Abstract:
|
Minimization has gained popularity as the preferred treatment allocation method in oncology where potent prognostic factors usually need to be accounted for, and where trials typically include a large number of centers within which some degree of treatment balance may also be desired. Minimization has raised three types of concerns. First, as for all dynamic randomization methods, the treatment allocations cannot be pre-determined as they depend on the characteristics of the patients who are sequentially entered in the trial. Second, balance is a characteristic of design optimality for the linear model, but more efficient designs have been developed for non-linear models. Third, the analysis must use a randomization test. We will discuss these concerns. We will show, in particular, that substantial gains in power can be achieved by both balancing prognostic factors that are highly correlated with the outcome of interest, and using randomization tests as the method of analysis. The appeal of minimization, besides simplicity, is that it can be used for all trial sizes and numbers of prognostic factors.
|
ASA Meetings Department
732 North Washington Street, Alexandria, VA 22314
(703) 684-1221 • meetings@amstat.org
Copyright © American Statistical Association.