Severe alcoholic hepatitis is associated with a high mortality and the presence of liver failure, manifested by jaundice, coagulopathy, and often encephalopathy. Whereas insights into the pathophysiology of this devastating disease have improved over the past years, clinical progress in the last two decades has been minor. For more than two decades, steroids have remained the only moderately effective treatment option. One key confounder to most therapy studies has been the use of the discriminant function to identify patients at highest risk of mortality in the absence of better scoring systems more accurately predicting the outcome of severe alcoholic steatohepatitis. The now reported Glasgow alcoholic hepatitis score might represent a substantial improvement in clinical phenotyping and could catalyse the development of new treatments in this disease.

The appearance of steatohepatitis is an important rate limiting step in the development of progressive alcoholic liver disease. One month mortality rates of 40–50% have been reported in patients hospitalised with acutely decompensated liver disease due to alcohol induced steatohepatitis.¹ Patients with severe alcoholic steatohepatitis typically present with fever, hepatomegaly, jaundice, and anorexia. The presence of liver failure manifested by coagulopathy, jaundice, and/or encephalopathy is an indicator of poor outcome, usually highlighting the presence of limited hepatic functional reserve. Approximately 40–50% of patients have ascites and tender hepatomegaly is common. Leucocytosis is frequent and correlates with the severity of hepatic injury.² Neutrophilic infiltration is commonly seen on liver biopsy and these cells may play an important role in the hepatic injury. Although the diagnosis can be confirmed by liver biopsy, clinical and laboratory features are often adequate for establishing the diagnosis. Absolute values for serum aspartate and alanine aminotransferases are almost always <400 IU/l and higher concentrations should raise suspicion of concurrent liver injury due to viral or other aetiologies. Combined cirrhosis and alcoholic hepatitis is often associated and has the worst prognosis.³

The severity of this disease is best correlated with serum bilirubin level and prothrombin time after vitamin K administration.⁴ As part of a seminal clinical study on corticosteroid therapy in severe alcoholic hepatitis, the discriminant function (DF) formula (which includes prothrombin time and bilirubin levels) was derived to predict disease severity and individual mortality risk in these patients.⁴ This DF was later modified in the context of a further placebo controlled corticosteroid trial.⁵ A modified discriminant function (mDF) of >32 and/or the presence of encephalopathy in placebo treated patients in this latter study was associated with a 65% 28 day survival. A recent re-analysis confirmed this finding and also demonstrated that patients with a score <32 had a survival of 93%.⁶ Therefore, this cut off value of 32 has been used and recommended as the threshold to consider corticosteroid treatment.⁷ More recently, the MELD score has been applied to alcoholic hepatitis. In a study of 34 patients with a MELD score of >11, there was a 45% 30 day survival while those with a score ⩽11 had a survival of 96%.⁸ The MELD score includes creatinine values as both blood urea and creatinine values increase and may reflect severity in severe alcoholic steatohepatitis. They predict the development of the hepatorenal syndrome. A polymorph leucocytosis is found (>15–20×10⁹/l) in proportion to the severity of the disease.⁹

DO WE NEED BETTER SCORING SYSTEMS IN ALCOHOLIC HEPATITIS?

(i) We need a simple test which can be calculated at the bedside

Calculation of mDF relies on the absolute value of the prothrombin time and there exists significant variation in the absolute values of prothrombin time obtained using different assays in different countries (for example, UK, USA). This may thereby affect the validity of the mDF score in countries such as the UK where a greater severity score will be generated based on the prothrombin time. This creates a definite inaccuracy in the mDF value and limits its translation between different facilities. Furthermore, the presence of encephalopathy has often been included when making a treatment decision, in addition to just calculating the mDF. This is problematic, as encephalopathy is very subjective in its milder forms. Kulkarni and colleagues¹¹ have also shown that addition of encephalopathy displayed a slight increase in sensitivity but a dramatic decrease in specificity compared with the mDF. Therefore, encephalopathy is not an attractive parameter influencing treatment decisions (although this reflects current practice). The MELD score is also difficult to calculate at the bedside and inclusion of creatinine in the MELD may also limit its usefulness as creatinine values will be underestimated in the context of hyperbilirubinaemia unless corrected. In addition, it has been demonstrated that the MELD score is not superior to mDF in this population.⁸ Recently, the MELD score has been demonstrated to be a useful score for a panel of chronic advanced liver diseases, including severe alcoholic hepatitis, albeit it is only equivalent to the Child-Turcotte-Pugh score in predicting survival.¹⁰ Both scoring systems (mDF, MELD) use highly relevant variables in severe alcoholic liver disease but ignore other aspects of this disease, such as inflammatory parameters. Therefore, a score including a more complex panel of clinical variables but which can be easily assessed at the bedside would be highly desired.

(ii) We need a score which identifies as many patients as possible with a high mortality risk

Apart from the positive reports on the relevance of an mDF cut off value of 32,⁶ its accuracy in predicting survival has been repeatedly questioned. Kulkarni et al recently reported that patients with a low mDF (⩽32) also had significant mortality, approaching 17%.¹¹ In this study, the mDF was only moderately sensitive (66.7%) and specific (61.7%). Considering the considerable mortality in patients with a DF ⩽32, it would be reasonable to consider treatment options for these patients also. In addition, even more importantly, re-evaluation of current criteria for assessing disease severity is needed to more accurately identify patients at risk for short and long term mortality.

Given the relatively poor ability of the mDF to predict short term mortality, a new indicator of mortality and severity was eagerly awaited. Forrest and colleagues¹² report in this issue of Gut the derivation of a score termed the Glasgow alcoholic hepatitis score (GAHS) based on a large patient population and its validation using a new and independent set of patients (see page 1174). Importantly, overall survival was studied in a large patient population (n = 241) until day 84, allowing more accurate judgement about the clinical course and not only short term survival (day 28). In their analysis, mDF was highly sensitive in the prediction of death from alcoholic hepatitis but lacked specificity. This was dramatic as it incorrectly predicted the outcome at 28 days after admission in 51% of cases. By multiple stepwise logistic regression they developed the key factors for later deriving and developing their GAHS. All of these factors (age, white blood cell count, blood urea, prothrombin time ratio, and bilirubin) are well known from other studies to affect mortality but have never been used in combination to calculate survival.^13–15 In their analysis, patients with an admission mDF ⩾32 had a survival of 64% and 52% at days 28 and 84, respectively, whereas patients with a GAHS ⩾9 had a survival of 46% and 40% at days 24 and 84, respectively. The GAHS at days 1 and 7 was significantly more accurate in predicting day 24 and 84 outcome than the mDF and had an impressive better overall accuracy. Furthermore, the GAHS was also significantly more accurate than the MELD score in predicting short and long term survival.

(iii) Future developments and aspects which might be considered in further new scores

Whereas the new GAHS for the first time includes an inflammatory parameter (white blood cell count) which is important for such a highly inflammatory disease, other inflammatory and immune parameters may also be helpful for future scoring systems. C reactive protein levels and plasma levels of tumour necrosis factor soluble receptors p55 and p75 have been demonstrated to correlate with short and long term survival.^16–18 On the other hand, extracellular matrix markers such as laminin might be accurate markers predicting survival in these patients.¹⁹ Apart from immune markers, early changes in bilirubin levels and major changes in portal flow are prognostic factors in patients with severe alcoholic steatohepatitis treated with corticosteroids.^20,21

Treatment of severe alcoholic hepatitis has hardly progressed in recent years. Lack of a scoring system with appropriate overall accuracy may have contributed to this poor situation by confounding clinical studies. Current clinical strategy favours a “watch and wait” strategy awaiting disease progression in those patients incorrectly identified as low risk by a low DF (⩽32). Such a strategy may be counterproductive in a severe disease with high mortality and is an uncommon strategy in modern medicine. In addition, patients with a DF >32 often have contraindications for steroid treatment and cannot receive this treatment. Therefore, the new score presented here raises hopes of a more accurate clinical tool to predict survival and therefore advance the field, initiating new studies with innovative treatment modalities reflecting current pathophysiology. In conclusion, the new Glasgow alcoholic hepatitis score might indeed be the long awaited progress in this field.