I noticed a great writeup by Michael Simpson, CEO of Skeptical Raptor, on this topic. Simpson does a great job of addressing the primary issues that come up when scientists attempt to do research in this area: First of all, the studies rely on self-reported measures. The scientists don’t follow people around 24/7. They just ask them “how much do you smoke,” “what do you eat,” etc. The problem with this is that someone who has poor health outcomes will assume they did something wrong and focus on what it is they did wrong. They will be ungenerous about their past “I smoked too much,” “ate too much,” “slept to little,” etc. So it’s a lot of speculation. Someone with the same diet and behavior who had good health outcomes will likely react in the opposite way, speculating why they’ve done so well. In either case it does become sort of speculative and not super informative.