Evaluating two strategies for the design of pediatric pharmacokinetic studies
Yuanxi Zou (1), 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
Objectives: Designing pediatric clinical trials are complicated by the large range of body sizes and developmental changes in the population and the need to limit sampling to a minimum. To justify a design of pediatric pharmacokinetic (PK) studies, a common criterion (standard) is that the power to achieve 95% confidence intervals (CI) within 60% to 140% of the geometric mean estimates of main PK parameters of interest in each sub-group should be at least 80%1. This approach focuses on parameter precisions. An alternative method based on the accuracy of dose selection (ADS) was previously proposed2. This work aimed to compare ADS with the standard approach including estimated power and sensitivity to different variables, using model-based simulation and re-estimation.
Methods: The comparison was based on a scenario with the anti-tuberculosis drug pretomanid and the design of a single-dose study in children with the objective to select doses for a subsequent long-term study.
A population PK model of pretomanid based on adult data was used to perform simulations (regarded as hypothetically “true”)3. The structure was one disposition compartment with transit-delayed absorption. Bioavailability increased with dose. To adjust for children, allometric scaling with weight was included on clearance and volume; metabolic maturation function with age was included on clearance.
In the base scenario of a simulated trial of the study, 36 patients above 4 kg and under 18 years were enrolled. Different doses were given by weight according to six dosing groups (4-6, 6-8, 8-12, 12-20, 20-40, >40 kg). The PK samples were taken for 2 days after the administration of the single dose. Limited by the tablet formulation, pretomanid was dosed only in multiples of 5 mg, up to 200 mg (approved dose for adults).
The ADS approach was conducted as previously described 2. In brief, an optimal group dose (GD*) that provided model-predicted exposures closest to the target exposure as in adults, was selected for all simulated individuals in each dosing group. After the model was re-fitted (including allometries and maturation parameters) to the PK data from a simulated trial, an estimated set of PK parameters was obtained, based on which the estimated group doses (GD̂) were selected in the same way as GD*. The ratio of GD̂ to GD* was calculated for each dosing group. This process was repeated with 500 simulated trials.
The standard approach was conducted based on the same simulated datasets as in the ADS approach. When re-estimating the PK parameters, parameter uncertainties were also computed using the covariance matrix. The 95% CI of clearance and volume from each simulated trial was obtained by bootstrapping from the covariance matrix.
The study power for each dosing group was summarized as the percentage of the 500 ratios in the ADS approach or 95% CIs in the standard approach within the 60-140% range.
To study a scenario with higher variance on both approaches, the interindividual variability of clearance (CV%) was doubled from 33 to 66%. For the ADS approach, the sensitivity to minimal available tablet sizes (from 0.1 to 25 mg) was also studied.
Results: Under the same base scenario, the ADS approach suggested that with the planned design, all the dosing groups could reach the power of 80%, except the lightest weight band having the least power of around 75%. This implies that the study was deemed sufficient for selecting the doses accurately at later steps. However, the standard approach showed that only dosing groups from 8 to 40 kg reached around 80% power for clearances, while the lightest group was as low as below 30% to provide the required precision in clearance, suggesting a much larger sample size was needed for better parameter precision. With higher variability of clearance, the standard approach suggested no-go for all the groups, while the ADS approach suggested that the study was still sufficiently powered for selecting accurate doses.
For the ADS approach, by varying the tablet sizes, there was no clear monotonous pattern in the change of the power. When the minimal tablet size was large enough, the lightest group reached almost 100% power. Decreasing the minimal tablet size to fine steps did not significantly worsen the power.
Conclusions: The ADS approach may be a good alternative for study power evaluation that allows lower sample size when the study is focused on determining doses using discrete tablet sizes.
References:
[1] Wang Y, Jadhav PR, Lala M, Gobburu J V. Clarification on Precision Criteria to Derive Sample Size When Designing Pediatric Pharmacokinetic Studies. The Journal of Clinical Pharmacology. 2012;52(10):1601-1606. doi:10.1177/0091270011422812
[2] Zou Y, Nedelman J, Karlsson MO, Svensson EM. A novel approach to evaluate the design of pediatric PK studies focused on accurate dose selection. In: Population Approach Group Europe (PAGE). ; 2021:I-72. Accessed March 5, 2023. https://www.page-meeting.org/default.asp?abstract=9800
[3] Salinger DH, Subramoney V, Everitt D, Nedelman JR. Population pharmacokinetics of the antituberculosis agent pretomanid. Antimicrob Agents Chemother. 2019;63(10). doi:10.1128/AAC.00907-19
