Why Specify Best Practices? The Case for Anarchy

• Activity Number: 257
• Type: Invited
• Date/Time: Tuesday, August 10, 2004 : 2:00 PM to 3:50 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Janet Wittes*+
• Companies: Statistics Collaborative, Inc.
• Keywords: Data Monitoring Committees ; clinical trials

Abstract

Bad practice in Data Monitoring Committees (DMCs) is easy to spot. Unclear tables, inaccurate data, biased reporting, failure to provide adequate timely information about serious adverse events, or obvious conflicts of interest quickly provide clues to the members of an experienced DMC that the information they are receiving may lead them to nonoptimal decisions. Thus, a first good step in a search for best practices is a statistical version of our medical colleagues' rule: first, do nothing wrong. Good practice is also easy to recognize ("I know it when I see it", as Justice Stewart said in another context). The search for "best practices," however, may be elusive--what is best depends on the disease under study, the state of knowledge of the biology, and the culture of the medical specialty. This paper explores a series of basic principles that should help guide the newcomer to DMCs. It takes the view that many quite different practices are acceptable statistically and ethically as long as all parties involved understand the purpose of the committee and give each other the latitude to perform their responsibilities efficiently.