|
Abstract:
|
We propose assessing an appropriate non-inferiority margin based on the stage one findings of a seamless trial designed to assess superiority of an experimental treatment vs. placebo and non-inferiority of the chosen experimental treatment vs. active control. We define a superiority/non-inferiority seamless trial as selecting the best experimental dose vs. placebo in Stage I; and assessing non-inferiority of that dose to active control, after adding subjects to yield adequate power, in Stage II. We simulate 10,000 trials consisting of three experimental doses, placebo, and an active control. We apply various adjustment methodologies and calculate the NI margin in Stage I, then test non-inferiority of the selected dose to the active control in Stage II. We assess type 1 error if inferiority exists and power if non-inferiority exists for various dose-response tends. We assess the need to adjust alpha in either stage, and the optimal approach for doing so, as well as optimal choice of NI margin. Seamless superiority/non-inferiority trials are possible and could result in less development time and fewer subjects needed to assess efficacy than separate trials.
|
