If it is noise/invalid/a mistake, it means nothing; the actual effect might actually be a DECREASE of CVD as far as we know.
And it is important to also look at all cause mortality; such studies have often shown a decrease in CVD but the same final number of deaths...
Some of those anyway. You can only account for confounders that you know about a priori.
And given they used Food Frequency Questionnaires, it's mostly wasted efforts imho.
https://www.cbc.ca/news/politics/rcmp-csis-cell-surveillance...
If it is noise/invalid/a mistake, it means nothing; the actual effect might actually be a DECREASE of CVD as far as we know.
And it is important to also look at all cause mortality; such studies have often shown a decrease in CVD but the same final number of deaths...