Article Text


Corticosteroids for severe alcoholic hepatitis: what constitutes response?
  1. R Hamid *1,
  2. H Lafferty1,
  3. E Watson1,
  4. J Winter1,
  5. A J Morris1,
  6. A J Stanley1,
  7. E H Forrest1
  1. 1Department of Gastroenterology, Glasgow Royal Infirmary, Glasgow, UK


Introduction Severe alcoholic hepatitis (AH) is associated with high short term mortality. Although still attracting some controversy, many hospitals will treat severe AH with corticosteroids (CS). Early identification of patients not responding to CS treatment is crucial as continuing treatment may be associated with side-effects and be considered futile. More importantly, identification of CS non-responders (NR) is essential as alternative therapies may be considered to improve their poor survival prognosis. The aim of the present study was to compare three previously published markers of CS response: Early Change in Bilirubin (ECBL); 25% fall in bilirubin from baseline (25%↓B); and the Lille Score.

Methods 83 patients with severe AH (Glasgow Alcoholic Hepatitis Score ≥9) treated with CS between 2004 and 2009 were analysed. Unit protocol was for discontinuation of CS treatment if there was not a 25%↓B, however some patients continued CS treatment at the discretion of the responsible consultant. ECBL, 25%↓B, and Lille Scores were calculated and outcome evaluated at day 28 and 84. Prognostic utility of the various scores was determined by generating a receiver operating characteristic curve (ROC) and assessed by comparison of area under ROC curves (AUROC).

Results Overall 28 and 84 day mortalities were 24% and 37% respectively. The AUROCs (95% CI) for each indicator of CS response are shown in table 1.

Table 1

PTH-090 AUROCs (95% CI) for each indicator of CS response for all patients (n=83)

For those patients who continued CS treatment irrespective of response, a 25%↓B to identify poor outcome at 28 days had a sensitivity of 78% and a specificity of 76%. The equivalent Lille Score cut off point of 0.45 had a sensitivity of 88% and a specificity of 69%. An ECBL had a sensitivity of 33% and a specificity of 96% (figure 1).

The 28 and 84 day mortalities for NR were as follows: ECBL NR (n=20) 55%, 80%; 25%↓B NR (n=44) 41%, 54%; Lille NR (n=47) 40%, 55%.

Conclusion Response to CS treatment is an important marker of prognosis in severe AH. Although the Lille Score has a slightly greater AUROC for outcome after CS treatment, this was not statistically significant. A 25%↓B retains reasonable sensitivity and specificity and is certainly easier to calculate than the Lille Score, making it a simple and practical marker in determining CS response in AH.

  • alcoholic hepatitis
  • corticosteroid
  • prognostic markers
  • treatment.

Statistics from


  • Competing interests None.

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.