A Prospective Study of Diet and Benign Breast Disease

Breast cancer is the most common cancer among women in America and throughout the developed world, but it occurs much less frequently in Asia (1). Women who move from countries with low breast cancer risk to countries with high risk acquire the higher breast cancer risk of their destination (2), suggesting that potentially modifiable, and, therefore, preventable, environmental factors, such as diet, influence breast cancer risk. Results from studies evaluating the relationship between adult diet and breast cancer are, however, inconsistent. Higher breast cancer incidence rates are consistently seen in countries with higher per capita consumption of fat (3). In contrast, case-control studies have generally reported only modest increases in risk of breast cancer associated with increased dietary fat intake (4), whereas prospective studies have found little evidence for an association (5). Attention has also focused on the potential role of antioxidants in breast cancer development, but again the data are largely inconsistent (5, 6).

One reason for the inconsistent results might be that dietary exposures may be more important at earlier stages in the disease process than are commonly considered in epidemiologic studies of cancer (7). Benign proliferative epithelial disorders of the breast, particularly those showing signs of atypia, are associated with approximately 1.5- and 4-fold increased risks of cancer, respectively, compared with women with non-proliferative breast tissue (8-10), suggesting that they may be a marker for subsequent breast cancer. Studies of the relation between diet and benign breast disease (BBD), thus, offer an opportunity to evaluate the role of diet at an earlier stage in the process of breast carcinogenesis.

Several studies that attempted to evaluate the relationship between dietary factors and BBD have reported conflicting results. Some studies found that fat intake was associated with increased risk of proliferative BBD (7), and particularly of atypical hyperplasia (AH) (11), but other studies found no association (12, 13). Increased intake of vitamin A supplements (14) and retinol and carotene (7) has been associated with reduced risk of BBD, although the latter effects were seen only in comparison to community controls (unbiopsied) and not compared with a second control group of women who underwent biopsy but were not found to have proliferative BBD. Other studies have reported no association between these nutrients and BBD (12, 13, 15). Reduced risks have also been reported for increased intake of vitamins C (12) and E (15), and vitamin E has been used for treatment of BBD, but again these effects have not been confirmed by others (16). Early reports suggested that methylxanthines (caffeine, theobromine, and theophylline) might influence risk of BBD (17, 18) and that caffeine restriction improved symptoms of BBD (19). However, this has not been supported by more recent studies (20) although in one study total methylxanthine intake was positively associated with risk of BBD with severe atypia in comparison to community controls (21).

Many of these studies were hampered by relatively small numbers of women with confirmed BBD. Furthermore, the majority of the previous studies were retrospective introducing the possibilities of both selection and recall bias. We evaluated the relationship between some dietary factors and subsequent diagnosis of BBD in a cohort of 58,628 U.S. women. Specific hypotheses were that higher total fat and saturated fat intake would be associated with increased risks of proliferative BBD but that higher intake of vegetable or monounsaturated fat, vitamin A, carotenoids and vitamins C and E would be associated with reduced risk. We also evaluated the relationship between caffeine intake and risk of developing proliferative disease with or without atypia.

The Nurses' Health Study II is a prospective cohort study that began in 1989 when 116,671 female nurses ages 25 to 44 years completed a mailed questionnaire about their medical history and lifestyle. Subsequent questionnaires requesting updated information on risk factors and medical events, including diagnosis of BBD, have been mailed to the women every 2 years. In 1991, the questionnaire included a 133-item food frequency questionnaire. The protocol for the study was approved by the Human Research Committees of the Brigham and Women's Hospital and the Harvard School of Public Health, Boston, MA.

Follow-up for the present study began in 1991 when diet was first measured. We excluded 18,864 women (16%) who did not complete the food frequency questionnaire and 2,361 (2%) who had incomplete dietary data (more than 70 food items blank) or implausible total energy intake (<800 or >4,200 kcal/d). We also excluded women who reported BBD (n = 32,217) or cancer other than non-melanoma skin cancer (n = 754) before 1991 as well as 3,079 women who were pregnant when they completed the dietary questionnaire because of concerns that the reported dietary data might not have reflected their usual intake. This left a baseline cohort of 59,396 women. A total of 768 women (1.3%) did not contribute any follow-up information and a further 817 (1.4%) and 1,696 (2.9%) were excluded in 1993 and 1995, respectively, because they had no further follow-up information after that point.

The food frequency questionnaire measures intake of a wide variety of nutrients and dietary factors and has shown good validity in similar study populations (correlations for energy-adjusted nutrients with dietary records are generally 0.60 to 0.70). The validity of a similar questionnaire was assessed among 191 older women in the Nurses' Health Study by comparison with two 1-week diet records (average r for all nutrients = 0.62; ref. 22). In this latter validation, correlations for energy-adjusted nutrients (without supplements) de-attenuated for week-to-week variation in the diet records were 0.57 for total fat, 0.68 for saturated fat, 0.48 for polyunsaturated fat, 0.58 for monounsaturated fat, 0.79 for total vitamin A, and 0.76 for vitamin C. Nutrients calculated from the food frequency questionnaire are correlated with their corresponding biochemical indicators: plasma β-carotene (r = 0.30-0.42; refs. 23-25), plasma vitamin E (r = 0.41-0.53; refs. 23-25), plasma folate (r = 0.35-0.51; refs. 26, 27), adipose linoleic acid (r = 0.35-0.37; refs. 24, 28), adipose trans fatty acid (r = 0.51; ref. 28), and adipose N-3 fatty acids (r = 0.48-0.49; refs. 24, 28).

In 1993, 1995, and 1997, the women were asked if, since the previous questionnaire, a physician had told them they had fibrocystic or other BBD and, if so, if this had been confirmed by tissue biopsy or aspiration. Two definitions of reported BBD were considered: firstly, all new reports of BBD and secondly a subgroup comprising only those cases reported as confirmed by biopsy. The date of diagnosis of BBD was estimated as the mid-point between the date of return of the questionnaire reporting incident BBD and the return date of the most recent questionnaire before that. Women who reported BBD and breast cancer on the same questionnaire were excluded from the case groups and from further follow-up because it was not possible to determine the relative timing of the two diagnoses.

Women who reported a diagnosis of biopsy-confirmed BBD on the 1993, 1995, or 1997 questionnaires were contacted to seek confirmation of the diagnosis and permission to obtain their pathology records. Of the 1,956 women who reported biopsy-confirmed BBD and who also had valid dietary data, 1,577 (81%) confirmed the diagnosis and granted permission for review of the pathology slides and records from their biopsy. The main reasons for non-response were that the woman could not be contacted (7%), did not confirm she had had a biopsy (7%), or did not give permission for her slides to be reviewed (6%). Adequate pathology material was obtained and reviewed for 1,409 women (89% of those who confirmed the biopsy and gave permission). Of these, 1,328 (94%) were confirmed to be eligible cases and a valid diagnosis was obtained. The main reasons for exclusion were that the pathology specimen did not contain breast tissue (n = 37) or the biopsy date was before the start of follow-up (n = 22). In addition, women were excluded if their biopsy date was after the date they reported BBD (n = 8) or if they had a diagnosis of breast cancer within 1 year of diagnosis of BBD (n = 3), or a diagnosis of carcinoma in situ (n = 5). Women without valid pathology material did not differ significantly from eligible women for whom material was available with respect to age, family history of breast cancer, body mass index (BMI), use of oral contraceptives (OC), or total fat or vitamin intake.

The benign breast lesions were initially classified by one of four pathologists (S.J.S., J.L.C., T.J., or G.P.) as normal or non-proliferative, proliferative without atypia, or AH. This classification was done after standard training and according to standard criteria (29). In a reliability study, a random sample of 53 cases (22 normal/non-proliferative, 31 proliferative) was re-reviewed in a blinded fashion by a second pathologist and the results were found to be highly reproducible (κ = 0.72). In addition, a proportion of biopsies were jointly reviewed by two pathologists to ensure consistency and any biopsy specimens that showed atypia or questionable atypia were jointly re-reviewed (by S.J.S., J.L.C., and T.J. or G.P.) and a consensus diagnosis reached. Specimens with intraductal papilloma, radial scar, sclerosing adenosis, fibroadenoma, fibroadenomatous change, or moderate to florid ductal hyperplasia in the absence of AH were classified as proliferative without atypia.

Each participant contributed person-time of follow-up from the time she returned the dietary questionnaire in 1991 until the first of any of the following: return date of the 1997 questionnaire, death from any cause, diagnosis of BBD, report of cancer other than non-melanoma skin cancer, or loss to follow-up. For dietary variables, energy-adjusted estimates of nutrient intake (30) were categorized by quartiles based on the distribution of intakes in the cohort. Dietary factors considered were: fat (total, animal, vegetable, saturated, monounsaturated, and polyunsaturated fats), fiber, vitamin A, retinol (equivalents of vitamin A), vitamin C, vitamin E (α-tocopherol), carotene, folate, and caffeine. Women were also classified as users or non-users of vitamin supplements (multivitamins and/or vitamins A, C, E, β-carotene, and folic acid).

For time-varying covariates, such as OC use and BMI, cases and person-time were updated every 2 years. Cox proportional hazards regression was used to estimate rate ratios (RR) and 95% confidence intervals (95% CI) compared with women in the lowest quartile of intake, after controlling for potential confounders. Estimates were adjusted for age (months), time period (three 2-year periods), history of breast cancer in mother or sister (yes/no), BMI (<21, 21-22.9, 23-24.9, 25-29.9, 30+), OC use (never, past, current), total energy intake (quartiles), and use of vitamin supplements (yes/no). Further adjustment for age at menarche, smoking status, menopausal status, and parity did not appreciably alter the effect estimates and these variables were not included in the final models. Tests for trend across quartiles of intake were conducted by assigning the median intake in each quartile to all women in that category and treating these values as a continuous variable.

In addition, case-case comparisons were conducted among the subgroup of women with histologically confirmed BBD to reduce the possibility of bias due to selective referral of women for biopsy. Logistic regression was used to calculate prevalence odds ratios and 95% CI after adjusting for confounders.

Between 1991 and 1997, the 58,628 women in the analysis contributed 321,973 person-years of follow-up. The distributions of selected characteristics of the population at baseline are shown in Table 1 according to intake of total fat, vitamin E and caffeine, and use of vitamin supplements. Women in the highest quartile of fat intake were more likely to be current smokers and less likely to be nulliparous than women in the lowest quartile of fat intake. They also had lower intake of vitamin E (and other vitamins, data not shown), were less likely to take vitamin supplements, and had higher caffeine intake. The opposite trends were seen for women with high and low vitamin E intake. High caffeine intake was positively associated with current smoking.

During follow-up, a total of 9,645 women reported a physician diagnosis of BBD and, of those, 1,956 women (20.3%) reported that this diagnosis was confirmed by tissue biopsy. After adjustment, total fat intake (RR = 1.06, 95% CI 1.00-1.12 for the highest versus lowest quartile, P for trend = 0.03), vegetable fat (RR = 1.07, 95% CI 1.01-1.13, P = 0.007), monounsaturated fat (RR = 1.07, 95% CI 1.01-1.13, P = 0.03), and polyunsaturated fat (RR = 1.11, 95% CI 1.05-1.18, P = 0.002) were all positively associated with any report of physician-diagnosed BBD, but the associations were weak. Intakes of total vitamin A (RR = 0.92, 95% CI 0.87-0.98, P = 0.002) and retinol (RR = 0.91, 95% CI 0.85-0.97, P = 0.001) were inversely associated with any report of BBD. However, these effects were no longer seen when the analyses were restricted to a more stringent case group, including only those cases of BBD that were reported as confirmed by tissue biopsy. In contrast, increasing intake of vegetable fat was associated with a statistically significantly lower rate of biopsy-confirmed BBD (RR = 0.84 for the highest versus lowest quartile, 95% CI 0.74-0.95, P for trend = 0.009). Use of vitamin supplements (multivitamins, vitamins A, C, E, β-carotene, and/or folate) was also associated with a lower rate of reported biopsy-confirmed BBD, regardless of the type of supplements used (RR = 0.89, 95% CI 0.81-0.97 for any supplement use versus no supplements). There was no association between intake of animal fat, saturated fat, fiber, caffeine, vitamins C and E, carotene, or folate and report of any BBD or biopsy-confirmed BBD.

Tables 2 and 3 show the relationship between the dietary variables and incidence of histologically confirmed BBD. The results are shown separately for non-proliferative BBD (475 cases), proliferative BBD without atypia (786 cases), and AH (67 cases).

As for reported biopsy-confirmed BBD, increasing intake of vegetable fat was associated with significantly lower rates of proliferative BBD without atypia; however, it was also associated with somewhat higher rates of AH (Table 2). None of the other types of fat were associated with any diagnosis of BBD. There was also no association between fiber intake and rates of non-proliferative BBD or proliferative BBD with or without atypia.

Table 3 shows the association between total intake of selected micronutrients (from dietary sources and supplements combined) and histologically confirmed BBD. Increasing intake of vitamin C was associated with a lower rate of non-proliferative BBD while increasing vitamin E intake was associated with lower rates of proliferative BBD without atypia; however, neither of these effects reached statistical significance (Table 3). Intakes of vitamin A, retinol, carotene, and folate were not associated with either non-proliferative or proliferative BBD. There was also no association with either α-carotene or β-carotene when these were considered separately (data not shown). Similar patterns were seen when only dietary sources of micronutrients were considered, but intake of dietary vitamin E was now also associated with a reduced rate of non-proliferative BBD (RR = 0.78, 95% CI 0.31-1.01 for the highest versus lowest quartiles, P for trend = 0.08).

Somewhat different patterns were seen when the case-group was restricted to the 67 women diagnosed with AH although interpretation of these results is difficult because of the small number of cases. There were no clear associations with total intake (or dietary intake only, results not shown) of individual vitamins (Table 3); however, use of vitamin supplements (multivitamins and/or vitamins A, C, E, β-carotene, or folate) was associated with an almost 50% lower rate of atypia (RR = 0.56, 95% CI 0.33-0.93). A similar effect was seen for multivitamin supplements alone (RR = 0.57, 95% CI 0.33-0.98). A significantly higher rate of AH was seen among women in the highest quartile of caffeine intake with a RR of 2.46 (95% CI 1.11-5.49, P for trend = 0.06) for women in the fourth quartile compared with the lowest quartile. Neither supplement use nor caffeine intake was related to either non-proliferative BBD or proliferative BBD without atypia.

To ensure that the associations seen for caffeine and supplement use and rates of atypia were not biased by selective referral of women for biopsy, we conducted additional case-case analyses comparing women with a histological diagnosis of proliferative BBD without atypia and women with AH to those with non-proliferative disease. As expected, there was no association with proliferative BBD without atypia but both caffeine (prevalence odds ratios = 2.43, 95% CI 1.04-5.72 for the highest versus lowest quartiles) and supplement use (prevalence odds ratios = 0.59, 95% CI 0.34-1.02) were associated with AH.

To explore these relations more thoroughly, additional analyses were conducted for food groups that contribute to fat and micronutrient intake. Foods considered included dairy foods, red meat, fruit and vegetables and also breakfast cereals because these are commonly fortified with multivitamins. We observed a weak positive association between increasing consumption of dairy foods (P for trend = 0.05), particularly high fat dairy foods (whole milk, cream, ice cream, high-fat cheeses, and butter, P for trend = 0.04) and rates of non-proliferative BBD. Proliferative BBD without atypia was unrelated to any of the food groups considered but the rate of AH was lower among women who ate one or more servings of breakfast cereal per day compared with those who ate it less than once a week (RR = 0.26, 95% CI 0.06-1.09).

These data provide no support for the hypothesis that high total fat or saturated fat intake is associated with development of proliferative BBD. This adds weight to the belief that these fats also do not play a major role in the development of breast cancer, although it is possible that they could influence breast cancer risk among women with BBD. Modest increases in the rates of non-proliferative BBD were seen with increasing consumption of high fat dairy foods but this type of BBD does not seem to be associated with an increased risk of breast cancer (8). Women in the highest total fat intake groups were no more likely to be diagnosed with proliferative BBD with or without atypia than women with lower fat intake. We did, however, observe lower rates of proliferative BBD with increasing consumption of vegetable fat and this association persisted after adjustment for vitamin E intake (vegetable fat is a major source of vitamin E). This inverse association is consistent with data suggesting that increasing intake of vegetable fat in adolescence may also be associated with a lower risk of proliferative BBD (RR = 0.73, 95% CI 0.55-0.96 for the highest versus lowest quartile (31).

The lack of association between increased total fat intake and BBD is consistent with several earlier reports (16, 32), including another prospective study conducted within a breast cancer screening study (13), although three studies had previously suggested a positive association (7, 11, 14). Fat intake was positively associated with BBD in a retrospective comparison of 383 women with biopsy-confirmed proliferative BBD and 192 women who underwent biopsy but were not found to have proliferative BBD (7) but this association disappeared after adjustment for total energy intake. Furthermore, in the same study, there was no association with fat when the women with BBD were compared with 383 unbiopsied community controls. In a second retrospective study conducted in Israel, 107 women with atypical lesions reported higher consumption of all foods than both hospital and community controls (11). This difference was attributed primarily to increased consumption of foods containing more than 10% fat although the authors were not able to adjust for energy intake. The third study reported an increased risk of atypia associated with highest consumption of meat and dairy fat; however, this was based on only 32 cases with atypia and the assessment of diet was based on reported intake of only 39 food items selected for their fat or vitamin A content (14).

Increasing fiber intake was not associated with risk of BBD overall or with the non-proliferative and proliferative subtypes considered separately. In contrast, high fiber intake in adolescence was associated with a significantly lower risk of proliferative BBD (31).

Our findings, along with those from case-control studies (7, 12) and another prospective study (13), provide little support for an association between vitamin intake and BBD overall. Women with higher intakes of vitamin E had a modest reduction in risk of proliferative BBD and women with higher intake of vitamin C were slightly less likely to be diagnosed with non-proliferative BBD, but neither of these associations reached statistical significance. The weak association between vitamin E and proliferative BBD is, however, consistent with a non-significant effect reported for vitamin E intake in adolescence and incidence of BBD (RR = 0.79, 95% CI 0.61-1.04; ref. 31). Experimental data have suggested that vitamin E can inhibit mammary tumors in rodents (33), and vitamin E has long been used for treatment of BBD (34), although randomized trials have not shown any benefits of vitamin E over placebo (35, 36). Two retrospective studies have evaluated the association between vitamin E and risk of BBD, one reporting no association (16) while the other, in contrast to the present results, found increasing intake was associated with a reduction in risk of AH but not proliferative BBD without atypia (10). Both this latter study and the present analysis included relatively few women with atypia, making it difficult to draw clear conclusions about this group. There is little evidence to support an association between vitamin E and breast cancer risk (37-39).

Use of vitamin supplements (multivitamins, vitamins A, C, E, β-carotene, or folate) was associated with a significantly reduced risk of reported biopsy-confirmed BBD. When the different histological subtypes of BBD were considered separately, the association was restricted to AH with no association seen for non-proliferative BBD or proliferative BBD without atypia. This apparent protective effect may be due either to the combination of multiple vitamins in the supplements or to some other component of multivitamins, or alternatively, use of vitamin supplements may be a marker for some other lifestyle factor that is associated with a reduced risk of BBD. The RRs were adjusted for age and family history of breast cancer, and allowing for other factors, including OC use, BMI and age at menarche, parity and menopausal status, did not appreciably alter the estimates. Furthermore, the association between most of these factors and BBD is weak, suggesting that the relation with supplement use is not due to uncontrolled confounding. The observation that a reduced risk of AH was also seen among women eating the highest levels of breakfast cereal, a food item that is commonly supplemented with multiple vitamins, strengthens belief that it might be the combination of vitamins that is important.

Initial reports from uncontrolled and unrandomized studies suggested that caffeine restriction improved BBD symptomatology. It was suggested that methylxanthines (caffeine, theobromine, and theophylline) inhibited cyclic adenosine monophosphate and guanylic acid and activated a protein kinase leading to overproduction of fibrous tissue and cystic fluid and, thus, BBD. Caffeine, from coffee, tea, and chocolate, is the major dietary methylxanthine. Results of studies investigating the association between caffeine and/or methylxanthine intake and BBD are, however, conflicting. Two early hospital-based case-control studies found no association between a discharge diagnosis of BBD and coffee consumption (40, 41) while a third found a significant positive association between coffee consumption and both a history and clinically diagnosed BBD among pairs of twins (42). More recent studies have required histologic confirmation of BBD and have evaluated the risk separately for different types of BBD. Of these, two studies reported significant positive associations that were strongest among women with high-risk types of BBD (17, 18) while a third found no association overall but an increased risk among the 40 women with severe atypia (21). Two other studies, however, did not find any association with BBD either overall or among the subgroup with atypia (20, 43).

To our knowledge, ours is the first prospective study to evaluate the relation between caffeine consumption and diagnosis of BBD and our data do not support the hypothesis that caffeine intake influences risk of BBD overall. However, consistent with three of five previous retrospective studies, we observed a significantly increased risk of AH associated with increasing caffeine intake.

A major problem in any study of BBD is that of identifying cases. It is possible that more health-conscious women with lower fat and higher vitamin intake might be more likely to identify and seek advice for a breast lump and, thus, be diagnosed with BBD, thereby biasing measures of effect upward. However, given the increasing awareness among women about breast cancer, it seems unlikely that members of this cohort of registered nurses would differ appreciably in their propensity to seek advice. Furthermore, the results were essentially the same when the analysis was restricted to women who had undergone some form of breast screening (mammography or examination by a physician) in the previous 2 years, and who, therefore, had similar opportunity for diagnosis of BBD. In addition, the positive association with caffeine intake and the inverse association with supplement use persisted when women with atypia were compared with women who were also biopsied for BBD but diagnosed with non-proliferative disease. It is, therefore, highly unlikely that these effects could be due to selective referral of high coffee drinkers and/or supplement users for biopsy.

Advantages of the present study include the high follow-up rates and the centralized review of pathology slides to ensure consistency. Another major strength is that diet was measured before women were diagnosed with BBD thereby ruling out the possibility for recall bias. However, risk factors may exert their effect many years before diagnosis of disease, thus, it is possible that the dietary exposures in the present study (measured a maximum of 6 years before diagnosis) may be too recent to be associated with incidence of BBD. Although it is likely that recent diet will reflect past diet, any misclassification is likely to bias estimates of effect toward the null. It is also possible that exposures much earlier in life may be important and this suggestion is supported by the observation that stronger apparent protective effects for vegetable fat and vitamin E were seen in an evaluation of adolescent diet and risk of proliferative BBD in adulthood (31) than in the present analyses.

In conclusion, the present data suggest that higher total fat and saturated fat consumption is not associated with an increased risk of BBD. This adds weight to the argument that high-fat diets also do not play a major role in the development of breast cancer although it is possible that they could influence breast cancer risk among women with BBD. The data also provide little evidence for a major role of micronutrients, including vitamins A, C, and E in development of breast disease, but do raise the possibility that high caffeine intake may increase, and use of vitamin supplements decrease risk of developing atypia. The suggestion that risk factors might vary for proliferative BBD with and without atypia is interesting; however, despite the large size of this cohort, only a relatively small number of women were confirmed histologically as having atypia making it difficult to draw clear conclusions about this group. This emphasizes the requirement for future studies to be large enough to have sufficient power to be able to evaluate risk factors separately for proliferative disease and AH.

We thank the participants of the Nurses' Health Study II for their dedication to this study and Dr. Gloria Peiro for her assistance with classifying the biopsies for the benign breast disease cases.