The Role of Statistical Thinking in Biopharmaceutical Research

AbstractAbstractThe development of new drugs has evolved dramatically over the past decade. Advances in technology enable scientists to generate "big data" faster than ever before. The availability of complex, high-volume data in turn creates demand for innovative quantitative solutions and tools in a rapidly evolving landscape. As a result, the role of the statistical scientist in collaborative research has never been more important. Reflecting on these changes, Cox (2012 Cox, D. R. (2012), "Comment on "Cornfield J (2012) Principles of Research"," Statistics in Medicine, 31, 2770. DOI: 10.1002/sim.5375.[Crossref], [PubMed], [Web of Science ®] , [Google Scholar]) wrote, "…[A]lthough the tactics of statistical analysis have been utterly changed… the strategy of research design and analysis has been much less affected…" In this article, we argue that the practice of statistics is built on the foundation of good statistical thinking and consists of a complex combination of problem-solving skills, the essence of what Cox meant by the "strategy of research." Although others have highlighted the role of statistical thinking in research design and analysis, in the age of data science, machine learning and artificial intelligence, it cannot be emphasized enough. We outline four general steps that contribute to good statistical thinking and illustrate them with five use cases ("vignettes") as well as a detailed case study discussion from a maintenance therapy clinical trial for depression.Keywords: Drug developmentGood statistical practiceInnovation cyclesStatistical sciences AcknowledgmentsWe thank the organizers, Bo Huang and Gene Pennello, for inviting us to present at the BIOP2021 workshop. We also thank two referees and the Associate Editor for their useful suggestions which greatly improved the manuscript. We are grateful to Mark Baillie, Lei Nie and Susan Mayo, who provided helpful comments on a previous version of this manuscript. We are also grateful to Mark Baillie, Robin Dunn, Enrico Ferrero, Allison Florance, Prasanti Goswami, Nathalie Fretault, Artem Gavryk, Malcolm Martiatu Franco, Peter Mesenbrink, Paul O'Connell, Konstantinos Sechidis and Marc Vandemeulebroecke for sharing and discussing with us the case studies in Section 5.Additional informationFundingThe author(s) reported there is no funding associated with the work featured in this article.