There has been a lot of selective reporting and ignoring of the methodological problems over the years. But probably the most astonishing example of selection bias was the near-complete suppression of the Minnesota Coronary Survey, which was an outgrowth of the National Diet Heart Study. Also funded by NIH, the Minnesota Coronary Survey is the largest-ever clinical trial of the diet-heart hypothesis and therefore certainly belongs on the list along with Oslo, the Finnish Mental Hospital Study, and the LA Veterans Trial, but it is rarely included, undoubtedly because it didn’t turn out the way nutrition experts had hoped.

Starting in 1968, the biochemist Ivan Frantz fed nine thousand men and women in six Minnesota state mental hospitals and one nursing home either “traditional American foods,” with 18 percent saturated fat, or a diet containing soft margarine, a whole-egg substitute, low-fat beef, and dairy products “filled” with vegetable oil. This diet cut the amount of saturated fat in half. (Both diets had a total of 38 percent fat overall.) Researchers reported “nearly 100% participation,” and since the population was hospitalized, it was more controlled than most—although, like the Finnish hospital study, there was a good deal of turnover in the hospital (the average length of stay was only about a year).

After four-and-a-half years, however, the researchers were unable to find any differences between the treatment and control groups for cardiovascular events, cardiovascular deaths, or total mortality. Cancer was higher in the low-saturated-fat group, although the report does not say if that difference was statistically significant. The diet low in saturated fat had failed to show any advantage at all. Frantz, who worked in Keys’s university department, did not publish the