Reproducibility of preclinical data: one man's poison is another man's meat

Abstract

Limited reproducibility of preclinical data is increasingly discussed in the literature. Failure of drug devel-opment programs due to lack of clinical efficacy is also of growing concern. The two phenomena may share an important root cause — a lack of robustness in preclinical research. Such a lack of robustness can be a relevant cause of fail-ure in translating preclinical findings into clinical efficacy and hence attrition, and exaggerated cost in drug develop-ment. Apart from the study design and data analysis factors (e.g., insufficient sample sizes, failure to implement blind-ing, and randomization), heterogeneity among experimental models (e.g., animal strains) and the conditions of the study used between different laboratories is a major contributor to the lacking of robustness of research findings. The flipside of this coin is that the understanding of the causes of heterogeneity across experimental models may lead to the identification of relevant factors for defining the responder populations. Thus, this heterogeneity within preclinical find-ings could be an asset, rather than an obstacle, for precision medicine. To enable this paradigm shift, several steps need to be taken to identify conditions under which drugs do not work. An improved granularity in the reporting of preclini-cal studies is central among them (i.e., details about the study design, experimental conditions, quality of tools and rea-gents, validation of assay conditions, etc.). These actions need to be discussed jointly by the research communities in-terested in preclinical data robustness and precision medicine. Thus, we propose that a lack of robustness due to the heterogeneity across models and conditions of the study is not necessarily a liability for biomedical research but can be transformed into an asset of precision medicine.