Good research practice in non-clinical pharmacology and biomedicine

Abstract

Pharmacologists and other experimental life scientists study samples to infer conclusions about how a molecule, cell, organ, and/or organism work in health and disease and how this can be altered by drugs. The concept of inference implies that whatever is reported based on the sample under investigation is representative for the molecule, cell, organ, or organism under investigation in general. However, this generalizability requires that two fundamental conditions are met: First, what is being reported must be a true representation of what has been found. This sounds trivial, but if data are selected, e.g., by unexplained removal of outliers or reporting is biased by focusing on the findings in support of a hypothesis, reported data become a biased rather than a true representation of what has been found. Second, what has been found must be robust, i.e., other investigators doing a very similar experiment should come up with similar findings. This requires a complete reporting of what exactly has been done. It also requires that biases at the level of sampling, measuring, and analyzing are reduced as much as feasible. These are scientific principles that have been known for a long time. Nonetheless, scientific practice apparently often ignores them—in part based on felt pressure to generate as many articles in highprofile journals as possible. While the behavior of each participant (investigators, institutions, editors, publishers, and funders) follows an understandable logic, the result is counterproductive for the greater aims of scientific investigation. Against this background, this volume in the series Handbook of Experimental Pharmacology discusses various aspects related to the generation and reporting of robust data. It is published 100 years after the first volume of the series, and this anniversary is a fitting occasion to reflect on current practice and to discuss how the robustness of experimental pharmacology can be enhanced. It has become clear in the past decade, that many if not most preclinical study findings are not reproducible. Governmental and nongovernmental funding bodies have realized that and demand that the scientific community improves its standards of scientific rigor. Particularly, the use of experimental animals in biomedical research creates the ethical imperative that they are utilized in robustly designed studies only.