Biomedical Engineering Reference
In-Depth Information
underpredicted at all time points. Similarly, Figure 9 of the Battelle
(2006) report (not shown) compares predicted and observed human
blood methanol levels during and after inhalation exposures using data
from Ernstgard et al. (2005). The data for the 100 ppm exposures are
substantially underpredicted at almost all time points.
Figure 10 of the Battelle (2006) report (not shown) compares
predicted and observed human blood methanol levels during and after
200 ppm inhalation exposures using data from Osterloh et al. (1996).
The data are substantially underpredicted during exposure even after
subtraction of the reported background concentrations. Similarly, Fig-
ure 11 of the Battelle (2006) report (not shown) compares predicted and
observed human blood methanol levels during and after 800 ppm
inhalation exposures using data from Batterman et al. (1998). The
data for all time points are substantially overpredicted during exposure
after subtraction of the reported background concentrations.
Given all of these instances of systematic bias of the Battelle (2006)
model predictions relative to the data that were employed in developing
and fitting it, perhaps it would be prudent to consider instead a
sequential fitting process in which data sets are incorporated one at
a time, thus permitting a determination of how the fit to previously
included data sets is improved or degraded as additional data are
included.
6.6 IS “VISUAL OPTIMIZATION” AN ADEQUATE
TECHNIQUE FOR ESTIMATING PBPK MODEL
PARAMETERS?
When PBPK models were first being developed, the software required
for rigorous statistical parameter estimation was completely lacking.
Initial values of many model parameters were often just “educated
guesses,” and parameter “optimization” was accomplished by trial and
error, with visual inspection of discrepancies between data and pre-
dicted time courses providing the sole basis for determining whether a
given model's “fit” to the data was acceptable. Clearly, this process
involved a great deal of subjectivity, leaving numerous opportunities for
