Risk Communication: A Broad Perspective on Cancer Risks from Low Exposures to Chemicals

Lois Swirsky Gold, Children’s Hospital Oakland Research Institute

The project aims to improve public understanding about potential human cancer risks from chemicals, at realistic human exposure levels. As with radiation, public misconceptions abound. In earlier work over 20 years, we have criticized the assumptions and methodology of standard risk assessments. Our criticisms are based on both theoretical considerations and our findings that a large proportion, more than half, the chemicals tested in high dose tests are carcinogenic. We have discussed that this high positivity rate was likely due to the effects of the high dose itself rather than the chemical per se for many reasons, including especially an increase in cell death, consequent increased cell division, leading to mutagencity and cancer. Therefore, risk assessments for humans require data on cancer mechanisms. In recent years there has been an important shift in cancer risk assessment methodology, particularly for nongenotoxic chemicals, away from qualitative "carcinogen vs. no carcinogen" and linear extrapolation, and toward a recognition that tumor rates can be increased in high-dose tests by modes of action that must be considered in quantitative risk estimation. Data on mechanism for individual chemicals have been incorporated into IARC classification, panels of experts have applied a Human Relevance framework, comparing each of multiple steps in the carcinogenic process in rodents to what might be plausible in humans. Current EPA Cancer Guidelines permit a nonlinear approach for risk assessment if supported by data on mechanism, and EPA has regulated some nongenotoxic chemicals that test positive in rodents, as unlikely to be carcinogenic to humans at levels that do not cause cytotoxicity and cell regeneration. We have developed a Margin of Exposure (MOE) graphic (see page 2) to convey our major findings and to use as a starting point for communicating recent developments in risk assessment to the public. The graphic emphasizes that in order to get a big picture about possible cancer hazards to humans from environmental contaminants or pesticide residues, consideration of human exposure levels is important at the outset.

Margin of Exposure for a chemical exposure is the ratio: Administered Carcinogenic Dose for 10% of Rodents (mg/kg/day) Average Human Exposure (mg/kg/day)

MOE values on 100 exposures range 100-billion-fold (shown on log scale). A value of 1 indicates that human exposure is equal to administered rodent carcinogenic dose; if <1, human exposure exceeds the rodent cancer dose. A value of 100,000 indicates the rodent dose was 100,000 times greater than human exposure, which approximates a default regulatory risk of one in a million using a linear model. Values are reported for all rodent carcinogens for which reliable estimates of both concentration and average chronic US exposure are available. The highest average human exposure is used for each chemical; drugs are at recommended doses. MOE values are presented as categories of exposure in different colors: occupational exposures (mainly prior to 1985), drugs (at recommended dosages), natural chemicals in food, air pollutants, synthetic food additives, and residues of synthetic pesticides in food or pollutants. Overall, the graphic indicates that some historically high exposures in the workplace were close to the rodent administered dose, including 3 marked with asterisks to indicate that epidemiological results were positive at those levels; the MOE for these is less than a factor of 2 of the rodent dose. Some pharmaceuticals are close to the rodent dose, while consumption of synthetic pesticide residues in food ranged historically from 6,000 to 1-billion-fold lower than the rodent dose. There is a large background of naturally occurring rodent carcinogens in average consumption of common foods (shown on the right) that ranges in MOE from alcohol consumption (close to the rodent dose) to heterocyclic amines formed by cooking at high temperature (50,000 to a million-fold lower than the rodent dose). The natural background of rodent carcinogens in the diet casts doubt on the importance of exposures to synthetic pesticides or water pollutants. We have estimated that 99.9% of the chemicals humans are exposed to are natural, such as "plant pesticides" or the products of cooking. Among chemicals tested in high-dose bioassays, however, less than a quarter occur naturally. Half the natural chemicals, like half the synthetic chemicals, are positive in such tests. A natural herbal supplement containing aristolochic acid was recently found to induce urothelial tumors in humans (asterisk in graphic) at a level within a factor of 2 of the rodent cancer dose.

For some nongenotoxic chemicals, sufficient toxicological data have been generated that panels of experts and regulatory agencies have incorporated the results into risk assessments, sometimes resulting in risk estimates that are based on nonlinearity or conclusions that the mode of action that induces tumors in rodents is not relevant to humans (see downward arrows in graphic. It seems likely that future evaluations for many chemicals will result in lower risk estimates. In the graphic, for example, rodent kidney and liver tumors are induced by chloroform (MOE in tap water 20 years ago=20,000) by a mode of action that involves oxidative metabolism leading to cytotoxicity and regenerative cell division. EPA concluded that without cytotoxicity, tumors were not expected to develop in humans. Another well delineated mode of action is for kidney tumors induced by d-limonene (MOE as a natural chemical in food=100, and MOE as a food additive=2,000); tumors are induced only in male rat kidney by the binding of the epoxide of d-limonene to a male rat-specific protein, a-2 urinary globulin, which leads to toxicity of the tubular cells, consequent cellular regeneration, and ultimately tumors. Humans can metabolize d-limonene to its epoxide but lack a similar protein to which the epoxide can bind; thus, scientific panels, including IARC, have concluded that this mode of action does not occur in humans.

» Return to the top