The July 6, 2011 issue of JAMA has an article from Chiuve et al who reviewed data from the Nurses Health Study (81,722 women in the US followed from 1984 to 2010). They found that following healthy habits such as not smoking, keeping a BMI under 25, exercising 30 minutes or more a day, and a healthy (Mediterranean) diet reduced the risk of sudden death.
How reduced? Doing just one of these cut the risk in half, while doing all four reduced the risk 92%!!! This study is limited by the typical problems of cohort studies, perhaps one of the most important being selection bias. This is where people who are generally healthy and well-educated are more likely to participate in such studies. In this case the study participants were compared to each other, so that makes it more likely that the effect is real. Furthermore, based on Bayesian probability theory, there is already a profound body of evidence that healthy behaviors are, well, healthy, so I am inclined to believe it. What is astonishing to me is the magnitude of the effect.
Of course the 92% estimate is just the middle of a range of possible "true" results. This range is referred to as the 95% confidence interval. For this study, the range is a 97% to 77% reduction (reported as a relative risk of 0.03 - 0.23). Roughly that means that there is a 95% chance that the "true" effect is between 97% and 77%. So, that is the statistics lesson for the day, time to go get some exercise, eat healthy, and not smoke.