Effect of Compliance on Analysis of Longitudinal Randomized Clinical Trials

• Activity Number: 139
• Type: Contributed
• Date/Time: Monday, August 10, 2015 : 8:30 AM to 10:20 AM
• Sponsor: Biometrics Section
• Author(s): Huaqing Zhao* and Susan Fisher
• Companies: Temple University and Temple University School of Medicine
• Keywords: Compliance ; intent-to-treat ; per-protocol ; longitudinal measurement ; randomized clinical trial ; adjustment

Abstract

Compliance in randomized clinical trials (RCT) is typically measured using self-report, pill counts and blood levels. In a longitudinal RCT the question arises as to how to define or analyze compliance when it may change over time. Intent-to-treat (ITT) is the primary approach to analyze longitudinal RCT. In this approach, all available data for randomized patients are included, regardless of compliance. In the per-protocol approach, only patients who were compliant with the protocol should be included. However, both ITT and per-protocol approaches may introduce bias in the longitudinal setting under various relations between compliance and response. A simple method is to define compliance on an observation level, rather than a patient level. A simulation study shows that results are varying among the different approaches to define and analyze compliance. In summary, if compliance is not related to outcome, or even moderately related to outcome, adjustment is not necessary in ITT analysis. If compliance is highly related to outcome, the adjustment is strongly suggested even when using the ITT population.