Combining different phases in the development of medical treatments withina single trial

In the development of medical treatments, identification of promising thera pies and inference on selected treatments are usually performed in subseque nt separate trials. An adaptive two-stage design is proposed for the situat ion of multiple treatments to be compared with a control, allowing integrat ion of both steps within a single confirmatory trial controlling the multip le level alpha. After the interim analysis, the trial may be terminated ear ly or is continued with a second stage, where the set of treatments may be reduced due to lack of efficacy or to safety problems. The procedure is hig hly flexible with respect to the distributional assumptions, stopping rules and selection criteria and allows a completely free recalculation of the s ample size for the second stage. Simulations show that the method may be su bstantially more powerful than classical one-stage multiple treatment desig ns with the same total sample size. As in conventional strategies with a se ries of separate experiments, a reasonable selection strategy has to be app lied in order to prevent proceeding with non-optimal treatments. Copyright (C) 1999 John Wiley & Sons, Ltd.