Glucocorticosteroids are probably ineffective in alcoholic hepatitis
ACP J Club. 1996 Jan-Feb;124:13. doi:10.7326/ACPJC-1996-124-1-013
Christensen E, Gluud C. Glucocorticoids are ineffective in alcoholic hepatitis: a meta-analysis adjusting for confounding variables. Gut. 1995 Jul;37:113-8. [PubMed ID: 7672658]
To determine whether glucocorticoids are effective in alcoholic hepatitis after adjustment for patient characteristics.
Studies were identified using MEDLINE and bibliographies of relevant papers and meta-analyses.
Studies were selected if they were published randomized controlled trials that included patients with alcoholic hepatitis and compared the short-term (< 3 mo) effect of glucocorticoids on survival. Studies had to include a placebo or no drug comparison and separate data on patient characteristics for both treatment groups.
Data extracted included study quality, survival, patient characteristics (age, sex, encephalopathy, ascites, and mean albumin and bilirubin levels), and drug studied, including doses and duration of follow-up. Other characteristics (fever; renal insufficiency; hematocrit; aspartate aminotransferase, alkaline phosphatase, hemoglobin, or leukocyte levels; or hospital days) were not studied because of missing data.
13 studies were included (659 patients). 8 studies evaluated prednisolone, 4 studies evaluated methylprednisolone, and 1 study evaluated prednisone. Data were not pooled because of heterogeneity. The therapeutic effect weighted for sample size was calculated for all trials. Large trials showed no association between the therapeutic effect and survival; a positive effect was seen only in small trials, suggesting publication bias. No association was found between the therapeutic effect and quality score or type of glucocorticoid, daily dose, or duration of therapy. The test of heterogeneity showed a large imbalance between the treatment and control groups for patient age and sex, encephalopathy, ascites, and bilirubin and albumin levels. Accounting for these multiple factors, prediction of the risk for death in each of the treatment and control groups showed that poor prognosis was associated with a high prevalence of encephalopathy and a high bilirubin level. The same analysis showed that adjustment for possible imbalance in prognostic variables found no significant interaction between glucocorticoids and mortality and that the outcomes did not differ for various patient groups. Evidence is not available to ascertain which subgroup of patients would benefit from glucocorticoids.
Published evidence does not support the routine use of glucocorticoids in patients with alcoholic hepatitis.
Source of funding: Not stated.
For article reprint: Dr. E. Christensen, Department of Medicine B, Bispebjerg University Hospital, Bispebjerg Bakke 23, DK-2400 Copenhagen NV, Denmark. FAX 45-35-313-966.
The meta-analysis by Christensen and Gluud and 3 related meta-analyses (1-3) are based on as many as 13 randomized controlled trials on the use of glucocorticoids in alcoholic hepatitis. 659 patients were studied in these trials, and 8 of the 13 trials included < 50 patients. 8 trials were negative, but few were large enough to exclude a false-negative conclusion. 1 of the earlier meta-analyses concluded that glucocorticoids reduce short-term mortality in patients who have acute alcoholic hepatitis but not acute gastrointestinal bleeding (2). The other 2 meta-analyses (1, 3) concluded, with qualified enthusiasm, that corticosteroid therapy may reduce short-term mortality in a subgroup of patients with severe alcoholic hepatitis but that better studies are needed in which the various subgroups are studied.
Why are the conclusions of this latest meta-analysis by Christensen and Gluud different than the other 3? The report by Christensen and Gluud includes 13 published trials, 2 more than any of the other meta-analyses. Also, of the 4 published meta-analyses, it is the best for addressing treatment and control imbalances and raises serious questions about publication bias among the trials. Nevertheless, these results will not put the issue completely to rest. Any benefit from glucocorticoid therapy in alcoholic hepatitis, if present, may be confined to patients with severe disease (4, 5). Christensen and Gluud did not identify encephalopathy or other factors as pretreatment predictors of response, nor did they stratify patients according to the presence or absence of gastrointestinal bleeding or infection. Rather than excluding a treatment effect, this meta-analysis points out the inadequacies of the published studies in characterizing patients with severe disease who might respond to glucocorticoids and confirms that further trials in this area are necessary.
Paul D. King, MD
University of Missouri HospitalColumbia, Missouri, USA