Delphine Martin, Roeline Jochemsen , Christian Laveille and Nicolas Frey
Institut de Recherches Internationales Servier, 6 place des Pleïades, 92415 Courbevoie cedex
INTRODUCTION AND OBJECTIVES
The compound (S-X) is characterised by good absorption but a low bioavailability due to a saturable first-pass hepatic effect, which is responsible for the non-linear pharmacokinetics. S-X is rapidly metabolised by cytochromes P450, 1A2 and 2C9 and only 0.010% of the unchanged compound is eliminated in the urine.
The objectives of this combined pharmacokinetic analysis are to estimate individual plasma pharmacokinetic parameters for S-X from all the plasma or saliva samples collected during the clinical development after repeated oral administration of SX in patients and to assess the influence of covariates (gender, dose, smoker, time of administration, ….) on these parameters.
METHODS
The first step: Development of a POP-PK model for phase I data
Plasma samples from six phase I studies with doses of 25 or 50 mg of compound S-X were pooled and analysed using NONMEM.
The second step: External qualification of the POP-PK model using phase II data
The plasma concentrations from a Phase II study were used for an external qualification test of the POP-PK model. The evaluation was assessed by using a maximum a posteriori Bayesian analysis (MAP Bayesian).
The third step: Use of the POP-PK model to analyse phase III data and to identify factors predicting between subject variability.
In phase III, only saliva samples were collected. Based on a linear relationship between plasma and saliva concentrations established previously from phase I data, plasma concentrations will be predicted and analysed using NONMEM.
RESULTS
The first step:
A one-compartment model was found to adequately describe the data. The analysis was performed on log-transformed data in order to reduce the residual variability estimate, and also to avoid biased PRED vs. DV plots. A mixture model was used for the input phase in order to describe the complex input profiles observed in a subset of patients. The input phase was hypothesised to be composed of two types of processes: a first-order input, or a zero order input followed by a first order-input. A specific variability on the disposition parameters was defined in order to estimate the interindividual-variability of the first-pass effect.
The second step:
The agreement between the observed and predicted concentrations issued from the MAP Bayesian analysis was good, suggesting that the model may be considered preliminarily qualified.
The third step:
The third step is in progress and the results will also be presented at the PAGE meeting.
Reference: PAGE 12 (2003) Abstr 451 [www.page-meeting.org/?abstract=451]
Poster: poster