Non-parametric methods for continuous and binary variables in population PK/PD.

A. Mallet

Departement de biomathematiques, INSERM U436, CHU Pitie-Salpetriere, 91 Bd de I'Hopital, 75634 PARIS CEDEX 13, France.

Non-parametric methods have been proposed in pharmacokinetic/pharmacodynamic population modelling; these methods aim at estimating population characteristics without imposing unnecessary structure on the distribution of the random effects in random- or mixed- effects models.

The main proposed approaches, either semi-parametric or purely non-parametric will be reviewed; advantages and limitations of these methods will be discussed. It will be shown how these methods allow estimation of the effects of covariates in population models, either based on fixed and random parameters, i.e. using second stage models relating PK/PD parameters to covariates, or based on random parameters alone, i.e. without specifying any second stage model.

These non-parametric approaches have been quite extensively used in pharmacokinetic/pharmacodynamic studies involving quantitative outcomes such as concentration levels or effects measured on continuous scales. It will be shown that these approaches may also help estimating inter individual variability and covariate effects in population models involving binary or time-to-event outcomes.

Reference: PAGE 8 (1999) Abstr 649 [www.page-meeting.org/?abstract=649]

Poster: oral presentation