A novel approach to evaluate the design of pediatric PK studies focused on accurate dose selection
Yuanxi Zou (1), Jerry Nedelman (3), Mats O Karlsson (1) and Elin M Svensson (1, 2)
(1) Department of Pharmacy, Uppsala University, Sweden; (2) Department of Pharmacy, Radboud Institute of Health Sciences, Radboud University Medical Center, The Netherlands; (3) TB Alliance, New York, USA
Objectives: A commonly used criterion to justify the design of a pediatric PK study is that the study must be powered to target a 95% confidence interval within 60% to 140% of the geometric mean estimates of the important PK parameters with at least 80% power [1]. However, that criterion does not directly investigate the clinically relevant target, namely, dose selection. This work proposes an alternative evaluation method based on the accuracy of dose selection.
Pretomanid, a newly developed nitroimidazole for treatment of tuberculosis (TB), is used to demonstrate this approach. Currently, pretomanid is being investigated for use in children. A population pharmacokinetic (PK) model was previously built based on adult data, suggesting that maximum concentrations are reached around 6 h after dose and that the half-life is about 15 h. This model was used to help design the first-in-pediatric single-dose clinical trial of pretomanid, an objective of which is to select doses for a subsequent safety and efficacy study.
Methods: A model-based simulation and re-estimation approach was outlined and tested using the pretomanid case.
Step 1: Population simulation. A pediatric population with TB with age uniformly distributed between 0 and 18 years old was simulated [2]. From this population, 30 000 children with weight above 4 kg were selected. Patients were assigned to six dosing groups by weight to enable different doses for different weight bands (4-6, 6-8, 8-12, 12-20, 20-40, >40 kg).
Step 2: Optimal dose calculation. The adult population PK model for pretomanid was scaled to children by incorporating a maturation function with age and allometric scaling with weight. This model with true PK parameters (θ*) was used to generate the true individual PK parameters (θ*i ) for each virtual patient given weight and age. Based on all individuals´ θ*i within a weight band, an optimal group dose (GD*) was selected (six GD* in total) to provide AUCinf close to that in adults at the clinical dose, AUCtarget, by minimizing the root mean squared error (RMSE) defined below. Doses were selected only among multiples of 5 mg from 5 mg to 200 mg because 5 mg is the technically supported minimum while 200 mg is the approved dose for adults.
RMSE = √((∑in=Ngrp(log(AUCdose,i)-log(AUCtarget))2)/n)
Ngrp: number of subjects in the group; i: the ith subject
Step 3: Data simulation. A proposed PK sampling schedule (1, 3, 6, 9, 24, 48 h post-dose) was tested with a sample size of 36. Patients were randomly sampled from the simulated population; patients received their GD* according to the RMSE; observations were simulated according to the sampling schedule using the model with θ*.
Step 4: Re-estimation. The model was re-fitted to the simulated data and the θ was re-estimated including allometric scaling factors, but maturation-related parameters were fixed (assuming that the maturity was well defined based on knowledge of the metabolic pathway). Based on the re-estimated θ ( θ^ ), six estimated GD (GD^) were selected. The ratio of the GD^ to the GD* was calculated for each group.
Step 5: Steps 3 and 4 were repeated 500 times.
Step 6: Power calculation. The power of the study design for each dosing group was summarized as the percentage of the 500 ratios within a specified limit of 80%-125%.
Results: The precision of the dose selection was estimated with above 80% power for all dosing groups except the group <6 kg. The groups >40 and <6 kg gave lower power than the groups in the middle. The power for the group <6 kg was reduced to less than 60%, due to the greater impact of a 5 mg difference at lower doses. For example, the ratio of a GD^ =150 mg to a GD*=155 mg would be 0.97, while the ratio of a GD^ =10 mg to a GD*=15 mg would be 0.67 even though the difference was only 5 mg. The relatively lower power for the group >40 kg was because much higher weights could be available in this group while only a maximum of 200 mg could be given, causing a limited choice of dose for more variable weights.
Conclusions: The proposed approach evaluates the precision in dose selection given the study design and enables interesting comparisons of power between different dosing groups, especially when those groups are limited by available dosage strengths. It provides a more relevant decision criterion for designing pediatric trials when the objective is dose selection than purely evaluating power based on parameter precision.
References:
[1] Wang Y, Jadhav PR, Lala M, Gobburu J V. Clarification on Precision Criteria to Derive Sample Size When Designing Pediatric Pharmacokinetic Studies. J Clin Pharmacol. 2012;52(10):1601-1606. doi:10.1177/0091270011422812
[2] Svensson EM, Yngman G, Denti P, McIlleron H, Kjellsson MC, Karlsson MO. Evidence-Based Design of Fixed-Dose Combinations: Principles and Application to Pediatric Anti-Tuberculosis Therapy. Clin Pharmacokinet. 2018;57(5):591-599. doi:10.1007/s40262-017-0577-6
