Actually, this might be an example of the "healthy user" bias in observational studies.[1] Although epidemiologists do try hard to include all of the relevant variables in their analyses, there are likely unidentified individual differences (perhaps including personality and lifestyle) that distinguish people with better health outcomes.

That's why double-blind, randomized, controlled trials provide more compelling evidence of cause and effect.

[1] Do we really know what makes us healthy? NY Times Magazine, Sept. 16, 2007. http://www.nytimes.com/2007/09/16/magazine/16epidemiology-t....