TL42: What do we do when the Sites are not Poolable?
*Jack Zhou, FDA CDRH 


For multicenter clinical trials, one way to demonstrate poolability is to test the treatment-by-site interactions. A large p-value from this test would indicate a consistent treatment effect between sites. However, what happens if a small p-value is shown for the treatment-by-site interaction test and the sites are not poolable? Any further analyses can be done besides stratifying the results by site? Please come share your opinions and stories.