Loes Maton 1,2, Pieter Colin 1,3, Jeroen Koomen 1,2, Douglas Eleveld 1, Efthymios Manolis 3, Roberto de Lisa 3
1 University Medical Center Groningen (Groningen, The Netherlands), 2 Medicines Evaluation Board (Utrecht, The Netherlands), 3 European Medicines Agency (Amsterdam, The Netherlands)
Introduction: Generation of evidence in the paediatric population is often hampered by difficulties in conducting clinical trials [1]. Extrapolation of efficacy based on demonstrating similarity in exposure between the target and reference population could provide an alternative to a confirmatory efficacy trial, when there is sufficient confidence that disease progression, exposure-response and treatment response are similar [2]. However, established methodological guidance for justifying sample sizes and assessing exposure similarity is currently lacking. One common approach to justify sample sizes is proposed by Wang et al. [3], who justifies sample sizes by targeting pre-specified parameter precision in paediatrics, but does not account for potential bias in these parameters. This project aims to benchmark current methodologies employed to justify sample sizes for paediatric pharmacokinetic trials submitted to regulatory authorities, and to propose a novel methodological framework balancing precision and bias in paediatric parameter estimations.
Methods: Sample size justifications and associated methodologies for exposure-matching trials were extracted from internal regulatory documentation. The precision and bias of these methodologies were evaluated by performing a simulation and re-estimation study using a theoretical population PK model. Paediatric trials were simulated with sample sizes derived using the method as proposed by Wang et al. [3], both with similar and 2-fold higher paediatric clearance compared to the adult clearance. Hereafter, the precision and bias were analysed using a stand-alone approach in which paediatric data is analysed separately, a pooled analysis approach combining paediatric and adult data, and our newly proposed difference parameter approach which is a pooled analysis with an additional subgroup factor. Additionally, simulations were performed using the proposed framework to determine the sample size required to detect prespecified differences between the paediatric and adult population and to evaluate the framework’s ability to quantify detectable effect sizes under feasibility limitations.
Results: Sample sizes were predominantly based on the acceptance criteria described by Wang et al. (n = 13, 44.8%) using either a pooled approach (n = 9, 69.2%) or a stand-alone approach (n = 4, 30.8%), or no method was described (n = 10, 34.5%). In paediatric trial simulations introducing a 2-fold higher clearance in paediatrics relative to the adult population, performance varied across methods and adult clearance uncertainty levels (<1%, 5%, 10%, and 15% RSE): the pooled approach yielded paediatric RSEs of 3.4, 5.7, 8.9, 15.5% with associated biases of 0.70, 0.72, 0.75, and 0.83; the stand-alone approach yielded paediatric RSEs of 9.1%, 10.0%, 10.9%, and 10.8% with biases of 0.95, 1.01, 1.00, 1.07; and the difference parameter approach yielded paediatric SEs of 7.1%, 6.5%, 9.6%, and 10.1% with biases of 1.05, 1.06, 1.04, and 1.05, respectively. Using the difference parameter approach, detecting the 2-fold higher paediatric clearance with 80% power required sample sizes of 2, 10, 20 and 44 paediatric subjects for <1%, 10%, 20%, and 30% adult RSE, respectively. Additionally, the minimal differences to be detected with 80% given a sample size of 6 were 22.3%, 60.6%, 72.7%, and 97.6% for the same adult %RSEs. Conclusions: Although the use of a pooled analysis approach to justify sample sizes achieved acceptable precision, the substantial bias, driven by dominance of adult data in the dataset, highlights its insensitivity to detect differences between the paediatric and adult populations. The stand-alone approach achieved both acceptable precision and bias, however, required a higher sample size to meet the criteria, reducing the efficiency of this design. In contrast, the difference parameter approach achieved both acceptable precision and bias in paediatric parameter estimations while needing less paediatric subjects. Furthermore, this analysis approach remained applicable in case of feasibility issues as it allows for determination of the minimal difference that can be detected given a predefined sample size. References: References: [1] CHMP. Guideline on the role of pharmacokinetics in the development of medicinal products in the paediatric population (EMEA/CHMP/EWP/147013/2004). [2] CHMP. ICHE11A Guideline on paediatric extrapolation (EMA/CHMP/ICH/20518/2022). [3] Wang Y, et al. Clarification on precision criteria to derive sample size when designing pediatric pharmacokinetic studies. J Clin Pharmacol. 2012 Oct;51(10):1601-6. Doi: 10.1177/0091270011422812.
Reference: PAGE 34 (2026) Abstr 11928 [www.page-meeting.org/?abstract=11928]
Poster: Methodology - Study Design