Byron Jones
Department of Medical Statistics, Faculty of Computing and Engineering, De Montfort University, The Gateway, Leicester LE1 9BH, UK
A brief review will be given of the available approaches to constructing optimal designs and an assessment of their usefulness in the context of pharmacokinetic (PK) trials will be attempted. The designs we consider are either made up of parallel groups of subjects or of cross-over sequences assigned to groups of subjects. Within each of these design types the aim is to find optimal sets of times at which to take blood samples. An additional problem with cross-over trials is that an optimal set of sequences must be found as well as an optimal set of sampling times.
The search for an optimal PK design is difficult because:
(i) the repeated samples taken from each subject are assumed to follow a nonlinear mixed model, (ii) the design criteria depend on the model fitting procedure, (iii) there are a number of alternative fitting procedures, (iv) the design must permit model checking, (v) practical constraints must be taken into account, (vi) the design criteria are computationally expensive to evaluate and often numerical integration is needed and finally (vii) local optimisation procedures may fail to converge or get trapped at local optima.
A brief review of current optimal design algorithms will be given and the possibility of using global optimisation procedures will be considered.
Reference: PAGE 8 (1999) Abstr 134 [www.page-meeting.org/?abstract=134]
Poster: oral presentation