Nearly every dramatic reversal in nutrition science has the same shape: an observational finding is reported as though it established cause, a trial is eventually run, and the trial disagrees. Understanding why is most of what Longevity Claim Verification does.
What each design does
Observational. Researchers watch what people already do and look for patterns. Cheap, large, long, and possible for things nobody could ethically assign. They establish association.
Randomised. People are assigned by chance to one condition or another. Expensive, smaller, shorter, and often impossible for lifestyle questions. They establish cause, within the population and duration tested.
Why association keeps misleading everyone
The problem is not that researchers forget the difference. It is that the people who do a health-associated thing differ from those who do not in dozens of ways nobody measured.
People who take supplements exercise more, drink less, sleep better, and see doctors more often. Statistical adjustment can account for the differences that were recorded. It can do nothing about the ones that were not, and the unrecorded differences tend to point the same direction, because the underlying driver is usually a general tendency to look after yourself.
The healthy user effect
The single most common explanation for an observational longevity finding that later fails a trial. People who adopt any health behaviour early tend to be healthier for reasons that have nothing to do with the behaviour.
What makes an observational finding stronger
Some are much more credible than others, and four things raise confidence: a dose-response relationship, where more exposure tracks with more effect; consistency across different populations and countries; a plausible mechanism established separately; and a large effect, since small ones are more easily produced by unmeasured differences.
A finding with all four is worth taking seriously. It is still not proof, and it is still the kind of thing that has been overturned before.
Why trials are not the final word either
Trials answer a narrow question: this dose, in these people, for this long, against this comparison. A trial finding nothing may have run too short for an effect that takes decades, or used a dose too low, or studied a population that had nothing to gain.
A null result is evidence, not a verdict, and the honest reading is usually "not shown in this population over this period" rather than "does not work".
How Longevity Claim Verification handles it
Study design is stated on every check, and a claim resting only on observational evidence is reported as associated rather than shown, regardless of how large the association is. Where trials and observational data disagree, both are reported along with what would explain the gap.