R. Bartolucci, P. Magni
Department of Electrical, Computer and Biomedical Engineering, University of Pavia, via Ferrata 5, Pavia, I-27100, Italy
Introduction: This work is focused on the use of Modeling and Simulation (M&S) techniques to optimize experimental protocols of paediatric clinical trials, using Macitentan as a case study. Macitentan has been approved for the treatment of pulmonary arterial hypertension (PAH) in adults, although PAH is a rare condition that can occur also in children [1]. A further investigation in the paediatric population is therefore required. However, paediatric clinical trials can have ethical and practical restrictions that need to be considered when the experimental protocol is defined and, for this reason, its optimization is required in order to maximize the information that can be exploited from collected data, minimizing the impact on patients [2]. M&S can help in this, for example, by looking for the optimal sampling schedule that can guarantee a certain level of precision for the parameter estimation.
Methods: A population PK model, describing the steady state kinetics of Macitentan and its metabolite in adults, was found in literature [3]. The model was translated to a paediatric population by a suitable scaling of the model parameters. To improve the model identifiability and then the optimization process, a constraint on the metabolism of Macitentan was applied, based on in vitro findings. Real data of age and weight of a paediatric population were collected from the public NHANES databases [4] and they were sampled to obtain a dataset of 40 subjects, from 2 to 6 years old, on which the protocol was optimized. The R package of PopED [5] was used to optimize the experimental protocol in terms of number of samples and sampling times. Different numbers of samples per patient were evaluated in order to select the minimum number that assures a certain level of precision in parameter estimation. The value of the Relative Standard Error (RSE%) calculated on the most important population parameters, such as volumes of distributions and elimination rates, was used for schedule comparison. The sampling schedule was optimized with a Ds-optimality, that maximize the determinant of the Fisher Information Matrix (FIM), obtained only with a subset of the model parameters. Finally, each scenario was evaluated twice: i) without additional constraints; ii) with a more realistic design, forcing the algorithm to use the same sampling schedule for the two observations, i.e., Macitentan and its metabolite.
Results: In both scenarios (with and without the constraint on the sampling of Macitentan and its metabolite), it appears that for a small number of samples per patient the algorithm is unable to invert the covariance matrix and cannot perform the optimization with reliable results. As the number of samples increased, the estimated precision of the parameters increased accordingly; however, some of the selected time points resulted very close to each other or even coincident. Although sample replicates can increase the model estimation performance, especially when dealing with inter-individual variability, these schemes are not easily implementable in clinical practice. A second observation is that the RSE% is slightly higher when the constraint is applied. However, this last case was considered more clinically relevant and it was used for further analysis. To test the results obtained with PopED, a third analysis was performed using MonolixSuite2018R2. Four scenarios, with 2, 3, 4 and 9 samples and same schedule, were compared estimating the parameters on 100 simulated datasets. An empiric RSE% and the ratio between estimates and the true value of each parameter were calculated. The estimates considered acceptable had a small RSE% and a ratio in the range [0.6, 1.4].
Conclusion: The optimization of the experimental protocol is an important phase of the clinical studies, especially for paediatric populations. PopED applied on a scaled PK model can be used to compare different designs and find the sampling schedule that maximize data exploitation under different practical constraints, however the user intervention remains essential to compare the results. The results obtained with Monolix in the estimation step show that the RSE% is small in all the considered scenarios, especially for CL1 and CL2, whereas for others, e.g. Ka and V2, is evident how the precision increases according to the number of samples. Therefore, the choice of the best number of samples per patient depends on which parameter is considered important for the trial.
References:
[1] M.G. George et Al. Pulmonary Hypertension Surveillance: United States, 2001 to 2010. Chest. 2014; 146(2): 476-495.
[2] F. Bellanti, O. Della Pasqua. Modelling and simulation as research tools in paediatric drug development. European Journal of Clinical Pharmacology. 2011;67(1):75-86.
[3] A. Krause, J. Zisowsky, J. Dingemanse. Modeling of pharmacokinetics, efficacy, and hemodynamic effects of macitentan in patients with pulmonary arterial hypertension. Pulmonary Pharmacology and Therapeutics. 2018; 49:140-146.
[4] Centers for Disease Control and Prevention. NHANESIII (1988-1994). (https://wwwn.cdc.gov/nchs/nhanes/nhanes3/default.aspx)
[5] M. Foracchia et Al. POPED, a software for optimal experiment design in population kinetics. Computer Methods and Programs in Biomedicine. 2004; 74:29-46.
Reference: PAGE 28 (2019) Abstr 9065 [www.page-meeting.org/?abstract=9065]
Poster: Study Design