• Margin of Exposure: The Big Picture
  • Methods
  • Results
  • Implication
• Cancer Risk Assessment Methods, 1960’s-1990’s
  • Proportion Of Chemicals That Are Carcinogenic in High Dose Rodent Experiments
  • More than 50% of chemicals tested in rodents at the Maximum Tolerated Dose (a near-toxic dose) cause tumors.
  • What may account for the high positivity rate?
  • Conclusion
• New Risk Assessment Paradigm
  • Human Relevance Framework
  • Human Relevance Framework Examples

Citation for this Web Page: Lois Swirsky Gold, Bruce N. Ames, Thomas H. Slone. Animal Cancer Tests and Human Cancer Risk: A Broad Perspective. http://potency.berkeley.edu/MOE.html, September 2008.

• Results are rarely available in humans that identify chemicals as a cause of human cancer. Therefore, to try and identify possible hazards, chemicals have been tested in rodents at high, near toxic doses daily for life, the Maximum Tolerated Dose (MTD), in order to maximize the chance of increasing tumors in a small number of animals. The MTD differs from one chemical to another by several million-fold, and human exposures to rodent carcinogens vary by more than a billion-fold. Both must be considered when evaluating possible cancer hazards. Therefore, for each chemical exposure we calculate how many fold lower the average human exposure is than the dose to give rodents cancer — the Margin of Exposure (MOE).
  
  
• The MOE for a chemical exposure is the ratio:
  

For example, a value of 1 means that the human exposure level is the same as the dose that gave tumors in rodent experiments. In a few unusual cases at the top of the graphic, the MOE is less than 1, which indicates that human exposures were so high that the rodent carcinogenic dose was actually lower than the exposures to humans (e.g., occupational exposure to Vinyl chloride in the 1950’s, MOE=0.01). A value of 300 indicates that the rodent carcinogenic dose is 300 times greater than the human intake (e.g., the naturally occurring chemical Safrole in spices in the total diet). A value of 7,000,000 indicates that the rodent carcinogenic dose is 7 million times greater than the human exposure (e.g., the synthetic pesticide Lindane in the total diet in 1990).

• The graphic on the right shows MOE values for different types of human exposure reported by colors that are defined in the box at top right. Human exposures to rodent carcinogens are ordered from greatest possible human cancer hazard at the top to least possible hazard at the bottom. Human exposures at the top are close to the rodent carcinogenic dose while those towards the bottom are tiny compared to the rodent carcinogenic dose; therefore, low MOE values at the top of the graphic (lowest MOE) are of greatest concern for possible cancer hazards.
  
  
• Symbols based on Human Relevance Framework
  ↓↓  Results in rodents are not relevant to human cancer risk.
  ↓  Results in rodents would only be relevant to humans at toxic levels.
  *  Carcinogenic to humans at this MOE.

A table in a popup window reports details on human exposure, rodent cancer dose and reference for each exposure in the graphic.

• The MOE for various human exposures ranges 100 billion-fold.
  
  
• Many foods contain naturally-occurring chemicals that have been shown to cause cancer in high dose rodent tests. Consumption of natural chemicals in the diet (green on the right) occurs in common foods at a range of MOE levels, and many are much closer to the rodent cancer dose than synthetic pesticide residues or pollutants (orange), which are far below the rodent cancer dose. Exposures to natural chemicals include the chemical constituents of fruits, vegetables or spices as well as the products of cooking, such as chemicals in roasted coffee or chemicals produced by cooking meat, fish, or French fries at high temperature.
  
  
• The common exposures to natural chemicals in the diet that cause cancer in rodent tests cast doubt on the importance for human cancer of synthetic pesticide residues or pollutants. We have estimated that 99.9% of the chemicals humans are exposed to are natural, and we find that they are as frequently positive in rodent cancer tests as synthetic chemicals (see Proportion Of Chemicals That Are Carcinogenic in High Dose Rodent Experiments). Many ordinary foods would not pass the health criteria that have been used to regulate human exposures to synthetic chemicals based on results of animal cancer tests.
  
  
• The graphic indicates that some historically high exposures in the workplace (dark blue) 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 (red) are also close to the rodent dose, but for clofibrate, phenobarbital, and gemfibrozil, the process of tumor development in rodents has been evaluated as not relevant to humans and there is no human risk (marked with ↓↓) (see New Risk Assessment Paradigm).

• Chemicals are tested in rodents at near-toxic doses daily for life, the Maximum Tolerated Dose (MTD).
  
  
• Species Extrapolation: Results in rodents are assumed to be relevant to humans, and therefore an increase in tumors indicates a potential human carcinogen.
  
  
• Dose Extrapolation: Tumor results in rodents at high dose (MTD) are assumed to be relevant to human exposures even at very low exposure levels, and the risk of cancer is proportional to the exposure level, i.e. linear extrapolation.

Natural chemicals are positive as often as synthetic, industrial chemicals. More than 99% of the human intake of chemicals is from naturally occurring chemicals.

Natural pesticides (the chemicals that plants produce to defend themselves against predators and are present in all the fruits and vegetables we eat) are as often positive as commercial pesticides.

1. High dose effects that are not relevant to low human exposures.
   
   
2. Carcinogenic processes in rodents that are not plausible in humans.
   
   
3. Bias in picking chemicals to test that were expected to be positive. Our research indicates that bias is a minor factor for several reasons, including:
   
   
   • Chemicals were selected primarily because of extent of human exposure, e.g. drugs, pesticides and workplace exposures.
     
     
   • In 2 separate prediction exercises, expert scientists predicted which of a set of chemicals that had not yet been tested would be carcinogens when tested and which would not. The experts differed from each other in their predictions. After the experiments were done, the accuracy of the predictions in both exercises averaged only slightly better than chance.

• The chronic, high dose rodent cancer tests is not adequate to understanding human cancer risk at the low doses of most human exposures. Tumor development in cancer tests is likely due, in part, to high dose effects or processes that may not be relevant to humans.
  
  
• No diet can be free of naturally occurring chemicals that cause cancer in rodents at high doses. [See green chemical exposures in graphic on right].

• Emphasis on biological processes leading to tumors, using new scientific methods to evaluate metabolism and effects of a chemical on cells.
  
  
• Qualitative Assessment: Compare key steps in rodent tumor development to plausibility in humans, using human physiology, metabolism, and epidemiology. If not plausible in humans, then no risk to humans.
  
  Quantitative Assessment: If the process in rodents is qualitatively plausible in humans then compare quantitative factors between species, such as dose to the target organ or quantitative differences in response to hormone imbalance.
  
  
• If the process of tumor development in rodents is plausible in humans or if evidence is not adequate to evaluate the process of tumor development, then for regulatory policy a risk assessment is necessary.
  
  
• Current EPA Guidelines provide for two types of default approaches to extrapolate from high dose rodent results to human exposure — one linear and the other nonlinear. EPA concluded, for example, that there is no cancer risk to humans from chloroform unless exposure is high enough to cause cell killing and cell proliferation; thus, there is a practical threshold in the dose-response relationship. Human exposure to chloroform as a by-product of disinfecting drinking water is far below that toxic dose (see Graphic).

• Formaldehyde: Damages DNA and therefore is evaluated as relevant to humans. It is also a nasal irritant when inhaled, and at high doses causes cell killing and cell proliferation in both rodents and humans. Risk declines markedly at doses that do not kill cells.

• Chloroform: The process of liver and kidney tumor development in rodents is plausible in humans, and requires high, toxic doses that kill cells. There is a threshold dose below which there is no cancer risk to humans.
  
  
• Clofibrate, Gemfibrozil: Not relevant to humans on quantitative basis. Peroxisome proliferation occurs in both rodents and humans, but there is a genetic difference — cell proliferation is seen only in rodents, but not in mice with the human gene for the peroxisome receptor alpha. Also, epidemiological studies of long-term use of these drugs show no increase in human cancer.
  
  
• d-Limonene: Not relevant to humans. The alpha₂-urinary globulin protein is specific only to male rats. Kidney tumors are induced by binding of the epoxide of alpha 2-urinary protein to kidney cells leading to toxicity and cell proliferation.
  
  
• Saccharin: Not relevant to humans. Bladder tumors are induced only at high dose, and only in rats. Tumor development is related to the physiology of the rat urinary system, including formation of a precipitate, irritation and cell killing.

Lois Swirsky Gold, Bruce N. Ames, Thomas H. Slone Carcinogenic Potency Project

This web page was supported by the Department of Energy, Low Dose Radiation Research Program. From 1980-2008 the Carcinogenic Potency Project has been supported by Department of Energy (DOE); National Institute of Environmental Health Sciences (NIEHS); National Toxicology Program (NTP); National Cancer Institute (NCI); Environmental Protection Agency (EPA); University of California, Berkeley, Dean’s Office of the College of Letters and Science.