I always wonder about survivorship bias in studies correlating exercise and health. Someone with joint trouble isn't able to run much, for example. It filters those people out. No doubt there are real benefits, but simple correlation will always have that problem.