Non-Inferiority and Equivalence Trials

A non-inferiority trial asks whether a new treatment falls short of an established one by less than a margin fixed in advance. The conclusion rests on where the confidence interval for the difference lies relative to that margin rather than on a p-value against zero, so non-inferiority is claimed when the whole interval sits on the acceptable side of it. An equivalence trial applies the same logic in both directions, the structure used for bioequivalence between formulations.

The margin is the entire design, and it is meant to be justified rather than chosen for convenience, preserving a defined fraction of the comparator's own established benefit over placebo while staying inside what a clinician would consider unimportant. Regulatory practice makes this concrete: after the 2008 FDA guidance on cardiovascular risk in type 2 diabetes therapies, new agents were expected to rule out a hazard ratio above 1.8 before approval and above 1.3 with fuller data.

Such a trial is worth running when the new option offers something other than greater efficacy: fewer injections, better tolerability, no refrigeration, lower cost. What it concludes is bounded, since a successful result says the new treatment is not worse by more than the margin, not that the two are the same, and it licenses no claim of superiority unless that was tested in a prespecified sequence.

The characteristic abuse is a margin wide enough to swallow a real loss of effect, reported as though the drugs were interchangeable. Biocreep is the compounding version, in which each new agent is licensed as non-inferior to the last until, after several rounds, a product may be no better than the placebo the original comparator once beat. Poor conduct also works in the sponsor's favour here, since non-adherence and measurement noise pull the arms together.