pedPower – An R package to aid pediatric clinical trial design and exposure evaluation
Daniel Lill (1), Anne Kümmel (1), Wangda Zhou (2), Jason Zhang (2), Juan Jose Perez-Ruixo, Oliver Ackaert (2)
(1) IntiQuan, Switzerland, (2) Janssen R&D, Belgium
Introduction
As one step in the pediatric development of drugs that have been evaluated in adults, pediatric trials are conducted to assess if the exposure following the selected doses matches with the adult exposure. Hence, these trials need to provide sufficient information on pharmacokinetic (PK) parameters in pediatric patients, facilitating adequate comparisons of the resulting pediatric exposures to target values. To this end, the adult population PK model is often utilized for pediatric dose selection based on an appropriate extrapolation to the pediatric population (e.g., by allometric scaling). Also, the population PK model can be leveraged to aid study design and to evaluate the dose selection after study conduct.
For the study design, different simulation-estimation-based approaches can be used to evaluate the expected PK information content in the trial: First, studies can be powered to preclude that the actual exposure differs from the target exposure by more than a limited extent (e.g. 20% or 50%). Second, the study can be powered to determine key PK parameters, such as clearance or volume of distribution, with sufficient precision (e.g., 95% confidence interval between 60%-140% of geometric mean estimates [1]).
For evaluating doses after study conduct, the adult model is re-evaluated using the pediatric study data to obtain individual model-derived exposures of the pediatric patients, which can then be compared to adult exposures.
To standardize and enhance the reproducibility of the analyses, an R package was developed to aid trial design and analysis of pediatric PK studies.
Objectives
- Standardize sample size assessment approaches and dose evaluation for pediatric clinical trials by providing re-usable functionality in an R package.
- Implement customizable workflows for the following analyses:
- Power calculations to detect differences in PK parameters for a given clinical trial design
- Power calculations to determine PK parameters in pediatric patients with sufficient precision
- Compare exposure observed in pediatric subjects following investigated doses to the exposure observed in adults
Methods
The presented R package “pedPower” builds the analyses on a previously established reference (adult) popPK model in NONMEM. “pedPower” can programmatically adjust this model to obtain NONMEM models for simulation, parameter estimation and uncertainty quantification as needed in the different analyses. For power calculations, the reference model is reused to simulate and estimate with the same structural model, with the option of using parameter estimates of the reference model as prior information (with original estimation method or method=Bayes). For exposure evaluation, pediatric exposures, using the adult population PK model as prior information, are compared with observed exposures in adults, with the option to stratify pediatric patients by selected covariates.
Results
The presented R package “pedPower” allows to inform sample sizes and their associated power as a function of the analysis method used, leveraging adult PK data as prior information. It also allows to evaluate exposures in pediatrics based on R scripts interfacing NONMEM. It supports a broad range of study designs and automatically derives NONMEM models for simulation, estimation and uncertainty quantification from the reference model.
The workflows are provided as Rmarkdown-scripts, showing the analyses based on synthetic data, which can readily be customized for a specific data set. Analysis results, e.g., type I and type II error rates, confidence intervals of estimated parameters, and comparison of pediatric and adult exposure can be reported in graphical and tabular form.
Conclusion
With pedPower, complex analyses for pediatric trial design and dose evaluation can be performed efficiently and in a standardized way, limiting the customized analysis code. With a focus on reproducibility, the analyses can be performed fully within R, but other software can be integrated into the analysis through simple interfaces, provided at each step of the analyses. While pedPower currently implements three different analysis workflows (see objectives) focusing on PK models using adult prior information, pedPower’s scope can readily be extended to other model types and applications due to the capability of manipulating NONMEM models in a modular way.
References:
[1] Food and Drug Administration, General Clinical Pharmacology Considerations for Pediatric Studies of Drugs, Including Biological Products. Draft Guidance. September 2022. https://www.fda.gov/media/90358/download