|
Abstract:
|
When several treatment arms are administered along with a control arm in a trial, dropping the non-promising treatments at an early stage helps to save the resources and expedite the trial. Drop-the-losers design (Sampson and Sill, 2005) and Winner design (Shun, Lan, and Soo, 2008) are such two-stage adaptive designs that select the treatment with the numerically highest mean response in the interim stage. We extend the interim treatment selection rule by introducing a flexible selection margin to allow the selection of more than one treatment arm if the difference in treatment effect falls within the margin. At the end of the study, the data from both stages are used to make inferences for hypothesis testing of the selected treatment arm(s) against control. The new design brings the flexibility to choose the experimental treatment with the best overall merit.
|
