Flaws in use of animal tests for new drugs
By Clive Cookson in York
September 14 2007
The use of animal tests in drug development is fundamentally flawed, according to new evidence presented at the BA (British Association for the Advancement of Science) Festival of Science in York on Friday.
As a result, many animal studies "overstate how effective drugs really are", said Malcolm Macleod, a consultant neurologist at Edinburgh university. These drug candidates then fail to work when given to humans in clinical trials - and could even put patients' health at risk. The main issue is experimenters' unconscious bias. Researchers do not routinely divide animals at random into treatment and control groups and, when they analyse results, they know which animals have been treated and which have not.
Dr Macleod heads an international consortium called Camarades, which analyses animal testing of treatments for stroke. This condition makes a good case study, because there is an appropriate "animal model" that mimics human stroke in rats and mice, and because researchers have published many comparable animal studies of the same stroke treatment.
"Over 800 drugs have been tested in animal models of stroke and about 500 of these seem to work in animals," said Dr Macleod. "Nearly 100 of these have been tested in people and - with one exception (tPA) - they don't work in patients with stroke."
Researchers who "randomised" animals before treatment - and were "blinded" to the allocation of animals - reported much less favourable results than the majority who failed to design their experiment in this way. A recent study of 100 animal studies of six stroke treatments found that only 36 per cent were randomised and 11 per cent were blinded. Yet randomised, blinded trials are routine in clinical trials with humans.
Dr Macleod also reported the results of an investigation into NXY-059, AstraZeneca's experimental stroke treatment. The company dropped NXY-059 when it showed no benefit in a clinical trial involving more than 3,000 patients, after very promising animal studies.
Overall, NXY-059 improved outcome in animals with stroke by 44 per cent, Dr Macleod said: "However, when we looked in more detail, disturbing patterns appeared."
For example, studies that did not randomise animals said NXY-059 improved outcome by more than 50 per cent; those that did not estimated the effect at only 20 per cent. Studies that did not blind the assessment of outcome said it improved that outcome by almost 50 per cent; those that did not estimated the effect at less than 30 per cent.
"We have reported similar findings for other interventions, but what is disturbing about the data for NXY-059 is that for a drug where most of the published work was funded by the drug manufacturers the impact of poor study quality was much more pronounced," said Dr Macleod. He emphasised that he was not opposed to animal experiments - indeed he uses them in his own work - but wanted them to be designed to the same high standard as human clinical trials.
Derek Fry, a senior Home Office animal inspector, and Simon Festing, head of the Research Defence Society, agreed there was scope to improve experimental procedures.
"Those of us defend animal research accept that there is a problem here," said Dr Festing.