Again, observational studies are going to have all kinds of confounding variables - if you're going to make an argument that one of them is important enough to reverse the conclusion, you should have some evidence and certainly more than a just-so-story.