C.Pobel* - E.Chatelut+ - B.Tranchand x - J.N. Parola* - S.Demare*
Service de pharmacie clinique SAINTES+ Centre C.REGAUD TOULOUSE + Centre Leon-BERARD LYON
An approach using neural networks (N.N.) was used to predict PK and PD observations in several population studies. Neural networks use an approach for prediction based on observation of the system to discover relationships from the systemÃs recorded behavior (including extended record of covariates). Neural computing is an attempt to build mathematical models that mimic the computing power of the human brain. In this way, this tool can be explored in the field of population PK and PD approach. In the examples presented here, this approach is compared with traditional deterministic approach:
– 1st: On CarboPt PK population which was used by CHATELUT and al. to built an apriori dose estimation formula using NONMEM.
– 2nd: On CarboPt aposteriori observed population of C.H. SAINTES and Croix Rousse Hospital (LYON) on which were applied a priori dose estimation formula of CALVERT, and CHATELUT.
– 3rd: On CarboPt population of CHATELUT to compare N.N. estimation and bayesian estimation of PK parameters and plasma concentrations.
– 4th: On a Melphalan PK and PD population (B.TRANCHAND, LYON) to compare N.N. estimation and bayesian estimation.
Bias and precisions were tested between each approach.
Key words: neural networks, NONMEM, pharmacokinetics, prediction.
Reference: PAGE 5 (1996) Abstr 582 [www.page-meeting.org/?abstract=582]
Poster: oral presentation