Ever wonder why the story about what kinds of food you're supposed to eat in order to have a healthy diet changes all the time? Why the stuff that you were supposed to eat 5 years ago turns out to be bad and the stuff you were supposed to avoid last year turns out to actually reduce the chances of some new disease? It's because the kind of studies that determine that sort of thing are pretty much always fatally flawed.
Want to know why? Here's a fantastic article in the New York Times that explains the problems. It uses hormone-replacement therapy as its core example, but it's really the same thing, as the studies used to tell you that eating more of food X causes disease Y are subject to all the same sorts of problems as the one that tells you that HRT is a good idea.
The article's really long, but it's worth reading all the way through. Just think of it as a condensed version of the kind of conversation I have with Joanna whenever one of these studies comes across her desk at work ;-)