Individual prior information in a physiological model of H-2(8)-toluene kinetics: An empirical Bayes estimation strategy

Physiologically-based toxicokinetic (PBTK) models are widely used to quanti fy whole-body kinetics of various substances. However, since they attempt t o reproduce anatomical structures and physiological events, they have a hig h number of parameters. Their identification from kinetic data alone is oft en impassible, and other information about the parameters is needed to rend er the model identifiable. The most commonly used approach consists of inde pendently measuring, or taking fom literature sources, some of the paramete rs, fixing them in the kinetic model, and then performing model identificat ion on a reduced number of less certain parameters. This results in a subst antial reduction of the degrees of freedom of the model. In this study, we show that this method results in final estimates of the free parameters who se precision is overestimated. We then compared this approach with an empir ical Bayes approach, which takes into account not only the mean value, but also the error associated with the independently determined parameters. Blo od and breath H-2(8)-toluene washout curves, obtained in 17 subjects, were analyzed with a previously presented PBTK model suitable for person-specifi c dosimetry. Model parameters with the greatest effect on predicted levels were alveolar ventilation rate Q(PC), fat tissue fraction V-FC, blood-air p artition coefficient K-b, fraction of cardiac output to fat Q(a/co) and rat e of extrahepatic metabolism Vmax-p. Differences in the measured and Bayesi an-fitted values of Q(FC), V-FC and K-b were significant (p < 0.05), and th e precision of the fitted values Vmax-p and Q(a/co) went from 11 +/- 5% to 75 +/- 170% (NS) and from 8 +/- 2% to 9 +/- 2% (p < 0.05) respectively. The empirical Bayes approach did not result in less reliable parameter estimat es: rather, it pointed out that the precision of parameter estimates can be overly optimistic when other parameters in the model, either directly meas ured or taken from literature sources, are treated as known without error. In conclusion, an empirical Bayes approach to parameter estimation resulted in a better model fit, different final parameter estimates, and more reali stic parameter precisions.