Regression Analysis of Sequentially Randomized Trials Through Artificial Randomization

• Activity Number: 73
• Type: Contributed
• Date/Time: Sunday, August 3, 2014 : 4:00 PM to 5:50 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Semhar Ogbagaber*+ and Abdus Wahed
• Companies: University of Pittsburgh and University of Pittsburgh
• Keywords: Artificial Randomization ; Adaptive Treatment Strategies ; SMART

Abstract

Adaptive treatment strategies (ATSs) are decision rules that take in inputs such as patient characteristics, covariate history, and previous treatments, and output a treatment option. ATSs are often compared via sequential multiple assignment randomized trials (SMARTs). Regression methods that allow for comparison of treatment strategies that flexibly adjust for baseline covariates are not as straight-forward. For instance, in a two-stage SMAR design, one may be tempted to compare the four strategies: A1B1,A1B2,A2B1,A2B2 using a regression model where strategy AjBk is defined as "if respond to Aj continue the same initial treatment, otherwise switch to Bk". Note that a patient responding to A1 is consistent with both strategies A1B1 and A1B2 which poses a challenge for data analysts as it violates basic assumptions of regression modeling of unique group membership. In this paper, we propose an "artificial randomization" technique to make the data appear that each subject belongs to a specific ATS. This enables treatment strategy indicators to be inserted as covariates in a regression model. The properties of this method are investigated analytically and through simulation.