Statistician and Clinician: Crosstalk on Missing Data

• Activity Number: 384
• Type: Topic Contributed
• Date/Time: Tuesday, July 31, 2012 : 2:00 PM to 3:50 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Janet Wittes*+
• Companies: Statistics Collaborative
• Keywords: missing data ; clinical trials

Abstract

When the Institute of Medicine's (IOM) report on missing data in clinical trials was published, the world must have heard a collective sigh of relief from the biostatistical community. At last we could put Last Observation Carried Forward and Completers' Analyses in the dustbin of clinical trial history. In fact, that is how I am using the report: a company or an investigator asks for LOCF and I now respond not with, "I don't do LOCF because." but with, "The IOM says not to use it." And not only did the IOM proscribe some methods, it gave us an officially sanctioned armamentarium of methods to handle missing data. But I struggle with the thought that, while we as statisticians can do analyses that deal with missing data in an unbiased way, perhaps we have defined the allowable questions in a manner that doesn't address what matters to clinicians: how does this drug affect patients who use it? This talk discusses designs, some of which are well known, that by their structure prevent missing data but at the same time address questions that are of clinical relevance.