Abstract

The use of animals in research depends on a favourable harm-benefit analysis. The 3Rs (replace, reduce, refine) serve to minimize the harm imposed on research animals. Unless the results of a study are scientifically valid and reproducible, however, the animals may be wasted for inconclusive research, no matter how little harm is imposed on them. Recent evidence indicates considerable risks of bias across at all levels of animal research, including the design, conduct, analysis and reporting of experiments. Thus, studies may be based on samples that are too small or idiosyncratic; they may lack independent replicates, violate good research practice (e.g. randomization, blinding), and use inappropriate statistics; or they may report results selectively or not at all (publication bias). All of this may compromise the scientific validity of results reported in the literature, thereby exacerbating the so-called “reproducibility crisis”. I will give an overview of risks of bias and measures to avoid them, and point out areas for future research.