News - May 31, 2018

‘You even have to repeat animal experiments’

Tessa Louwerens

Nadia Vendrig understands that people want animal experiments kept to a minimum. But she sees a problem too. Animal experiments tend to produce small data sets in which chance plays a big role. Repeat studies are therefore essential, believes the PhD candidate.

11-portretjes Stelling.jpg

Nadia Vendrig graduated with a PhD on 25 May for her study of statistical methods of analysing animal behaviour data. 

Proposition: ‘The emphasis on novelty of research hampers scientific progress, also in research involving animal experiments’

‘It is a condition for applications for research approval by the Animal Experiment Committee that the research should be new. That sounds logical because if we already know something, there is no need to repeat that experiment. You don’t want to use any more lab animals than necessary. But in animal experiments too, statistically significant effects can come about by chance.

It is good to use no more lab animals than strictly necessary, but that does give experiments less statistical weight. It only takes something to go wrong – two rats falling ill, or something like that – and the group is too small to draw firm conclusions. Many other factors play a role too. For example, it could be that the results are only valid for that one mouse strain at a particular age.

If you don’t repeat that study, it will also take longer before you find out that something is not right. Meanwhile, follow-up experiments are being done, so they are based on false results. The pressure to publish plays a role in this too. If you don’t find any correlations, your results don’t usually get published. That makes people go on looking until a statistically significant correlation turns up somewhere.

The aim of my PhD research was to optimize the statistical analysis of automated home-cage experiments. In these experiments, animal behaviour is recorded by cameras placed in their own cages. The advantage of this is that you can collect much more data per animal than with traditional animal experiments, in which the animals have to be moved for the experiment and then observed by people. This increases your chances of demonstrating an effect without having to use more animals.’