A history and critique of our understanding of the relationship between dietary fat and breast cancer risk
In the 1980s national policies directed at lowering fat consumption emerged. One of the many reasons for recommending fat restriction was the apparent association between high dietary fat and breast cancer incidence—an association based on epidemiological and case control data. We now know that epidemiological studies and case controlled studies are poor predictors of how dietary patterns influence cancer risk. Furthermore, newer, prospective studies no longer support the association between dietary fat and breast cancer except in a small subset of cancers.
In the 1980s national policies directed at lowering fat consumption emerged. One of the many reasons for recommending fat restriction was the apparent association between high dietary fat and breast cancer incidence—an association based on epidemiological and case control data. We now know that epidemiological studies and case controlled studies are poor predictors of how dietary patterns influence cancer risk. Furthermore, newer, prospective studies no longer support the association between dietary fat and breast cancer except in a small subset of cancers. This paper reviews the evolution of the evidence on fat intake and breast cancer risk.
History of the Hypothesis
The idea that high animal fat in the diet causes breast cancer goes back at least to 1950 when Silverstone and Tannenbaum reported that the more fat they fed mice, the more likely the mice were to spontaneously develop breast cancer.1 This idea gained traction in 1975 when a key paper by Armstrong and Doll was published comparing local dietary habits with cancer incidence and mortality rates in a range of countries.2 When it came to breast cancer, the higher the national per capita consumption of animal fat, the higher the rate of breast cancer mortality.
A 1978 study by Miller et al gave further credence to the idea that animal fat consumption increased breast cancer risk. Miller conducted a case control study that matched 400* cases of breast cancer against matched controls. Through dietary history questionnaires, the researchers claimed that increased total dietary fat was associated with greater risk for breast cancer.3 In Miller’s data the association between breast cancer and fat did not actually reach statistical significance. Nevertheless, the idea that dietary fat raised the risk of breast cancer was accepted. In the words of the authors, “Reasons why a weak association might have been anticipated are discussed, and it is concluded that in reality the association is stronger. Furthermore, its consistency with other evidence, both experimental and international, suggests that it is causal.”
Even as the NCI launched the public campaign to reduce fat consumption, research was casting doubt on this theory.
In 1982 the Committee on Diet Nutrition and Cancer, a division of the National Research Council, came out recommending that Americans reduce fat consumption from an average of 40% of daily calories to 30%.4 By 1984, the National Cancer Institute (NCI) had turned these recommendations into nationwide public policy. From that point on, this hypothesis was assumed to be true.
Even as the NCI launched the public campaign to reduce fat consumption, research was casting doubt on this theory. Graham et al reported in 1982 that having compared the diets of 2,024 breast cancer patients with those of 1,463 control patients without cancers who were seen at the Roswell Institute between 1958 and 1965, they found no differences in fat consumption patterns.5
In 1990, Howe et al countered Graham’s negation of the association in a meta-analysis of 12 studies. While the studies individually had not demonstrated a significant positive association, the combined data did. Howe’s group also reported an inverse association between fruit and vegetable intake and breast cancer risk, data that led to their famous conclusion: “If these dietary associations represent causality, the attributable risk (i.e., the percentage of breast cancers that might be prevented by dietary modification) in the North American population is estimated to be 24% for postmenopausal women and 16% for premenopausal women.”6
Their conclusion was influential and many healthcare professionals based dietary recommendations on this pronouncement.
However, there was a major fault in Howe’s meta-analysis. The studies had not compared diets on an iso-caloric basis, that holding total caloric intake constant. An iso-caloric approach is essential when seeking etiologic relationships between diet and disease. When these calculations were actually included, the fat and cancer association dropped down to a few percentage points difference—even when there was a significant 10% increase in calories derived from dietary fats.7
These 2 competing narratives, whether fat intake did or did not affect risk of breast cancer, remained a subject of debate, with various case controlled studies being held up as new proof by one side or the other. In the end the various weaknesses inherent to case control studies and dietary recall questionnaires made it impossible to reach a firm conclusion. The definitive data had to come from prospective studies.
Prospective studies have a number of advantages and are less prone to error than case control studies that rely on honesty, memory and selection of representative control subjects.
In 1983 Phillips and Snowdon reported on breast cancer mortality during a 21-year prospective study of 21,295 Seventh Day Adventists (vegetarians) versus meat eating non-Adventists. They reported no significant trend toward increasing risk for breast cancer with greater meat consumption.8
After examining the same data, Mills et al reported no association between, meat, milk, or egg consumption and breast cancer death. Oddly enough, women who adopted a vegetarian diet early in life were at higher risk of death from breast cancer.9
In 1987 Willet et al published an analysis of data from the Nurses Health Study Cohort that included 89,538 nurses in the United States. There was a trend toward a decreased relative risk of breast cancer among women in the highest quintile of calorie-adjusted total fat intake as compared with women in the lowest quintile, a statistically nonsignificant 18% decrease.10
Researchers have continued to monitor this same cohort. The most recent analysis of their data, this time by Michele Holmes, was published in 2014, 34 years after the start of the study. Holmes et al reported that 1,529 women from the original group had died from breast cancer, noting that “higher total fat intake was associated with a slightly lower lethal breast cancer risk (top vs bottom quintile hazard ratio [HR] 0.85; 95% CI 0.72, 1.01; P trend=0.05).”11 [Note: this was a trend and did not reach significance.]
In the late 1980s and early 1990s, 5 other prospective studies were published that are listed and summarized in the table below.
Table: Prospective studies of total and saturated fat intake and risk of breast cancer
|Study||Study size||Follow-up years||Number of cases||RR†: total fat (95% CI ‡) [high vs low]||RR: saturated fat [high vs low]|
|Nurses Health Study (Willet 1992)12||89,494||8||1,439||0.86 (0.67-1.08)||0.86 (0.73-1.02)|
|Canadian Study (Howe 1991)13||56,837||5||519||1.30 (0.90-1.88)||1.08 (0.73-1.59)|
|New York State Cohort (Graham 1992)14||17,401||7||344||1.00 (0.59-1.70)||1.12 (0.78-1.61)|
|Iowa Women's Study15||32,080||4||408||1.13 (0.84-1.51)||1.10 (0.83-1.46)|
|Dutch Health Study (Van den Brandt 1993)16||62,573||3||471||1.08 (0.73-1.59)||1.39 (0.94-2.06)|
|Adventists Health Study (Mills 1989)17||20,341||6||193|| ||1.21 (0.81-1.81)|
† Relative risk
‡ Confidence interval
In none of these separate trials did the association between total fat and breast cancer incidence reach statistical significance.
A 1996 meta-analysis by Hunter et al combined the data from these 6 cohorts plus an additional study by Wolk et al from Sweden18 to create a pooled database of 4,980 cancer cases from prospective studies that included a total 337,819 women. When women in the highest quintile of energy-adjusted total fat intake were compared with women in the lowest quintile, the multivariate pooled relative risk of breast cancer was 1.05 (95% CI: 0.94-1.16). In simple terms, risk for breast cancer did not vary significantly with fat intake.19
The Swedish study by Wolk et al is of interest in light of the recent PREDIMED data on breast cancer that reported consumption of extra virgin olive oil was associated with significant drops in breast cancer incidence. Wolk’s group analyzed dietary patterns of a total of 61,471 Swedish women, among whom 674 cases of invasive breast cancer occurred. No significant association was seen for total fat intake and breast cancer risk, but associations were seen by type of fat. Monounsaturated fat (eg, olive oil) was inversely associated with breast cancer risk, while polyunsaturated fat had a positive association.
Total fat made no difference. “The RR for each 10-g increment in daily intake of monounsaturated fat was 0.45 (95% CI, 0.22-0.95), whereas the RR for a 5-g increment of polyunsaturated fat was 1.69 (95% CI, 1.02-2.78).”
[Note: these increments correspond to approximately 2 standard deviations of intake in the population.]
Two other notable prospective clinical trials on diet and breast cancer deserve mention.
The Women’s Healthy Eating and Living (WHEL) study
Between 1995 and 2000, the WHEL study randomly assigned 3,088 women who had been diagnosed and treated for breast cancer to an intensive diet intervention or to a comparison group and followed them through 2006. This trial was conducted to determine if women gained benefit from a diet that consisted of 5 vegetable servings plus 16 oz of vegetable juice, 3 fruit servings, 30 g of fiber, and fat intake lowered to 15% to 20% of total caloric intake.20 Over the mean 7.3-year follow-up, 16.7% of the women in the intervention group vs. 16.9% of the women in the control group experienced an invasive breast cancer event. Rates of death were equally close: 10.1% in the intervention group vs 10.3% in the control group. Neither of these between-group differences approached statistical significance.21 While diet had no significant effect, we should mention that the women who were most active at baseline had a 53% lower mortality risk compared to the least active women and adherence to activity guidelines provided to all participants in the study was associated with a 35% lower mortality risk.22
When Harvard professor Willard Willet, PhD, lectured at the National Cancer Institute’s annual Cancer Prevention Symposium in 2012, this fat and breast cancer theory was already history. In Willett’s words, regarding the idea that fat in the diet is a major cause of cancer, “There was never any strong evidence for this idea, but it was repeated so often that it became dogma in the 1980s and 1990s ... The hypothesis that the percentage of calories from fat in the diet is an important determinant of cancer risk, at least during midlife and later, is not supported by the data.”23
The Women’s Intervention Nutrition Study (WINS)
WINS may have been the largest and most ambitious nutritional intervention study ever attempted. A great deal of time and expense was invested in trying to prove that a low-fat diet would benefit those with a history of breast cancer.
The WINS researchers carried out a randomized, prospective, multicenter clinical trial to see if getting women who had been treated for breast cancer to lower their dietary fat intake might improve their disease prognosis.
Between February 1994 and January 2001, 2,437 women were randomly assigned to 1 of 2 groups in a 40:60 ratio either into a dietary intervention group or into a control group.24 This trial’s initiation was 2 years before Hunter et al’s meta-analysis was published; researchers still thought fat consumption was correlated to breast cancer risk.
The WINS diet attempted to drop dietary fat down to 15% of daily calories and required determination and commitment on the part of both participants and researchers. Twelve months into the study, participants in the intervention group had dropped their fat consumption from 51.3 g/day to 33.3 g/day (from 29.6% of total calories to 20.3%).24
A 2006 interim follow-up of the WINS cohort reported that women in the low-fat diet group now weighed about six pounds less than women in the control group. Of the women in the diet group, 9.8% had experienced a relapse or new breast cancer while 12.4% of those in the control group had. These differences did not reach statistical significance.24
A final data summary from this WINS cohort was presented at the San Antonio Breast Cancer Symposium in December 2014:
“There was still no overall survival benefit in the whole population—mortality rates were 17.0% in the intervention arms and 13.6% in the control arm [a non-significant difference] … however, there was a fairly striking impact on survival in the ER-negative and ER/PR-negative groups.”
Estrogen Receptor Status and Fat Intake
In a conference report at the 2014 San Antonio Breast Cancer Symposium, Rowan Chlebowski, MD, PhD, one of the WINS investigators, noted that after more than 15 years of follow-up, the low-fat intervention was effective at improving survival among women with hormone receptor–negative tumors. In women with estrogen receptor (ER)-negative tumors, a statistically significant 36% reduction in deaths was seen with the dietary intervention. The benefit was even greater—a 54% reduction in deaths—among women whose tumors were negative for both ER and progesterone receptors (PR).25
These data are worth noting. There was no overall survival benefit in the group as a whole; mortality rates were 17.0% in the low-fat groups and 13.6% in the control group, a nonsignificant difference. However, for the 478 ER-negative patients, the differences were significant, with the low-fat group surviving 13.6 years compared to 11.7 years (HR=0.64, P=0.045). For the 362 women with ER and PR negative cancers, survival times were even better: 14.0 years (HR=0.46, P=0.006).26
Between the results from the WINS and of the WHEL trials the researchers have failed to prove that the dietary interventions they instituted after breast cancer diagnosis improved prognosis for the cohorts as a whole. However, these subgroups of women who had ER-negative or ER/PR-negative tumors may prove to be more responsive to fat restriction diets.27
Fat Intake or Resultant Weight Loss?
By the end of the WINS study, women who had followed the low-fat diet had lost an average of 5% of their body weight. The perceived benefits of this low-fat intervention seen for these hormone-negative cancers may be due to this weight loss.25
Chlebowski is quite aware that exercise and weight loss may improve breast cancer survival.28 Gaining weight after initial treatment is associated with higher incidence of breast cancer recurrence.29 Weight gain is significantly associated with increased risk of triple-negative breast cancer (TNBC). Obesity (BMI ≥ 30) is strongly associated with ER-negative and PR-negative breast cancer incidence in women less than 50 years old.30
The idea that weight loss may have a greater impact on ER- and PR-negative breast cancers was emphasized by Vitolins et al: “The detrimental relationship between body size and breast cancer recurrence may be more pronounced among women with estrogen receptor (ER)/progesterone receptor (PR)-negative breast cancer.”
In June 2014, Vitolins reported on a pilot study of obese women who were survivors of double-negative breast cancer and who with dietary interventions lost an average of 6.3 KG (~14 pounds), about 7.5% of their baseline weights, in just 12 weeks. The long-term impact of this intervention is still unknown, but obviously these researchers are hoping it proves useful.31
Lipid paramenters are also implicated in TNBC prognosis. A May 2015 paper reported that low HDL cholesterol was correlated with worse recurrence-free survival and overall survival in TNBC. Women with TNBC are also more likely to have elevated blood sugar than the women with other types of breast cancer.32 If glucose control is the goal, weight loss would be more likely to improve such control than being on a low-fat diet.
If the goal is to lose weight and at the same time increase HDL/total cholesterol ratios, a low-fat diet is still not the most effective way to achieve this. A September 2014 study compared 12 months on a low-carbohydrate diet (n=59) vs a low-fat diet (n=60) and reported that the low-carbohydrate diet was superior at achieving weight loss and increased HDL.33
Returning to the data Chlebowski reported in San Antonio, this suggestion that hormone receptor–negative breast cancer patients should follow a low-fat diet is not unreasonable, though it is more likely that weight loss and the metabolic changes that weight loss triggers in the body are responsible for the improved statistics that Chlebowski reported. When this study began, no one thought to ask other questions.
Extra Virgin Olive Oil
There may be good reason to hesitate encouraging women to limit fat consumption at least in the case of extra virgin olive oil (EVOO). Two recent studies suggest high levels of EVOO may provide protection, in particular against hormone receptor negative cancers.
In 2014 Spanish researchers analyzing data from the EpiGEICAM Trial reported that adherence to a Mediterranean-style diet pattern was associated with lower risk of breast cancer, especially in contrast to a Western-style diet that “was characterised by high intakes of high-fat dairy products, processed meat, refined grains, sweets, caloric drinks and other convenience food and sauces and by low intakes of low-fat dairy products and whole grains.”34
EpiGEICAM researchers compared dietary data from 1,017 breast cancer cases with 1,017 matched controls. Adherence to the Western dietary pattern was associated with a higher risk of breast cancer [odds ratio (OR) for top vs bottom quartile=1.46; 95% CI: 1.06-2.01], especially in premenopausal women (OR=1.75; 95% CI: 1.14-2.67). In contrast, the Mediterranean pattern was related to a lower risk (OR for top quartile vs bottom quartile=0.56; 95% CI: 0.40-0.79). The protective effect of the Mediterranean pattern was stronger for triple-negative tumors (OR=0.32; 95% CI: 0.15-0.66). While the study did not specifically quantify the amount of EVOO consumed by either group, EVOO is a fundamental component of the Mediterranean diet.
The recently published analysis of data from the PREDIMED trial also suggests adherence to a Mediterranean style diet and in particular the consumption of EVOO is associated with lower risk of breast cancer. In this large prospective trial, participants were allocated to a Mediterranean diet supplemented with EVOO (n=1,476), a Mediterranean diet supplemented with mixed nuts (n=1,285), or a control diet (n=1,391) in which participants received dietary counseling encouraging a low fat diet.
Participants in the 2 intervention groups were given either free EVOO (1 liter/wk) or free mixed nuts (30 g/day) according to their randomization group. In the primary analysis adding either nuts or EVOO significantly reduced risk of cardiovascular events.35 EVOO addition to the diet was associated with a 68% decrease in ER-positive breast cancer diagnosis over a 4.8-year period. Adding nuts was associated with a 41% drop.36
Rather than attempting to control or limit fat consumption, it may be more prudent for women to strive to increase EVOO and nut consumption.
Fats from Dairy
A May 2013 paper in the Journal of the National Cancer Institute argued that high intake of dairy fat is associated with worse breast cancer survival, likely due to the estrogenic hormones derived from these foods. These researchers had analyzed data from 1,893 women in the Life After Cancer Epidemiology (LACE) study diagnosed with early-stage invasive breast cancer from 1997 to 2000, of whom 349 had a recurrence and 372 died (189 from breast cancer) during the nearly 12-year follow up. Those women consuming larger amounts of high-fat dairy products exhibited a trend toward higher breast cancer mortality (≥1.0 servings/day: HR=1.49; 95% CI: 1.00-2.24; P trend=0.05), higher all-cause mortality (P trend<0.001), and higher non–breast cancer mortality (P trend=0.007). The relationship with breast cancer recurrence was positive but not statistically significant.37
In response to Kroenke’s paper, Zucchetto et al analyzed data already in their possession from an Italian cohort of 1,453 women who had been diagnosed with breast cancer between 1991 and 1994 and who had been enrolled in a case controlled study. Among this group there had been 503 deaths, 398 from breast cancer. Zucchetto found no association between dairy consumption and all-cause or breast-cancer mortality. They did find a reduced mortality associated with cheese consumption (all-cause mortality for ≥1 servings/day vs <0.5 serving/day: HR=0.74; 95% CI: 0.56-0.99).38
Kroenke et al hypothesize the differences in their findings and those of the Italians are due to possible differences in Italian dairy farming practices that might lessen estrogen content of Italian milk products, though they provided no specific details about what these differences were.39
In a 1984 Italian case-control study, Talamini and coworkers had reported a positive association with intake of milk and dairy products, but not meat in a comparison between 368 women with breast cancer and 373 age-matched controls.40 In a 1986 French case-control study of 1,010 breast cancer cases and 1,950 controls, Lê and colleagues reported positive associations between cheese and full-cream milk, but not butter or yogurt, and breast cancer.41
These case-control studies fail to provide definitive answers. Is the difference in results due to the population studies or the production methods of the milk consumed as Kroenke suggests, or simply the inherent weaknesses in case-control studies?
Publication of a study in a peer-reviewed journal does not mean the conclusions reached are true. Past predictions based on epidemiology studies that have suggested specific nutrients would be cancer preventive have proven remarkably unreliable in predicting benefits in prospective clinical trials.42-46
It is imperative that every clinician learn to discern the type of evidence we are relying on to make decisions. We should be cautious about depending on conclusions from case-control studies related to diet and cancer. In regard to fat and breast cancer, it was the case-control studies that provided “evidence” in support of a hypothesis that was eventually disproven.
Case-control studies have several inherent weaknesses:
- Retrospective data. Collecting information retrospectively allows for collection bias, errors in memory, and other limitations.
- Recall bias. Patients in either group may provide answers that they think the researcher wants.
- Lack of defined questions. Some studies start as “fishing expeditions” looking for associations between a range of dietary or lifestyle behaviors with cancer incidence rather than seeking answers to a clearly defined hypothesis. On the other hand, these studies may be so focused on a particular hypothesis that they ignore potential confounders (eg, the weight loss seen in the WINS participants).
- Selection of controls. People motivated to participate in long-term, in-depth studies may not be representative comparators. They may desire to show off what an exemplary life style they have and may not be representative of the norm.
- Number of controls. It is now thought better to have multiple controls per actual case to improve accuracy (though more than 4 controls provides no additional benefit).47
Future case-control studies may provide more accurate predictions as the techniques of conducting them become more stringent. Currently greater attention needs to be directed at study methodology in order to judge the reliability of findings.
Current evidence does not support the idea that decreasing dietary fat will decrease risk of breast cancer. The possible exception is Chlebowski’s report that there may be benefit in lowering recurrence risk in hormone-negative cancers.
Evidence does suggest that a low–simple carbohydrate, high-EVOO, and high-nut diet may lower recurrence rates in women with a history of breast cancer, especially for women with a history of double- or triple-negative breast cancer.
We are already aware that epidemiological associations have been poor predictors of eventual outcomes in human clinical trials. We can take this a step further and say that studies that use case-control comparisons also have been poor predictors of prospective human trial outcomes.
* While the PubMed abstract for this study represents that it included only 40 cases of breast cancer, the original abstract lists that number as 400. The publishers have verified that the number 400 is correct.