|
Abstract:
|
Considerable attention has been paid in recent years to the topic of estimands, and philosophical discussions have arisen as to the purpose of clinical trials and what is/are the most important quantities to estimate in a clinical trial. While this problem seems like it may have obvious answers, it is riddled with complexities relating to treatment effects, treatment policy effects, the handling of post-randomization events that perturb the original design/randomization (e.g. treatment discontinuations, study discontinuations, use of rescue medications).
This paper will investigate an in-depth application of the tripartite estimand approach proposed by Akacha et al (Stats in Medicine, 2016) by analyzing a diabetes clinical trial using this approach. The application to an actual clinical trial dataset will help uncover the pros and cons of the approach as well as other considerations that may be useful in the analysis and comprehensive interpretation of a treatment effect. Comparison will be made to other approaches to this complex issue.
|
