Many researchers wouldn’t buy it. Jacob Yerushalmy, who ran the biostatistics department at the University of California, Berkeley, and Herman Hilleboe, the New York State commissioner of health, co-authored a critique of Keys’s hypothesis, noting that Keys had chosen only six countries for his comparison though data were available for twenty-two countries. When all twenty-two were included in the analysis, the apparent link between fat and heart disease vanished. Keys had noted associations between heart-disease death rates and fat intake, Yerushalmy and Hilleboe pointed out, but they were just that. Associations do not imply cause and effect or represent (as Stephen Jay Gould later put it) any “magic method for the unambiguous identification of cause.”

This is an irrefutable fact of logical deduction, but confusion over the point was (and still is) a recurring theme in nutrition research. George Mann, a former director of the famous Framingham Heart Study, called this drawing of associations between disease and lifestyles “a popular but not very profitable game.” When the science of epidemiology was founded in 1662 by John Graunt, a London merchant who had undertaken to interpret the city’s mortality records, Mann noted, even Graunt realized the danger of confusing such associations with cause and effect. “This causality being so uncertain,” Graunt wrote, “I shall not force myself to make any inference from the numbers.”

The problem is simply stated: we don’t know what other factors might be at work. Associations can be used to fuel speculation and establish hypotheses, but nothing more. Yet, as Yerushalmy and Hilleboe noted, researchers often treat such associations “uncritically or even superficially,” as Keys had: “Investigators must remember that evidence which is not inherently sound cannot serve even for partial support.” It “is worse than useless.”

Ironically, some of the most reliable facts about the diet-heart hypothesis have been consistently ignored by public-health authorities because they complicated the message, and the least reliable findings were adopted because they didn’t. Dietary cholesterol, for instance, has an insignificant effect on blood cholesterol. It might elevate cholesterol levels in a small percentage of highly sensitive individuals, but for most of us, it’s clinically meaningless.*5 Nonetheless, the advice to eat less cholesterol—avoiding egg yolks, for instance—remains gospel. Telling people they should worry about cholesterol in their blood but not in their diet has been deemed too confusing.