Friday morning at the BA Festival of Science in York saw a whole morning devoted to improving the robustness, relevance and transferability of data from animal experiments which are intended to evaluate potential clinical interventions.
Interesting points were raised by several speakers, particularly about the importance of properly designed experiments. Not only should sample sizes be correctly calculated to ensure that the data produced are statistically significant, but randomisation(1) and double-blinding(2) should be standard practice.
Investigations into the published literature on stroke showed that the majority of studies have not fulfilled these criteria, and it is likely that other areas of research are similarly afflicted.
This is similar to the state of human clinical trials 10-15 years ago, and we can only hope that experimental design is improved as quickly as possible. Recognition of the issues is the key first step. The speakers at the event are doing the vital job of highlighting these problems, whilst stressing exactly what this does and does not mean.
Just as historical bad design did not render the concept of clinical trials as being flawed, the speakers noted that this did not mean that animal models were not a valuable tool in biomedical research:
"This isn’t to say that animal models of stroke are bad… it’s that we can do them better and we must make substantial efforts to improve the quality and reporting of those studies.”
- Dr Malcom Macleod, CAMARADES
Dr Ian Roberts, who spoke on the important of statistical analysis, recognised in a BMJ article that analysis of information is currently limited to stroke research. Based on such limited information he wrote,
"it would be inappropriate to make general statements about the value of animal research"
RDS supports every effort to improve the quality of data gained from animal studies.
-------------------------------------------------
(1) Where allocation of which animals get the trial substance and which get the placebo is randomly determined.
(2) Where those giving the substances as well as those monitoring the results do not know which animals have received the trial substance or placebo.
