Anne Dubois and Marylore Chenel
Department of Clinical Pharmacokinetics, Institut de Recherches Internationales Servier, Suresnes, France
Objectives: When performing a pharmacokinetic (PK) study, it is important to define an appropriate design, which has an important impact on the precision of parameter estimates and the power of tests. This is of most interest, especially when the number of PK samples is limited, as in paediatric studies. To optimise such designs, methods based on the Fisher information matrix in nonlinear mixed effects modelling (NLMEM) can be used [1]. A development plan for the use of a Servier drug S in the paediatric population is underway. Our objective was to propose a design for the future paediatric PK study and to evaluate the influence of the number of children and of the bodyweight (BW) distribution on the precision of the parameter estimates.
Methods: Adult PK data were modelled using NLMEM by a two-compartment model with first-order absorption and elimination. We assumed a BW effect on all disposition parameters. This model was implemented in PopDes software [2]. We assumed there were 2 age groups ([2; 11[ and [11; 18[ years), as required by the regulatory agency, with the same number of children per group. We evaluated a rich (38 samples per subject) and a sparse (6 samples per subject) designs assuming five different BW distributions. For each case, we computed NSNRSE, the total number of subjects needed (NSN) to obtain suitable relative standard errors (RSE<30% and 50% for the fixed effects and variance parameters, respectively), and NSNpower, the NSN to demonstrate a significant BW effect on all disposition parameters assuming a 5% type I error and a 80% power.
Results: The predicted RSE of all variance parameters were similar for all tested BW distributions whereas the RSE of the fixed effects parameters (typical values and BW effects) depended on the BW distribution assumption. For the sparse design, NSNRSE was above 40 children for all tested BW distributions. For both designs and all tested BW distributions, NSNpower was lower than NSNRSE and lower than 40 children.
Conclusions: Optimal design trough Fisher information matrix approach is a powerful tool to help planning paediatric PK studies. The assumption on the expected BW distribution is important to choose a number of children sufficient to demonstrate a significant BW effect on the PK parameters but the precision of the parameter estimates should not be neglected.
References:
[1] Mentré F, Mallet A and Baccar D. Optimal design in random-effects regression models. Biometrika. 1997, 84: 429-442.
[2] Ogungbenro K, Gueorgueiva I and Aarons L. PopDes – A Program for Optimal Design of Uniresponse and Multiresponse, Individual and Population Pharmacokinetic and Pharmacodynamic Experiments
Reference: PAGE 22 () Abstr 2885 [www.page-meeting.org/?abstract=2885]
Poster: Paediatrics