The Integration of Allometry and Virtual Populations to Predict Clearance and Clearance Variability in Pediatric Populations over the Age of 6 Years

Abstract

Background and Objectives

Pharmacokinetics play an integral role in the pediatric drug development process. The determination of pharmacokinetic parameters, particularly clearance, in different age groups directly informs dosing strategies for subsequent efficacy trials. Allometric scaling for prediction of pediatric clearance from the observed clearance in adults has been used in this effort. Clinical trial simulation, a powerful tool used to inform clinical trial design, requires both an estimate of clearance along with an estimate of the expected pharmacokinetic variability. The standard deviations (SD) of individual clearance values for adults are typically used and may lead to inaccurate predictions by not taking into account the more widespread distribution of factors such as body weight in the pediatric population. The objective of this study was to assess the accuracy of allometric prediction of drug clearance as well as methods for predicting clearance variability in children 6 years of age and older.

Methods

US Food and Drug Administration (FDA) clinical pharmacology reviews of pediatric studies conducted from 2002 onwards were reviewed to collate adult and pediatric clearance and clearance variability for studies including children 6 years of age and older. A set of 1,000 virtual adults {A} and a set of 5,000 virtual children (aged 2–17) {P} were generated on the basis of the White American NHANES database. Pediatric clearances were predicted in method 1 by using the geometric mean adult clearance from the in vivo study and calculating pediatric clearance for each virtual child within a subset {P′} of {P} that contained only those children that were within the age range of the in vivo pediatric study. In method 2, adult clearance values were randomly generated from the geometric mean adult clearance and standard deviation and assigned to each virtual adult in {A}. For each adult, allometric clearance scaling was completed with each virtual child within {P′}. The prediction error for the predicted and observed clearance and the clearance variability metric, coefficient of variation (CV), was calculated. The prediction accuracy as a function of the lowest age range (2 years and older) included in the study was also assessed.

Results

Thirty-nine unique drugs were included in the study. For both method 1 and method 2, 100 % of predicted pediatric mean clearances were within 2-fold of the observed values and approximately 82 % were within a 30 % prediction error. There was a significant increase in the prediction accuracy of CV using method 2 vs. method 1. There was a major bias towards underprediction of pediatric CV in method 1 whereas method 2 was precise and not biased. Clearance and CV prediction accuracy were not a function of the age range included in the in vivo studies. The observed CV between the adult and pediatric study groups was not significantly different although, on average, the observed pediatric CV was 32 % greater than that from adult studies.

Conclusions

Allometric scaling may be a useful tool during pediatric drug development to predict drug clearance and dosing requirements in children 6 years of age and older. A novel methodology is reported that employs virtual adult and pediatric populations and adult pharmacokinetic data to accurately predict clearance variability in specific pediatric subpopulations. This approach has important implications for both clinical trial simulations and sample size determination for pediatric pharmacokinetic studies.