In clinical trials, patient allocation to treatment groups is generally fixed and determined a priori in order to obtain statistical estimates of treatment differences. With this traditional design, data obtained during the trial do not influence the randomization probabilities. While this randomization strategy is ethically attractive, has the potential to reduce sample size, shorten drug development time, and save money and resources; it introduces statistical and logistical complexities. This presentation will describe a case study of Bayesian outcome adaptive randomization design employed in a Phase 2 trial in ovarian cancer at Biogen Idec. The discussion will address the adaptive design, statistical methods, logistical issues, and lessons learned.