Associate Director, Clinical Pharmacology Arcus Biosciences Inc. BURNSVILLE, Minnesota, United States
Disclosure(s):
Jordon A. Johnson, PharmD: No relevant disclosure to display
Objectives: Pediatric dose selection is a major challenge in drug development. Pharmacokinetic (PK) data is mainly collected from adults, with population PK (PPK) modeling used commonly for pediatric dose extrapolation simulations. Existing methodologies do not account for age-related correlations with renal clearance. We present a novel approach of a data-driven framework for pediatric renal clearance prediction combining the National Health and Nutrition Examination Survey (NHANES) and Centers for Disease Control and Prevention (CDC) growth charts, which provide free, publicly available sources of age-appropriate covariate data for children.
This approach allows simulation of age-dependent pediatric covariate distributions (e.g., serum creatinine (SCr), creatinine clearance (CrCL)) and their impact on drug exposure.
Methods: A virtual pediatric population was generated across three age brackets (≥2 to < 6, ≥6 to < 12, and ≥12 to 17 years), 500 virtual patients per age bracket per sex were generated. A random normal distribution for subject weight was generated by age and sex using mean and standard deviations extracted from the CDC growth charts. Linear regression models were constructed to generate pediatric covariate data for height and SCr, respectively, using lab data obtained from the NHANES database. CrCL was calculated using either the Schwartz method (≥2 to 6 years) or a maturation method (≥12 to 17 years). Visual predictive checks (VPCs) were used to qualify the generated covariate distributions. An extrapolated pediatric PPK model, based on adult PK data, was used to simulate the steady state exposure for each virtual subject and summarized by each age bracket or by weight (≤40 kg or >40 kg). The PPK model was a 2-compartment model with body weight and CrCL as covariates on central volume and clearance, respectively.
Dose selection criteria were based on efficacy, safety, and convenience. Pediatric dosing regimens were selected based on how well the predicted exposure fit the upper and lower bounds of the 95% prediction interval for the time-course PK profile in adults. Convenience was defined as a preference for flat dosing regimens and Q2W over Q1W dosing frequency.
Results: Lab data from the NHANES database, years 1999-2018, was extracted for height, weight, and SCr. Height and weight data was available for 32,930 subjects (ages 2-17 years) and SCr data was available for 11,341 subjects (ages 12-17 years). Height and SCr data for virtual subjects were generated via fitted linear regression models (Ht = -16.47 + 43.23*ln(Wt) and SCr = 0.529 + 0.0025*Wt, respectively). VPCs found the virtual subject covariate distributions for each age group to be within expectations. The final pediatric dosing recommendations using the real-world PPK model was a flat dosing regimen of 100 mg Q2W in patients ≥6 to 17 years and 50 mg Q2W in patients ≥2 to 6 years old.
Conclusions: A framework for using free and publicly available resources to generate virtual pediatric populations and predict pediatric renal clearance and overall PK was developed. This approach supports data-driven, age-appropriate dose selection in pediatric patients aged 2 years and older via extrapolation from adult PK models for compounds with significant renal clearance.
Citations: NA
Keywords: Pediatrics, Population Pharmacokinetics, Simulation, Model-informed Drug Development
