Phase III Design for Continuous Responses with an Adaptive Tuning Parameter

• Activity Number: 404
• Type: Contributed
• Date/Time: Tuesday, July 31, 2012 : 2:00 PM to 3:50 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Aleksandra Stein*+ and Aleksandra Stein
• Companies: University of Nebraska and University of Nebraska
• Keywords: adaptive designs ; biased coin designs ; tuning parameter

Abstract

Equal-allocation Randomized Clinical Trials (RCTs) are the current "gold standard" in medical research, but they may no longer be the most appropriate designs. Balanced RCTs promote statistical concerns; ethical concerns are better addressed by sequential estimation and adaptive designs. Such studies assign future patients based on previous results. Recent research testifies that response-adaptive randomization schemes are becoming "well-understood" and many offer better individual care to enrolled subjects with minimal impairment to statistical properties of the trials or benefit to future patient populations.

In this paper we focus on an adaptive design introduced by Bandyopadhyay and Biswas in 2001 (BB). We begin with a broad treatment of adaptive biased coin designs, narrowing our focus to the BB design. In particular, we note that the fixed tuning constant T from this design does not meet any optimal criteria. We propose to replace the constant T with an adaptive function of sample size and parameter estimates. We present asymptotic and simulated results of the new designs. Finally, we make recommendations for the new designs and future research.