|
Abstract:
|
FD&C Act requires manufacturers of drug products to establish a drug's effectiveness by "substantial evidence" and FDA has interpreted that at least two adequate and well-controlled studies, each convincing on its own, are required. Therefore most often two pivotal studies are required for a new drug approval. While the conventional approach for claiming drug success is based on p values, the substantial evidence could also be based on Bayesian posterior probability approach. The program success criteria could be based on each individual dataset or pooled dataset from both pivotal studies. We explored several program success criteria, such as using meta-analysis by pooling the two studies with a more stringent cut-off (p values or posterior probabilities of treatment difference) vs. each individual dataset with the conventional cut-off. Information from early stage trials may be available as a prior to be borrowed in the Bayesian approach. Simulations are conducted to evaluate the impact of prior and program success criteria on the probability of program success in terms of power and type I error. Challenges of using Bayesian posterior probability will also be discussed.
|
