Hazard Assessment and Contaminated Sites - Dealing with Contaminated Sites: From Theory Towards Practical Application - page 551

Additional limitations of the threshold approach include: the NOAEL is often
perceived as a biological threshold, whereas it is a threshold limited by the experi-
mental protocol; risk is expressed as a fraction of the guidance dose (e.g., ADI); it
makes limited use of the dose-response slope; the choice of safety factors has been
arbitrary to some extent and the process does not generate a range of estimates of
risk, but rather a single estimate of a dose below which no adverse effects are likely
to be produced.
Dose selection in non-threshold models has been discussed by Lovell and
Thomas ( 1996 ) who suggest that the estimate of q 1 ∗ (the 95% upper confidence
limit of the slope estimate used for the linear multi-stage model used by the US
EPA) is so dependent on the doses selected that it is almost independent of, or at
least insensitive to, the actual tumour incidences in the dose groups. Specifically,
the highest dose in an animal bioassay has overwhelming influence on the esti-
mate of q 1 ∗ , thus leading to the overestimation of risk at very low doses, with the
extent of overestimation increasing as the environmental exposure becomes lower.
Typically, the highest dose in a carcinogenicity bioassay is the maximum tolerated
dose (MTD), a dose that causes no more than a ten percent decrease in body weight
and no other overt toxicity. The MTD is very much greater than doses expected
from non-occupational environmental exposures. Therefore, the dose which is the
least relevant to environmental Risk Assessment has the greatest influence on low
dose risk estimates.
Non-threshold models currently in use are inflexible and generally do not take
account of the complexities of the events between exposure to a contaminant and the
induction of a neoplasm. Risks estimated at doses below the range of experimental
data can vary considerably depending on the model used, even though the various
mathematical models used generally fit the experimental data equally well (Crump
1985 ; Paustenbach 1995 ). The numerical expression of the estimated level of risk
falsely gives the impression that it represents an exact measure of actual risk. This
numerical expression provides little or no information on the uncertainties related to
the estimated level of risk, nor does it allow comparison with values for non-cancer
health effects.
Low-dose linearity assumes a positive slope of the dose-response curve at zero
dose and implies that a single, irreversible genetic event at the initiation stage of
carcinogenesis leading to transformation of a cell, is sufficient by itself to lead to
the development of cancer. The major difficulty in this debate is the impossibility of
testing experimentally the shape of the dose-response curve at extremely low doses
(Purchase and Auton 1995 ).
A transformed cell which has acquired the potential to develop into a tumour,
will probably realise that potential only rarely (US EPA 1996 ), most likely because
of the natural large scale repair of DNA damage and other defence mechanisms of
the body (DOH 1991 ). Furthermore, whilst it is generally accepted that mutagens
and mutations play a role in the development of cancer, carcinogenesis is more than
mutagenesis, with a number of non-mutagenic as well as mutagenic events taking
place during the process (Bishop 1991 ). The shape of the dose-response curve at
any one of these steps, not just the mutagenic events, can influence the shape of the