Treatment Discontinuation and Missing Data in Clinical Trials: Identifiability and Ignorability Issues

• Activity Number: 311
• Type: Invited
• Date/Time: Tuesday, July 31, 2012 : 10:30 AM to 12:20 PM
• Sponsor: Social Statistics Section
• Author(s): Roderick Little*+
• Companies: University of Michigan/U.S. Census Bureau
• Keywords: Dropout ; Nonresponse ; Randomized Trials ; Principal Stratification ; Missing at Random

Abstract

I consider the analysis of clinical trials when participants discontinue treatments, because of side-effects, lack of efficacy or other reasons. This form of "treatment dropout" is distinguished from "analysis dropout" where follow-up measures on the participant are not recorded. In many clinical trials these concepts are related, in that follow-up measures are not attempted when a participant discontinues treatment and hence is no longer following the study protocol. Treatment dropout can be regarded as creating a missing data problem, in the sense of creating missing values for a principal stratifier indicating whether participants would discontinue under each of the compared treatments. We consider approaches to identifying treatment effects under treatment dropout, both when follow-up measures are obtained for treatment dropouts and when they are not. Extensions to situations where there is treatment dropout and analysis dropout are also considered.