"the late, great Ancel Keys, and his equally late (but maybe not as great) role in the history of heart disease research. The oft-repeated tale goes something like this:
Once upon a time, a scientist named Ancel Keys did an awful thing. He published a study about different countries that made it look like heart disease was associated with fat intake. But the truth was that he started out with 22 countries and just tossed out the ones that didn’t fit his hypothesis! When other researchers analyzed his data using all the original countries, the link between fat and heart disease totally vanished. Keys was a fraud, and he’s the reason my mom made me eat skim milk and Corn Chex for breakfast instead of delicious bacon and eggs. LET HIS SOUL BURN. BURN! BUUUUUURN!
Depending on who tells the story, some of the details (and wishes for eternal hellfire) may differ. But in many cases, Keys’ infamous cherry-picking is attributed to his Seven Countries Study, a landmark project that helped sculpt our common beliefs about fat. Even the Seven Countries Study page on Wikipedia—the first hit when you Google “Seven Countries Study”—says that Keys shamelessly erased the data he didn’t like:
The study began with a great many more countries … but Keys deleted the countries whose results did not match his pre-conceived conclusions, leaving him with only Japan, Italy, Great Britain, Australia, Canada and the US. Full disclosure would have made a great deal of difference....
No use in beating around the bush. After statistic-ifying all 22 countries, Yerushalmy and Hilleboe found that not only was “fat as percent of total calories” still associated with heart disease (r = 0.59), but animal fat was clearly driving that correlation. In fact, plant fat had a negative association with heart disease (-0.47) while animal fat was uber positive (0.68). And to rub salt into the wounds of omnivores everywhere, the animal protein/plant protein division was equally stark: animal protein as a percent of total calories had a correlation of 0.64 with heart disease, while plant protein had an inverse correlation of -0.65.
In number-free language, this means the countries eating more animal foods were—as a general trend—reporting more deaths from heart disease.
As with any observational data, this doesn’t tell us diddly squat about cause and effect. Drawing correlations between countries is particularly risky because of massive confounding that’s almost impossible to account for. But if we’re going to be honest about these specific numbers, a heart disease/animal food relationship is very much there.
Oh, the irony. The fat Keys focused on for his 1953 graph was basically a reflection of meat and dairy. We’ve lambasted him for not using all the available data, but if he had, he might’ve turned his “correlation is causation” laser-gaze onto animal foods and plumb gone vegan"
Read more...