Deferiprone sampling optimisation in a pharmacokinetic bridging study including children with β-thalassaemia
Francesco Bellanti (1), Meindert Danhof (1), Oscar Della Pasqua (1, 2)
(1) Division of Pharmacology, Leiden University, Leiden, The Netherlands; (2) GlaxoSmithKline, London, United Kingdom
Objectives: Practical and ethical constraints impose special requirements for clinical trials in children. The application of population PK analysis to sparse data allows reducing the burden in such a vulnerable population [1]; yet it is important to optimise the quality of the information gathered. The aim of this analysis is to optimise sampling times for the evaluation of deferiprone PK in children in a prospective clinical investigation in order to subsequently optimise the dosing regimen in the same population.
Methods: A one-compartment PK model with first order oral absorption has been developed on adults’ data using a non-linear mixed effects approach, as implemented in NONMEM VII. Two covariate models have been used to optimise sampling times in children, namely M1 (body weight with a linear correlation on CL/F and Vd/F), and M2 (fixed allometric scaling on the same parameters). Uncertainty (20%) in CL/F and Vd/F estimates has been accounted for in the optimisation procedures. The study consisted of a parallel design with three dose levels randomised across 18 children (aged between 2 and 10 years). The final sampling scheme (maximum of 5 samples per subject) has been selected based on the outcome of four scenarios in PopED 2.12. The accuracy and precision of parameters estimates were estimated for primary and secondary (i.e., AUC and Cmax) PK parameters. Predicted AUC and Cmax estimates were compared with simulated data using frequent sampling (n=12) according to the trapezoidal rule.
Results: Given practical constraints, the selected sampling scheme was the result of a compromise between full optimisation and feasibility in a real clinical trial. The accuracy of primary PK parameters estimates was below 10% except for KA (-11%); whereas precision, as expected, was slightly lower given the small sample size (> 30% for Vd/F and KA). AUC values (mean and standard deviation) were found to be 33.37 (19.24) and 35.61 (20.22) mcg/ml.h and Cmax values 10.17 (6.05) and 10.94 (6.68) μg/ml in sparse and frequent sampling respectively.
Conclusions: The results of our analysis illustrate that despite feasibility issues, study characteristics can be optimised using ED-optimality concepts. Predefined sampling schemes and sample sizes do not warrant accurate model structure and parameter identifiability. Of particular importance is the accurate estimation of the magnitude of the covariate effects, as they may determine the final dose recommendation for the population of interest.
References:
[1] Anderson BJ, Allegaert K, Holford NH, (2006) Population clinical pharmacology of children: general principles. Eur J Pediatr, 2006 165:741–746.
