Practical Considerations for Building Priors for Confirmatory Studies
Being able to borrow information from previous studies is an appealing and exciting feature of Bayesian methods. However, use of historical information in confirmatory trials always poses a challenge because verification of the underlying assumption of exchangeability is difficult. This chapter will discuss considerations of integrating historical information into confirmatory trials from a practical perspective, using the development of treatment for spinal muscular atrophy as an example. Section 3.1 will introduce the clinical program and different data sources available at the planning stage and the possible impact of such data sources on designing the phase III trials. Section 3.2 will review the corresponding statistical models that are applicable for analyzing the historical and the phase III data together through Bayesian borrowing techniques. Section 3.3 will present the analyses results and the operation characteristics of some selected simulation scenarios, and a discussion of further thoughts will be in Section 3.4.