The mechanism of action of medical devices is typically physical and local, therefore the treatment effect could often be predictable based on the information from the previous generations of a device or engineering data. Prior information can also be available for the device in oversea studies or for the control in historical studies. Bayesian method is a useful approach in combining prior information and current information to make an inference for the treatment effect. In our presentations, we will discuss the topics of ‘what to borrow’ and ‘how much to borrow’ using the Bayesian approach. We will also provide examples to illustrate the challenges in IDEs submissions with Bayesian Design and our recommendations for improving the efficiency of the communication between the companies and FDA.