Clarity in Reporting Parameter Variance Needed to Improve Use of Published Models for Simulation Applications

Conference: ASCPT
Division: Cognigen

Abstract

Background: Since published pharmacokinetic and pharmacodynamic models are often used by others for the purpose of simulations, enhanced clarity in reporting and clear statements regarding assumptions will improve the reproducibility of modeling and simulation results and allow for accurate re-use of models and modeling findings.
Methods: To illustrate the importance of this issue, simulations were performed using the pharmacokinetic model for paroxetine by Feng et. al.1 , a manuscript that did not report which method was used for %CV calculation. This paper was selected in part because reported variability was in excess of 70 %CV for several parameters. Two simulations of 1000 individuals were performed using ω2 parameter estimates that were calculated from the reported %CV values, one based on the method of %?? = 100 ∗ exp ?2 − 1 and one one based on the method of %?? = 100 ∗ ?2. Simulations were performed using the R package mrgsolve. 2 Results of the simulations were then used to calculate the magnitude of between-subject variability (%CV) for each run and compare it to the original value.
Results: A 13.3-27.8% difference in %CV of the simulated distribution of the VM parameter was observed with the %?? = 100 ∗ ?2method when the other method was assumed to be used for reporting, and a 15.6-24.5% difference was observed when the reverse was assumed. Calculated % difference increases as the true ω2 increases, with ω2 = 0.1 yielding a 0.81% difference and ω2 =1.25 yielding a 46.0% difference.
Conclusions: Accurate reporting of either the variance (ω2 ) estimates in parameter tables or the method used to calculate %CV is important, especially as between-subject variance estimates increase.

By James Clary, Jill Fiedler-Kelly, Joel S Owen

ASCPT 2019 Annual Meeting, March 13-16, 2019, Washington, DC