Pharmaceutical Statistics Using SAS

14.6. Sequential Designs in Clinical Trials

Sequential design and analysis plays a natural role in clinical trials, primarily because of the ability it gives of stopping early either if the treatment under test does not deliver the anticipated effect, or because it is more efficacious than was thought initially. Most sequential approaches are based on frequentist statistics (see, for example, Armitage (1975) and Whitehead 1997), and normally carried out group sequentially, as data are reviewed and monitored on a periodic basis. There is a long tradition of criticizing such frequentist approaches, firstly because by focussing on the consequences of the stopping rules on the Type I error rate they contradict the likelihood principle (Anscombe, ...