody) in patients with acute alcoholic hepatitis.19 The exclusion criteria were similar to many previous large studies on alcoholic hepatitis, including hepatitis B, hepatitis C, human immunodeficiency virus, hepatocellular carcinoma, recent gastrointestinal bleeding, hepatorenal syndrome, and current or recent infection, to name only a few. In the United States, this would exclude a large number of patients seen clinically. Indeed, in this multicenter trial from France, 109 patients were screened to randomize a total of 36. The study was designed so that the infliximab group received an infusion of 10 mg/kg at weeks 0, 2, and 4. This high dose of anti-TNF antibody (double the standard dose used for Crohn's disease) was given at relatively short intervals. Both groups received prednisolone. The French study's design reflected the investigators' view that it would be unethical to withhold “standard of care” from either group. The major endpoint was reduction in 2-month mortality. Secondary endpoints were prevalence of bacterial infections at 2 months, reduction of Maddrey's score at 1 and 2 months, and reduction in 1- and 12-month mortality. In a subset of patients, neutrophil counts and immune function were evaluated. After randomization of 36 patients, this study was terminated because the probability of being dead at 1 and 2 months was higher in Group A (27.8% ± 11% and 39% ± 11%, respectively) than in Group B (5.9% ± 6% and 18% ± 9% respectively). Moreover, severe infections were more frequent in the infliximab group (10 episodes in 8 patients) compared to prednisolone alone (1 episode in 1 patient). The authors speculate that the increased incidence of infection may have been due to impaired neutrophil function, which was documented in a subset of the infliximab treated group. Another possibility, which was not examined in this study, is the effect of infliximab on activated T lymphocytes. Infliximab has the potential to affect T lymphocyte survival and function and it has been implicated in signaling via the transmembrane TNF-α, and it may induce apoptosis in activated T lymphocytes.20 Thus, infliximab may not only bind circulating TNF but also may destroy TNF-producing immune cells. Loss of activated T lymphocytes may also be a relevant aspect underlying immune suppression and consequent increase in infections in the infliximab group, possibly explaining infliximab's longer pharmacodynamic effect compared to that of other agents, such as etanercept.
This study highlights the difficulties in designing clinical trials for patients with alcoholic hepatitis. Dysregulated cytokine metabolism was described in alcoholic hepatitis long before it was recognized in inflammatory bowel disease and rheumatoid arthritis. Yet, anti-TNF therapy has been an FDA-approved, highly effective therapy for the latter diseases for several years. An initial concern with anti-TNF therapy in alcoholic liver disease came from early observations that low “basal” amounts of TNF were important for liver regeneration.21 Thus, many investigators took the tack that downregulating TNF production but not totally blocking TNF activity would be a preferred therapeutic intervention. Moreover, many investigators feel it is unethical to withhold prednisone therapy from patients with alcoholic hepatitis. One option is to evaluate potential incremental benefits of therapy in addition to corticosteroids, as was done in this study. Another problem with many studies in alcoholic hepatitis is the exclusion of large numbers of patients in the “real world” who still need some form of therapy. This study points out the importance of having a strong data safety monitoring plan and a Data Safety Monitoring Board, and this particular research team acted in a highly responsible fashion. This study clearly demonstrates that the combination of infliximab plus corticosteroids in an immunocompromised host can lead to infections of diverse etiology. In retrospect, the combination of prednisolone plus high-dose infliximab may have been an overly aggressive approach.
Where should we proceed from here concerning anticytokine therapy in alcoholic hepatitis? A Phase II randomized, double-blind, placebo-controlled study in patients with moderate to severe alcoholic hepatitis is being sponsored by the National Institutes of Health (NIH). This trial is based on a recently published pilot protocol suggesting the safety of one such drug in alcoholic hepatitis.22 The principal investigator, Dr. Vijay Shah from the Mayo Clinic, selected etanercept, a TNF-soluble receptor, as a therapeutic intervention, and no patients (active treatment or control) will receive prednisone. This anti-TNF therapy may be more appropriate in that the duration and strength of action are more limited. Pentoxifylline has shown promising efficacy in a broad spectrum of patients with alcoholic hepatitis including those with hepatorenal syndrome.23 Pentoxifylline is a nonselective phosphodiesterase inhibitor. Pentoxifylline treatment increases intracellular concentrations of adenosine 3′,5′-cyclic monophosphate and guanosine 3′,5′-cyclic monophosphate, a potential mechanism for inhibiting TNF production. Additionally, pentoxifylline has the potential to block reactive oxygen species–mediated regulation of nuclear factor-κB independently of the phosphodiesterase inhibitory activity.24 One of us (D.H.) has an ongoing NIH-funded pilot study using a prostaglandin analog (misoprostol) in alcoholic hepatitis. This trial was based on our early observations that a prostaglandin analog was able to downregulate TNF production and attenuate hepatotoxicity in a rat model of alcoholic hepatitis.10 Similar to pentoxifylline, prostaglandin analogs also increase 3′5′ adenosine cyclic monophosphate, one mechanism for inhibiting TNF production. Importantly, we performed ex vivo dose-finding and pharmacodynamic studies to document anti-TNF and anti-interleukin-8 activities with this agent and compared it to prednisone as a “gold standard.”7 Thus, patients were given either prednisone or misoprostol for 2 days, and their blood was collected and allowed to sit for 24 hours with or without low-dose endotoxin stimulation. At the end of 24 hours, plasma was removed, then assayed for TNF and interleukin-8 activity (Fig. 1). In this fashion, we were able to determine whether the amount of drug in the bloodstream was able to effectively inhibit basal- or endotoxin-stimulated cytokine production.
There are now a host of drugs and complementary and alternative medicine agents that have inhibitory effects on cytokine production. The above-described dose-finding and pharmacodynamic studies should optimize the success of future large, multicenter trials. Moreover, it would be valuable to have animal models to screen potential therapeutic agents for efficacy in alcoholic liver disease. The intragastric ethanol feeding model has been a valuable tool in documenting that TNF plays an etiological role in the development of alcoholic liver disease. However, few studies have attempted to use this model to test the efficacy of therapeutic agents once liver injury has been established. If one could fine-tune this model to develop inflammation and fibrosis within 4 to 6 weeks, new drugs could be tested to determine the efficacy in halting or reversing injury. Clearly, as highlighted by the difficulties in the study by Naveau et al., animal protocols are much easier to perform than human trials, and this intragastric infusion model could be a preclinical screening technique.