A Simple Way to Expand Bias Coin Randomization to Unequal Allocation

• Activity Number: 626
• Type: Contributed
• Date/Time: Thursday, August 4, 2011 : 8:30 AM to 10:20 AM
• Sponsor: Biopharmaceutical Section
• Author(s): Yevgen Tymofyeyev*+ and Olga M. Kuznetsova
• Companies: Merck Sharp and Dohme Corp. and Merck Sharp and Dohme Corp.
• Keywords: Biased coin ; minimization ; covariate-adaptive allocation ; unequal allocation ; randomization test ; selection bias

Abstract

An expansion of the biased coin randomization [Efron, Biometrika (1971)] and minimization to unequal allocation to two or more treatment groups that preserves the target allocation ratio at every allocation step is offered. The expansion termed Allocation Ratio Preserving Biased Coin (ARP BC) Randomization (Minimization) allows to avoid selection bias possible in double-blind trials when the allocation ratio is not preserved, observer bias, and the accidental bias. ARP BC randomization (minimization) is asymptotically protected against the shift in the re-randomization distribution identified by Proschan et. al. [Biometrics (2010)] for some examples of unequal allocation. The asymptotic shift in the re-randomization distribution of the difference in treatment means for an arbitrary unequal allocation procedure is explicitly derived in the presentation. Simulations that show that with ARP BC randomization the power of randomization test is similar to the power of ANOVA are presented. Balancing properties of the ARP BC randomization (minimization) are described through simulations.