Abstract

The aim of this study was to perform a meta-analysis of controlled clinical trials of glucocorticoid treatment in clinical alcoholic hepatitis, adjusting for prognostic variables and their possible interaction with therapy, because these trials have given appreciably different results. Weighted logistic regression analysis was applied using the summarised descriptive data (for example, % with encephalopathy, mean bilirubin value) of the treatment and control groups of 12 controlled trials that gave this information. Despite evidence of publication bias favouring glucocorticoid treatment, its overall effect on mortality was not statistically significant (p = 0.20)--the relative risk (steroid/control) was 0.78 (95% confidence intervals 0.51, 1.18). There was indication of interaction between glucocorticoid therapy and gender, but not encephalopathy. Thus, the effect of glucocorticoid treatment may be different (beneficial or harmful) in special patient subgroups. These results do not support the routine use of glucocorticoids in patients with alcoholic hepatitis, including those with encephalopathy. Whether other subgroups may benefit needs further investigation using the individual patient data from the published trials and testing in new randomised trials.