Letter to the Editor: Survival in placebo-treated patients with severe alcohol-associated hepatitis.

Abstract

To the editor, With great interest, we read the recent article of Szabo et al,1 reporting a negative randomized controlled trial (RCT) in patients with severe alcoholic hepatitis (sAH) and the accompanying editorial of Lucey and Thursz.2 We share the concern for future trial design in sAH regarding the sample size needed to demonstrate the survival benefit of a specific therapy. Insight into the survival of placebo-treated patients, as a surrogate for the natural history of the disease, is crucial herein. Therefore, we performed a systematic review and meta-analysis of all RCTs in sAH that included a placebo-controlled group (>10 patients). A systematic search on Pubmed, Embase, Cochrane library, and Web of Science (until March 2021) retrieved 4874 studies. Eleven RCTs reporting 30-day survival (568 placebo-treated patients) and 2 studies reporting 90-day survival (269 placebo-treated patients) were included (Figure 1). Data were analyzed at study-level using a linear mixed model (DerSimonian and Laird), with the arcsin-transformed survival rates as outcome.FIGURE 1: Pooled 30-day and 90-day survival in placebo-treated patients with severe alcohol-associated hepatitis. Data analyzed using a linear mixed model (DerSimonian and Laird) with arcsin-transformed survival rates as outcome. Median with 95% CI plotted.Pooled 30-day survival was 72.93% (95% CI: 67.09%–78.39%) and pooled 90-day survival was 73.61% (95% CI: 68.18%–78.69%) in placebo-treated sAH patients. A more recent year of publication was positively associated with 1-month survival (regression coefficient 0.0053, p=0.0037). Notably, the last 2 Western trials (n=289), demonstrated a 30-day survival of 80% and 83%, respectively, of placebo-treated sAH patients. If a given therapy would reduce mortality (30 or 90-day) in sAH-patients by 25% (comparable with the 28.7% used in the power analysis of the STOPAH trial), this would be of great added value to clinical practice. However, with a 27% mortality rate, this would mean that 622 patients need to be included in each treatment group in a 2-armed study (power calculated with a log-rank test, power=80%, alpha=0.05), which results in a study population of 1244 patients without taking into account drop-outs. This number of patients is larger than that reported in any RCT in sAH so far, including the STOPAH trial where patients were included in 65 different hospitals over 3 years.3 Not even to mention the study patient number we would need if the 30-day mortality rate is less than 20% as in the recent RCTs. In conclusion, the shown survival figures of placebo-treated sAH patients have major implications for the required sample size of future trials in sAH, questioning the feasibility of demonstrating superiority over placebo in a classical RCT design. Therefore, we fully agree with the conclusion of Lucey and Thursz that alternative trial designs should be explored to identify effective new treatments for this extremely vulnerable patient population.2