Jeudi 26 novembre 1998 de 14 à 15 heures, .
Population pharmacokinetic/ pharmacodynamic approaches are increasingly
used during drug development (1). These analyses rely upon nonlinear regression mixed
effects models. In pharmacokinetics, the population design is defined by the total number
of individuals and the several elementary designs to be performed within groups of
individuals. Our objective is to find the optimal population design for a given maximal
cost of the experiment and practical constraints.
We proposed an approach based on the D-optimality criterion: the maximization of the
determinant of the Fisher information matrix of the population parameters (2). We derived
this matrix for a normal distribution of the random effects and using a first-order
linearization of the regression model about the mean. An optimization algorithmis also
proposed to find the design maximizing the determinant, given a set of sampling times and
a maximal cost for the experiment.
This approach was applied to two examples from real pharmacokinetic data of drugs in
development: one example in rodents, one example which is the prospective design of a
population pharmacokinetic study in children . The results on these examples illustrated
that optimal population designs may differ from standard intuitive ones with an important
increase in efficiency especially in case of sparse elementary designs.
(1) Vozeh S et al. The use of population pharmacokinetics in drug development. Clin Pharmacokin, 1996, 30:81-93.
(2) Mentre F, Mallet A, Baccar D. Optimal design in random-effects regression models. Biometrika, 1997, 84:429-42.