Environmental Working Group
Chemical Industry Archives
home | what's new | the inside story | fact and fiction | about | search options | 

Fact and Fiction


Fiction #6: 'Animal studies can't predict whether a chemical will harm people.'

To the contrary, animal studies are quite reliable predictors of human harm.

In fact:

  • Every single chemical pollutant found to cause cancer in humans was confirmed or predicted by animal studies. (NRC 1993)
  • Nearly every single chemical (98 percent) found to cause birth defects in humans was confirmed or predicted by animal studies (Klaassen 1996).

Animal studies are a basic stepping stone of toxicology. Those who say otherwise typically don't like the results of the test in question. Every single regulatory agency in the world uses animal studies as the basis for regulating chemical use and protecting human health. In fact, industry is effusively supportive of animal studies when they support a claim of safety for one of their products.

There are two basic rules of thumb with animal studies:

  • When a series of animal studies replicate a clear effect, the chemical is virtually guaranteed to produce the same effect in humans.
  • On the other hand, negative results from a single, and sometimes repeated, animal studies are not reliable predictors of safety. For instance, 25 percent of the time when rat studies are negative for birth defects, the chemical was ultimately found to cause birth defects in humans (Klaassen 1996).

This is because:

Animal studies use genetically uniform rodents to help elicit a consistent response to chemical exposure from a relatively small number of animals. This is exactly the opposite of the human population where millions of highly variable individuals are exposed to widely different doses at different times in life.

Animal tests can easily miss effects that occur in less than 1 percent of the population, mostly due to limits on the number of animals in the test. To find effects in a small percentage of the population requires an unmanageable number of animals. According to Casarett & Doull's Toxicology, the standard desk reference in the industry:

"A 0.5 percent increase in cancer incidence in the United States would result in over 1 million additional cancer deaths each year -- clearly an unacceptably high risk. However, identifying with statistical confidence a 0.5 percent incidence of cancer in a group of experimental animals would require a minimum of 1000 test animals, and this assumes that no tumors were present in the absence of exposure (zero background incidence)." (Klaassen 1996)

Animal studies can produce a false sense of safety. A chemical that causes cancer in 8 percent of the exposed human population, might fly through the standard cancer study and produce no effects at all.

To quote again from Casarett and Doull's Toxicology,

"In a chronic bioassay with 50 animals per test group, a tumor incidence of about 8 percent could exist even though no animals in the test group had tumors." (Klaassen 1996)

If the entire U.S. population were exposed to this chemical, as a food additive, or example, this would translate into 22.4 million people with a cancer that did not even show up in an animal study.

Animal tests are notoriously bad at picking up learning and behavioral problems. Animal studies with lead, mercury and PCBs missed the toxic dose in humans by from 100, to 10,000 times (Rice et al 1996). This is largely because the human brain is much more complex and vulnerable than the rat brain, and learning tests for rats are just not as easy to devise as learning tests for children. The toxic dose for these compounds was only determined after decades of human injury.

Pound for pound, humans can get higher doses of many chemicals than laboratory animals - simply because of basic differences in biology and metabolism. For the drugs oxazepam, primidone, and phenolphthalein humans get higher doses by factors of 10, 40, and 100, respectively, relative to rodents. Toxicity tests on chemicals, when they are done, almost never consider this possibility (Bucher 2000).


Bucher, JR. 2000. Doses in rodent cancer studies: Sorting fact from fiction. Drug Metabolism Reviews. 32(2), 153-163.

Kimmel, Carole, et al., 1992, Animal models for assessing developmental toxicity, in Similarities & Differences Between Children & Adults, ILSI Press, Washington, DC, pp 43-65.

Klassen, CD, ed. 1996. Casarrett & Doull's Toxicology. The Basic Science of Poisons. Fifth ed. The McGraw-Hill Companies, Inc.

National Research Council, 1993, Pesticides in the Diets of Infants and Children, National Academy Press: Washington, DC.

Rice D., A. Evangelista de Duffard, R. Duffard, et al., 1996, Lessons for neurotoxicology from selected model compounds: SGOMSEC joint report, Environ. Health Perspect., 104, 205-215.

« prev page

last updated: march.27.2009

Home  |  What's New  |  The Inside Story  |  Fact and Fiction  |  About  |  Search  |  Contact  |  Donate  |  Site Index

The Chemical Industry Archives is a project of the Environmental Working Group.
Many documents in the archives require Adobe Acrobat reader (free download).