Caroline Petit (1), Adeline Samson (2), Satoshi Morita (3), Jérémie Guedj (4), Vincent Julien (5), Emmanuelle Comets (4, 6), Sarah Zohar (1)?
(1) INSERM, UMRS 1138, CRC, Team 22, Univ. Paris 5, Univ. Paris 6, Paris, France ; (2) LJK, UMR CNRS 5224, Univ. J. Fourier, Grenoble, France ; (3) Department of Biomedical Statistics and Bioinformatics, Kyoto University Graduate School of Medicine, 54 Kawahara-cho, Shogoin, Sakyo-ku, Kyoto 606-8507, Japan ; (4) INSERM, IAME, UMR 1137, F-75018 Paris, France; Univ Paris Diderot, Sorbonne Paris Cité, F-75018 Paris, France ; (5) Pharmacology Department, Univ. Paris 5, Sorbonne Paris Cité, Inserm U1129, HEGP, Paris, France ; (6) INSERM, CIC 1414, Univ. Rennes 1, Rennes, France
Objectives: Dose-finding (DF) studies are challenging to calibrate, particularly in a paediatric population. Settings are usually chosen by usage or expert advice, without clear underlying methodology. We investigated calibrating an existing DF model for a paediatric population using adult observations, such as pharmacokinetics (PK), toxicity and efficacy, in order to choose (i) the parameters of the DF model, i.e. the initial guess of toxicity for each dose (working model WM) and the prior distribution of parameters, and (ii) the dose-range given to the paediatric population in the study. We evaluated the resulting method on a drug used against cancer, erlotinib.
Methods: Our approach relies on the bivariate continual reassessment method (bCRM) [1], which evaluates both efficacy and toxicity with a power model and a Bayesian estimation. For calibration, we assumed that target exposures, measured by AUC, were the same in both populations. We extrapolated paediatric clearance from the adult value by allometry and maturation [2, 3]. We obtained a paediatric model (bCRMpaed) by calibrating the WM assuming similar toxicities as those observed in adults. The variances of prior distributions were set using the ESS criterion [4]. We built a dose range based on extrapolation processes from adult doses with maturation and allometry. For comparison purpose, we also used a dose range with the current practice, that is a dose range linearly adjusted from adult dose on weight. The different approaches were compared in a simulation study aiming to design a DF study for the anticancer drug erlotinib in a population of children aged 2 to 5. Performances for each scenario were assessed by compting the percentage of simulated trials that selected the right dose.
Results: For erlotinib, the maturation and allometry dose ranges yielded doses different from the current practices. For every dose range, the bCRMpaed method selected the appropriate dose in a higher percentage of simulated datasets (10%-15%) than the bCRM. In scenarios where paediatric doses were different from adults (misspecification of bCRMpaed), results were similar to bCRM.
Conclusion: Designing a paediatric DF study based on former adult information can significantly improve the chance of selecting the appropriate dose for a drug in a paediatric population. Moreover, the use of extrapolation may improve the choice of evaluated doses in paediatrics.
References :
[1] Zohar S. and O’Quigley J. Identifying the most successful dose (msd) in dose-finding studies in cancer. Pharmaceut. Statist., 5:187–199, 2006.
?[2] Anderson B.J. and Holford N.H.G. Mechanistic basis of using body size and maturation to predict clearance in humans. Drug. Metab. Pharmacokinet., 24:25–36, 2009. ?
[3] Petit C., Jullien V., Samson A., Guedj J., Kiechel J.R., Zohar S., and Comets E. Designing a paediatric study for an antimalarial drug including prior information from adults. Antimicrob. Agents Chemother. Accepted manuscript posted online 28 December 2015. ?
[4] Morita S., Thall P.F., Müller P. Determining the effective sample size for a parametric prior. Biometrics, 64 :595-602, June 2008.
Acknowledgment: Caroline Petit was supported during this work by a grant IDEX from the university Sorbonne Paris Cité (2013, project 24).
Reference: PAGE 25 (2016) Abstr 5866 [www.page-meeting.org/?abstract=5866]
Poster: Methodology - Study Design