Balancing Treatment Allocation and Randomization with Combinatorics

• Activity Number: 598
• Type: Contributed
• Date/Time: Wednesday, August 7, 2013 : 2:00 PM to 3:50 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Ruji Yao*+ and Norman Ying Yao
• Companies: Merck and Department of Physics, Harvard University
• Keywords: Combinatorics ; Randomization by site ; Treatment balance ; Latin-squares ; Clinical trial design ; Allergy studies

Abstract

For a randomized clinical trial, it is a basic requirement that the actual allocation among treatment groups is similar to that which is specified in the original trial design. This maximizes the power and efficacy of the trial and helps to prevent statistical biases. For example, in a multi-site clinical trial (when randomized by site), each site will assign blocks of treatment codes. Since the total number of enrolled subjects is ultimately random, each site may not be balanced over all treatments and such imbalances can be accumulated over the entire trial. One of the post popular approaches to this challenge is known as centralized randomization; this approach is most effective when the number of sites is large and each site typically has a small number of subjects (relative to the number of treatment groups). However, when randomization by site is also necessary (e.g. in an allergy study), finite size imbalances can become a serious limiting factor. Here, we present a modified treatment allocation method based upon a combinatoric approach.