A new way to design experiments could reduce the overall number of animals used in biomedical research, a German study has suggested.

Calculating the exact number of animals needed in a research study – the sample size – is usually worked out using historical data, but the article in LabAnimal argues that this can often mean either too many, or too few animals are used, leading to ureliable results.

In the article, Helene Richter, of the department of behavioural biology at the University of Münster, Germany, believes there is increasing evidence that animal sample sizes estimated by researchers are actually too small for any valid conclusions to be made. 

“The consequences are twofold: first, many studies are dramatically underpowered in animal research, involving too few animals per group to correctly identify true treatment effects.”

In a challenge to scientific practice, Richter calls instead for a ‘mini-experiment design’ approach combined with Bayesian statistical theory that splits an experiment into several parts and allows analysis at each stage to determine the optimum number of animals that should be used next.

While the new approach might lengthen research times and require training, Richter believes it could reduce overall animal use, increase the validity of the studies, and improve the chances of an experiment being reproduced accurately in the future.