Clinical studies of dynamic treatment strategies

• Activity Number: 103
• Type: Invited
• Date/Time: Monday, August 5, 2013 : 8:30 AM to 10:20 AM
• Sponsor: Biometrics Section
• Author(s): Michael Baiocchi*+ and Tze Lai and Jane Paik
• Companies: Stanford University and Stanford University and Stanford University
• Keywords: sequential analysis ; adaptive design

Abstract

The demand for rigorous studies of dynamic treatment regimens is increasing as medical providers increasingly face patients with both multi-stage disease states and chronic care issues. One difficulty with running clinical trials in a chronic care situation is that the sequential nature of the treatments require exponentially expanding space to explore to find the most efficacious treatment regime (e.g., if two treatments are available at any given time t, then the space of possible sequential treatment regimens grows at 2^t). There are many constraints on the design of such trials, two of the most pressing are: (1) researchers have a finite number of volunteers for the trial and (2) there is an ethical imperative to provide patients with the best outcomes, while still collecting information on relative efficacy. We propose an adaptive design methodology for identifying near-optimal treatment paths, wherein the most number of patients receive near-optimal treatment while still allocating enough observations to exploration of the possible treatment space to find a near-optimal regime. We also describe methods for valid inference at the conclusion of the trial.