Causal Effects of Observed Error-Prone Exposure Measures in Randomized Clinical Trials

• Activity Number: 381
• Type: Invited
• Date/Time: Wednesday, August 10, 2005 : 10:30 AM to 12:20 PM
• Sponsor: General Methodology
• Author(s): Els J. Goetghebeur*+ and Stijn Vansteelandt
• Companies: Ghent University and Ghent University
• Keywords: causal inference ; compliance ; measurement error ; clinical trials ; structural models ; potential outcomes

Abstract

While most scientists appreciate the value of a simple and robust intent-to-treat analysis in clinical trials, many challenge the wisdom of relying solely on the ITT summary for major decisions facing a complex, time-sensitive drug delivery system. In practice, exposure to an experimental drug varies over time, as compliance with prescribed drug dosing regimens tends to vary within and between subjects---even in clinical trials. This affects the relevance of a measured intent-to-treat effect for future patient horizons. Hence, a stream of research efforts has focused on estimation of the causal effect of observed exposure patterns in clinical trials. Randomization-based methods in the framework of potential outcomes have proved very useful to answer such questions. Structural mean models regress the causal effect of randomized treatment on treatment patterns actually received. In doing so, they silently assume exposure was accurately measured. Unfortunately, there typically remains a margin of error in exposure measurement, even with today's highly sophisticated electronic monitors of drug intake.