Ray Greek has been held up by the animal rights community as a standard bearer for the “animal research doesn’t work” movement. While his arguments appear credible at first glance, they quickly fail under scientific scrutiny. In this post we take a broad look at the arguments made in his key work, Animal Models in the Light of Evolution.
Those willing to play dirty in the service of animal rights activism have employed many tactics over the years, from breathless but groundless “reports” describing the ‘barbarism’ of the animal lab to threats of physical violence and beyond.
However, the modern era has been marked by the rise of pseudoscience which argues from a perspective that animal research is ineffective, rather than immoral. It is often underpinned with an assertion that there are “modern alternatives” that should be used instead, if only scientists would take off their blinkers!
The response of the scientific community thus far has largely been to refute such claims via sending papers to what are (to the public) obscure scientific journals, or maybe a stiff letter to Scientific American. For instance Adrian R. Morrison, writing in Perspectives in Biology and Medicine, Volume 45, Number 4, Autumn 2002 illustrates how animal rights pseudoscientists misuse references, often quoting passages out of context so they contradict the wider conclusions of a paper.
However, we must do more to argue openly and publicly against these scientific distortions, and it might be helpful to first consider what elements go into a misleading proposition.
Perhaps a good place to start is by looking at some of the work of Ray Greek, who will be familiar to readers of this site, but has just started manifesting in the UK with his work used as the basis of a Parliamentary Motion and as the “scientific” rationale behind pseudoscience organisations like For Life on Earth.
There are two aspects here. Firstly, Greek is careful to qualify his work as being limited in its scope, although even he must admit that some of his sources aren’t used honestly. Secondly, the narrow focus of Greek’s work is often extrapolated across all animal research by protesters, so although he argues, along with Niall Shanks, in Animal Models in the Light of Evolution (AMLE) that:
“We are primarily concerned with the practise of using animals as predictive models of human and biomedical phenomena. […] [However], there can be no doubt that studies of animals have contributed greatly to our scientific understanding of life, and there is little doubt that these studies will continue to illuminate these matters in the future” (page 30)
Their intended audience has clearly misunderstood their claims and extrapolated their narrow evidence base to all animal research. So, it’s arguable that their harmless (if lucrative) enterprise is becoming dangerous in the same way that the novel of Frankenstein is harmless until it inspires an angry mob to burn down a lab, or, more likely, an undereducated politician influenced into making a silly decision.
Greek and Shanks – Animal Models in the Light of Evolution (2009)
Of course, animal models in most cases aren’t being used to “predict” anything. They are being used to test a hypothesis (in which sense there is no such thing as a failed experiment – only an incorrect hypothesis) or to discover or understand a process. Such descriptions cover much basic research, as well as many other areas. Greek acknowledges that there are valid uses of animals and instead focuses on animals as models of human biology, particularly drug development and safety testing.
Where animals are used in drug testing it is to suggest that an effect observed in animals may also be observed in humans. When an effect is observed across multiple mammalian species, the probability of it also having that effect in humans can be sufficiently high to allow the compound to proceed to human trials. It should be noted that the system is primarily screening for acute undesirable effects (such as severe vomiting or seizure) which would prevent a drug candidate from proceeding to human trials. The tests cannot possibly predict all adverse drug reactions, and many will only be known about when a compound is administered across a large number of human patients across the globe. Nor are they intended to – they are intended to prevent the illness or death of phase 1 clinical trial volunteers – something animal tests achieve remarkably well.
Prediction in this context is about probability. An analogy might be if I told you that if you walk into the road right now you have a 9 in 10 chance of getting run over. Note I am not saying you will be run over, but there’s a good chance you might be. This is how ‘prediction’ is understood in this context, not as a form of precognition or the outcome of a complex model where all variables are known.
Initially, then, the enterprise has been to frame the concept of prediction in a particular way which will of course make it unachievable. What comes next is mathematically correct but largely irrelevant – a tactic that crops up again in his discussions of complexity[i]. This is in turn complemented by cherry-picked references that seem to support the argument, in many cases from papers describing out-of-date and far less effective toxicology than we have today.
Elsewhere, AMLE’S companion “FAQs about the use of animals in science” (FAQ) quotes heavily from “Animal Toxicity Studies: their relevance for man”, an out-of-print book of essays from 1990 in which a selection of toxicologists consider how toxicology models could be improved. Greek gleefully cherry-picks from their examples, which are necessarily focused upon examples where the models of the day weren’t working well. Nonetheless he regularly ignores the overriding points that they make – that animal research remains important and that the animal models need improving not removing. Greek is happy to support his case with quotes such as:
“In one small series in which the toxicity in clinical trials led to the termination of drug development it was found that in 16/24 cases toxicity was not predicted in animals  “
 Lumley C: Clinical toxicity: could it have been predicted? Premarketing experience. In Animal Toxicity Studies: Their Relevance for Man. Edited by Lumley C, Walker, S: Quay; 1990: 49-56.
This appears on Page 53 of his FAQ (and his marketing excerpt of the book), but curiously does not appear in either the chapter of the book quoted, nor anywhere else in Lumley’s (et al.) book. On the other hand the chapter of Lumley’s book that Greek points to does include:
“Dayan, stated that the rarity of predictive toxicity in man demonstrates the success of the animal toxicity procedure at preventing toxic compounds from reaching the clinic. The pilot study confirms earlier findings that only a very small proportion of compounds evaluated in volunteers and/or patients are terminated due to clinical toxicity.”
So there are a few red flags here: old or irrelevant references, suspect definitions, cherry-picked data, a certain amount of “blinding them with science”. These effects are compounded since neither his Animals Models in Light of Evolution, nor FAQs about the use of animals in science are peer reviewed (since they are books not journal submissions).
Arguing that animal models are not predictive because life is complex, when there hasn’t been a single death during a phase 1 clinical trial in the UK in over 30 years (and only one serious incident, due to human error), is a bit like arguing that airplanes can’t fly whilst being seated on an airplane at 30,000 feet.
As far as I can see, features of Greek’s pseudoscience include:
- Narrowly defining the terms (probability = strict definition of prediction, now).
- Limiting the data set – Greek is not alone here, animal rights activists like Victoria Martindale have created myths entirely out of limiting the dataset.
- Quoting out of context – Discombobulated quotes from old papers are common, especially when the paper is making recommendations on how to improve the quality of animal research (similar tactics used by Akhtar when quoting Macleod)
- Demonstrating how the limited data set doesn’t satisfy the narrow definition.
What is most troubling is it’s hard to tell if such writers have got a point, or could contribute usefully to refining experiments or reducing animal use, because their arguments are wrapped up with poor evidence, geometry, vague or invalid sources of information, faulty statistical methodology and easily-disproved claims. I have written elsewhere about one of the citations used by Greek being based on a “very rough estimate” of the predictive value of toxicity testing of 45 random drugs in 1978. It’s time to stop the mischief, stick to appropriate methodology and, when the science still supports animal research, attempt to criticise animal experiments from an “ethical” perspective (although it’s a slim ethical argument at best).
There is no great conspiracy on the part of pharma companies and regulators to use animals unnecessarily, and this patronising notion that scientists are so thoughtless, institutionalised or psychotic that they use animals when there are easily-accessed alternatives shows the highest arrogance. If animals are poor models for humans, then let it be proven properly and scientifically, hand in hand with regulators and industry, neither of which have an interest in animal research continuing unnecessarily. Regulators are using well-established statistical models as diagnostic tools, and if they’re wrong then it should be possible to prove it to them without statistical trickery. Otherwise the intellectually honest thing to do is abandon the claim.