"The reason for the change was more thorough epidemiology. The earlier studies tended to be "retrospective," relying on people to remember dietary details from the distant past."
So what? Which are the studies that are invalid? If none of them are invalid, then why disregard any information they might give us? 4000 studies with similar observations is not "scant evidence", unless you can rule out every single one. And are these 4000 the only studies which have similar conclusions? Probably not.
"These results were often upended by "prospective" protocols, in which the health of large populations was followed in real time."
In a "retrospective" study, you look at the past - for instance, you interview people and ask them what they recall eating over the last year, and see how that tracks with their current health condition. But memories are faulty and selective and easily influenced, so you might be operating from bad data. And the associations you find could be spurious for other reasons. So you can quickly reach some conclusions with this type of study, but they might just reinforce what people already believed rather than tell you what is actually true.
In a "prospective" study, you look at the present - for instance, you could ask a group of people to keep a food journal, measure their health at regular intervals, encourage some of them to make some changes, and see if the advice ends up making them healthier. The advantage is that you have MUCH more accurate data. The disadvantage is: (a) you have to wait many years before you can reach any conclusions, (b) most of the time the conclusion you discover is that the advice didn't help.
Retrospective studies can generate hypotheses but you have to TEST those hypotheses with prospective studies.
"The reason for the change was more thorough epidemiology. The earlier studies tended to be "retrospective," relying on people to remember dietary details from the distant past."
So what? Which are the studies that are invalid? If none of them are invalid, then why disregard any information they might give us? 4000 studies with similar observations is not "scant evidence", unless you can rule out every single one. And are these 4000 the only studies which have similar conclusions? Probably not.
"These results were often upended by "prospective" protocols, in which the health of large populations was followed in real time."
I don't know what this means.