Phase II designs play an important role in drug development. The main purpose of early phase II trials is to determine whether a new treatment demonstrates sufficient efficacy to warrant further investigation. Often decisions are based on single-arm, open-label studies using short-term endpoints such as the tumor response in cancer studies. Two-stage designs are commonly employed to allow for early stopping due to inactivity of the new treatment based on the interim results observed at the end of the first stage. However, two-stage designs do not provide a formal mechanism to stop trials before predetermined sample size at the end of the first stage is reached, and adjustment to statistical inference is necessary when actual trial conduct deviates from the proposed design. An alternative approach to two-stage designs is a flexible Bayesian predictive probability design that allows for continuous monitoring of the response rate (Lee and Liu, 2007). The new treatment is shown to be efficacious or futile based on the predictive probability for declaring that the treatment is promising at the end of the study given interim observed data. We utilize this approach to generate the futility boundary in the design of a Phase II cancer trial for a rare disease by using the prior information observed in a Phase I trial. The operating characteristics of the Bayesian design and the Simon’s two-stage design are investigated and compared under various design scenarios.