Hazard Assessment and Contaminated Sites - Dealing with Contaminated Sites: From Theory Towards Practical Application - page 550

Numerical estimates of risk probabilities are generated by fitting one or more
mathematical models to the data in the experimental dose range and extrapolating to
the low environmental exposure doses. For example, low-dose extrapolation using
a linear model is a default approach for cancer Risk Assessment in the USA (US
EPA 1996 ) and is one approach which has been used by the WHO for genotoxic
carcinogens in deriving drinking water guidelines (WHO 1993 ).
12.5.5 Threshold Versus Non-Threshold Approaches
This area of scientific debate largely centres on the management of carcinogens.
The important conceptual distinction between non-threshold methods and those
which derive an acceptable exposure from the NOAEL using a safety factor is that
the safety factor approach makes no attempt to determine a finite level of risk at low
exposures whereas the linear methods make an estimate of the risk at low exposures.
The NOAEL is assumed to be the threshold dose for the effect. Both approaches
have advantages and disadvantages.
The advantages of the threshold approach are that the NOAEL is relatively easy
to determine, and the process is simple to use, easy to understand and allows the
use of expert judgement. In the few cases where epidemiological data have become
available, the ADIs derived by this method have been validated (Lu and Sielken
1991 ). Additionally, the approach has been applied seemingly in a consistent fash-
ion by the WHO in the last three decades in deriving ADIs for pesticides (Lu 1995 ).
The safety factor approach remained essentially unchanged until 1994 (WHO 1994 ),
although a number of articles were published suggesting modifications or improve-
ments (e.g., Calabrese and Baldwin 1994 ; Calabrese and Gilbert 1993 ;Crump 1984 ;
Johannsen 1990 ; Lewis et al. 1990 ; Lu and Sielken 1991 ; Zbinden 1979 ).
Because it provides numerical estimates of risk at all doses, the non-threshold
approach, in principle, has the potential advantages (if the estimates are correct)
of allowing: computation of comparative risks in the sub-experimental range, which
may be a useful tool in Risk Management and communication; potency comparisons
between contaminant at a particular risk level; and estimates of the increased risks
if a particular dose is exceeded. It has been argued (McMichael 1991 ) that risk
estimates by this approach approximate those seen in humans in some cases and
where there are disparities they are overestimates of the risks.
Both the threshold and non-threshold methods, however, are likely to be unduly
influenced by the selection of doses. The choice of the NOAEL is limited to one of
the doses included in the experimental design. The biological no effect dose may
occur at this dose or at a dose not included in the study. The closeness with which
the selected NOAEL truly reflects the actual no effect dose has an obvious impact
on the degree of protectiveness in the derived ADI, PTWI or RfD. Furthermore, the
NOAEL is influenced by the biological effects monitored, the number of animals
in the test groups, the spontaneous incidence of the adverse effect, and the criteria
used to determine when the incidence in a test group exceeds that in the controls
(Renwick and Walker 1993 ).