High-fat diets have long been known to increase the occurrence of mammary tumors in rodents. However, the interpretation of animal data is controversial. Fat is the most energy-dense macronutrient; therefore, it is sometimes difficult to disentangle the effects of fat intake from those of energy intake in studies of cancer. In studies specifically designed to determine the independent effects of fat and energy intake, the effect of fat was either weak in relation to that of energy intake 2 or it was nonexistent 3. Furthermore, in these experiments rodents are given high doses of carcinogens, and the relevance to human cancer is questionable. The clearest message from the animal data is the importance of total energy intake.

The dietary fat hypothesis is largely based on the observation that national per capita fat consumption is highly correlated with breast cancer mortality rates 4. However, per capita fat consumption is highly correlated with economic development. Also, low parity and late age at first birth, greater body fat, and lower levels of physical activity are more prevalent in Western countries, and would be expected to confound the association with dietary fat.

Both per capita fat consumption and breast cancer incidence rates increased substantially in the USA during the 20th century. However, estimates of per capita fat consumption are based on 'food disappearance' (the food available rather than the amount actually eaten). Surveys based on measures of actual individual intake indicate that consumption of fat as a percentage of energy has actually declined during the past several decades.

Data from populations with distinct dietary patterns are potentially valuable. These groups often follow diets prescribed by religious rules, and may represent a more stable long-term exposure than is found in most adults. However, these populations often have unusual distributions of other risk factors such as alcohol consumption, smoking, and reproductive behavior. Therefore, care must be taken in attributing differences in cancer rates to diet alone.

Seventh-Day Adventists, who consume relatively small amounts of meat and other animal products, have substantially lower rates of colon cancer but only slightly lower breast cancer rates than do other US women 5. Breast cancer rates among UK nuns who ate little to no meat were similar to rates among single women from the general population 6, also suggesting that there is no substantial association between animal fat intake and risk for breast cancer.

Howe and coworkers 7 performed a meta-analysis of 12 case–control studies including a total of 4312 cases. The overall pooled relative risk (RR) for a 100 g increase in daily total fat intake was 1.35, with greater risk among postmenopausal women (RR = 1.48) and no association in premenopausal women (RR = 1.13). However, the average total fat consumption is about 70 g/day in US women. A reduction in fat intake as large as 100 g would not be feasible, but the RR for readily achievable changes in total fat intake would be relatively small. For example, there would be only a 10% decrease in risk among postmenopausal women corresponding to a decrease from 40% to 29% of calories of fat intake. RRs of this magnitude in case–control studies may easily be due to selection or recall bias 8.

The association of childhood dietary fat intake with breast cancer risk has been assessed retrospectively because of the difficulty in performing prospective studies. One study including 1647 cases of breast cancer in young women 9 reported a slightly increased risk for breast cancer with adolescent intake of high-fat meat. In a retrospective cohort study of 47,355 women based on the Nurses' Health Study II, intake of total fat during high school was not associated with breast cancer risk, but intake of vegetable fat was associated with decreased risk 10.

In a cohort (prospective) study selection bias and recall bias should not occur. The results for postmenopausal breast cancer from prospective studies with at least 200 incident cases of breast cancer are shown in Table 1. Not one study reported a significant positive association with total fat intake. A collaborative pooled analysis that included most of the studies shown in Table 1 comprised a total of 4980 breast cancer cases among 337,819 women 11. Overall, no association was observed between intake of total, saturated, monounsaturated, or polyunsaturated fat and risk for breast cancer. No reduction in risk was seen even for fat intakes as low as 20% of energy.

In the Nurses' Health Study, additional analyses were conducted with 14 years of follow-up (2956 cases) 12; up to four assessments of fat intake were available, which substantially improves the measurement of long-term dietary intake. The RR for a 5% increase in percentage of energy from total fat was 0.97 (95% confidence interval [CI] = 0.94–1.00); the overall weak trend was actually statistically significant (inverse), and no suggestion of any reduction in risk was seen for fat intakes even lower than 20% of energy. In addition, results were no different when using only the baseline diet assessment: RR (95% CI) = 0.99 (0.87–1.01). Thus, the prospective studies provide strong evidence that no major relation exists between total dietary fat intake over a wide range during midlife and breast cancer incidence. Some researchers hypothesize that systematic and random measurement error occurs when using food frequency questionnaires in cohort studies, which has obscured an association between fat intake and breast cancer 13. However, the theoretical biases needed to account for a lack of association are extreme, and were not supported by empirical evidence 14.

Some have suggested that the relation between dietary fat and breast cancer can only be established by randomized trials of fat reduction. The Women's Health Initiative sponsored by the US National Institutes of Health randomly assigned women to one of two groups 15: in one group (n = 19,541) the women reduced their total fat intake to 20% of calories and in the other group (n = 29,294; control group) the women consumed their usual diet. In addition, two ongoing trials (Women's Intervention Nutrition Study and the Women's Healthy Eating and Living Study) are being conducted in women with breast cancer to determine whether a low-fat dietary intervention will prevent recurrence and improve survival 1617.

However, it is difficult to maintain compliance with a diet that is very different from prevailing food consumption habits. Also, a secular decline in fat consumption is already underway. These two facts may reduce the difference in fat intake between the intervention group and controls, and may compromise the ability to detect a difference in cancer rates. Furthermore, women in the dietary intervention group of the Women's Health Initiative will be counseled to adopt a dietary pattern high in fruits, vegetables, and grains, as well as low in fat 18. Thus, the trial would be unable to distinguish between a decrease in risk due to increased intake of fruits, vegetables and grains, and a decrease due to lower fat intake. Also, that trial will not determine whether dietary fat reduction at an earlier age may reduce breast cancer risk.

Specific types of fat could differentially affect risk for breast cancer. In most animal studies diets high in polyunsaturated fat, but typically at levels beyond human exposure, have clearly increased the occurrence of mammary tumors. A positive association has not been found in prospective epidemiologic studies 19.

In a pooled analysis of cohort studies 19, saturated fat (compared with carbohydrate) was weakly associated with higher risk for breast cancer (RR for 5% of energy = 1.09; 95% CI = 1.00–1.19). In a recent prospective study conducted in premenopausal women in the Nurses' Health Study II 20, intake of animal fat and high-fat dairy foods was associated with a 33–36% increase in risk for breast cancer for the highest compared with the lowest quintile of intake. Total fat per se was not associated with breast cancer risk, suggesting that other constituents of dairy foods consumed early in adult life may increase breast cancer risk.

As noted earlier, energy restriction powerfully reduces mammary tumors in rodents. The World War II Norwegian famine provided a natural human experiment 21. Female adolescents who lived through the famine have had a slightly lower breast cancer risk at all later ages – 13% reduction in breast cancer risk. However, the timing and duration of energy deprivation may be important. The risk of breast cancer was increased among women exposed to the short but severe 1944–45 Dutch famine, and the elevation in risk was strongest (double compared with unexposed) for those exposed between the ages of 2 and 9 years 22.

Energy-deprived children do not attain full height, which may be used as a proxy for childhood energy intake. Most observational studies of height and breast cancer risk suggest a modest positive association 23242526. In a pooled analysis of large cohorts (4385 cases), the RRs for an increment of 5 cm in height were 1.02 (95% CI = 0.96–1.10) for premenopausal and 1.07 (95% CI = 1.03–1.12) for postmenopausal women (Fig. 1) 27.

Age at menarche – an established breast cancer risk factor – also indicates childhood energy balance. In prospective studies, weight, height, and body fat predict age at menarche 2829. The potential for energy balance to influence breast cancer risk through age at menarche is great. Although the average age at menarche in Western countries is 12–13 years, in rural China the typical age has been 17–18 years 30, which is similar to that in Western countries 200 years ago.

Attained weight and weight change in adults provide sensitive measures of energy balance. Two findings have been particularly enigmatic. First, in affluent Westerners, body fat has been inversely related to premenopausal breast cancer risk. Second, body fat has been only weakly related to increased postmenopausal risk, despite strong links with endogenous estrogen levels.

The inverse relation between body weight (typically body mass index [BMI] calculated as kg/m²) and premenopausal breast cancer has consistently been seen in observational studies 2731. Heavier premenopausal women have more irregular menstrual cycles and anovulatory infertility 32, suggesting their lower risk may be due to lower ovarian hormones 33.

In observational studies, the association between BMI and postmenopausal breast cancer risk has been only weakly positive or nonexistent 72327. This is surprising because obese postmenopausal women have endogenous estrogen levels nearly double those of lean women 34. This lack of association appears due to two factors. First, the early adult reduction in breast cancer risk due to being overweight appears to persist 3135, opposing the adverse effect of elevated estrogens after menopause. Thus, weight gain should be more strongly related to postmenopausal breast cancer risk than attained weight. Indeed, this was true in both case–control 3637 and prospective studies 31353839. Second, use of postmenopausal hormones obscures the variation in endogenous estrogens due to adiposity and elevates breast cancer risk regardless of body weight 3139. Among never-users of postmenopausal hormones, those gaining at least 25 kg after age 18 years had double the breast cancer risk of women who maintained their weight (Fig. 2) 31.

In a recent pooled analysis of eight prospective studies of postmenopausal women including 624 cases 40, the authors reported that breast cancer risk increased with increasing BMI, and that this elevated risk was substantially attenuated by adjustment for serum estrogen concentrations.

An expert panel of the International Agency for Research on Cancer of the World Health Organization estimated that there is a 20–40% decrease in breast cancer risk among the most physically active women, regardless of menopausal status, type, or intensity of activity 41. A mechanism for this decrease in risk might be the lower levels of circulating ovarian hormones found in physically active women 4243.

In summary, energy balance appears to play an important role in human breast cancer. Rapid childhood growth rates accelerate the age at menarche and result in greater stature, both of which are associated with increased risk. Overweight is associated with a lower incidence of breast cancer before menopause. However, weight gain during adulthood is associated with a graded and substantial increase in postmenopausal breast cancer risk, which is seen most clearly in the absence of hormone replacement therapy.

Vitamin A consists of preformed vitamin A from animal sources, and carotenoids found primarily in fruits and vegetables. Many carotenoids are potent antioxidants and may provide a defense against reactive oxygen species that damage DNA. Vitamin A also regulates cell differentiation, and may thus prevent carcinogenesis.

In a cohort of Canadian women (519 cases) 46, a marginally significant protective association between total vitamin A intake, preformed vitamin A and β-carotene, and breast cancer was seen. With 14 years of follow-up in the Nurses' Health Study (2697 cases) 56, an inverse association with total vitamin A was seen only among premenopausal women. This inverse association was primarily accounted for by intakes of β-carotene and lutein/zeaxanthin, and was strongest among women with a family history of breast cancer. However, in an extended follow-up of the Canadian cohort (1452 cases) and in a Swedish cohort (1271 cases), little overall association was seen between intake of carotenoids and breast cancer 5758.

An alternative to the dietary assessment of vitamin A intake is the measurement of vitamin A compounds in blood. In the two largest studies based on blood samples collected before diagnosis 5960, low levels of β-carotene and other carotenoids were associated with an approximately twofold increase in risk for breast cancer.

Thus, available data from observational studies suggest a possible protective effect of vitamin A intake, particularly carotenoids, on breast cancer risk, particularly in premenopausal women. Ideally, the effect of vitamin A supplements should be evaluated in randomized trials. However, the β-carotene arm of the Women's Health Study (a breast cancer prevention trial conducted in 40,000 women) was terminated in 1996 after reports that β-carotene supplements appeared to increase the risk for lung cancer among smoking men. Thus, data from randomized trials on specific carotenoids and breast cancer risk may never be available.

Vitamin E has inhibited mammary tumors in rodents in some experiments 61. However, none of the published prospective studies has reported a significant inverse association 46565862.

Vitamin C (ascorbic acid) can block the formation of carcinogenic nitrosoamines. In prospective studies no significant overall association between intake of vitamin C and breast cancer was observed 46566263, including with long-term use of vitamin C supplements 56.

Selenium, an important component of the antioxidant enzyme glutathione peroxidase, inhibits cell proliferation and in animal studies protects against a variety of cancers, although usually at high levels of intake 64. Individual selenium intake cannot be measured accurately by dietary assessment because the selenium content of foods depends on the geographic area in which the foods were grown. However, selenium levels in tissues such as blood and toenails do reflect intake 65. In prospective studies 6667, no association between toenail selenium and risk for breast cancer was observed.

There is substantial evidence that alcohol consumption increases breast cancer risk. In a pooled analysis of the six largest cohort studies with data on alcohol and dietary factors 68, the risk for breast cancer increased monotonically with increasing intake of alcohol (Fig. 3). For a 10 g/day increase in alcohol, breast cancer risk increased by 9% (95% CI = 4–13%). Adjustment for other breast cancer risk factors had little impact. Beer, wine, and liquor all contribute to the positive association, strongly suggesting that alcohol per se is responsible for the increased risk. In intervention studies, consumption of approximately one to two alcoholic drinks per day increased estrogen levels in premenopausal and postmenopausal women 6970, suggesting a mechanism by which alcohol may increase breast cancer risk.

In several large prospective studies, high intake of folic acid appeared to mitigate completely the excess risk for breast cancer due to alcohol 717273. This relationship was recently confirmed using plasma folic acid levels 74.

The public health recommendations for alcohol are complicated because consumption of one to two alcoholic beverages per day probably protects against cardiovascular disease. Because cardiovascular disease is the leading cause of death among women, moderate drinking is associated overall with a modest reduction in total mortality 75. However, avoiding alcohol appears to be one of relatively few methods for reducing breast cancer risk, whereas many methods exist to reduce risk for cardiovascular disease. For women choosing to consume alcohol regularly, use of a multivitamin to ensure adequate folic acid intake may decrease breast cancer risk.

Speculation that caffeine may increase breast cancer risk followed a report that women with benign breast disease experienced symptom relief after eliminating caffeine from their diet. In prospective studies no increase in breast cancer risk has been seen 7677.

Phytoestrogens in soy products have attracted attention because they are highly consumed in Asian countries that have low rates of cancer 78. These isoflavone compounds, which include daidzen and genistein, can bind estrogen receptors but are much less potent than estradiol, and may act like tamoxifen by blocking the action of endogenous estrogens to reduce breast cancer risk.

Studies in Western populations with low soy intake have been unable to show an association with breast cancer risk 79. In some case–control studies in Asian populations 808182, soy intake, particularly during adolescence, was associated with lower breast cancer risk. However, in two others 8384, and in one prospective study from Japan 85, little relation was seen. However, a recent prospective study from Japan including 179 cases 86 reported a decreased risk for breast cancer when comparing the highest with the lowest quartile of isoflavone consumption (RR = 0.46, 95% CI = 0.25–0.84). Similar results were found for intake of miso soup, but not soy foods, and results were strongest for postmenopausal women. More prospective studies are needed.

Conceivably, high-dose phytoestrogens could even increase overall estrogenic activity among postmenopausal women. In a cross-sectional study 87 it was found that women consuming more soy had more mammographically dense breasts, which is a known risk factor for breast cancer. In addition, randomized controlled feeding trials have found that women consuming soy had increased markers of cell proliferation in their breasts 8889.