L. Claret, A. Iliadis.
Laboratoire de Pharmacocinétique, Faculté de Pharmacie. 27 Bd Jean Moulin 13385 Marseille cedex 05.
The aim of a population Study [1] is to describe the variability and to detect particular subsets by establishing the relationships between PK parameters and easily measurable subject characteristics, the covariates (age, body weight, …). The usually proposed methods are based on linear regression equations relating PK parameters to the covariates [2].
We propose a new approach : to measure these dependencies through the non-parametric joint density between PK parameters and covariates, as estimated by a kernel method [3]. An Index of the quantity of information measuring uncertainty was used to determine :• the necessary number of individuals to describe the variability [4] ;• the measure of bivariate non-parametric dependence [5].
This approach was used to obtain the non-parametric conditional density function of the PK parameters given the covariates. This function supplied a prior information in a bayeslan estimator. The global approach was validated by simulation study in which non-linear relations link covariates and PK parameters. The performance of this new estimator using covariates was compared with usual bayesian estimation.
References
[1] M. Rowland. L. Aarons, New Strategies In drug development and clinical evaluation : The population approach, Commission of European Communities, Luxembourg (1992)
[2] S. L. Beal, LB. Sheiner. Estimating Population Kinetics, CRC Biomed. Eng., 8:195-222 (1983)
[3] B. W. Silverman, Density Estimation for Statistics and Data Analysis. Chapman & Hall, (1956)
[4] A. Iliadis, Population Pharmacokinetics : Methodological concerns in applications, in : Topics in pharmaceutical sciences, D.J.A. Cronmelin K.K. Midha (Eds), Medpharm Stuttgart, pp 441-460 (1991)
[5] Joe, Relative Entropy of Multivariate Dependence, J. Am. Stat. Ass. 84, 405:157-164 (1989)
Reference: PAGE 3 (1994) Abstr 847 [www.page-meeting.org/?abstract=847]
Poster: poster