Don’t believe the mice
Experiments using mice are often heavily publicised – but very, very few of them translate into humans. Anthony King reports on why animal models are of questionable value.
When you read that a lab animal with a human disease has been cured with a new drug candidate, do not get your hopes up. The stats for converting these successes into human patients are appalling. Results in animals are often the opposite of those seen in humans. For example: corticosteroids were shown to treat head injuries in animals, but then increase deaths in new-born babies in trials.
This is a big deal. A staggering 95% of drugs tested in patients fail to reach the market, despite all the promising animal studies that precede their use in humans.
“There are lots of reasons why, but in essence we are not 70 kilogram rats and we are not inbred strains,” says Thomas Hartung, a toxicologist at Johns Hopkins University in the US.
Two industry studies showed that many key findings that triggered drug development could not be repeated.
Mice are the most popular lab animals, but their brains and biology are quite different from our own. Surprisingly, rats and mice predict each other for complex measures with only 60%. Different animals, different effects.
Newspapers headlines heralding cures for Alzheimer’s to autism, on the back of rodent studies, can be taken with a pinch of salt. Neurodegenerative disorders such as Alzheimer’s were one of the first areas to turn against the animal models, says Hartung.
“It was shown that the animal tests were misleading with respect to what is a cure and what is not,” he says.
After hundreds of human trials for promising treatments for Alzheimer’s, almost none helped patients.
This is a colossal waste of money. Industry has noticed.
“The pharma industry is now using about one-sixth the number of animals that they used in the past for drug studies,” says Hartung. “They go very late into these models.”
In a look at animal experiments, Hartung and colleagues found that pharma continues to reduce animal testing in Europe, despite rising R&D spend. From a stable 12 million used in Europe, the industry’s share dropped from 31% in 2005 to 23% in 2008, and then to 19% in 2011.
Disease researcher John Ioannidis at Stanford University in California has written that the safety and effectiveness of interventions in humans can only “be speculated from animal studies”.
Speaking at the EuroScience Open Forum (ESOF) in Toulouse, France, earlier this year, he said that “industry doesn’t want to waste money taking academic papers that promise that they have found a drug target and spend billions of dollars to develop it, and then come up with nothing”.
He pointed to just six of 53 landmark studies in cancer being repeatable and lamented that too many basic scientific discoveries are wrong.
One problem is that scientists often take a simple approach to mimicking a disease in mice, by just finding a gene that when knocked out stamps the mice with hallmarks of the human disease.
This is how the first Alzheimer’s disease mouse was created, but the animal did not reflect the true Alzheimer’s condition of most patients.
“Single gene mouse models are different from the illness that we experience in humans,” says neuroscientist Malcolm MacLeod at the University of Edinburgh, UK, who describes mouse models for stroke, high blood pressure, Parkinson’s and more as failing to reflect the complexity of the human disease.
“This has been a failed strategy,” he warns, in terms of finding therapies.
Hartung too has warned about the hype about these genetically modified animals.
Sometimes scientists discover therapies to cure mice, but not people. The record for inflammatory disease is especially striking. More than 150 trials have tested agents to block inflammation in critically ill patients. The candidates worked in animals, but all failed in patients.
With this in mind, Ronald Davis, at Stanford Genome Technology Centre in California, decided to compare how all genes in mice and all genes in people react when they encounter trauma, burns or bacterial toxins. There was almost no connection whatsoever. Mice genes did one thing; human genes did another.
The immune systems of mice and people are that different.
“Mice eat garbage,” says Davis. “Their habitat is extremely exposed to microorganisms that they eat.”
Our immune system is far more sensitive. For example, between five and 25 milligrams of endotoxin, per kilogram of body weight, will kill mice. Ten thousand times less can cause humans to go into life-threatening shock.
Davis initiated the statistical analysis after the journal Nature Medicine rejected a research paper with human results because it did not demonstrate the same effect in mice.
“It was almost as if the focus was in trying to treat mice, not humans,” he recalls. Mouse studies are valuable, but we always need to move to humans, he argues.
“We can cure cancer in mice pretty effectively, but the agents don’t work in humans in most cases,” says Davis. “These are complicated diseases and we live far longer than mice and evolutionarily we are far apart.”
He says many immunologists, who mostly use mice, criticised his findings, but industry shrugged its shoulders.
“The pharma industry said it was obvious,” he explains. “One person said we’ve known this for years, but they didn’t publish it.”
He recommends that science funders should give larger grants to those studying in humans – because it is more expensive. Another issue is that funders measure academic success by counting how many research papers a scientist publishes.
“The yardstick funders use is publications,” he says. “Whether you develop a route to curing a disease is irrelevant.”
He says more data is collected from people now, though, since it is possible to get more and more insight from a blood sample or even just a few human cells. This, at least, is promising.
Another issue is that inbred mice, often all the one age and sex, are usually used for tests.
“People are completely genetically diverse,” says Hartung. “We are different sizes, eat differently, have a disease history. This is not, and cannot, be reflected in animals.”
The animals, thus, only take us so far. Often a company will only realise a drug can cause side-effects in the liver, sometimes in one in 10,000 cases, after it goes into patients.
A final issue with animal studies is how many are carried out, often by trainee PhD scientists or lead researchers looking to publish interesting results. Sometimes outliers in results can be cherry-picked and written about.
“You then build a story of how you logically came to this result, but this is a fairy tale,” says Hartung.
He has argued that the quality of clinical trials in humans is monitored far better than in academia, so that statistics from industry are more reliable. But if the human trial is built on shaky animal experiments, then the trials will fail.
And there is a cost to failure.
“If all the money spent on biomedical research in the last 20 to 30 years had been spent instead on public health, stopping smoking and alcohol control, it would have had a greater impact on the incidence and severity of Alzheimer’s disease,” says MacLeod.