BEMP: Methodological Research for Pre-Clinical Studies

Biostatisticians serve as a resource for statistical questions that arise for in vitro and animal studies regarding study design, data analysis, and interpretation of results. Most therapeutic advances for human populations are initially tested in at least one animal model. Appropriate design and analyses are crucial for these animal experiments, especially when there is a focus on moving advances rapidly from the laboratory to the clinic. It is particularly important in the context of a translational research program that biostatistical collaboration is involved in the "three Rs" of refinement, reduction, and replacement in animal experiments such that pre-clinical studies are optimally designed to inform future human experimentation. Much of the methodology developed for clinical studies is directly applicable to studies in model systems. The importance of novel study designs to meet the standards for use of animals in research has been recognized in both biostatistics and pharmacology. Rodent toxicology studies that involve the identification of some treatment construct as the "best" among those tested are an example where statistical methodology could reduce the expected number of animals used and improve efficiency of therapeutic development.

Statistical ranking and selection procedures, including gatekeeping methodology, could be used in adaptive designs to quickly screen out less promising therapies. The same paradigm would also be applicable in studies designed to identify therapeutically relevant genetic variants among multiple strains of a particular species. There is a substantial literature on these methods in various fields and they have been applied to the design and analysis of clinical trials. Other methods familiar within clinical trials also have direct application to optimizing animal experiments. For example, common sequential methods and adaptive randomization could be used to minimize expected sample numbers. A translational program that incorporates multiple therapeutic animal models might also be suited to the application of Bayesian methods for both study design and analysis, as discussed further below. In many cases, the challenge will be to adapt the application of these methods to the needs of particular animal studies, addressing issues such as joint housing or limitations on procedures per day. In other cases, additional work is needed to expand these methodologies to issues such as serial sacrifice unique to laboratory experimentation.