Environment, Health, News May 6, 2013

The Experiment Is on Us: Science of Animal Testing Thrown into Doubt

by Jonathan Latham

by Pat Dutt and Jonathan Latham, PhD

New scientific research has cast grave doubt on the safety testing of hundreds of thousands of consumer products, food additives and industrial chemicals.

Everyday products, from soft drinks and baby foods, to paints, gardening products, cosmetics and shampoos, contain numerous synthetic chemicals as preservatives, dyes, active ingredients, or as contaminants. Official assurances of the safety of these chemicals are based largely on animal experiments that use rabbits, mice, rats and dogs. But new results from a consortium of researchers and published in the Proceedings of the National Academy of Sciences suggest such assurances may be worthless (Seok et al. 2013).

The results of these experiments challenge the longstanding scientific presumption holding that animal experiments are of direct relevance to humans. For that reason they potentially invalidate the entire body of safety information that has been built up to distinguish safe chemicals from unsafe ones. The new results arise from basic medical research, which itself rests heavily on the idea that treatments can be developed in animals and transferred to humans.

Laboratory Rat
A Laboratory Rat (Photo:ressaure)

The research originated when investigators noted that in their medical specialism of inflammatory disease (which includes diabetes, asthma and arthritis), drugs developed using mice have to date had a 100% failure rate in almost 150 clinical trials on humans.

According to Kristie Sullivan, Director of Regulatory Testing Issues at the Physicians Committee for Responsible Medicine (PCRM), this is not unusual “about 90% of all pharmaceuticals tested for safety in animals fail to reach the market, or are quickly pulled from the market”.

Wanting to understand why this might be so, the consortium decided to test the effects of various treatments that lead to inflammation, and systematically compare results between mice and humans. This postulated correlation across different animal species is sometimes known as the concordance assumption.

In a first set of experiments the researchers looked at acute inflammation in mice brought on by various stimuli. These stimuli were bacterial toxins (endotoxaemia), trauma, and burns. To measure responses the authors quantified positive or negative changes in gene activity for thousands of individual genes. The researchers found that changes in activity of a particular mouse gene after treatment typically failed to predict changes in activity in the closest related human gene. This was not the expected result. If humans and mice are meaningfully similar (i.e. concordant) then gene activity changes in mice should have closely resembled those in humans after a similar challenge. But they did not.

In further experiments, the researchers identified another difference. While humans responded with similar patterns of gene changes to each of the three different challenges (trauma, burns, and endotoxaemia), mice did not. The three treatments in mice each resulted in a distinct set of gene activity changes. This confirmed the initial results in the sense that mice and humans responded differently. It also implied that the differences in gene response between mice and humans are attributable not so much to a lot of detailed ‘noise’ but to fundamental differences in the physiology of mice and humans in dealing with these challenges.

Next, the researchers examined the activity of specific biological signaling pathways after similar treatments. These too were highly divergent between mice and humans. Surprised by the consistently poor correlations between the two species, the authors then tested other human/mouse models of inflammatory diseases. Again, the similarity between mice and humans was low.

In summary, repeated experiments confirmed that, when it comes to inflammation, mice and humans have little in common, a finding important enough in itself given the prevalence of inflammation-related diseases in humans. These include allergies, celiac disease, asthma, rheumatoid arthritis, and autoimmune diseases.

Perhaps these results should not be a surprise. Concordance has been questioned by numerous researchers, some of whom have noted that mice are separated from humans by 120 million years of evolutionary change (Stoloff 1992; Greek and Swingle Greek, 2003; Mestas and Hughes, 2004; Knight, 2007). And, unlike humans, mice also suffer from different diseases, lack a gall bladder, have no menstrual cycle, have multiple births, differ in immune systems, lifespan and size, to name only a few dissimilarities.

Thus the Seok study is not the first to conclude that mice are poor models for human disease, but it is notable for being by far the most comprehensive. Combined with results of previous experiments, its conclusions suggest researchers should expect that mouse, and probably other animal testing, is of little use in advancing the treatment of human illnesses, including heart disease and cancer.

In other words, the public is probably being badly served by much of the money spent on medical research. According to PCRM’s Kristie Sullivan, “the National Institutes of Health is giving researchers billions of dollars every year for research on animals”. While missing out on potential cures, the public is also likely being exposed to dangerous or ineffective pharmaceuticals. Animal testing nearly prevented the approval of valuable drugs such as penicillin and subsequent antibiotics, but it did not prevent the thalidomide disaster of the 50s and 60s (Greek and Swingle Greek, 2003).

This finding of non-concordance need not mean the end of medical research. It could even herald a more promising and scientific era. Sullivan believes that medical researchers “simply take for granted that animal models are useful” even though other, and possibly better, techniques for studying human disease are available. These include greater emphasis on human clinical observation and making better use of cell cultures for research.

But wasteful and unproductive medical research is arguably a sideshow besides the misplaced confidence in the safety testing of environmental and household chemicals. While medical failures affect the unwell, chemical toxins have potential repercussions for everyone.

If animals are not useful predictors of important disease responses in humans it is unlikely they are useful as test subjects for toxicological safety. In other words, lack of concordance means that the synthetic chemicals that are found in industrial products, incorporated into food, and otherwise spread throughout the environment, are essentially untested. The regulatory process through which they passed was never a scientifically validated and evidence-based system, but now the evidence shows it to have been functioning as a system of random elimination. “We are not protecting humans” says Kristie Sullivan, noting that “even a National Academy study agrees that many toxicological tests are not human-relevant.”

There are potential alternative toxicological tests, but despite multi-billion dollar grants, and even a human on a chip, the science is still incomplete. Michael Hansen, Senior Scientist at the Consumers Union, has been contributing to recent discussions over replacing animals for the purposes of regulatory toxicology. He acknowledges that “we should be moving towards in-vitro cell-based models” for chemical risk assessments. But how this can be done is not yet clear. Hansen points out that not only is “there a technical problem of how to incorporate them into an overall risk assessment”, but also that “in-vitro alternatives have yet to be validated”. Nevertheless, he still believes specific uses for animal research remain: “for carcinogenicity, for example, mice are appropriate models”.

An interesting question, when an estimated 100 million mice are sacrificed each year for medical research and in toxicology, is why it took so long to test this fundamental assumption. The answer is that it has been tested before, though not nearly as rigorously as it could have been. And the results have, in the view of many, not supported the idea that animals reliably model human physiology (Knight, 2007; Dressman, 2007).

A different kind of answer is that animal research is now big business. One genetically engineered mouse can cost $100,000 while a mouse treadmill can set taxpayers back $9,600 (Greek and Swingle Greek, 2003). For medical researchers, animal research offers a steady income and a successful career pathway regardless of whether, as in the field of inflammation, experiments deliver practical benefits to patients. These are just some of the entrenched interests maintaining the animal testing system. Other prominent beneficiaries include the food and chemical industries which profit from the public perception of safety derived from animal testing.

Going back to the time of ancient Greece, we have used animals to teach us about the human body; however, it was not until 1937 — after 100 people died from taking Elixir Sulfanilamide — that Congress mandated drug safety testing on animals. Since then, literally billions of mice and other mammals have been sacrificed in a Faustian bargain—that their suffering was preventing human experimentation. Seemingly, that calculation was misguided from the start.

The failure of animal experiments to predict human responses and the inability of alternatives to replace them leaves few options. Individuals can to a limited extent protect themselves through avoiding packaged, processed and non-organic food and buying goods made from traditional materials. But ultimately, chemical exposure and chemical pollution are a collective responsibility.


Dressman HK et al, 2007. Gene expression signatures that predict radiation exposure in mice and humans. PLoS Med 4:4.
Greek CR, Swingle Greek, J (2003). Specious science: Why Experiments on Animals Harm Humans.  The Continuum International Publishing Group, Ltd, London.
Knight A (2007) Systematic reviews of animal experiments demonstrate poor human clinical and toxicological utility. ATLA 35: 641-659.
Mestas, J and Hughes, CCW, (2004) Of mice and not men: differences between mouse and human immunology, The Journal of Immunology, 172: 5.
Seok, J Shaw Warren, H et al, (2013) Genomic responses in mouse models poorly mimic human inflammatory diseases. PNAS February 11, 2013 online edition.
Stoloff L (1992) An analysis of the 1987 list of IARC-identified human carcinogens and the correlated animal studies. Regulatory Toxicology and Pharmacology 15: 10–13

Postscript (added May 8th)

Readers may find it useful to get an idea of the prior debate over concordance. Below are some of the scientific papers that have debated concordance. We covered this paper (Seok et al. 2013) because we believe it exemplifies a pattern and not so much because it is new.

David Horrobin (2003) Modern biomedical research: an internally self-consistent universe with little contact with medical reality? Nat Rev Drug Discov. 2: 151-4.

P Pound, S Ebrahim, P Sandercock et al. (2004) Where is the evidence that animal research benefits humans? BMJ.  328: 514–517.

A good place to gain access to this literature is at http://www.afma-curedisease.org/


If this article was useful to you please consider sharing it with your networks.

Print Friendly, PDF & Email
Comments 8
  • This is disquieting, but not entirely unexpected. So what do we do? Are there new thoughts on risk assessment for GMOs?

    • Aruna
      There seems no good solution, at least in the short term. Animal testing (for chemicals) isn’t supported by the evidence in the sense that it’s not anywhere near reliable enough. Sometimes animal experiments give similar results, sometimes not at all, and for toxicology that is no use. On the other hand there are potential alternatives, but they have not been validated. Indeed, they may never be validated. A ‘lung cell’ on a chip is not a lung and it may not behave like one. We should also remember that for animal testing, even if it were 100% reliable, many endpoints need to be examined for it to reliably protect us, and the environment (and they are not). Testing also needs to be removed from the hands of corporations or commercial testing agencies who seem mainly (there are exceptions) to want to please their regular customers and not the public. Concordance, therefore, is only one of many problems within the wider system. All of them have to be working well if the public is to be protected, and at the moment none of them are.
      These arguments all apply to GMOs as well as chemicals, but GMO testing on animals has the added problem that one can’t administer high doses of eg GE tomatoes, as one does with chemicals, to try to generate evidence of toxicity in a small experiment.
      One can hope that new and better methods will be developed but they will also need to be applied retroactively, to products already on the market. Again, this is only one problem in the bigger political problem, which is that our economy is now designed around these chemicals, which no one knows are safe. GMOs can easily be banned without negative ramifications. Banning all known xenobiotics tomorrow would cause the economy to collapse.

      • I should also add, since you may be thinking about evidence of harm from GMOs, that (in toxicology) uncertainty resulting from lack of concordance affects positive results rather differently from negative results, so long as you believe in the precautionary principle. (Anyone who looks before crossing the road believes in the precautionary principle). If a researcher finds a result of “no harm” in a rat but that is only 50% likely to apply to humans, it is not very reassuring as a finding on which to base an approval. However, if a researcher finds evidence of harm (eg the recent Seralini study), then 50% concordance is not likely to affect your view very much (you probably still wouldnt want to eat the product). 50% concordance has been commonly found. However, this new study, finds concordance to be much lower than 50% in that gene expression similarities between mice and humans were close to random. Random presumably would be 0% concordance. There is, by the way, no precise definition of concordance or scale on which it is measured. Presumably that’s because (within science) no one likes to talk about it much.

  • Good Night

    Physicians Committee for Responsible Medicine (PCRM)

    is nota a medical association.

    Yet, a Animal Rights group linked to PETA.

    This claim is Bad science.

    • What about all the studies the article cites to prove its claims?
      Also, modern animal testing has been going on since the 1800’s, and with a vengeance since the end of World War II. Where are all the cures for all the diseases?

  • This morning I’ve been reading about how laboratory mice get somewhat fat and inactive and for that reason are not necessarily good pharmacology/behavioural models. It was suggested that this was due to ad libitum feeding (which requires less human intervention), and because there’s nothing for them to do in their cages.

    This, coupled with the fact that having accidentally confronted images of burnt mice I’m in a situation of temporary desensitisation to the idea of deliberately burning animals, I’ve come back to finish the article. It was interesting and important. Five years ago I was reading into allergy – I remember talking with an allergy researcher who confirmed the need to be mindful about particular reverse responses in mice compared to humans when I was reading papers.

    I looked at Seok very quickly, but spent some time on the correlation table (Fig1). I wondered about the human-human high correlations compared with the mouse-mouse exceptionally low correlations. Why would that be?

    Recognising the animal testing industry mentioned in this article, and seeing the J. Pires comment above, I thought there would probably be criticisms of the Seok paper. There were two letters and replies but they weren’t on open access. Google scholar listed 137 citations of the paper – looking at a sample of open access papers it seems the findings of this study are being rapidly included into planning and outcome considerations. I did see a comment that there was a closer human-mouse correlation between main response genes.

    Thinking about the GM company dossiers containing acute toxicity tests and short term chronic toxicity tests, about all the studies that report on inflammation et al variables in GM and control fed animals, suddenly this study is not just interesting but quite big. It’s as though I’d have to put a ‘however’ statement after every study. Because Seok could impact on so much I want to know its exact specificity and what it could mean. [sigh]

  • Animal testing has shown very little actual proof of being effective enough. It is pathetic to say in a lot of studies, flipping a coin would be more accurate. Animals are not humans, they have different bodies built for different things, and do not deserve to be harmed in any case. Humans are no better than animals. We have greater intelligence, and not much else. Other animals are better than us in different ways. We do not own them, nor this planet, just because we can build cars.

    Yes, human lives do matter, but not at the expense of other lives.

  • is this article peer reviewed?

Leave a comment