The therapeutic use of corticosteroids in AH has to be viewed from the perspective of trials in which their effect was evaluated. These trials have numerous limitations, including the heterogeneity of patients with AH, presence of nonresponders to corticosteroids, and exclusion of patients with infections, GI bleeding, or renal insufficiency.

AH is associated with significant inpatient mortality that can reach up to 80 % in the most severe forms [78]. The main causes of death are liver failure, sepsis, hepatorenal syndrome, and gastrointestinal bleeding. Evaluation of treatment effect on short-term survival requires identification of patients with significant risk of death. Until the Discriminant Function became available [78, 79], no objective and reproducible criterion existed to predict early mortality. The heterogeneity of patients with AH that were enrolled in corticosteroid trials in pre-DF era resulted in a wide range of survival in the placebo-treated arms, ranging from 81 % if patients with mild AH were included to 0 % survival in trials with patients with the most severe AH [76–83].

The DF described by Maddrey et al. identifies patients with a high risk of early (1–2-month) mortality [79, 81]. It is calculated as 4.6 (prothrombin time − control prothrombin time [in seconds] + bilirubin [in μmol/L]/17). The presence of DF > 32 and/or encephalopathy is used as a definition criteria for acute AH, which, in the absence of treatment, has spontaneous survival between 50 and 65 % [78, 79, 81]. After acceptance of DF > 32 as inclusion criteria of patients with AH to the clinical trials, the short-term survival of glucocorticoid-treated patients has remained remarkably constant, with a 2-month survival rate of approximately 80 % [73, 79, 84–86].

The variable severity of AH in patients enrolled into clinical trials was likely the strongest contributor to the “controversial evidence” of the efficacy of steroids in AH, advocated by some authors [87–89]. This skepticism was further deepened by the widespread belief in the 1970s and 1980s that corticosteroids caused duodenal ulcers [90] and by the belief that corticosteroids increase the risk of infection [91]. The latter myths have been recently disproved by two randomized controlled trials ([91, 92] and below), and the “steroid controversy” has been the subject of multiple meta-analyses [73, 74, 89, 93, 94] and of an ongoing prospective trial [88].

The attempts to clarify whether corticosteroids improve survival in AH followed two approaches. The first approach relied on the classical methodology of meta-analysis utilizing published results of randomized controlled trials [89, 93, 94]. The second approach was based on a meta-analysis of primary, raw data derived from randomized controlled trials but restricted only to patients with severe alcoholic hepatitis (i.e., those having DF > 32 and/or encephalopathy) [73, 74].

In 1990, Imperiale et al. [93] studied 11 randomized trials from 1966 to 1989 and found that patients with alcoholic hepatitis treated with steroids were associated with a 36 % (CI 95 %, 28–50 %) reduction in mortality. In this meta-analysis, the mortality benefit was only evident in patients with hepatic encephalopathy, and the influence of admission DF on therapeutic effect of steroids was not evaluated in this study [93]. In 1994, Christensen et al. [94] examined 13 randomized controlled trials published between 1971 and 1992 and found no survival benefit of therapy of corticosteroids. In this meta-analysis, Christensen et al. did not perform a subgroup analysis of the data [94], but a review of the studies included in that meta-analysis indicates the survival benefit of steroids in all three studies in which Maddrey’s DF was used as inclusion criteria (studies [78, 79, 81] within meta-analysis [94]). This study was followed by another meta-analysis from the same group, published in 2008 under the label of the Cochrane Hepato-Biliary consortium [89] that examined 15 randomized trials with a total of 721 patients. The study found that corticosteroids did not reduce mortality of the entire population of patients with alcoholic hepatitis, compared with placebo, but mortality was statistically reduced in subgroups that were either experiencing hepatic encephalopathy or could be characterized as having severe AH with DF > 32 [89]. In spite of the beneficial effect of corticosteroids in patients with severe AH (as opposed to patients with mild or moderate AH), both meta-analyses [89, 94] lead Christensen et al. to conclude that corticosteroids are ineffective in patients with AH [94].The appropriateness of this conclusion was challenged by Imperiale, O’Connor, and McCullough [95] who emphasized the beneficial effect of corticosteroids in patients with severe alcoholic hepatitis.

In 2001, Mathurin et al. [73] used pooled primary data from three placebo-controlled trials to compare corticosteroids with placebo in patients, all of whom had severe AH (DF > 32). In this group of patients, at 28 days, patients who received corticosteroids had a significantly higher survival (85 % vs. 65 %) compared to those who received placebo, revealing that steroid treatment of severe alcoholic hepatitis is associated with a survival advantage, with a number needed to treat of 5. Increasing age and serum creatinine were found to be independent prognostic factors for death from severe AH in this study [73]. Nine years later, Mathurin et al. published another meta-analysis [74] based on individual patient data by adding two additional randomized controlled trials [65, 96] to their previous analysis [73]. Again, only data for patients with DF > 32 on admission were included in meta-analysis. Combination of all five RCTs [65, 79, 81, 96, 97] confirmed beneficial effects of corticosteroids on short-term survival, demonstrating an 80 % survival of patients treated with steroids at one month, compared to 65 % survival in patients treated with placebo [74].

About one-fourth of patients with severe AH (DF > 32) will not respond to corticosteroids [85]. These patients have a grim prognosis, with fewer than 25 % surviving 6 months. The benefits of early identification of nonresponders to steroid treatment would be that these patients could be spared of side effects of unnecessary corticosteroid treatment and that alternative therapies, such as biological treatment or liver transplant, advocated by some authors [98, 99], could be considered.

Nonresponsiveness to steroids can be predicted in an assay that tests responsiveness to prednisone in peripheral blood mononuclear cells isolated from patients with AH [100]. In this assay, termed DILPA (dexamethasone suppression of lymphocyte proliferation test), lymphocytes isolated from the blood of patients with severe AH are stimulated with phytohemagglutinin in vitro with or without the presence of dexamethasone [100]. The assay duration is about 48 h and requires ³H thymidine for the testing of lymphocyte proliferation. A lack of response to steroids is defined as suppression of lymphocyte proliferation by less than 60 % of the maximal proliferation count. The accuracy of the DILPA assay in predicting 6-month survival in patients with severe AH is 0.86 (as determined by area under the ROC) [101]. The DILPA assay has not been widely accepted in clinical practice.

Out of clinical parameters that would predict responsive to corticosteroids, the early change in bilirubin levels (ECBL, defined as bilirubin level at 7 days lower than bilirubin level on the first day of treatment) has shown high predictive value [85]. In their 2003 prospective study involving 238 patients, Mathurin et al. [85] set out to identify clinical and laboratory markers that would predict lack of response to corticosteroids in patients with severe AH (DF > 32). They used 6-month survival as an end point because of the rule requiring 6 months for listing alcoholic patients for liver transplantation. All patients were treated with corticosteroids. The overall 1-month survival was 85 %, consistent with previous data on survival in corticosteroid-treated patients with AH [73, 74], and the 6-month survival was 64 %. An ECBL at 7 days was observed in 75 % of patients. At 6 months, survival of patients with ECBL (83 %) was significantly higher than in patients without ECBL (23 %). ECBL was by far the strongest predictor of 6-month survival (odds ratio 7.1 [95 % confidence interval 4.0–11.1]) for 6-month survival compared to patients with no ECBL. The study suggested that evolution of bilirubin level at 7 days may be useful in identification of severe AH patients with poor prognosis, and the predictive value of ECBL was higher than that of PT or DF.

The ECBL has been incorporated as a dynamic component into the Lille model predicting the 6-month survival in patients with severe AH [102]. Other clinical predictors included in the Lille model are age, the presence of renal insufficiency, albumin level, and prothrombin time. The Lille score calculator (available at http://​www.​lillemodel.​com/​score.​asp) generates a score that ranges between 0 and 1. The ideal cutoff of 0.45 can be used to define responders to corticosteroids (Lille score < 0.45) and nonresponders (Lille score > 0.45). The therapeutic response to steroids can be further refined by defining three populations: complete responders (Lille score < 0.16), partial responders (Lille score 0.16–0.56), and null responders (Lille score > 0.56) [98]. This sub-stratification has prognostic value, as demonstrated in two subsequent studies by Mathurin’s group [92, 103]: the 6-month survival is about 90 % in complete responders, 45–78 % for partial responders, and less than 45 % in null responders. DF, MELD, or Glasgow score have significantly worse accuracy to predict 6-month survival, compared to the Lille score (are under the ROC of 0.73, 0.72, and 0.67, respectively, vs. 0.85 for Lille score) [102]. In patients with Lille score > 0.45 (nonresponders), the EASL guidelines recommend to discontinue corticosteroids after 7 days of treatment [12].

Infection has long been considered a contraindication to glucocorticoid therapy in patients with severe alcoholic hepatitis. It is estimated that the presence of infection as exclusion criterion in patients with severe AH may have barred 25 % of otherwise eligible patients from consideration of treatment [91]. It is thought that liver failure with resulting immune paralysis is a major factor driving increased susceptibility to bacterial infections in patients with advanced liver cirrhosis or severe AH [104, 105].

In their prospective study, Louvet et al. [91] performed screening for infection in 246 patients admitted for severe AH. All patients underwent chest X-ray and their blood, urine, and ascites were cultured. Out of all patients, 63 (25 %) were found to have an infection (spontaneous bacterial peritonitis, urinary tract infection, and pneumonia being the most prevalent infections) and received antibiotic treatment. Corticosteroids were started only after a priori defined criteria for antibiotic treatment response were met. Patients with severe AH and with infection responding to antibiotic treatment had similar 2-month survival compared to patients with severe AH and no infection (71 % vs. 72 %). Data on 6-month survival were not provided in the study [91].

However, there was a significant difference in survival when patients with infection treated prior to initiation of steroids (above) were compared to patients who developed infection while on treatment with steroids. After initiation of corticosteroids, 57 patients (24 %) developed an infection after a median time of 14 days, and there has been substantial proportional increase in pneumonia (40 % of infected patients). Patients infected after corticosteroid treatment had significantly lower 2-month survival (46 %) compared to noninfected patients (78 %). Only the Lille score and MELD score were independent predictors of survival in these patients, whereas ascites, encephalopathy, Maddrey DF, leukocytosis, or C-reactive protein levels were not. Importantly, the presence of infection after corticosteroid initiation proved to be a negative determinant of survival only in patients who responded to corticosteroids (i.e., Lille score < 0.45), but not in nonresponders (Lille score > 0.45) where it is thought that it was the lack of steroid response but not infection determining unfavorable prognosis in nonresponders [91].

This study suggested that it is safe to use corticosteroids in patients with severe AH if infection is identified and treated prior to initiations of corticosteroids. The study also showed that survival benefit of antibiotic therapy in patients who develop infection after initiation of corticosteroid treatment can be expected only in responders to steroids [91].

Both AASLD and EASL guidelines specify that patients with severe AH with recent upper GI bleed are not ideal candidates for corticosteroid treatment [3, 12]. The reason for this may be a carryover from clinical trials performed in the 1970s and 1980s in which patients with upper GI bleeding were excluded due to the belief that corticosteroids caused gastroduodenal ulcers and also because no effective treatment of upper GI bleeding existed at that time [90]. Since then, however, much has changed, including the advent of proton pump inhibitors, endoscopic treatment of variceal bleeding, and transjugular portosystemic shunting [106]. Also, antibiotics given for 7 days after GI bleeding in patients with cirrhosis reduce infections and increase survival [107].

Rudler et al. [92] conducted a retrospective analysis of survival among patients with severe AH who presented to a hospital with upper GI bleed and compared them with patients with severe AH without GI bleeding. A total of 48 patients with upper GI bleed and 47 patients without GI bleed were analyzed. The two groups did not differ in the presence of AH on biopsy (approximately 80 %) and DF or MELD score. After stabilization and effective bleeding control per Baveno V recommendations [108], both groups were started on corticosteroids. The 6-month survival was similar in both groups (74 % vs. 70 %). The probability of developing an infection after starting corticosteroids was lower among subjects with upper GI bleed (24 %) as compared with subjects without upper GI bleed (45 %). This was attributable to antibiotic therapy mandated in patients with acute GI bleed and could have improved survival in GI bleeders in the study. If validated in prospective trials, this data indicate that GI bleed does not worsen survival in AH patients treated with corticosteroids [92].

Based on recent experimental evidence on the effect of alcohol on the gut, interventions that prevent alcohol-induced increase in gut permeability and microbial translocation and/or restore alcohol-induced disturbance of the gut microbiome could be reasonable interventions. It remains to be evaluated whether these components should be targeted individually or collectively to achieve improvement in alcoholic liver disease or in alcoholic hepatitis [118, 119]. Animal and human studies identified the benefit of zinc on gut permeability [120]. Targeting the microbiome composition by using VSL3 or lactobacillus GG may have benefits in alcoholic liver disease [121, 123]. Fecal transplantation has proven benefit in Clostridium difficile infection, and it is in early clinical trials in alcoholic liver disease [122]. Additional potential targets based on animal studies include miR-155 inhibition [124]. Farnsenoid receptor (FXR) activation inhibited inflammation and preserved the intestinal barrier in inflammatory bowel disease, and its potential role in alcoholic liver disease is yet to be explored [125] (Fig. 15.1).

Alcohol-induced steatosis in hepatocytes is an early effect of alcohol that is sustained during chronic alcohol use. Of the many signaling mechanisms regulating hepatocytes fat accumulation, activation of the PPAR-alpha was shown to have benefits on murine alcohol-induced liver disease [126]. PPAR-gamma and PPAR-delta agonist treatment improved hepatic insulin resistance in another study [126].

Recruitment of macrophages and neutrophils to the liver in alcoholic liver disease is mediated by chemokines [127]. MCP-1 (also named CXCL2) is produced by alcohol-exposed hepatocytes and immune cells are increased in early alcoholic liver disease and they recruit monocytic cells into the liver. In addition, MCP-1 induces fat accumulation in hepatocytes [128, 129]. In mice deficient of MCP-1, chronic alcohol feeding resulted in attenuated steatosis and inflammation suggesting that inhibition of MCP-1 may provide benefits in ALD. A dual CCR2/CCR5 antagonist cenicriviroc, is currently in clinical trial in human nonalcoholic steatohepatitis and testing of this antibody in ALD awaits investigation.

Targeting pro-inflammatory cytokines in acute alcoholic hepatitis has been the target of previous and current investigations. Clinical trials using anti-TNF-α blocking antibodies or inhibiting TNF receptor I had unsuccessful outcomes due to increased infections [130–133]. While TNF has complex effects on the immune system and the liver, studies with anti-TNF-α are perfect reminders about the impaired immune system of patients with alcoholic liver disease. Alcohol compromises both innate and adaptive immunity and the increased pro-inflammatory cascade activation in acute alcoholic hepatitis occurs together with impaired immune responses in these patients [134]. It remains to be seen whether anti-inflammatory strategies alone or in combination with other treatment could be beneficial. Currently, clinical trials have been initiated with the combination of recombinant IL-1 receptor antagonist combined with zinc and pentoxifylline in severe acute alcoholic hepatitis. Another study will test the effects of an anti-IL-1 antibody in combination with steroids in severe alcoholic hepatitis (Table 15.2).