In my view, the questionable use of animals in toxicology studies for the regulation of devices, medicines and agrichemicals is more of a concern than the inappropriate use of animal models in research (see I. A. S. Olsson and N. H. Franco Nature 523, 35; 2015).

Testing animals' reactions to commercial products still serves all too often as a formality, rather than as a considered attempt to evaluate hazards to people or the environment (see, for example, T. Hartung Nature 460, 208–212; 2009). Studies using rodents for their full lifetime continue, despite evidence that 90-day tests have the same predictive value (S. M. Cohen Toxicol. Pathol. 38, 487–501; 2010).

Irreproducibility in regulatory studies is a major problem that makes risk prediction unreliable (C. Berry Toxicol. Res. 3, 411–417; 2014). This, combined with a tendency to invoke a precautionary approach in identifying putative hazards from poorly designed regulatory studies, has encouraged adherence to an established framework of testing that has stultified thinking about experimental design.

Olsson and Franco suggest that animal models are more acceptable in research if the results are relevant to humans. That is not the case in much of regulatory toxicology — a huge consumer of laboratory animals.