All Times EDT
Keywords: Bayesian, power prior model, propensity score
Bayesian methods that borrow strength from good prior information can lead to more efficient regulatory decision making. For medical devices, clinical data from previous generations of a device, virtual patient data based on computational modeling, data from oversea studies, or reliable and relevant registry data may serve as good prior information. Recently, Medical Device Innovation Consortium (MDIC) proposed a dynamic borrowing method using modified power prior methodology where the similarity between the prior data and the current data is considered to discount the prior data. However, the use of outcome data in determining how much to borrow is controversial. Wang et (2018) proposed a propensity score-based power prior model to determine how much to borrow, which is outcome free. This talk will describe these innovative methods for leveraging prior information to support medical device regulatory submissions. The pros and cons of these two models will be discussed.