In reply:

Abstract

Dr. Carson disagrees with a qualified conclusion for the noninferiority trial by Rehrer et al.1Rehrer M.W. Liu B. Rodriguez M. et al.A randomized controlled noninferiority trial of single dose of oral dexamethasone versus 5 days of oral prednisone in acute adult asthma.Ann Emerg Med. 2016; https://doi.org/10.1016/j.annemergmed.2016.03.017Abstract Full Text Full Text PDF PubMed Scopus (18) Google Scholar In this study, the 95% confidence intervals for the outcome overlapped the prespecified noninferiority threshold, and therefore Dr. Carson argues that noninferiority is rejected and that the conclusion should simply state this negative result. Specifically for this study, single-dose dexamethasone should not be considered as effective as 5 days of prednisone for asthma. Dr. Carson represents the behaviorist approach to classic statistical analysis, in which one relies on hypothesis testing and arbitrary thresholds to accept or reject a hypothesis.2Schriger D.L. Problems with current methods of data analysis and reporting, and suggestions for moving beyond incorrect ritual.Eur J Emerg Med. 2002; 9: 203-207Crossref PubMed Scopus (18) Google Scholar Unfortunately, research outcomes are not always so neatly positive or negative. There are periodically high-quality trials with borderline results. The study by Rehrer et al1Rehrer M.W. Liu B. Rodriguez M. et al.A randomized controlled noninferiority trial of single dose of oral dexamethasone versus 5 days of oral prednisone in acute adult asthma.Ann Emerg Med. 2016; https://doi.org/10.1016/j.annemergmed.2016.03.017Abstract Full Text Full Text PDF PubMed Scopus (18) Google Scholar is a classic example, with the relapse size of effect being 2.3%, with 95% confidence intervals (–4.1% to 8.6%) that barely cross their prespecified 8% margin. As editors, should we unconditionally reject noninferiority according to this trivial 0.6% overlap? Should we not recognize that, for some readers, 9% might be a reasonable clinically important margin—a scenario in which a similarly rigid interpretation would abruptly reverse the study conclusion, or that other readers are interested in the general magnitude of the outcome without feeling forced to call the study results positive or negative? Given these principles, we encouraged the authors to openly communicate the equivocal nature of their outcome. Our Editor’s Capsule Summary stated that “[a] single dose of oral dexamethasone 12 mg is either similar to or slightly inferior to a 5-day course of prednisone 60 mg for asthma.” A black-and-white interpretation of what are often nuanced research findings is not a necessary or desirable component of evidence-based medicine. In fact, many argue that modern evidence-based medicine must go beyond frequentist statistical approaches to overtly adjust for study bias and selection of study bias.3Greenland S. Multiple-bias modeling for analysis of observational data (with discussion).J R Stat Soc. 2005; 168: 267-308Crossref Scopus (361) Google Scholar, 4Turner R.M. Spiegelhalter D.J. Smith G.C.S. et al.Bias modelling in evidence synthesis.J R Stat Soc. 2009; 172: 21-47Crossref Scopus (210) Google Scholar A Randomized Controlled Noninferiority Trial of Single Dose of Oral Dexamethasone Versus 5 Days of Oral Prednisone in Acute Adult AsthmaAnnals of Emergency MedicineVol. 68Issue 5PreviewOral dexamethasone demonstrates bioavailability similar to that of oral prednisone but has a longer half-life. We evaluate whether a single dose of oral dexamethasone plus 4 days of placebo is not inferior to 5 days of oral prednisone in treatment of adults with mild to moderate asthma exacerbations to prevent relapse defined as an unscheduled return visit for additional treatment for persistent or worsening asthma within 14 days. Full-Text PDF Clarification: Editorial Oversight of Results Reported in AnnalsAnnals of Emergency MedicineVol. 69Issue 4PreviewThe letter to the editor by Carson1 and the reply by Green and Schriger2 raise some important questions about the nature of noninferiority studies and the meaning of these results in the article by Rehrer et al.3 In my opinion both sides are correct. Noninferiority studies have recently enjoyed resurgence by examining previously validated and investigational therapies.4 The key to interpreting these studies lies in two domains, defining the noninferiority margin and interpreting the statistical outcomes. Full-Text PDF