A. J. Sankoh (Wyeth-Ayerst Research, USA)
Non-superiority Clinical Trials and Multiplicity: An Interpretation Issue
Interpreting the efficacy results from active control clinical trials is not easy. This difficulty is compounded when drug efficacy is demonstrated on the basis of clinical evidence from none traditional superiority clinical trials. This is because such efficacy interpretation depends on the quantification of a clinically and statistically acceptable minimal margin of inferiority d by which the effectiveness of the new drug can be reduced and still be viewed clinically relevant and statistically significant compared to no treatment. The quantification of d requires a clear understanding of the data upon which the approval of the active comparator R was based. The general premise for such quantification is that the clinical trials that formed the basis for the approval of the reference active comparator were placebo controlled randomized clinical trials. In other words, the quantification depends on the validity of the assumption that there was a clinically meaningful or sizable and statistically significant treatment difference DR-P between the reference (R) and placebo (P) treatments in the clinical trials on which the approval of R was based. Efficacy interpretation becomes hopelessly even more complicated when such clinical trials designed to demonstrate non-superiority drug effect have multiplicity components (due to multiple endpoints and/or multiple comparisons) in them. Multiplicity due to multiple comparisons could arise when more than one active experimental dose is included in the design and/or more than one clinical efficacy objective is being investigated. We discuss in this presentation the difficulty in interpreting the efficacy results of non-superiority clinical trials in the presence of multiplicity with a special focus on type I error rate and power of the tests.