|
Abstract:
|
Non-inferiority (NI) studies are to show that any difference between the new drug and the active control is small enough to allow a conclusion that the new drug has at least some effect that is not too much smaller than active control. [FDA guidance, March 2010] This concept can also be applied on the safety endpoint. In efficacy NI trials, the NI margin is considered to be the effect of active control relative to placebo. However, the concept of NI margin of test drug versus active control recommended in the FDA guidance may not be able to be applied to the safety endpoint like SAE rate. In addition, in a disease area where there is no commonly known active control, the standard of care depends on physician's choice and geographic area where the patient is located. The control is neither placebo nor a single known active control. It is difficult to estimate the NI margin and the prior. Simulations using the hierarchical Bayesian approach proposed by Thall et al. as well as other Bayesian approach to evaluation the sample size for above mentioned situation were conducted. The operating characteristics equivalent to Type I error and power are evaluated.
|
