Issues in Noninferiority Trials: The Evidence in Community‐Acquired Pneumonia

Open Access
Publisher Website
Google Scholar
Abstract
When investigators hypothesize that experimental interventions provide advantages other than improved effectiveness, they can use noninferiority (NI) trials to determine whether one can rule out the possibility that those interventions have unacceptably worse effectiveness than the standard "active control" regimen. To conduct valid NI trials, there must be evidence from historical studies that provides reliable, reproducible, and precise estimates of the effect of the active control, compared with that of placebo, on the specific outcomes investigators plan to use in the NI trial; this effect should be of substantial magnitude, and the estimates of the active control's effect from historical studies must represent its effect in the planned NI trial had a placebo group been included. These conditions allow formulation of an NI margin such that, if the NI trial establishes that the effectiveness of the experimental intervention is not worse than the effectiveness of the active control by more than the NI margin, then one can conclude that the experimental regimen (1) preserves a substantial fraction of the effect of the active control and (2) will not result in a clinically meaningful loss of effectiveness. After general discussion of NI trial design issues, we consider the design of NI trials to evaluate antimicrobials in the treatment of community-acquired pneumonia. We present an extensive literature review, allowing estimation of the historical effect of active control regimens in community-acquired pneumonia primarily on the basis of evidence related to use of sulfonamides or penicillin. This review allows formulation of NI margins that are specific to age and bacteremia status of patients.