Caroline Petit (1), Vincent Jullien (2), Adeline Samson (3), Jérémie Guedj (4), Jean-René Kiechel (5), Sarah Zohar (1), Emmanuelle Comets (4,6)
(1) INSERM, UMRS 1138, Univ. Paris 5, Paris, France ; (2) Pharmacology Department, Inserm U1129, HEGP, Paris, France ; (3) LJK, UMR CNRS 5224, Univ. J. Fourier, Grenoble, France ; (4) INSERM, IAME, UMR 1137, Paris, France ; (5) Drugs for Neglected Diseases initiative, Geneva, Switzerland ; (6) INSERM, CIC 1414, Univ. Rennes 1, Rennes, France
Objectives: To investigate designing a pharmacokinetic (PK) study using adult prior information for a case-study on mefloquine and to evaluate robustness of the optimal design to different model misspecifications, comparing it with the empirical design.
Methods: PK data for adults and children were obtained from two different randomized studies for treatment of malaria with the same artesunate-mefloquine combination regimen, given once daily over 2 or 3 days. A recommended design for paediatric study on mefloquine was obtained by design optimisation on an extrapolated model built from adult data. The adult PK parameters were estimated using SAEM algorithm [1] with the software MONOLIX 4.2.2 [2]. Paediatric PK parameters were then obtained by adding allometry and maturation [3] to the adult model and employed for designing the paediatric population study. Optimisation of the design is based on the Fisher information matrix [4] and was performed with PFIM 3.0. [5]. Robustness for the recommended design was evaluated in terms of the relative bias and relative standard errors (RSE) of the model parameters by simulating the paediatric population, keeping the distribution of doses and covariates from the actual study. Varying the parameters used in the simulation in four scenarios assessed the robustness, and the performance of the optimal design was compared to that of the empirical design.
Results: A two-compartment model with absorption was shown to best describe the adult data. When the children data was used as an external evaluation, differences between the two populations were apparent, especially in the early days after the beginning of treatment. The optimised design for children with 5 sampling times showed that early concentrations were needed to estimate the absorption phase accurately, recommending to collect the first sample 2 hours after the first dose and then during days 1, 5, 14 and 57. It gave good results in terms of bias and RSE and was robust across various model modifications, in stark contrast to the empirical design from the paediatric study.
Conclusion: Using prior information combined with allometry and maturation can help provide robust designs for paediatrics studies.
Acknowledgment: Caroline Petit was supported during this work by a grant IDEX from the university Sorbonne Paris Cité (2013, project 24).
We thank the Drugs for Neglected Diseases initiative for making their datasets available for this project.
References:
[1] E. Kuhn and M. Lavielle. Maximum likelihood estimation in nonlinear lixed effects models. Comput. Statist. Data Anal., 49:1020–1038, 2005.
[2] Lixoft. Monolix methodology. Available at http://www.lixoft.com/wp- content/resources/docs/monolixMethodology.pdf, March 2013. Version 4.2.2.
[3] B. J. Anderson and N. H. G Holford. Mechanism-based concepts of size and maturity in pharmacokinetics. Annual Review of Pharmacology and Toxicology, 48:303–332, 2008.
[4] S. Retout, E. Comets, A. Samson and F. Mentré. Design in nonlinear mixed effects models: Optimization using the Fedorov-Wynn algorithm and power of the Wald test for binary covariates. Statistics in Medicine, 26:5162–79, 2007.
[5] PFIM group IAME UMR1137, INSERM, and Université Paris Diderot in Paris France. Pfim user guide. Documentation available at http://www.pfim.biostat.fr/, August 2014. Version 4.0.
Reference: PAGE 24 (2015) Abstr 3412 [www.page-meeting.org/?abstract=3412]
Poster: Drug/Disease modeling - Paediatrics