A review of issues involving non-inferiority trials: A review of study design features, sample size, and analytic issues.

摘要

Randomized, non-inferiority (abbreviated hereafter as NI) trials directly compare the new experimental therapy to an existing effective treatment without a placebo or with no treatment arm. With advancement in medicine, non-inferiority trials with active control are being utilized more often when the classic placebo-controlled superiority trial is considered unethical due to existing effective therapy and to test the efficacy of the alternative treatment. Non-inferiority trials aim to demonstrate that the new test treatment is not inferior to the standard therapy. Oftentimes the new therapy claims to have some advantages over standard treatment such as less toxicity, simpler regimen or lower cost. For example, voriconazole may provide better adverse event profile than does amphotericin B deoxycholate in the treatment of invasive aspergillosis. Also, for community acquired pneumonia, a new quinolone is more convenient to administer than penicillin.;If the non-inferiority is established, then the new treatment will be accepted as a valid alternative with better benefit to risk ratio. Due to complexities inherent to non-inferiority trials, there are many challenges associated with this design. This paper will review and discuss important design features (i.e. choosing an appropriate NI margin, assessing 'assay sensitivity' and 'constancy assumption' that we define later), sample size estimation and analytic issues around non-inferiority trials (i.e. intent to treat versus per protocol and fixed margin versus synthesis approach to determine non-inferiority). Non-inferiority trials can be anticonservative in nature. This means that non-inferiority trials are more likely to yield desired results than superiority trials thereby more vulnerable to biases. Hence the assumptions underlying the NI design must be made explicitly and the resultant conclusion should be reviewed rigorously for the new test treatment to be clinically accepted and marketed to the general public.