Fruits, Vegetables, and Micronutrients in Relation to Breast Cancer Modified by Menopause and Hormone Receptor Status

Whether the consumption of fruits, vegetables, and antioxidant micronutrients is associated with a reduction in the incidence of breast cancer remains unresolved, with even the recent reviews or meta-analyses reaching conflicting conclusions (1-6). Several issues have been put forth as possibly contributing to the variability in results across studies, including whether the subjects were drawn from case-control or cohort studies (7); whether portion size (8), subgroups of fruits and vegetables (9), or vitamin supplementation (9) was considered in ascertaining dietary intake; whether subjects were categorized by menopausal status (1-6); whether the clinical characteristics of the cases' disease were considered, such as stage of disease and hormone receptor status (10); or whether chemotherapy had been initiated by the time of the subject's interview (11).

The analysis reported here, which uses data from a large population-based case-control study, was undertaken to address some of these issues. A food frequency questionnaire (FFQ) was used to assess diet, which inquired about portion size, as well as frequency of fruit and vegetable intake, and fruit juice consumption, along with vitamin supplement use. The large sample size permitted consideration of whether the association between intake of fruits, vegetables, and antioxidants varied with menopausal status, or with clinical characteristics of the cases' disease, including a joint measure of estrogen receptor (ER) and progesterone receptor (PR) status.

The Long Island Breast Cancer Study Project (LIBCSP) was a population-based case-control study that was conducted among adult female residents of Nassau and Suffolk counties, Long Island, NY (12). Institutional review board approval was obtained from all participating institutions, and signed informed consent was obtained for all respondents. In brief, English-speaking cases were newly diagnosed with a first primary in situ or invasive breast cancer between August 1, 1996 and July 31, 1997. Study personnel identified cases through rapid case ascertainment via active contact with local pathology departments. The case's physician was contacted to verify details of her diagnosis and permission to contact her. Of the 2,030 eligible cases, physician's consent was obtained for 1,837 (90.5%). Physician refusal was commonly due to illness of the patient.

English-speaking controls, who were residents of the same two counties as the cases but did not have a history of breast cancer, were randomly selected by random digit dialing for women ages <65 years at reference and by Health Care Finance Administration rosters for women ages ≥65 years. Controls were frequency matched to the expected distribution of cases by 5-year age group.

In-home interviews, conducted by trained interviewers, were completed by 1,508 (82.1%) of eligible case women (235 with in situ breast cancer and 1,273 with invasive breast cancer) and 1,556 (62.7%) of eligible controls. The interview covered such topics as reproductive history, physical activity, height, weight, alcohol use, active and passive cigarette smoke exposure, and exogenous hormone use. The medical records of case women were abstracted for ER and PR status. Among women in the dietary analysis, ER/PR data were available for 965 (66.0%) women. As reported previously (12), odds ratios (OR) for breast cancer were elevated in relation to late age at first birth, lower parity, little to no breast-feeding, and family history of breast cancer.

A total of 98.2% (n = 1,481) of cases and 97.6% (n = 1,518) control participants completed a scannable self-administered FFQ. Completion of the FFQ took an average of 36 minutes, which did not vary by age of the participants (12). The modified Block FFQ (13) included questions regarding usual frequency and portion size of ∼100 foods and beverages over the 12 months prior to interview. Intake of 13 fruits and fruit juices and 16 vegetables (excluding French fries) and use of multiple vitamins (i.e., one-a-day, therapeutic, and Stresstab type) and single supplements of vitamin A, β-carotene, vitamin E, calcium or Dolomite, vitamin C, and iron were also included in the FFQ. For food items and supplements, eight options of frequency ranged from <1 per month to 2+ per day. For food items, responses for portion sizes were relative to a stated average serving size with options of less, average, or more.

The frequency and portion sizes data were translated to daily intakes of nutrients from both dietary and supplement sources using the National Cancer Institute's DietSys version 3. Frequency and portion size of each food item were converted to a common denominator of 0.5 cup servings per week. Individual fruits and vegetables (see Appendix 1) were categorized into nine food groups: any fruits, fruit juices, and vegetables; any fruits and fruit juices; any fruits; fruit juices; citrus fruits; any vegetables; leafy vegetables; yellow vegetables; and cruciferous vegetables. These food groups were chosen based on the current breast cancer literature (13, 14). We also examined the dietary plus supplemental sources of β-carotene, vitamin C, and vitamin E and the dietary sources of α-carotene, β-carotene, cryptoxanthin, lutein plus zeaxanthin (further referred as lutein), lycopene, vitamin C, and vitamin E. For the main effect of fruits and vegetables, the cumulative frequencies of each food group and antioxidant nutrients were categorized based on quintiles of the control distribution (both premenopausal and postmenopausal women). The quintiles for premenopausal controls were slightly lower than for postmenopausal controls (data not shown); however, the ORs for the menopause-specific cut points (data not shown) did not differ from the ORs using cut points for all control women. Therefore, the latter cut points were used to facilitate comparison between premenopausal and postmenopausal women. To gain precision in stratified analyses, we created dichotomous variables by collapsing the lowest three fifths and the highest two fifths based on the similarity of the ORs in each of those groups.

To facilitate the results of our comparison with other studies, participants with daily energy intakes above or below 3 SDs of the log-transformed mean (16 cases and 22 controls) were dropped from the analyses (1). The case-control distribution of established risk factors for breast cancer among 1,463 cases and 1,500 controls included in these diet analyses (data not shown) were not substantially different from those identified among all LIBCSP participants (12).

Unconditional logistic regression (15) was used to estimate ORs and 95% confidence intervals (95% CI) for the association of selected fruit and vegetable groups and the antioxidants and breast cancer incidence. Age at reference date (defined as date of diagnosis for cases and date of identification for controls) and dietary energy intake (continuous) were included in all models. Factors considered as potential effect modifiers included physical activity from menarche to reference date, body mass index [weight (kg) / height (m)²] at reference, passive/active cigarette smoking, lifetime alcohol intake, family history of cancer, and linoleic acid. Factors considered as potential confounders included menopausal status, active and passive smoking, lifetime alcohol intake, menopausal status, physical activity from menarche to reference date, family history of cancer, season of interview, education, use of oral contraceptives, use of hormone replacement therapy, lactation, race, age at first birth, parity, body mass index at reference, body mass index at age 20 years, and single and multiple vitamin supplements.

Effect modification on the multiplicative scale was evaluated using the log likelihood ratio test to compare logistic models with and without the interaction term. The interaction term for menopausal status [which was defined based on information provided by the subject on her date of last menstrual period, prior surgical information on hysterectomy or oophorectomies, cigarette smoking status, and use of hormone replacement (12)] was statistically significant (P < 0.05). Confounding was evaluated separately for premenopausal and postmenopausal women, starting with a full model and using backward elimination. None of the potential confounders altered the estimates of either the combined fruit and vegetable intake or the micronutrient intake by >10% among premenopausal or postmenopausal women (data not shown). To test for potential confounding among the exposures of interest, final models were further adjusted for intake of fruits, vegetables, and antioxidants and use of antioxidant vitamin supplements; estimates from these models were not substantially different from excluding these covariates (data not shown). Tests of trend were conducted using the continuous values for nutrients and food groups. To test for potential heterogeneity by ER/PR status and stage of disease (invasive versus in situ cancer), stratified analyses were done for premenopausal and postmenopausal women. To assess for potential recall bias among case women who have already begun chemotherapy, as described by Potischman et al. (11), cases were stratified by whether chemotherapy had been initiated by the time of interview and by the interval of time from diagnosis to interview (above and below the mean of 96 days for cases).

As shown in Table 1, among postmenopausal women, breast cancer risk was decreased in relation to intake of fruits, fruit juices, and vegetables. For example, the ORs (95% CIs) were reduced by more than a third among postmenopausal women in the highest fifth of intake of any vegetables [0.63 (0.48-0.86)] or leafy vegetables [0.66 (0.50–0.85)] compared with women in the lowest fifth. Weaker inverse associations were found for yellow vegetables, cruciferous vegetables, and any intake of fruits and fruit juices. Citrus fruits were not related to breast cancer among premenopausal or postmenopausal women. In contrast, among premenopausal women, there was a suggestion of increase found for intake of any fruits, juices, and vegetables and any vegetables that seemed driven by the association with cruciferous vegetables (highest fifth of intake: OR, 1.76; 95% CI, 1.18–2.61). Consumption of any fruits and fruit juices, citrus fruits, any vegetables, leafy vegetables, and yellow vegetables was not consistently associated with premenopausal breast cancer.

There was no clear or consistent association between antioxidant nutrients and premenopausal breast cancer, as shown in Table 2. Among postmenopausal women, however, there was a consistent inverse association for most carotenoids (α-carotene, β-carotene, and lutein), particularly for lycopene (highest fifth of intake: OR, 0.66; 95% CI, 0.48–0.90; P for trend = 0.03). Postmenopausal women consuming ≥413.4 μg/d of dietary α-carotene had a 27% reduction in breast cancer incidence (OR, 0.73; 95% CI, 0.54–0.99) compared with women in the lowest fifth. Results for dietary intakes of vitamin C and vitamin E did not show a consistent inverse pattern.

Most cases, 89.2% of postmenopausal women and 82.0% of premenopausal women, were interviewed prior to the initiation of chemotherapy. The menopause-specific ORs for breast cancer in relation to fruit and vegetable intake did not vary substantially when cases were stratified by chemotherapy status (data not shown) or when cases were stratified by the timing of the interview relative to diagnosis (data not shown) compared with all controls.

When breast cancer cases were stratified by whether they were diagnosed with in situ or invasive disease, ORs in relation to the intake of fruits, vegetables, and micronutrients did not differ substantially from those for all women combined (data not shown). Among case women with information on diet and hormone receptor status, 562 (59.2%) were diagnosed with an ER+PR+ tumor, 136 (14.3%) had an ER+PR− tumor, 49 (5.2%) had an ER−PR+ tumor, and 220 (23.2%) had an ER−PR− tumor. The OR estimates among premenopausal women are not shown because they are based on limited numbers and did not allow for meaningful interpretation. Among postmenopausal women, the inverse associations between intake of fruits and vegetables were more pronounced among women with either ER+PR+ or ER+PR− tumors (Table 3), although this pattern was not evident for micronutrient intake (Table 4). For example, high intake of fruits and vegetables was associated with a 36% decrease in the OR (95% CI) among women with ER+PR+ tumors [0.64 (0.48-0.83)], whereas no relationship was found for women with ER−PR− tumors. Further, our results based on categorizing the case women based only on the ER status of their tumor (regardless of PR status) illustrate that the reduced risk is more pronounced among women with ER+ tumors. For example, the OR (95% CI) for intake of any fruits, fruit juices, and vegetables among women with ER+ tumors was 0.65 (0.51–0.82); among women with ER− tumors, the corresponding OR (95% CI) was 0.92 (0.64–1.32; ratio of the OR, 0.71; 95% CI, 0.48–1.05).

Fifty percent of controls used an antioxidant vitamin supplement in the previous 10 to 15 years. Estimates for combined supplemental and dietary sources of β-carotene, vitamin C, and vitamin E (data not shown) did not appreciably differ from those results presented in Table 2, which were based on dietary intake of these nutrients only. Looking at supplements alone, use of any antioxidant vitamin supplement was not associated with the incidence of breast cancer (OR, 0.93; 95% CI, 0.78–1.11) among all women; however, there was a suggestion of a decrease in breast cancer among multiple vitamin supplement users (OR, 0.80; 95% CI, 0.62–1.04) compared with nonusers. Multiple vitamin supplements, particularly one-a-day–type supplements, were the most common supplement among users. Results stratified by menopausal status were similar. Neither single antioxidant supplements, carotenoids, vitamin C, and vitamin E nor use of multiple and single supplements were associated with breast cancer (data not shown).

Inverse associations for intake of fruits and vegetables and consumption of antioxidant nutrients were found among postmenopausal but not premenopausal women. Among postmenopausal women, the inverse relationship was particularly strong for high consumers of all vegetables and leafy vegetables, α-carotene, β-carotene, lutein, and lycopene. In addition, postmenopausal women consuming the highest levels of lycopene had 34% lower incidence of breast cancer than women with the lowest intakes.

To the best of our knowledge, this is the first study to examine the influence of diet on breast cancer stratified by a joint measure of ER/PR status. The overall inverse association of fruit and vegetable intake was found among postmenopausal women, which was more pronounced among women with either ER+PR+ or ER+PR− tumors.

Potter et al. (10) has hypothesized that ER/PR status describes subtypes of breast cancer that may have different etiologies. Further, in a comparison of Danish women to Japanese women, Yasui and Potter (16) found that among Danish women the incidence of breast cancer for women with ER+PR+ tumor continues to increase with age after menopause, whereas the incidence for women with other tumor subtypes declines after menopause; in contrast, the incidence of breast cancer among Japanese women declines after menopause regardless of their ER/PR status. Several studies (17-24) have examined intake of fruit, vegetable, and micronutrients, stratified by ER status with inconsistent results. However, most studies were based on small numbers of subjects, which would limit their ability to detect a subtle difference between breast tumor subtypes, and none examined whether there was variation with ER and PR status concurrently. In our study, the inverse associations for fruits, vegetables, and micronutrients were stronger for those with ER+PR+ and ER+PR− tumors than they were for the other tumor types. Although our study is based on a large number of cases, the cell sizes for ER+PR− and ER−PR+ cases were still small.

Vitamin C, vitamin E, and the carotenoids, α-carotene, β-carotene, cryptoxanthin, lutein, and lycopene, act as antioxidants, which protect DNA from oxidative damage (25). In addition, each has other chemopreventive actions (25). α-Carotene, β-carotene, and cryptoxanthin, found in orange vegetables and fruits, may inhibit cell proliferation and inhibit cellular growth via conversion into retinols in the intestine (26, 27). A pooled analysis of 12 early case-control studies found that women in the highest fifth of β-carotene consumption had a 19% reduction in breast cancer incidence among postmenopausal women (1). A meta-analysis of literature from 1982 to 1997 found a similar result (5). However, only one of four recent cohort studies (28) found an inverse relationship for dietary β-carotene and only among premenopausal women (29). We found inverse associations for both β-carotene and α-carotene among postmenopausal women only.

Other carotenoids have been understudied in epidemiologic research of breast cancer, which may be due to difficulties associated with obtaining a comprehensive assessment with a FFQ. A large case-cohort study of dietary lycopene consumption (30) did not find an association with breast cancer even after stratification by menopausal status. Similarly, a review of observation studies of tomatoes and tomato-based products (31) did not support an association with breast cancer, although not all studies stratified by menopausal status. However, some but not all studies of high serum concentrations of lycopene have found inverse associations (32), although levels of lycopene in breast tissue were found in one study to be poorly correlated with dietary intake (31). Both α-carotene and lutein have been inversely associated with breast cancer in serum studies (32) and among premenopausal women in one prospective study of dietary intake (29).

In addition to its antioxidant potential, vitamin C is also critical for immune function (25). A meta-analysis of case-control studies found a 37% reduction in breast cancer risk among postmenopausal women in the highest fifth of consumption of vitamin C; the OR for premenopausal women was closer to the null (0.84, P = 0.26). A review (3) of five prospective studies found no association with breast cancer among both premenopausal and postmenopausal women. We found an inverse association, albeit not statistically significant, for the highest fifth of vitamin C intake, which was unchanged after the addition of supplemental sources of vitamin C.

In animal models, vitamin E has been found to inhibit tumors and reduce cell proliferation (25); however, available epidemiologic data do not support this effect (3). Our results also support a null association between dietary and supplemental sources of vitamin E and breast cancer.

We observed an inverse association for vegetables but not for fruits among postmenopausal but not premenopausal women. Other studies, but not all, have found results consistent with ours (1-6). A combined analysis of Italian breast cancer studies (4) and a meta-analysis (5) of 23 studies concluded that vegetable intake reduced breast cancer by 20% to 25%, although there was no association with fruit intake. However, in a pooled analysis of seven North American and European cohorts, neither vegetable nor fruit consumption was found to be associated with the risk of breast cancer regardless of menopausal status [for 100 g/d intake increments: relative risk (RR), 1.00; 95% CI, 0.97–1.02 and RR, 0.99; 95% CI, 0.89–1.00, respectively (7)].

On further analysis of individual vegetables, Smith-Warner et al. (7) found that broccoli and Brussels sprout consumption was associated with a 14% reduction in risk (for 0.5 cup increments: RR, 0.86; 95% CI, 0.72–1.02 and RR, 0.67; 95% CI, 0.35–1.27, respectively), which is consistent with our result for cruciferous vegetables among postmenopausal women. In addition, Smith-Warner et al. (7) found that increasing consumption of spinach (0.5 cup cooked increments: RR, 0.61; 95% CI, 0.33–1.15) was associated with decreasing risk of breast cancer. We found a 34% decrease in the OR with intake of nine or more 0.5 cup servings per week among postmenopausal women.

Each fruit or vegetable contains its own combination of numerous plausible nutritive and nonnutritive compounds that exert chemopreventive effects through stimulation of cell differentiation, cessation of cell division, antioxidant potential, antiestrogenic effects, induction of metabolic detoxification, and enhancement of immune function (25). The high correlation between antioxidant nutrients and their food sources makes it difficult to decipher whether antioxidants and/or other compounds are the preventive agent in fruits and vegetables. In our study, the inverse association between fruits and vegetables and breast cancer, even after adjustment for each antioxidant under study and supplement use, suggests that there may be additional chemopreventive agents in fruits and vegetables that may perform outside the oxidative stress pathway. However, our results of a pronounced reduction of ER+ breast cancer associated with fruits and vegetables among postmenopausal women seem to support the hypothesis that endogenous estrogen levels may contribute to oxidative stress (33). The biological mechanism for the effect measure modification by menopausal status observed in our data and by others is not clear. Given the hypothesized chemopreventive effects of fruits and vegetables, consumption of them would be expected to be protective for both postmenopausal and premenopausal women, which makes our findings for the later group particularly perplexing. Further research based on large numbers of both premenopausal and postmenopausal women is needed to help clarify this issue.

There may be noncausal explanations for our results. Measurement error in a FFQ, such as our inability to accurately capture intake of all fruits and vegetables, is an issue for all study designs. FFQs have been shown to rank individual's relative intake of micronutrients (8, 33). In addition, validation (8), intervention (34), and feeding studies (35) have shown that the FFQ may be an adequate, albeit not perfect, instrument for measuring relative fruit and vegetable intake. However, the nondifferential misclassification errors observed in the FFQ likely attenuate the estimated RR toward the null; consequently, true RRs are probably greater than they seem (8, 36). In our data, use of alternate cut points based on menopausal status did not alter our estimates; this suggests that our categorization of women and our results are robust.

Case-control studies have additional biases related to the potential for differential recall, which may be due to the cases' cancer diagnosis. In addition, recall bias may also be related to use of chemotherapy among the cases. Potischman et al. (11, 13) found that breast cancer cases, who currently reported that chemotherapy had been initiated by the time of the interview, had higher intakes of macronutrients and calories than did cases not receiving chemotherapy. In our study, great effort was taken to interview cases prior to chemotherapy. The distribution of intake of total energy and micronutrients did not vary substantially by chemotherapy status at the time of interview or by number of days from diagnosis to interview.

Residual or uncontrolled confounding may bias the association between consumption of fruits and vegetables and risk of breast cancer. Many lifestyle factors that may potentially confound the relation between breast cancer and fruit and vegetable intake are highly correlated with the exposure of interest (37, 38), making it difficult to firmly establish independent effects of fruits and vegetables on cancer risk (39). Our questionnaire had extensive assessment of exposures over the life course including recreational physical activity from menarche to age at diagnosis, lifetime active and passive smoking exposure, lifetime alcohol consumption, and vitamin supplement use. However, when we controlled mutually or individually for these factors, our results were not substantially altered.

Our study supports an inverse association between intake of fruit, vegetable, and carotenoids and breast cancer among postmenopausal women that is more pronounced among women with ER+PR+ tumors. Increasing intakes of leafy vegetables were associated with decreasing incidence of breast cancer regardless of ER/PR status. A modest inverse association was found between high intakes of lycopene and postmenopausal breast cancer. Despite the conflicting results from epidemiologic studies, the public health significance and the biological plausibility for a beneficial effect of fruits and vegetable suggest that further research should be conducted, particularly with respect to breast cancer, for which few modifiable risk factors have been identified. Future dietary studies, particularly those with large numbers of premenopausal women, should incorporate gene-diet interactions and stratify by tumor characteristics.

We thank the following for their valuable contributions to the LIBCSP: members of the Long Island Breast Cancer Network; the 31 participating institutions on Long Island and New York City, NY; our NIH collaborators, Gwen Colman, Ph.D., National Institutes of Environmental Health Sciences; G. Iris Obrams, M.D., Ph.D., formerly of the National Cancer Institute; members of the External Advisory Committee to the population-based case-control study: Leslie Bernstein, Ph.D. (Committee Chair); Gerald Akland, M.S.; Barbara Balaban, M.S.W.; Blake Cady, M.D.; Dale Sandler, Ph.D.; Roy Shore, Ph.D.; and Gerald Wogan, Ph.D.; as well as other collaborators who assisted with various aspects of our data collection efforts, including Regina Santella, Ph.D.; Gail Garbowski, M.P.H.; H. Leon Bradlow, Ph.D.; David Camann, B.S.; Martin Trent, B.S.; Ruby Senie, Ph.D.; Carla Maffeo, Ph.D.; Pat Montalvan; Gertrud Berkowitz, Ph.D.; Margaret Kemeny, M.D.; Mark Citron, M.D.; Freya Schnabel, M.D.; Allen Schuss, M.D.; Steven Hajdu, M.D.; and Vincent Vinceguerra, M.D.