R Package Development for Evaluating Misclassification Effects on Single Sequential Treatment and Dynamic Treatment Regimens in Sequential Multiple Assignment Randomized Trial

• Activity Number: 441 - Bayesian (and Other) Clinical Trials Designs
• Type: Contributed
• Date/Time: Thursday, August 12, 2021 : 4:00 PM to 5:50 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Jun He* and Jason Sinnwell and Abraham D. Eyman Casey and Donna K McClish and Roy T Sabo
• Companies: The Mayo Clinic and The Mayo Clinic and The Mayo Clinic and Virginia Commonwealth University and Virginia Commonwealth University
• Keywords: sequential multiple assignment randomized trial ; misclassification; single sequential treatment; dynamic treatment regimens; R package

Abstract

The sequential multiple assignment randomized trial (SMART) design includes multiple stages of randomization, where participants are randomized to an initial treatment in the first stage and then subsequently re-randomized between treatments in the following stages. Due to potential number of reasons (e.g. measurement error, biomarker accuracy), participants in a SMART study could be misclassified at the end of the first stage based on inaccurate response from intermediate outcome, potentially resulting in re-allocation to treatments that may be harmful or inappropriate. Therefore, a tool of assessing the potential misclassification effect before carrying out a trial is needed. Based on previous analytic research, we developed a R package “SMART” to evaluate the misclassification effect on single sequential treatment (SST) and dynamic treatment regimens (DTRs) in the two-stage SMART design by visualizing the mean/variance and relative bias, and power comparison between two SSTs or between two DTRs under misclassification. Demonstrations of the package’s usefulness in study design is presented.