Abstract
For ethical reasons, testing the efficacy of new treatments for chronic diseases such as osteoporosis has hit a brick wall. Furthermore, existing trial designs, although satisfying requirements for drug registration, typically produce results with limited generalizability. Compounding the problem, these results often serve as the basis for treatment guidelines, which, in turn, are assembled by policy-makers and analysts who often do not understand the biology of the systems concerned. In this perspective, the details of these problems are briefly described and the broad outlines of some solutions suggested.