Reducing versus Embracing Variation as Strategies for Reproducibility: The Microbiome of Laboratory Mice

SIMPLE SUMMARY: The rate at which biomedical animal experiments fail to reproduce expected results is of raising concern in the animal research community. One of the explanations of irreproducibility is that animals used in repeated experiments are not identical to the original animals: there is variation in the animals’ bodies, even if they are genetically similar. For example, there might be substantial differences between the microorganisms inhabiting the experimental groups of mice. For animal researchers, it is essential to know how to deal with this variation in animal experiment design. Both reducing and embracing variation are suggested as strategies for reproducibility. In this theoretical review, we use mouse microbiome variation as an example to analyze this ongoing discussion, drawing on both animal research and philosophy of science. This analysis helps articulate options for animal researchers to deal with variation in the future design of animal experiments, contributing to reproducibility. ABSTRACT: Irreproducibility is a well-recognized problem in biomedical animal experimentation. Phenotypic variation in animal models is one of the many challenging causes of irreproducibility. How to deal with phenotypic variation in experimental designs is a topic of debate. Both reducing and embracing variation are highlighted as strategies for reproducibility. In this theoretical review, we use variation in mouse microbiome composition as an example to analyze this ongoing discussion, drawing on both animal research and philosophy of science. We provide a conceptual explanation of reproducibility and analyze how the microbiome affects mouse phenotypes to demonstrate that the role of the microbiome in irreproducibility can be understood in two ways: (i) the microbiome can act as a confounding factor, and (ii) the result may not be generalizable to mice harboring a different microbiome composition. We elucidate that reducing variation minimizes confounding, whereas embracing variation ensures generalizability. These contrasting strategies make dealing with variation in experimental designs extremely complex. Here, we conclude that the most effective strategy depends on the specific research aim and question. The field of biomedical animal experimentation is too broad to identify a single optimal strategy. Thus, dealing with variation should be considered on a case-by-case basis, and awareness amongst researchers is essential.