ABSTRACT
In this chapter, we lay out methods for estimating country-specific treatment effects and borrow more information from similar countries and less from dissimilar countries. As stated in ICH E17 Guideline, the estimation of the local treatment effect may need to borrow information from the other regions based on similarity in intrinsic or extrinsic factors if the sample size in a region is small. The shrinkage estimate approach improves efficiency by incorporating data from other countries but treats all other countries equally thus ignoring similarity to the country of interest. We summarize methods for estimating local treatment effects in an MRCT, with focus on borrowing information from neighborhood. The neighborhood is defined by a tree built from the similarity score in a continuous scale or a categorical scale. By letting treatment effects from countries as random effects following a Brownian diffusion process (BDP) along the tree, we demonstrate that the local treatment effect estimate borrows more information from countries that are similar but less from countries that are different. A clinical example is provided to illustrate these methods.
