|
Abstract:
|
Bayesian methods are increasingly popular in medical device trials, as evidenced by the recently released FDA Guidance Document. Bayesian methods offer several key advantages, including easy incorporation of information from earlier studies, the ability to borrow strength across studies or centers, and a framework well-suited to adaptation (say, of the patient allocation ratio) that results in more ethical treatment of patients. However, such trials retain a strong frequentist flavor, due largely to the continued insistence by FDA that all trials have good frequentist operating characteristics (power and Type I error). We will discuss this situation and attempt to find a middle ground between frequentist long-run protection and Bayesian flexibility.
|
