Objectives: Clinical Pharmacology studies typically conducted in healthy adult volunteers provide robust pharmacokinetic (PK) information. However, due to recruitment challenges in pediatric participants, studies conducted in pediatric population usually involve sparse sampling leading to limited PK data. Nevertheless, it’s always very challenging in determining the appropriate number of participants and sampling time points for any pediatric studies to provide a robust set of PK data needed for full characterization of drug pharmacokinetics and PopPK model optimization.
An early dose-finding pediatric study was conducted on a target molecule and a preliminary pediatric PopPK model was developed to inform study design of future pediatric studies. Herein, we present SSE, a Monte Carlo method more commonly used to investigate relations between simulation model and estimation models based on Perl-speaks-NONMEM1. In our unique case here, we utilized SSE to provide perspective on study design (sample size and PK time points) needed for an optimal PK parameter analysis.
Methods: SSE was performed to provide perspective on the sample size and PK time points in the pediatric study needed for parameter analysis and model-based PK bridging to support extrapolation. Briefly, concentrations were simulated from a pediatric PopPK model using a simulation template dataset consisting of 2/4/8/10/12/16 pediatric patients in each age cohort sampled at 6 time points to create 1000 datasets, each representing a single trial. The same simulation was run for 10/12/14/16 patients in each age cohort at 4 time points for additional comparison. Then, specific parameters including but not limited to clearance, volume of distribution and absorption were estimated from fitting the model to each simulated dataset, while fixing parameters that were not expected to be well informed (e.g. formulation, etc). The results were summarized to determine the precision and accuracy of prediction using relative bias and root-mean-square deviation (RMSE), and different study designs were compared based on parameter statistics.
Results: Our analysis showed that both N=10 per age cohort at 6 time points and N=14 per age cohort at 4 time points provided robust sets of PK data that can adequately characterize PK parameters (clearance, volume, absorption) with reasonable precision and bias ( < 20%). One important note is the results were based on a preliminary model intended to provide perspective on study design, and the values of precision and bias would be expected to be lower when combining with additional pediatric PK data from future studies. Overall, the SSE here supports a mechanistic study design and ensure adequate PK datasets from future studies needed to fully characterize target molecule PK and optimize the popPK model.
Conclusions: Herein we successfully utilized SSE to mechanistically design a pediatric study and ensure adequate PK dataset needed for model optimization.
