Many clinical trials compare the efficacy of a new treatment to no treatment, or to a placebo. However, for diseases where a standard drug or treatment is already established to provide a clinically important reduction in the risk of irreversible morbidity or mortality, a trial might compare the novel treatment to the standard treatment. Often, trials of this type ask whether the new treatment is as safe and has efficacy that is not unacceptably worse than that of the standard treatment – this type of trial is called a non-inferiority trial. Designing a non-inferiority trial poses important challenges. VIDD member Dr. Tom Fleming and colleagues considered guidelines for designing such trials, particularly regarding the determination of the “non-inferiority margin,” or the minimum threshold for what would constitute unacceptable loss of efficacy.
The researchers outlined two major considerations. One relates to adjustments needed to account for the level of uncertainty about the true effect of the standard treatment. If the non-inferiority trial does not also include a placebo group, the standard treatment’s efficacy in that trial must be determined from past data, and this may induce risk of bias for a number of reasons. For example, if the standard treatment is an antibiotic, resistance levels to that drug in the patient population might be different at the time of the non-inferiority trial than when the antibiotic was first evaluated. The scientists suggest that trial design should carefully take into account bias when determining the baseline efficacy of the standard treatment. The second consideration the researchers discuss is the need to select the non-inferiority margin in a manner to achieve preservation of an appropriate percentage of the standard treatment’s efficacy, so that novel treatments clinically less effective than standard treatments will not be pursued further.
Fleming TR, Odem-Davis K, Rothmann MD, Li Shen Y. Some essential considerations in the design and conduct of non-inferiority trials. Clin Trials. 2011 Aug;8(4):432-9.