Toxicology of Industrial Compounds

170 EXTRAPOLATION OF TOXICITY DATA AND ASSESSMENT OF RISK
factors or different default factors have recently been published (Lewis et
al., 1990; Renwick, 1991, 1993).
The second approach (for non-threshold effects) also relies mainly on
default assumptions for dose-response extrapolation and cross-species
extrapolation. Especially cancer risk assessment has been the subject of
much debate and there are a number of extrapolation methods reviewed
recently by Park and Hawkins (1993) and Hallenbeck (1993). The default
methodology in the . has been summarized by Frederick (1993). In
principle, the risk assessment is based on a chronic rodent bioassay
conducted at or near the maximum tolerated dose (MTD). The lifetime
constant dose rates and the tumour incidence data for the individual dose
groups are used to determine the dose response by fitting the data with a
computer program. The linearized multistage cancer model (LMS) is often
used to perform this step. The LMS model extrapolates the rodent tumor
data observed at the MTD to a dose with a predefined risk and the 95 per
cent upper bound on the dose-response curve is calculated. The interspecies
extrapolation to humans is performed by a correction factor based on body
weight or body surface. Subsequently, the dose is determined that
corresponds to a maximum allowable calculated upper bound on risk. The
resulting number does not describe the actual human risk under low-level
environmental exposure, but provides an upper bound to human risk that
is assumed not to be exceeded. The actual risk may be in the range between
0 and the upper bound. In the process described, the dose is defined as
administered dose or inhaled concentration. As a result, the lowdose
extrapolation does not take into account non-linearities in tissue dosimetry
and response. In addition, the interspecies extrapolation is performed using
a default approach that does not account for mechanistic species
differences.
Use of PBPK models in risk assessment
General description
Physiologically based pharmacokinetic (PBPK) models have been used
increasingly over the past decade to improve several aspects of the
assessment of risk associated with human exposure to chemicals. Examples
are PBPK models for styrene (Ramsey and Andersen, 1984; Csanády et al.,
1994), dichloromethane (Andersen et al., 1987), 1,4-dioxane (Reitz et al.,
1990a), chloroform (Reitz et al., 1990b), ethyl acrylate (Frederick et al.,
1992), methanol (Horton et al., 1992) and 1,3-butadiene (Johanson and
Filser, 1993). Recent reviews of the use of PBPK models in risk assessment
have been published by several authors (Frederick, 1993; Travis, 1993;
Wilson and Cox, 1993; Andersen and Krishnan, 1994).