This sounds similar to a project I worked on for the NIDDK.  They created a repository of clinical data for completed trials sponsored by NIDDK.  What was realized is that there is a thin line between sharing data sets and sharing information about data sets.  In order to get access to the data, as it seems the case here, there was an "approval process."  So in order to minimize resources, we created the equivalent of a Sears Catalog of NIDDK studies.  One could look up the study, papers, but also see a breakdown of many of the variables (e.g., number of men and number of women).   We used a controlled vocabulary as to make studies comparable for many variables.  I assume this method, or a similar will be put in place at BMS or DCRI.