Multistep Unified Models Using Prior Knowledge for the Prediction of Drug Clearance in Neonates and Infants

Abstract

Allometric approaches are widely used for interspecies scaling for the prediction of pharmacokinetic (PK) parameters during drug development. The concept of allometry can also be extended to predict PK parameters from adults to children. Three methods for extrapolating pediatric clearance were developed and evaluated using the clearance values of 4 drugs. The first method was established using a simple allometric (SA) model with estimated coefficient and exponent based on data ranging from children older than 2 years to adult. Then we developed a unified multistep single-exponent (MSE) and multistep body-weight-dependent exponent (MBDE) models. The major steps in these 2 new methods include generating pseudopredicted clearance for unobserved new populations such as preterm neonates, term neonates, and infants. Subsequent steps involve incorporating the pseudopredicted clearance with the actual PK data from older children and adults. All 3 models were then used to predict drug clearance in children ≤2 years old (N = 278). Drug clearance was predicted with mean absolute error of 29.6, 14.2, and 12.9 using SA, MSE, MBDE, respectively. The root mean square error was 65.9, 29.8, 24.7 for SA, MSE, MBDE, respectively. Approximately 41%, 72%, and 74% of the children's clearance data were within 0.5 to 1.5-fold of the observed values when drug clearance was extrapolated using SA, MSE, and MBDE models, respectively. The present multistep unified extrapolation approaches improved the prediction of clearance from preterm neonates to 2 years of age and may have practical use for first-in-pediatric dose selection.

Keywords: dosing and clearance; extrapolation; nonlinear model; pediatrics.