Translational neuroscience requires better design and analysis of preclinical studies

Abstract

Animal models are often used to obtain a better understanding of psychiatric, neurodegenerative, and neurodevelopmental disorders. Despite many years of research, these models have not led to many novel therapies or treatments. Translating results between species will always be difficult, and it is argued that inappropriate statistical analyses, failure to identify the experimental unit, lack of random assignment to treatment conditions, and unblinded assessment of outcomes contribute to the low rate of translating preclinical in vivo studies into successful therapies. It is known that these shortcomings can generate biased estimates, too many false positives and false negatives, and unreproducible results. These issues have been raised repeatedly, but have largely gone unheeded by scientists. Two recommendations are made to improve the situation.