Comparative analysis of software for physiologically based pharmacokineticmodeling: Simulation, optimization, and sensitivity analysis

Historically, a number of different software packages running on a variety of hardware platforms have been used for model simulation. SimuSolv found w ide use because of its broad capabilities, including optimization, statisti cal analysis, and formalized sensitivity analysis as well as the capacity t o incorporate user-supplied subroutines. However in the early 1990s, SimuSo lv development ceased and a final version was released in 1999. Thus SimuSo lv will not be developed for newer platforms and operating systems. In this article, we compare and contrast the use of SimuSolv and Matlab (The MathW orks, Natick, MA) for physiologically based pharmacokinetic (PBPK) model im plementation with respect to parameter estimation (optimization) and sensit ivity analysis using a PBPK model for trichloroethylene (TCE). In both pack ages, it is possible to code PBPK models, run simulations, estimate paramet ers, and do sensitivity analysis. The hey difference is the additional prog ramming required in Matlab. Since Matlab does not have built-in estimation and sensitivity routines, it was necessary to write them for the Matlab TCE model. Additionally, Matlab handles flow control differently from SimuSolv , so the model code is written in a different order than for SimuSolv. In s pite of the additional coding requirements, Matlab is a well-supported and mathematically oriented simulation software package that is clearly suitabl e for application to PBPK modeling. All of the modeling tasks done in SimuS olv could also be done readily in Matlab. Most of the comparisons made to S imuSolv also carry over to ACSL-Tox, however ACSL-Tox calculates some sensi tivity coefficients very differently from the way they are defined in SimuS olv. Future development of art interpreter for Matlab would make modeling, sensitivity analysis, and parameter estimation less programming-intensive.