GCSF stood out
The review's clearest finding was a signal for GCSF, which was the only treatment in the placebo comparison to show a statistically significant improvement in survival 90 days after treatment. The result came from four studies involving 186 people. Its risk ratio was 0.49, meaning the estimated risk of death was about half that in the placebo comparison. The 95% confidence interval ran from 0.26 to 0.90, and the reported p-value was 0.023. The analysis also reported an absolute risk reduction of 0.24 and a number needed to treat of 4.10.
The finding makes GCSF a candidate for future Phase III evaluation, rather than a settled answer for everyone with the disease. The review's stated aim was to compare treatments on 90-day survival and identify which ones should move into larger trials. In that sense, the result is a research signal, not a recommendation for individual care.
The corticosteroid result was different
Corticosteroids gave a different result. Across 29 studies and 2,358 participants, the pooled risk ratio was 1.12, with an interval from 0.69 to 1.81 and a p-value of 0.640. The review did not find a statistically significant association with lower 90-day mortality. That does not establish that corticosteroids have no benefit in every subgroup; it describes what this pooled analysis showed.
Death at 90 days was the primary outcome. The review also examined one-month survival, including 28-day and 30-day mortality, and the incidence of infection during treatment. Those were secondary outcomes, so the headline comparison was the 90-day result.
A broad review with a narrow evidence base
The researchers searched from database inception through 12 February 2025. They included randomized studies of adults with alcohol-associated hepatitis receiving medical treatments that did not involve transplantation, limited the evidence to full English papers, and excluded non-randomized, retrospective and observational studies. The protocol was uploaded to PROSPERO under CRD 420251010768.
Screening began with 11,094 titles. Of these, 11,028 were rejected, leaving 64 full manuscripts for closer review. Sixteen studies did not have extractable 90-day outcomes. The final network contained 35 studies, 5,045 participants and 24 treatments; 13 studies were placebo-controlled.
The researchers used a network meta-analysis, a method for combining evidence from a connected set of treatment comparisons. They pooled random-effects risk ratios with 95% confidence intervals and ranked interventions by P-score, a summary score used to order the treatments. Most comparisons used placebo as the reference, while corticosteroids served as the reference in a subgroup. Planned analyses also separated studies published since 2000, studies judged at low risk of bias, and severe disease defined by a Model for End-Stage Liver Disease score greater than 25.
Why the signal remains provisional
The size of the individual studies puts limits on how much weight to give the pooled result. The median trial arm contained 36 patients, with the middle half ranging from 21 to 66. Only eight studies averaged more than 100 patients per arm. Risk-of-bias assessment classified 18 studies as low risk, 16 as having some concerns and two as high risk.
Across the main 90-day network, overall survival was 71%, but study estimates ranged from 29% to 76%. Between-study variance was 0.25, classified as moderate, indicating that results did not line up perfectly from study to study. A funnel plot showed low asymmetry, and the Egger test was not statistically significant, with a p-value of 0.41.
Taken together, these findings leave GCSF as a candidate for Phase III testing, not a settled treatment effect. The central question is whether the 90-day signal survives a larger randomized study with stronger safeguards against bias.
Paper data and sources
Original title: Network Meta‐Analysis: Treatments for Acute Alcohol‐Associated Hepatitis
Authors: Gallaher C, Aithal G, Dhanda A et al.
Journal/Repository: Liver international : official journal of the International Association for the Study of the Liver
Status: Peer-reviewed
First online: 2026-10-11
DOI: Not available
Original paper · Full text