Modeling Treatment Compliance in Randomized Clinical Trials

• Activity Number: 461
• Type: Contributed
• Date/Time: Wednesday, August 6, 2014 : 8:30 AM to 10:20 AM
• Sponsor: Mental Health Statistics Section
• Author(s): Joseph Rausch*+ and Chad Shenk
• Companies: Cincinnati Children's Hospital Medical Center and Penn State
• Keywords: RCT ; treatment compliance ; treatment adherence ; mediation ; moderation

Abstract

The intention-to-treat approach to analyzing randomized clinical trials (RCTs) is popular in the RCT community and is beneficial for simplifying RCT analyses. Nevertheless, researchers should be aware of (a) the implications of treatment non-compliance in RCTs (see, e.g., Sheiner & Rubin, 1995) and (b) the importance of using statistical methods such as the Complier Average Causal Effect (CACE) approach to fully examine treatment efficacy (Imbens & Rubin, 1997). We argue here that treatment compliance requires more attention from the scientific community to fully explicate the manner in which it can and should be used within RCTs (cf. Kraemer, Wilson, Fairburn, & Agras, 2002; Jo, 2008; Frangakis & Rubin, 2002; Pearl, 2001). We draw from various sources to examine current frameworks for modeling treatment compliance in RCTs. Furthermore, we employ illustrative examples from multiple disciplines to better identify the implications for data analysis across a variety of RCTs. We end with our conclusions concerning the viability of current frameworks and discuss whether new approaches are needed to capitalize on the unique nature of treatment compliance in RCTs.