The keyword for people to look for here is “association,” as in, “we found an association between X and Y” where when X moves in a certain direction, Y moves as well. Often the claim is that the direction and magnitude of the moves suggest that one may have a role in causing the other, but there is no way to prove or disprove an associative relationship such as this because these studies are observational/retrospective, not controlled.
I think “junk epidemiology” might be a mild overstatement, as it’s these observational studies that can be used to justify doing a controlled study that might actually isolate causal relationships. But relying on associations found by such studies to inform how we eat is probably unwise.
Yep. And in many cases, such studies rely on food questionnaires, often administered long after the time period being asked about. How reliably could you tell someone what you ate last month, six months ago, or a year ago?
Trouble is that interventional studies are tremendously expensive, so we’re not likely to see a significant increase in those kinds of studies. So we’re left with observational, and have to do the best we can to interpret the results, along with our own n=1 experiments.