Determining optimal sequence allocations for crossover designs in pharmacokinetic studies
by
Tim Waterhouse
University of Queensland
Coauthors: John Eccleston, Stephen Duffull

A commonly used design in drug trial experiments is a balanced crossover design where all subjects receive all treatments, and the sequences in which treatments are received are designed so that all treatments follow all others the same number of times ( > 1). While balanced crossover trials have their advantages, they are often not the most efficient for the nonlinear models which occur in pharmacokinetic studies. The problem of allocating different dose levels of a fictive drug to different subjects is considered. The response in this case is considered to be a Bernoulli random variable, e.g. either the subject still has a headache, or the headache is gone. D-optimal crossover designs for this model are found which are more efficient than balanced designs. We hope to extend this study to more complex nonlinear pharmacokinetic models in the future.

Date received: April 3, 2002