Non-Inferiority Studies: What do the numbers mean, how does the FDA interpret them?

Non-inferiority studies are used to investigate whether a new drug treatment is no worse than a reference (comparator drug) treatment. Simply put, non-inferiority trials are less risky than superiority trials where often times the comparator is placebo (which in most infections is not considered ethical). When planning a non-inferiority trial, a drug company (along with the regulatory agency) establishes the non-inferiority margin (NIM) which is the difference in the lower end of the 95% confidence interval [CI] (lower end 95% CI for new drug under study – lower end 95% CI for comparator drug). The US FDA has established a NIM of 12.5% while the European Medicines Agency (EMEA) uses 10%. This means that the lower end of the 95% CI for the new drug treatment needs to be no more than 10-12.5% lower (in terms of some efficacy measure for antibiotics) than the reference treatment to be considered non-inferior. Non-inferiority studies generally use the “per protocol” population for analysis, that is, only patients for whom the investigator followed the protocol without major violations are included in the analysis.

The US FDA in September 1998 issued a guidance entitled “Statistical Principles for Clinical Trials” where they stated that it is the lower equivalence margin (lower end of the difference of the 95% CI) that is used to determine non-inferiority. The upper end of the 95% CI difference does not have to reach or cross 0 since it is only the lower end that is examined.

Non-inferiority studies are generally designed to tell only 2 things: 1) A new drug is non-inferior to the comparator or 2) a new drug is inferior to the comparator. Non-inferiority studies generally are not designed to examine superiority. Superiority suggestions from a non-inferiority study should define that goal a priori (that is prior to the study), and use all enrolled patients (an intention to treat or ITT) analysis. Thus, claims of superiority from a non-inferiority study analyzed without those a priori definitions are fraught with problems.

Taking the example of iclaprim above, the 95% CI difference was -8.4, -1.0%, with the lower limit (-8.4%) meeting both the US FDA and EMEA NIM definitions (12.5% and 10% respectively). Thus iclaprim is non-inferior to linezolid in this study, regardless of whether or not the lower 95% CI limit difference reached or crossed 0. No claims of superiority can be made as this study was not designed a priori to examine these. Thus, unless there is a change in the opinion at the FDA in terms of interpreting this data, this study suggests non-inferiority of iclaprim to linezolid.