http://www.slate.com/id/2150354

This link does an excellent job of pointing out the pitfalls of evaluating the benefit of a drug.

So, if a drug company, (lets take Biogen as an example), says that there are 47% less attacks on Tysabri, this article points out how that tells you very little about the benefit of Tysabri. It can mask the real data which might show that it is far less effective than that.

I hope you find the article as good as did I. I had known this before, but these folks put it in terms that are easy for non-statistics folks to comprehend.