ABSTRACT
The effects of therapies (e.g., the active control and experimental therapies in a non-inferiority trial) may vary across meaningful subgroups. A noninferiority inference involves concluding that the effectiveness of the experimental therapy is both superior to placebo and not unacceptably worse than the active control. To formally make such an inference, or to check for consistency in those inferences, across subgroups would require an understanding of the effect of the active control relative to placebo in the investigated subgroups along with the estimated difference in effects between the active control and experimental therapies from the non-inferiority trial(s). There are various scenarios in which the effects (relative to placebo) of the active control therapy and/or the experimental therapy, as well as the differences in their effects, may vary across subgroups. It may also be the case that different subgroups may have different “non-inferiority margins” due to varying effects of the active control.
