Why does drug development using animals need a rethink?
I think we’ve got ourselves into a mess right now, with lab mice in particular. The benefits to humans, in certain diseases and mouse models, have shrunk to such low levels that it’s time we found better ways to work with animals in medical research.
How have the benefits to humans shrunk?
Advertisement
Of the drugs that get past the animal testing phase and into human trials, only about 1 in 9 actually make it to the market, and that’s dropping all the time. It costs about $2 billion to bring a single drug to market, largely because of failed human trials. And they usually fail simply because the drug doesn’t work, or not as well as animal testing predicted.
Why isn’t research translating well from animals to humans?
The idea is that the animal model will tell you something about how this drug is going to perform in humans. But we have to ask ourselves, are these animals truly modelling the human disease? Increasingly, they are not. So we end up learning a great deal about how mice respond to various compounds, but it’s irrelevant to humans and an enormous waste of money.
We need to do animal research that actually translates well. A knock-on effect of that will be that these drugs fail earlier in the development process, before the cost of failure skyrockets.
So what’s the underlying issue?
There are several, but in drugs for cognitive conditions, the single biggest issue is the use of the word “like”. We describe mouse models as OCD-like, autism-like, Alzheimer’s-like, and so on. This happens for well-intentioned reasons: scientists are flagging up that they’re not sure whether these traits in their animals really represent the condition in humans. But it only takes four or five other research groups to follow suit and, lo and behold, although we still don’t know if this OCD-like mouse model usefully represents OCD, you are not allowed to point that out any more – the emperor has no clothes.
Taking autism as an example, why doesn’t research in mice translate well?
To be diagnosed with autism, you have to show some behaviours and, crucially, not others. It’s largely a diagnosis of exclusion: kids with autism have problems with imaginative social play; they have real trouble with theory of mind – that is, being able to understand the private emotional lives and knowledge of other people – and they often don’t make eye contact.
Now, it’s pretty hard to figure out whether a mouse is making eye contact, there is no behaviour in mice that we really understand as being play, and mice are not on the very short list of species shown to have a theory of mind. So it’s bonkers to think that the social issues of autism are measurable in mice, yet researchers claim to be finding them. Put simply, you can’t model the absence of something in a species that doesn’t have it in the first place.
How has genetic engineering altered our animal models?
We see researchers “disorder-shopping”: they genetically engineer a mouse with an unusual behavioural or physical trait, then tout it around to researchers until someone bites. For example, there’s a GM mouse model called D1CT-7, which has been heavily marketed. These mice have spontaneous tics and they are classed as Tourette’s-like, not to mention anxiety-like and OCD-like. Yet it isn’t a true model of Tourette’s syndrome because the part of the mouse brain causing its Tourette’s-like behaviour is not the part of the brain that is central to Tourette’s in people.
Do you take issue with experimental methods more generally?
Certainly. Just think about what it would mean if a human experiment were comparable to a mouse experiment. We would be using a population of nothing but 42-year-old white males that live in identical ranch homes in some small town somewhere with identical diets, identical wives, identical children, identical furniture; they eat the same thing every day for every meal, the thermostat is locked, and the gardener is an extremely scary-looking Tyrannosaurus rex that pulls the roof off their house once a week and destroys their Facebook account and all of their social information. Just take it as far as you want to go because it is such a crazy way to design an experiment that’s meant to shed light on humans.
Good experimental design involves controlling for confounding factors, but the only place where we try to do that by making every individual identical is lab animal work, and particularly mouse work. These are Stepford mice! It’s the only field of biology or psychology where we don’t recognise that variation is what makes us special, what makes us interesting – and what makes us ill.
“It’s bonkers to think that the social issues of autism are measurable in mice“
How has this line of thinking changed the way you conduct your own experiments?
It’s little use manipulating animals, by genetic engineering or otherwise, so that they simulate the symptoms of conditions they don’t really have. So what’s the antidote? The secret is to really start looking for animals that spontaneously get ill or display unusual behaviour, and then ask why. My collaborative research on autism has led us to work with very, very large colonies of free-living primates, identifying the socially unusual individuals and asking, are these individuals behaving in ways that make sense in terms of autism? Are they showing the same biological markers that we have found in children with autism? The answer is “watch this space” – I can’t give specifics until the work is published.
What about the work you do with mice?
Mice are furry little time machines. I can actually follow a mouse from before its conception through to its death in a year or two, and I can see which animals become spontaneously ill and which ones don’t, and measure the processes that I think might be involved. That illness might be diabetes, mammary cancer, compulsive hair-pulling, or whatever condition I happen to be interested in. Because the minute you start looking for spontaneous illness in this way, you’re suddenly just in the world of human research methods, your animals are happier, research is more likely to translate to humans, and a lot of the problems associated with animal research simply disappear. That’s good science, good animal welfare, and good business too.
Profile
Joseph Garner at Stanford University, California, with a focus on why most drugs (and basic science findings) fail to translate to human outcomes
This article appeared in print under the headline “The autistic mouse and other unlikely stories”
