Good Calories, Bad Calories

Considerable evidence also argued against the hypothesis. As John Higginson, founding director of the International Agency for Research on Cancer, noted in 1979, the international comparisons were as contradictory as they were confirmatory. In urban Copenhagen, breast-cancer rates were four times higher than in rural Denmark, but fat consumption was 50 percent lower. Large-population studies in Framingham; Honolulu; Evans County, Georgia; Puerto Rico; and Malmö, Sweden, had all reported low cholesterol levels associated with higher cancer rates. Since low cholesterol is allegedly the product of low-fat diets, it was “difficult to reconcile” this evidence, as the Framingham investigators noted in 1981, with the hypothesis that high-fat diets cause cancer.

The publication of the National Academy of Sciences report Diet, Nutrition, and Cancer in 1982 prompted the National Cancer Institute and the NAS to make funding available to test the hypothesis. A critical test would come from the Nurses Health Study, led by the Harvard epidemiologist Walter Willett, which began tracking diet, lifestyle, and disease in nearly eighty-nine thousand nurses around the country in 1982. Such a prospective study is no substitute for a randomized clinical control trial, but it constitutes the best that observational epidemiology can do. Willett and his colleagues published his first report on fat and breast cancer in January 1987 in The New England Journal of Medicine. Over six hundred cases of breast cancer had appeared among the eighty-nine thousand nurses over the first four years of the study. If anything, the less fat the women confessed to eating, the more likely they were to get breast cancer. In a New York Times article reporting the results of the study, Peter Greenwald, director of the National Cancer Institute Division of Cancer Prevention, said that the Nurses Health Study was “a good study, but not the only one,” and so NCI would continue to recommend—despite what was then, by far, the best evidence available—that Americans eat less fat to prevent breast cancer. Eight months later, NCI researchers themselves published the results of a study similar to the Nurses Health Study but smaller, also suggesting that eating more fat and more saturated fat correlated with less breast cancer. The NCI study went virtually unnoticed, as Science later noted, “perhaps because no one wanted to hear the message that a promising avenue of research was turning into a blind alley, and perhaps because it swam against the ‘medically politically correct’ idea that fat is bad.”

In 1992, Willett published the results from eight years of observation of the Nurses cohort. Fifteen hundred nurses had developed breast cancer, and, once again, those who ate less fat seemed to have more breast cancer. In 1999, the Harvard researchers published fourteen years of observations. By then almost three thousand nurses had contracted breast cancer, and the data still suggested that eating fatty foods (even those with copious saturated fat) might protect against cancer. For every 5 percent of saturated-fat calories that replaced carbohydrates in the diet, the risk of breast cancer decreased by 9 percent. This certainly argued against the hypothesis that excessive fat consumption caused breast cancer.

Despite this accumulation of contradictory evidence, Peter Greenwald and the administrators at NCI refused to let their hypothesis die. This was Rose’s philosophy at work. After Willett’s publication of the first Nurses Health Study results, Greenwald and his NCI colleagues had responded with an article in JAMA entitled “The Dietary Fat–Breast Cancer Hypothesis Is Alive.” The NCI administrators argued that any study that generated evidence refuting the hypothesis could be flawed. The existence of any positive evidence, they argued, even if it came from admittedly rudimentary studies—in other words, studies that almost assuredly were flawed—was sufficient to keep such a critical hypothesis alive.