|
|
| |||||||||||||||||||||||||||||||||||||||
Background Although it has been hypothesized that large intakes of the antioxidant vitamins C, E, and A reduce the risk of breast cancer, few prospective data are available.
Methods We prospectively studied 89,494 women who were 34 to 59 years old in 1980 and who did not have diagnosed cancer. Their intakes of vitamins C, E, and A from foods and supplements were assessed at base line and in 1984 with the use of a validated semiquantitative food-frequency questionnaire.
Results Breast cancer was diagnosed in 1439 women during eight years of follow-up. After multivariate adjustment for known risk factors, the relative risk among women in the highest quintile group for intake of vitamin C as compared with the risk among those in the lowest quintile group was 1.03 (95 percent confidence interval, 0.87 to 1.21); for vitamin E, after vitamin A intake had been controlled for, the relative risk was 0.99 (95 percent confidence interval, 0.83 to 1.19). In contrast, among women in the highest quintile group for intake of total vitamin A the relative risk was 0.84 (95 percent confidence interval, 0.71 to 0.98; P for trend = 0.001). Among women in the lowest quintile group for intake of vitamin A from food, consumption of vitamin A from supplements was associated with a reduced risk (P = 0.03). The significant inverse association of vitamin A intake with the risk of breast cancer was also found on study of data based on the 1984 questionnaire and four years of follow-up.
Conclusions Large intakes of vitamin C or E did not protect women in our study from breast cancer. A low intake of vitamin A may increase the risk of this disease; any benefit of vitamin A supplements may be limited to women with diets low in vitamin A.
Vitamin C has been postulated to decrease the risk of cancer in general10 and breast cancer in particular11. Vitamin E is a potent, lipid-soluble antioxidant12 and inhibits mammary tumors in rats in some13,14 but not all15 experiments. Clinically, vitamin E is sometimes recommended for the treatment of benign breast disease,16 itself a risk factor for breast cancer. The term "vitamin A" refers to a combination of preformed vitamin A (retinol, retinyl esters, and related compounds, hereafter referred to as preformed vitamin A) and carotenoids with provitamin A activity (hereafter referred to as carotenoids), which can be partially converted to retinol17. The total intake of vitamin A (preformed vitamin A and carotenoids) has been inversely associated with breast cancer in several case-control studies11 and in one small prospective study18. To assess these relations further, we examined the associations between intake of vitamins C, E, and A and breast cancer in the Nurses' Health Study, a prospective investigation of a cohort of American women.
Methods
Study Cohort
In 1976, 121,700 female registered nurses 30 to 55 years old who were living in 11 large states completed a mailed questionnaire about known and suspected risk factors for cancer19 and cardiovascular disease20. Every two years, follow-up questionnaires have been sent to update information on menopausal status, parity, and other potential risk factors as well as to identify newly diagnosed cases of cancer and other diseases. In 1980, the questionnaire was expanded to include an assessment of diet. The cohort described in the present analysis consisted of 89,494 women who completed dietary questionnaires and had no previous diagnosis of cancer (other than nonmelanoma skin cancer).
Food-Frequency Questionnaire
A detailed description of the semiquantitative food-frequency questionnaire used in 1980 (61 food items) and documentation of the reproducibility of its results and its validity have been presented elsewhere21,22,23. In 1984, we used an expanded, 121-item, semiquantitative food-frequency questionnaire23. For both questionnaires, nutrient intake was computed by multiplying the frequency of consumption of each unit of food by the nutrient content of the specified portions. The vitamin composition of the foods was obtained from the Department of Agriculture24. The questionnaires also included detailed questions on the use of specific vitamins and brands of multivitamins as well as the dose and duration of use. We have developed a comprehensive data base on multivitamin preparations that provides the dose of vitamin A (in the early 1980s, mostly from preformed vitamin A), vitamin C, and vitamin E in each preparation. We adjusted nutrient values for total energy intake by regression analysis25. We also computed indexes for the major food groups by summing the intakes of each component food.
Validation of Questionnaire
Nutrient intakes calculated from the 1980 questionnaire were compared with those recorded by 173 Boston-area women who kept diet diaries covering a four-week period21,22. The correlation coefficients for nutrient intake (adjusted for energy intake) calculated from the two methods were r = 0.49 for vitamin A and r = 0.75 for vitamin C. Because the data on the nutrient composition of vitamin E in the diet-record data base were inadequate, it was not possible to determine the validity of data on the intake of this vitamin. The estimate of vitamin E intake from the questionnaire did, however, predict the plasma level of alpha-tocopherol in two other studies (r = 0.34,26 r = 0.5227). Similarly, the estimate of dietary levels of carotenoids from the questionnaire correlated with the levels of total carotenoids in plasma (r = 0.35,26 r =0.4427).
Cases of Breast Cancer
Follow-up questionnaires asking whether breast cancer had been diagnosed were mailed to all participants in 1982, 1984, 1986, and 1988, and we attempted to interview nonrespondents by telephone. The response rates were 97 percent in 1982, 90 percent in 1984, 93 percent in 1986, and 92 percent in 1988; the overall follow-up rate for nonfatal outcomes was 96 percent of the total possible person-years. Deaths in the cohort were identified by reports from family members, the Postal Service, and a search of the National Death Index; we estimate that 98 percent of all deaths were reported to us28.
When a case of breast cancer was identified, we asked the participant (or her next of kin if she had died) for permission to obtain hospital records and pathology reports. These were obtained for more than 95 percent of the cases and confirmed the participants' reports in more than 99 percent of these cases. After the few cases not confirmed by records were excluded, 1566 cases remained. We also excluded 127 cases of breast cancer recorded as in situ lesions, which left 1439 cases. We included 80 cases for which no pathology reports could be obtained, because the accuracy of the participants' reports was high and because analyses limited to confirmed cases yielded results virtually identical to those of analyses including these 80 cases.
Statistical Analysis
Participants were categorized according to their quintile of nutrient intake adjusted for energy intake, as calculated from the 1980 and 1984 questionnaires. Quintile cutoff points for nutrient intake unadjusted for energy intake were similar to energy-adjusted values. In addition, we classified participants according to their use of multivitamins (dose and duration) and their use of specific vitamin A, C, or E supplements, as well as their dose of vitamin A from specific supplements and multivitamins combined.
For each participant without a diagnosis of breast cancer, follow-up time equal to the number of months between the return of the 1980 and 1982 questionnaires was allocated to each dietary variable recorded in 1980. Similarly, for each subsequent two-year interval, additional person-months were allocated according to the 1980 dietary status. For each participant who died, follow-up time was allocated up to the date of death. Participants were reclassified according to their menopausal status and other covariates at the start of each two-year interval if their status had changed.
Incidence rates were calculated by dividing the number of events by the number of person-months of follow-up. The relative risk was calculated by dividing the incidence rate in an exposure category by the corresponding rate in the reference category. Age-adjusted rates were calculated with the use of five-year age categories29. Analyses to control for age and other variables simultaneously were conducted with proportional-hazards models30; in these models, menopausal status and other covariates were updated at two-year intervals. For all relative risks, we calculated 95 percent confidence intervals31. All P values were two-tailed.
Results
During eight years of follow-up (692,675 person-years), 1439 incident cases of invasive breast cancer were diagnosed. To determine whether known risk factors were associated with intake of vitamin A, C, or E in 1980, we calculated the age-adjusted percentages of women with each risk factor, according to their category of vitamin intake (described elsewhere*). We observed no material associations between the intake of vitamin C or vitamin E and parity, maternal age at first birth, age at menarche, history of breast cancer in the participant's mother or a sister, menopausal status, or body-mass index (the weight in kilograms divided by the square of the height in meters). Slightly more women in the lowest quintile group for vitamin E intake than in the highest quintile group drank 5 g of alcohol or more daily (41 percent vs. 34 percent). More women in the highest quintile groups for vitamin C and vitamin E intake had a history of benign breast disease (27 percent for both vitamins) than in the lowest quintile groups (23 percent for both vitamins), presumably because vitamin supplements were taken for this condition. Sixteen percent of the women in the lowest quintile group for carotenoid and total vitamin A intake had body-mass indexes of more than 29, as compared with 12 percent of the women in the highest quintile group, but vitamin A intake was not materially associated with other risk factors for breast cancer.
Vitamin C intake based on the 1980 questionnaire had no appreciable association with the incidence of breast cancer from 1980 to 1988 (Table 1), and intake based on the 1984 questionnaire (completed by 71,312 women) was unrelated to the incidence from 1984 to 1988 (Table 2). The results for vitamin C intake without adjustment for energy intake were similar, as was the association with vitamin C derived from food only (i.e., when intake of this vitamin from supplements was excluded).
|
|
In multivariate analyses controlling for known risk factors, the incidence of breast cancer was significantly reduced among women in each of the three highest quintile groups for total vitamin A intake, as compared with women in the lowest quintile group, and the test for trend gave a significant result (chi for trend = -3.32, P = 0.001) (Table 1). The foods that contributed most to vitamin A intake were (in descending order) spinach, carrots, liver, and sweet potatoes. When the intake of vitamin A from food alone was considered (i.e., when intake from supplements was excluded), the relative risk in the highest quintile group as compared with the lowest was exactly the same as for total vitamin A intake including intake from supplements (relative risk, 0.84; 95 percent confidence interval, 0.71 to 0.98). When the intakes of vitamins C and E were added to the model, the inverse association of total vitamin A intake with the incidence of breast cancer was strengthened (relative risk, 0.78 for women in the highest quintile group as compared with those in the lowest; 95 percent confidence interval, 0.65 to 0.95; chi for trend = -3.52, P = 0.001). Women in the highest quintile group for intake of preformed vitamin A including intake from supplements were at significantly reduced risk, and the trend across quintile group for the intake of preformed vitamin A was also significant (P = 0.003). The relative risk for women in the four highest quintile groups for the intake of preformed vitamin A, as compared with those in the lowest quintile group, was 0.91 (95 percent confidence interval, 0.80 to 1.03), and the risk in the four highest quintile groups for total vitamin A intake was 0.80 (95 percent confidence interval, 0.71 to 0.91). As compared with the risk in the lowest quintile group for carotenoid intake, the risk was reduced in each of the four highest quintile groups; however, the test for linear trend did not yield a significant result (P = 0.08). The four highest quintile groups for carotenoid intake combined had a relative risk of 0.86 (95 percent confidence interval, 0.76 to 0.98), as compared with the lowest quintile group. Models controlling for dietary intake of fat and fiber gave essentially the same results; models using vitamin A values unadjusted for energy intake revealed slightly but not significantly stronger inverse associations with risk.
No important differences emerged when these results were examined according to menopausal status. The relative risk of breast cancer among women in the highest quintile group for vitamin C intake as compared with women in the lowest quintile group was 1.05 for premenopausal women and 1.07 for postmenopausal women; for the respective quintile groups for vitamin E intake, these risks were 0.99 for premenopausal women and 0.89 for postmenopausal women. The inverse association of total vitamin A intake with risk was slightly stronger among premenopausal women than among postmenopausal women; the relative risk among premenopausal women in the highest quintile group for total vitamin A intake was 0.81 (95 percent confidence interval, 0.62 to 1.05; chi for trend = -2.69, P = 0.01), as compared with 0.90 among postmenopausal women (95 percent confidence interval, 0.72 to 1.12; chi for trend = -1.67, P = 0.09). Controlling the analysis for the use of estrogen by postmenopausal women did not materially alter these results.
The intake of total vitamin A in 1984 was significantly inversely associated with the risk of subsequent breast cancer through 1988 (chi for trend = -2.21, P = 0.03) (Table 2), and the magnitude of the relative risk was similar to that for the period 1980 to 1988. As compared with the women in the lowest quintile group for the intake of preformed vitamin A, those in the two highest quintile groups were at significantly reduced risk of breast cancer, and the multivariate trend was significantly inverse (chi for trend = -2.27, P = 0.02). We observed little association between the intake of carotenoids calculated from the 1984 questionnaire and the incidence of breast cancer.
Use of Vitamin Supplements
Overall, the use of multivitamins and specific vitamin supplements was not significantly associated with breast cancer (Table 3). We observed no significant association with either the dose or the duration of use of specific supplements of vitamin C or vitamin E, even when taken in high doses or for long periods. Women with the highest dose intakes of vitamin A derived from specific supplements ( 
23,000 IU per day) had no significant reduction in risk.
|
|
With the exception of alcoholic beverages, no single food in 1980 was significantly associated with the incidence of subsequent breast cancer. We also observed no apparent association between the intake of red meat, fish and chicken, or fruit and the incidence of breast cancer from 1980 to 1988. However, we found a significant inverse trend between the index of vegetable consumption and the risk of breast cancer (P = 0.04). As compared with the lowest quintile group (<0.9 serving of vegetables per day; relative risk, 1.0), the fourth highest quintile group (0.9 to 1.2 servings per day) had a multivariate relative risk of 0.79 (95 percent confidence interval, 0.62 to 0.99); the third highest quintile group (1.3 to 1.6 servings per day), a risk of 0.81 (0.64 to 1.01); the second highest quintile group (1.7 to 2.1 servings per day), a risk of 0.72 (0.57 to 0.91); and the highest quintile group ( 2.2 servings per day), a risk of 0.83 (0.66 to 1.03).
Evaluation of Possible Ascertainment Bias
To evaluate the possibility that these findings might be due to differences in the rates of screening mammography among women with higher vitamin consumption, we compared the dietary intakes of women who reported on the 1988 questionnaire that they had undergone mammography with the intakes of those who had not. The women with mammograms were slightly more likely to be in the highest quintile group for the intake of preformed vitamin A (20.1 percent) than in the lowest quintile group (19.3 percent), and to be in the highest quintile group for carotenoid intake (20.7 percent) than in the lowest (19.4 percent).
Discussion
In this large prospective study, we observed a small inverse association between vitamin A consumption and the incidence of breast cancer during eight years of follow-up. Both the consumption of preformed vitamin A and that of carotenoids appeared to reduce the risk of this disease, and the association was not due to an elevated prevalence of known risk factors or excess mammographic screening among women with a low vitamin A intake. Consumption of vitamin A supplements was not associated with a reduced risk except among women in the lowest quintile group for vitamin A intake from food. Neither vitamin C intake nor vitamin E intake had any important inverse relation to the incidence of breast cancer, even among women with a long-term intake of high doses of supplements.
In our study, the relation between total vitamin A intake and breast cancer was not strictly linear; among the women in the highest four quintile groups for the intake of total vitamin A, the risk of breast cancer was approximately 20 percent lower than among women in the lowest quintile group. This suggests that only women with a low vitamin A intake may be at an increased risk of breast cancer. We found no evidence that the elevation in risk among women in the lowest quintile group of total vitamin A intake was due to a higher prevalence of known risk factors for breast cancer among these women. More data are needed to assess the nature of this dose-response relation.
The prospective design and high follow-up rates suggest that a biased assessment of diet or a difference in loss to follow-up is unlikely to explain these findings. The estimates of vitamin intake from the food-frequency questionnaire that were used in the study have been previously shown to be reasonably valid21,22,23,26,27; any misclassification of calculated intake is likely to be random, leading to an underestimate of any true associations32. The absence of any association of breast cancer with vitamin C consumption is unlikely to be due to such misclassification, since the range of vitamin C intakes was large and this vitamin was one of the nutrients most accurately measured by the questionnaire21. We cannot exclude the possibility, however, that very low intakes of vitamin C may be associated with an increased occurrence of breast cancer.
The apparent weak inverse association between vitamin E intake and the incidence of breast cancer essentially vanished after vitamin A intake was controlled for. Like vitamin C, vitamin E was consumed by a relatively high proportion of women in the cohort, either as a specific supplement or as part of a multivitamin preparation; thus, it is unlikely that we missed any important benefit that may have accrued from increasing the intake of this vitamin by the use of supplements.
Vitamin A induces cellular differentiation,17 protects cells from injury by free radicals,33 decreases the expression of certain oncogenes in vitro,34 inhibits the growth of human breast-carcinoma cells in vitro,35 and reduces the incidence of breast cancer in rodents36,37,38. Treatment with synthetic retinoids reduces the occurrence of some skin39 and head and neck40 cancers.
The association between the intake of vitamin A and breast cancer has been examined in relatively few observational studies. Among the case-control studies that have reported specific data on vitamin A and breast cancer, an inverse association was seen in several,11,41,42,43 whereas no relation was seen in two44,45. Three studies examined the intake of carotenoids and preformed vitamin A separately: in two studies11,46 carotenoids had a stronger effect, and in the third47 both components of vitamin A were protective. In the single previously reported prospective study, which included 123 cases of breast cancer,18 women in the highest third for vitamin A intake were at reduced risk (relative risk, 0.82) as compared with women in the lowest third. According to our data, both preformed vitamin A and carotenoids were associated with lower risk; however, the inverse association of total vitamin A intake with risk was even stronger. Of the food groups we studied, only the intake of vegetables was inversely associated with the risk of breast cancer.
Only a small proportion of the study cohort took specific vitamin A supplements; however, a large number of women took vitamin A as part of a multivitamin preparation. We observed no consistent evidence of a protective effect of a long duration of use or a high dose of vitamin A supplements in the overall population. Among women with the lowest dietary intake of vitamin A, however, the use of vitamin A from supplements was significantly associated with a lower risk of breast cancer. This is consistent with a threshold effect, in which only women with a low intake of vitamin A are at increased risk, and suggests that only women with a low intake of vitamin A from food may benefit from vitamin A supplements.
In summary, we found no evidence of a protective influence of vitamin C or vitamin E on the incidence of breast cancer in this prospective study. In contrast, we observed a significant inverse association of vitamin A intake with the risk of this disease. However, our data suggest that vitamin A supplements are unlikely to influence the risk of breast cancer among women whose dietary intake of this vitamin is already adequate. The effect of vitamin A supplements should be evaluated in randomized trials, although a protective effect might occur only in the subgroup with a low dietary intake of vitamin A. Especially in view of the known potential adverse effects of high-dose supplements of preformed vitamin A, particularly among women who may become pregnant,48 recommendations to use vitamin A supplements to prevent breast cancer are premature. These prospective data, in conjunction with previous evidence, do suggest that vitamin A derived from food may have some role in inhibiting the development of breast cancer.
Supported by research grants (CA-40356 and CA-50597) from the National Institutes of Health and in part by a Faculty Research Award (FRA-398) from the American Cancer Society (to Dr. Colditz).
Source Information
From the Channing Laboratory, Department of Medicine, Harvard Medical School and Brigham and Women's Hospital (D.J.H., J.E.M., G.A.C., M.J.S., C.H.H., F.E.S., W.C.W.), the Department of Preventive Medicine, Harvard Medical School (B.R., C.H.H.), and the Departments of Epidemiology (D.J.H., G.A.C., M.J.S., W.C.W.), Nutrition (W.C.W.), and Biostatistics (B.R.), Harvard School of Public Health -- all in Boston.
Address reprint requests to Dr. Hunter at Channing Laboratory, 180 Longwood Ave., Boston, MA 02115.
References
pyrene-induced mammary carcinogenesis by retinyl acetate. J Natl Cancer Inst 1981;66:559-564.
| |||||||||||||||||||||||||||||||||||||||
Related Letters:
Vitamins and Breast Cancer
Ennever F. K., Paskett E. D., Hunter D. J., Stampfer M. J., Willett W. C.
Extract |
Full Text
N Engl J Med 1993;
329:1579, Nov 18, 1993.
Correspondence
This article has been cited by other articles:
HOME | SUBSCRIBE | SEARCH | CURRENT ISSUE | PAST ISSUES | COLLECTIONS | PRIVACY | TERMS OF USE | HELP | beta.nejm.org Comments and questions? Please contact us. The New England Journal of Medicine is owned, published, and copyrighted © 2009 Massachusetts Medical Society. All rights reserved. |
Previous
