Written by Kimiko Dunbar
Spoon Feed
Bayesian analysis of the use of EpiDex in bronchiolitis demonstrates a reduced probability of hospitalization for bronchiolitis, although highly skeptical clinicians may require additional evidence.
More is more?
National guidelines recommend supportive care for bronchiolitis, advising against adjuncts such as nebulized epinephrine and steroids. Prior data have demonstrated a synergistic effect in the combined use of nebulized epi and oral dex (EpiDex) in other respiratory illnesses such as asthma. This study re-analyzed data from the CanBEST trial using Bayesian methods. The original study demonstrated a significantly decreased rate of hospitalization in adjusted models but not in unadjusted models, leading to confusion in how to interpret the results. Authors utilized an alternate method of re-analysis – in other words, the Bayesian approach allows for the inclusion of pre-existing evidence or clinical expertise into the analysis. Using various prior beliefs, researchers found that there is ≥98% probability that EpiDex reduces hospitalizations compared to placebo, unless clinicians are highly skeptical. Even under skeptical assumptions, there remained a 90% chance of ≥10% reduction. The study is limited by varying definitions for bronchiolitis, possibly including infants with a mixed picture.
How does this change my practice?
I’m one of the “highly skeptical” clinicians; clinical practice guidelines routinely advise against the use of adjunct therapies in bronchiolitis. While compelling, I don’t have enough familiarity with the Bayesian approach to change my interpretation of the initial study, which did not show a difference in readmission rates in adjusted models. I suppose this study has softened me to the idea of EpiDex a bit, although I won’t be reaching for it unless there are alternate clinical indications for the use of EpiDex, such as stridor or “croup-ilitis”.
Source
The probability of reducing hospitalization rates for bronchiolitis with epinephrine and dexamethasone: A Bayesian analysis. PLoS One. 2025 May 16;20(5):e0318853. doi: 10.1371/journal.pone.0318853. PMID: 40378135
