Statistical Computing Environments and the Practice of Statistics in the Biopharmaceutical Industry

A. Hopkins,Susan P. Duke,S. Dubman
DOI: https://doi.org/10.1177/009286151004400104
2010-01-01
Abstract:A structured statistical computing environment (SCE) enhances rigor in operational implementation of statistical analyses of clinical studies through process transparency, allowing reproducibility of results by independent reviewers. Desirable features and associated benefits of an SCE system are described. Minimum SCE requirements discussed in detail consist of a structured programming environment, an operational analysis data repository, and a metadata-driven architecture containing information about data and status of various processes. The metadata provide a foundation for connecting multiple processes and systems, thereby allowing the creation of tools that largely automate the analysis process. Standards drive productivity enhancement for creating statistical deliverables based on metadata obtained from the development plan, protocols, and analysis plans. Not all the features discussed are available today in commercial systems. In the future, nearly all information about clinical trial analytics can be driven by a standards-based, metadata-driven architecture. To accomplish this goal, metadata need to be available about all the processes used to collect, transform, and analyze the patient data. Further standards development will be necessary to fully describe the entire statistical analysis process. Learning Objectives Upon completion of this article, participants should be able to do the following: Describe the elements of good statistical practice that contribute to establishing the credibility of clinical trial results. Describe the fundamental concept of statistical computing environment (SCE); the SCE as a programming environment; and as a clinical data platform and repository driven by a metadata architecture. Target Audience This article is informative for medical doctors working in the pharmaceutical industry; biostatisticians, statistical programmers, clinical data managers, and IT professionals.
What problem does this paper attempt to address?