Nature. 14 December 2006
Animal studies: a good guide for clinical trials? Study
reveals animal experiments often fail to predict outcomes in
Animal-rights activists have long claimed that differences
between humans and animals make experiments in species such
as rats of little use. Most scientists disagree, saying that
drug development would be impossible without initial tests in
animals (see our Animal research special).
Now a team of medical researchers has published in the BMJ1
results from what they say is the first attempt to produce a
scientific, quantitative answer to this question. The results
provide food for thought for any scientist who works with
The authors say they have highlighted serious problems with
the way in which animal research is translated into human
trials. Only half of the small sample of tests analysed by
the team so far produced the same results in animals as they
did in people.
The team stresses that this is not an argument against doing
animal studies. Even so, the paper is likely to be seized on
The researchers started by taking six sets of clinical trials
that had produced definitive answers as to whether specific
treatments for conditions such as stroke and head injury are
useful or not. They then assessed whether the prior animal
research had given similar results to the human trials. The
results, published today, say that the sets of data matched
in only three of the six cases.
When the team investigated the reasons for this, they exposed
a series of problems with the animal data. Many studies did
not allocate animals randomly to control and treatment
groups, a problem that is known to introduce bias into
clinical trials. Comparison of experiments on a stroke
treatment also suggested that studies showing negative
effects are more likely to go unpublished, skewing
impressions of efficacy.
Work on a model of head injury was undermined by the use of a
model that did not match the later clinical trials. Rodents
were injured and then treated five minutes later, says Ian
Roberts, an author on the study and an epidemiologist at the
London School of Hygiene and Tropical Medicine. In the
clinical trials, which are based on hospital admissions,
patients are typically treated within three hours of being
Although the results could be seen as evidence that animal
work does little to inform clinical studies, the authors
stress that their results show only that more time needs to
spent thinking about how to translate research between the
two spheres. Better-designed models and an awareness of
publication bias are two priorities, they say.
Other researchers question whether the results are really as
alarming as they sound. Robert Lechler, an immunologist at
Kings College London, says that designers of clinical trials
are already well aware of limitations in animal models. He
says that scientists apply an "intelligent filter" when
looking at such tests, and that crude comparisons between the
outcomes of animal and human studies do not capture this.
But Peter Sandercock, a neurologist at the University of
Edinburgh and an author on the study, says the paper shows
that more needs to be done on this front. If such a filter
was applied, he points out, the animal models of head injury
would have been repeated using a better design before human
studies started. "This is not a polemic against animal
research," stresses Sandercock. "But we need to be aware that
there are biases in the animal trials."
BMJ. 16 December 2006,
Just how useful are animal studies to human health?
(Comparison of systematic reviews of animal trials with
Animal studies are of limited usefulness to human health
because they are of poor quality and their results often
conflict with human trials, argue researchers in a study on
Before clinical trials are carried out, the safety and
effectiveness of new drugs are usually tested in animal
models. Some believe, however, that the results from animal
trials are not applicable to humans because of biological
differences between the species.
So researchers compared treatment effects in animal models
with human clinical trials.
They used systematic reviews (impartial summaries of evidence
from many different studies) of human and animal trials to
analyse the effects of six drugs for conditions such as head
injury, stroke and osteoporosis.
Agreement between human and animal studies varied. For
example, corticosteroids did not show any benefit for
treating head injury in clinical trials but did show a
benefit in animal models. Results also differed for the drug
tirilazad to treat stroke - data from animal studies
suggested a benefit but the clinical trials showed no benefit
and possible harm.
Some results did agree. For instance, bisphosphonates
increased bone mineral density in both clinical trials and
animal studies, while corticosteroids reduced neonatal
respiratory distress syndrome in animal studies and in
clinical trials, although the data were sparse.
Animal studies are generally of poor quality and lack
agreement with clinical trials, which limits their usefulness
to human health, say the authors. This discordance may be due
to bias, random error, or the failure of animal models to
adequately represent clinical disease.
Systematic reviews could help translate research findings
from animals to humans. They could also promote closer
collaboration between the research communities and encourage
an interative approach to improving the relevance of animal
models to clinical trial design, they conclude.
Contact: Professor Ian Roberts, London School of Hygiene and
Medicine, London, UK Email: firstname.lastname@example.org