Published in:
01-01-2019 | Original Research Article
Choosing the Allometric Exponent in Covariate Model Building
Authors:
Jaydeep Sinha, Hesham S. Al-Sallami, Stephen B. Duffull
Published in:
Clinical Pharmacokinetics
|
Issue 1/2019
Login to get access
Abstract
Background
Allometric scaling is often used to describe the covariate model linking total body weight (WT) to clearance (CL); however, there is no consensus on how to select its value.
Objectives
The aims of this study were to assess the influence of between-subject variability (BSV) and study design on (1) the power to correctly select the exponent from a priori choices, and (2) the power to obtain unbiased exponent estimates.
Methods
The influence of WT distribution range (randomly sampled from the Third National Health and Nutrition Examination Survey, 1988–1994 [NHANES III] database), sample size (N = 10, 20, 50, 100, 200, 500, 1000 subjects), and BSV on CL (low 20%, normal 40%, high 60%) were assessed using stochastic simulation estimation. A priori exponent values used for the simulations were 0.67, 0.75, and 1, respectively.
Results
For normal to high BSV drugs, it is almost impossible to correctly select the exponent from an a priori set of exponents, i.e. 1 vs. 0.75, 1 vs. 0.67, or 0.75 vs. 0.67 in regular studies involving < 200 adult participants. On the other hand, such regular study designs are sufficient to appropriately estimate the exponent. However, regular studies with < 100 patients risk potential bias in estimating the exponent.
Conclusion
Those study designs with limited sample size and narrow range of WT (e.g. < 100 adult participants) potentially risk either selection of a false value or yielding a biased estimate of the allometric exponent; however, such bias is only relevant in cases of extrapolating the value of CL outside the studied population, e.g. analysis of a study of adults that is used to extrapolate to children.