They take the few people that had some form of heart disease (37 out of 1104 over ten years) and noted that none of them except one could do 40 pushups. By some magic, this is translated into a 96% reduction in chance of a heart attack.
If that's really the reasoning, could I not just pick any metric (i.e. words typed per minute over 120, or IQ over 130) which is likely to eliminate almost all of the participants and conclude that it is reducing risk of heart disease?
Am I missing something? Is this kind of reasoning not complete bogus?