点击显示 收起
1 From the Cancer Prevention, Detection, and Control Research Program, Department of Community and Family Medicine, Duke University Medical Center, Durham, NC (PGM) and the National Institute of Environmental Health Sciences, Epidemiology Branch, Research Triangle Park, NC (PDT)
2 Address reprint requests to PD Terry, National Institute of Environmental Health Sciences, Epidemiology Branch, PO Box 12233 MD A3-05, Research Triangle Park, NC 27709. E-mail: terry2{at}niehs.nih.gov.
ABSTRACT
Differences in eating patterns and breast cancer rates across countries suggest that several dietary components, including dairy products, could affect breast cancer risk. However, dairy products are a diverse food group in terms of the factors that could potentially influence risk. Some dairy products, such as whole milk and many types of cheese, have a relatively high saturated fat content, which may increase risk. Moreover, milk products may contain contaminants such as pesticides, which have carcinogenic potential, and growth factors such as insulin-like growth factor I, which have been shown to promote breast cancer cell growth. In contrast, the calcium and vitamin D contents of dairy products have been hypothesized to reduce breast cancer risk. We reviewed the current epidemiologic literature on the relation between dairy product intakes and breast cancer risk, focusing primarily on the results of cohort and case-control studies. Most of the studies reviewed showed no consistent pattern of increased or decreased breast cancer risk with a high consumption of dairy products as a whole or when broken down into high-fat and low-fat dairy products, milk, cheese, or butter. Measurement error may have attenuated any modest association with dairy products. The available epidemiologic evidence does not support a strong association between the consumption of milk or other dairy products and breast cancer risk.
Key Words: Breast neoplasms diet dairy products dietary fats conjugated linoleic acids vitamin D epidemiology
INTRODUCTION
Breast cancer is the most commonly diagnosed cancer in women and is the leading cause of cancer mortality in females in the world (1, 2). The estimated annual incidence of breast cancer worldwide is 1 million cases, with 200 000 cases in the United States and 320 000 cases in Europe (1, 3, 4). A comparison of breast cancer data across geographic regions shows tremendous variation in both incidence and mortality rates, with >5-fold differences observed between low-risk and high-risk areas (2, 5). In the United States, breast cancer incidence rates have been rising slowly over the past 2 decades (6).
On the basis of epidemiologic studies conducted in several populations, established determinants of breast cancer include age, relative body weight, change in weight over time, the number and timing of reproductive events and lactation, exogenous and endogenous hormone concentrations and metabolism, history of benign breast disease, exposure to radiation, alcohol consumption, and family history of breast cancer (7, 8). The striking geographic differences in breast cancer incidence rates may be attributable to genetic differences between populations and differences in lifestyle or environmental exposures, including diet. Migrant studies have found that breast cancer rates increase in persons who move from an area with a low incidence to one with a high incidence of breast cancer (9). Higher rates, compared with the country of origin, have been observed both among the migrants themselves (after 15 y in their adopted country) and among their descendants (9). These studies are highly suggestive that the observed geographic differences in breast cancer incidence are due, at least in part, to differences in environmental exposures. Dietary factors are thought to be most responsible for the change in incidence rates among migrants (10), although correlated lifestyle changes, such as obesity and physical activity, may also play a role.
Differences in eating patterns across countries suggest several possible dietary components that could affect breast cancer risk. The focus of this review is on dairy products, although many dietary factors have been investigated in relation to breast cancer, including the consumption of fat or meat, fruit and vegetables, and soy products (11, 12). Dairy products are a diverse food group in terms of the factors that could potentially influence cancer risk. Some dairy products, such as whole milk and many types of cheese, have a relatively high saturated fat content, which may increase breast cancer risk. In contrast, the calcium and vitamin D contents of dairy products are hypothesized to reduce breast cancer risk.
Our aim here is to review the current epidemiologic literature on dairy product consumption and risk of breast cancer. We obtained relevant articles through searches of the MEDLINE and CANCERLIT databases (both from the National Institutes of Health, Bethesda, MD) and through published reports by cross-matching the references of relevant articles (13). We first describe some of the leading hypotheses that have been put forth to link dairy products with risk of breast cancer and then summarize the evidence from epidemiologic studies of dairy products and breast cancer risk.
HYPOTHESES CONCERNING DAIRY PRODUCTS AND BREAST CANCER RISK
There are several postulated mechanisms through which dairy products could influence breast cancer risk in either a positive or inverse fashion. The major hypotheses that have been put forth to suggest an increased risk of breast cancer risk associated with the consumption of dairy products include the following: 1) a high consumption of dairy products may reflect an overall high dietary fat intake, particularly saturated fat, which in turn has been associated with breast cancer risk; 2) milk products may contain contaminants, such as pesticides, that are potentially carcinogenic; and 3) milk may contain growth factors, such as insulin-like growth factor I (IGF-I), which have been shown to promote breast cancer cell growth. Hypotheses suggesting that dairy products protect against breast cancer have focused on the anticarcinogenic properties of the dairy product constituents calcium, vitamin D, and conjugated linoleic acid (CLA).
Dairy products account for a substantial proportion of the total fat and saturated fat intakes in the Western diet. Studies of dietary sources of nutrients estimate that the proportion of saturated fat in the diet that comes from dairy products is 31% in the United States and 50% in Sweden (14, 15). Although not all dairy products have a high fat content, a high consumption of dairy products may be associated with overall high dietary fat intakes (16). Total fat consumption has been thought to increase breast cancer risk by increasing circulating estrogen concentrations, although evidence for this hypothesis is weak (17, 18). The dietary fat hypothesis is supported by experimental data in rodents and cell lines (19, 20); however, epidemiologic studies, especially prospective cohort studies, have failed to show that dietary fat increases breast cancer risk (21).
Another hypothesis suggesting an increased risk of breast cancer with high dairy intake focuses on contaminants in milk products, with a particular emphasis on pesticides. In Israel, 3 pesticides were present in dairy products in concentrations up to 100 times those in US dairy products before the use of these pesticides was outlawed in Israel in 1986 (22). Subsequent decreases in breast cancer mortality rates in Israel have been attributed by some to the pesticide ban (22). Other investigators have tried to link dietary variables to serum concentrations of organochlorines, environmental contaminants sometimes described as "endocrine disruptors." However, the evidence linking organochlorine concentrations to dairy product consumption is inconsistent (23-25), and most epidemiologic studies do not support an association between circulating concentrations of organochlorines and breast cancer (26). Therefore, it seems unlikely that the presence of organochlorines in dairy products could plausibly be linked to breast cancer.
It has also been suggested that dairy product consumption could increase breast cancer risk because milk and milk products contain hormones and growth factors. Outwater et al (27) hypothesized that IGF-I, which is present in human and cow milk, may be a link between dairy product consumption and breast cancer. IGF-I has been shown to promote breast cancer cell growth (28). In addition, experiments have shown that IGF-I is likely to be involved in cell transformation because removing or blocking IGF-I receptors from the cell membrane can abolish viral or cellular oncogene-induced malignant transformation (28). These investigators have argued that bovine growth hormone (bGH), which is sometimes administered to dairy cattle to increase milk production, results in increased concentrations of IGF-I in cow milk (29). Nonetheless, the US Food and Drug Administration approved the use of bGH in 1993, noting that the compound appears to be inactive when administered orally (30). In response to this, Outwater et al noted that IGF-I does not appear to be destroyed during pasteurization and may not be degraded in the gastrointestinal tract during digestion. However, little evidence exists to support the intact absorption of IGF-I after oral ingestion. The results of in vitro and in vivo studies of IGF-I absorption in animals have been inconsistent; to our knowledge, no studies have been conducted in humans. Given the current lack of evidence of the substantial absorption of intact IGF-I after oral administration, the hypothesis that IGF-I in milk is related to breast cancer remains less than compelling.
Other hypotheses suggest an inverse relation between dairy product consumption and breast cancer risk. These hypotheses have focused on the anticarcinogenic effects of vitamin D and calcium. Dairy products have a high calcium content and are also a major dietary source of vitamin D in countries where milk and other dairy products are fortified, such as the United States. The active form of vitamin D3, 1,25-dihydroxyvitamin D [1,25(OH)D], affects multiple processes related to cell growth and development (31-33). In normal mammary tissue, 1,25(OH)D concentrations are increased during pregnancy and lactation, which suggests a role for vitamin D in the differentiation of the mammary gland (33). It has also been shown in breast cancer cell lines that vitamin D exerts antiproliferative effects by causing arrest in phase G0/G1 of the cell cycle (32, 33). This leads to the down-regulation of several growth promoting factors, such as IGF-I and the up-regulation of negative growth regulators such as transforming growth factor ß. Beyond its antiproliferative effects, 1,25(OH)D causes morphologic and biochemical changes indicative of apoptosis, including cell shrinkage, chromatin condensation, and DNA fragmentation (32).
The cellular functions of vitamin D are closely linked to calcium. Calcium is a pivotal regulator of a wide variety of cellular functions, including cell proliferation and differentiation, in its role as a major second messenger (34-36). Vitamin D is one of several regulators of the absorption and metabolism of calcium (31). Much of the evidence suggesting that calcium protects against cancer relies on its interrelation with vitamin D. However, experimental evidence from several model systems suggests that an increased level of calcium alone is a sufficient signal to induce apoptosis (37). Experimental data in rodents support the hypotheses that vitamin D and calcium have anticarcinogenic effects. Several investigators have shown that animals fed diets deficient in calcium and vitamin D develop mammary hyperplasia and hyperproliferation (31). Furthermore, animal studies have shown that supplementation with calcium and vitamin D reduces the risk of mammary tumors in animals fed a high-fat diet and prevents the development of mammary tumors in animals treated with the carcinogen 7,12-dimethylbenz[a]anthracene (38-40).
A third potential mechanism to suggest that dairy products may reduce breast cancer risk involves CLA. CLA is a common term used to describe a mixture of positional and geometric isomers of linoleic acid (41). The primary dietary sources of CLA are dairy products and meat from ruminants (animals that chew cud). Animal studies suggest that CLA confers protection against the development of mammary tumors (42). It is interesting to note that tumor formation was inhibited in animals fed CLA, regardless of the type or amount of fat in their diets. Postulated mechanisms of CLA include effects on oxidative behavior in cancer cells, effects on the metabolism of linoleic acid, and induction of apoptosis (43, 44).
EPIDEMIOLOGIC STUDIES OF DAIRY PRODUCTS AND BREAST CANCER RISK
When evaluating the epidemiologic studies that have examined the relation between dairy product consumption and breast cancer risk, it is important to keep in mind both the diversity of dairy products and the multiple mechanisms through which these products could influence the development and proliferation of breast tumors. An evaluation of the evidence to date indicates considerable variation in how data on the consumption of dairy products were collected and reported. Some studies report the overall consumption of dairy products, whereas others break down dairy product consumption into various categories such as milk products, cheese, and fermented milk products. Some studies also assessed low-fat dairy products separately from high-fat products. Other studies evaluated the consumption of dairy products during childhood and adulthood separately.
In Tables 1 and 2 we summarize cohort studies and case-control studies, respectively (ordered by date of publication), that evaluated dairy product consumption in relation to breast cancer (41, 45-89). Three cohort studies (45, 49, 52) and 9 case-control studies (55, 61, 67, 71, 72, 79, 87, 88) reported on the relation between breast cancer and dairy products as a whole. A statistically significant inverse relation between breast cancer and dairy product consumption was reported in each of the cohort studies, although the effect was limited to premenopausal women in the report by Shin et al (45). Results from the case-control studies were inconsistent, an inverse association was reported in 2 studies (71, 87), no association was reported in 5 studies (55, 60, 61, 67, 72), and positive associations were reported in 2 studies (79, 88). Because dairy products are a heterogeneous group of foods, it may not be appropriate to consider them a single exposure in relation to breast cancer. Evaluation of dairy products as a single category could mask the true effects, either positive or inverse, of different classes of dairy products.
View this table:
Table 1. Summary of 10 cohort studies that evaluated the association between dairy product consumption and breast cancer risk1
View this table:
TABLE 2. Summary of 36 case-control studies that evaluated the association between dairy product consumption and breast cancer risk1
Many of the published epidemiologic studies reported on specific categories of dairy products. Two of the cohort studies (46, 49) and 10 of the case-control studies (41, 70, 73, 75, 76, 78, 79, 82, 83, 85) investigated the association between breast cancer and butter consumption. If dairy products affect breast cancer risk primarily because of their saturated fat content, one might expect that the effect would be most prominent in analyses focused on butter intake. However, no consistent pattern was observed with reported butter intake. In a cohort study conducted in Finland (49), an inverse association that was not statistically significant was reported, whereas a slight positive association was reported in a cohort study in the Netherlands (46). In the case-control studies (41, 70, 73, 75, 76, 79, 78, 82, 83, 85), odds ratios both > and <1.0 were reported, but generally differences between cases and controls were not statistically significant. However, because butter is often added to other foods during the manufacturing process and is an ingredient in many recipes (15), this food item may be particularly difficult to measure in epidemiologic studies.
Milk is the major component of dairy intake (90), and most of the epidemiologic studies listed in Tables 1 and 2 provided estimates of the relative risk for breast cancer associated with milk intake. The results are not strictly comparable across all studies because some investigators examined all types of milk combined and others examined whole milk separately from low-fat or skim milk. Of the studies that reported on the risk associated with total milk intake, the results were inconsistent. Inverse associations among premenopausal women were reported by Shin et al in the US Nurses Health Study (45), by Hjartåker et al in a Norwegian population (47), and among both pre- and postmenopausal women by Knekt et al (49) in a Finnish population. The results of these studies contrast with findings from 2 other cohorts, both of which were conducted in Norwegian women (51, 53), in which nonsignificant positive associations with milk intake were reported. In 2 other cohort studies (46, 50), no association was found. Case-control studies present a similarly inconsistent picture of the relation between milk consumption and breast cancer, with results ranging from significantly reduced risk (62) to no significant association (66, 69, 73-75, 81), to significantly increased risk (85, 88).
Several studies distinguished between whole milk and reduced-fat milk in their analyses (45, 46, 58, 76, 77, 79, 82). Within most of these studies, however, the distinction between low-fat and whole milk made little difference in the reported relative risks. Several investigators reported inverse relations for both low-fat dairy and total dairy products (45, 82), whereas positive associations were reported for both low-fat and regular milk products in other studies (51, 76, 79). The fact that no significant differences between the effects of low-fat and those of higher-fat dairy products were observed in most of the studies argues against the hypotheses that the fat content of milk products may be a major determinant of breast cancer risk. Various studies have evaluated other specific categories of dairy products, including ice cream, yogurt, and cheese. None of these products showed a consistent relation to breast cancer risk.
Because the breast is thought to be particularly sensitive to carcinogenic exposures during adolescence (91), some investigators have evaluated the association between dairy product consumption during childhood and breast cancer (47, 57, 59, 80, 86). No association was reported in a study by Potischman et al (59); however, inverse associations between breast cancer and milk or dairy product consumption during childhood or adolescence were reported in 4 other studies (47, 57, 80, 86). Two of these studies (80, 86) stratified by menopausal status of the breast cancer cases and found similar relations for pre- and postmenopausal breast cancer. In all of these studies, exposure assessment was retrospective; the adult study subjects reported details of their diet during adolescence, which raised concerns about misclassification resulting from the recall of dietary factors from decades earlier.
The studies presented in Tables 1 and 2 exemplify some of the difficulties in interpreting epidemiologic studies of dietary factors and cancer. Although case-control studies are an efficient study design for studying relatively rare diseases with long latency periods, such as cancer, they have distinct limitations when dietary data are used because of the possibility that cases and controls recall their past diets differently or that cases may have made changes in their diet around the time of diagnosis. Prospective cohort studies, which are generally considered to provide more compelling evidence for an association between an exposure and disease, usually require very large numbers of participants and many years of follow-up before adequate numbers of cases are available for analysis. As shown in Table 1, even when the cohort sizes numbered in the tens of thousands, most of the prospective studies had fewer than 350 breast cancer cases, except for the Nurses Health Study cohort in which there were nearly 3500 cases (45) and the Netherlands Cohort Study in which there were >900 cases (46). It is notable that different effects were observed for premenopausal and postmenopausal women in the Nurses Health Study (45), although the mechanisms underlying these differences remain unclear. For example, we can speculate that if the consumption of dairy products early in life is more important with respect to breast cancer risk than is dairy product consumption later in life, then misclassification with respect to exposure early in life would be greater in postmenopausal women whose diets had changed over time and that this might have obscured the association in these women. Given the relatively low number of cases, other studies were much more limited in their ability to stratify by menopausal status.
To overcome some of the sample size limitations in the published data, there have been 2 published reports that have combined data from multiple studies (92, 93). Boyd et al (92) conducted a meta-analysis of 10 studies published between 1981 and 1990. The summary relative risk for milk consumption was 1.17 (95% CI: 1.04, 1.31), which suggests a modest increase in risk. Analyses stratified by menopausal status were not reported. A second combined analysis was performed by the Pooling Project (93), in which data from 8 prospective cohort studies conducted in North America and Western Europe were used to evaluate relations between consumption of meat and dairy products and breast cancer risk. The combined dataset included data for 351 041 women, among whom 7379 were diagnosed with invasive breast cancer during follow-up. The studies included 2 of the cohort studies listed in Table 1 (52, 54) as well as other studies of breast cancer that had earlier examined dietary factors other than dairy products (14, 94-98). Results were presented for various categories of dairy products, based on food type or fat content. The reported relative risks for each 100-g/d increment in consumption ranged from 0.97 to 1.03 for all types of dairy products examined, with the exception of a relative risk of 1.16 for cheese products. All of the 95% CIs around the relative risks included the null value of 1, which indicates no statistically significant relations between breast cancer and consumption of any type of dairy product. Odds ratios associated with the consumption of total dairy fluids or total dairy solids did not vary substantially by menopausal status.
Several studies specifically addressed the hypothesis that dairy products affect breast cancer risk because they are a primary source of CLA (41, 46, 99). In the first study by Aro et al (41), postmenopausal breast cancer cases reported lower dietary intakes of CLA and had lower serum CLA concentrations than did the controls. The odds ratio for the comparison of the highest with the lowest quintile of serum CLA, after adjustment for established breast cancer risk factors, was 0.2 (95% CI: 0.1, 0.6). This strong inverse association was not confirmed in 2 subsequent studies, however (46, 99). Voorrips et al (46) reported a weak positive association between CLA intake and breast cancer that was not statistically significant. Chajès et al (99) compared CLA concentrations in breast adipose tissue between breast cancer patients and women with benign breast conditions. CLA concentrations were higher in the breast cancer patients, but were not significantly different from those in women with benign conditions. Overall, these studies do not provide consistent support for the hypothesis that CLA is a protective factor against breast cancer.
DISCUSSION
The published epidemiologic data do not provide consistent evidence for an association between the consumption of dairy products and breast cancer risk. When interpreting these data, however, several factors must be considered. Most importantly, assessment of dietary factors in relation to cancer risk is notoriously difficult and subject to many potential biases. Various methods used in epidemiologic studies, including food-frequency questionnaires and diet records or food diaries, have shown only moderate reliability, and some misclassification of intake is unavoidable (100, 101). Moreover, different dietary assessment methods may yield different results. For example, a recent study within the European Prospective Investigation on Cancer (101) compared dietary information obtained from food diaries with that obtained with the use of a food-frequency questionnaire and found a positive association between saturated fat intake and breast cancer risk only when using data from the food diaries. Assuming that errors in the reporting of dairy intake are nondifferential between women with breast cancer and women without breast cancer, the misclassification would tend to bias results toward the null (101, 102). This underestimation of the strength of the association would exist whether the exposure was associated with risk positively or negatively. It is generally believed that most dietary factors have relatively small effects on cancer risk and, therefore, the inevitable misclassification of dietary variables would make it even more difficult to detect modest increases or decreases in risk associated with consumption of dairy products.
Another challenge when evaluating dairy products in relation to disease risk is the correlation among nutrients in the diet. Persons with a high consumption of butter, cheese, and other high-fat dairy products may also be more likely to consume large amounts of meat or other high-fat foods that could also contribute to an increased risk of breast cancer (16). Even when total energy intake is controlled for, it may be impossible to completely separate the effects of dairy intake from that of other dietary factors, including intake of various types of fat. Conversely, persons may consume low-fat milk and other dairy products as part of an overall healthier diet that is high in fruit and vegetables and low in fat (16). Once again, it may be difficult to separate the effects of dairy products from those of other nutrients that alter breast cancer risk. It is noteworthy that relatively few studies have controlled for dietary factors other than alcohol consumption and total energy intake in their analyses of dairy products, which makes it difficult to compare studies according to the dietary variables considered in the analysis. Nonetheless, the results of studies that adjusted their estimates for a wide range of potentially confounding variablesincluding alcohol consumption, cigarette smoking, oral contraceptive use, parity, family history of breast cancer, history of benign breast disease, age at menopause, age at first birth, body mass index, and parity (45, 46, 81)did not differ systematically from the results of the studies that did not adjust for many or all of these factors. Adjustment for covariates within studies often does not alter appreciably the crude estimates for dairy product consumption (45, 46, 52).
Variation in the reported levels of consumption among populations is another important consideration when evaluating epidemiologic data. No standard method is currently available for categorizing dairy product intakes, and many studies have very reasonably made comparisons based on quantiles of intake within their specific study population. Average intake varies considerably between populations such that a level of consumption that is considered "high" in one population might be considered "low" in another population. For example, in a study conducted in Japan, women reporting daily consumption of dairy products were in the highest exposure category (71). In contrast, in a US population, the reference category (lowest exposure level) consisted of women who reported consuming 1 serving/d and the highest exposure level consisted of women who reported >3 servings/d (45). In some cases, the investigators did not report the level of consumption within each quantile, which made it impossible to compare effects at similar levels of consumption across studies. Furthermore, as mentioned above, dietary changes over time may be an important source of misclassification of diet and the degree to which this misclassification occurs may vary across populations.
Beyond the challenges that are common to any study of nutritional epidemiology, there are specific challenges related to evaluating dairy products. A leading hypothesis suggesting that dairy products may reduce breast cancer risk is based on the vitamin D content of these products. Few foods naturally contain significant amounts of vitamin D, however, and the vitamin D content of dairy products is mostly the result of fortification (103). Although the fortification of dairy products, cereals, and other foodstuffs is common, the types of products fortified and the amount of vitamin D added varies between countries (104). In the United States, fortification is voluntary, but most manufacturers add vitamin D to many products, including milk and margarine. In contrast, the level of fortification is much less than that in other countries. In the United Kingdom and Finland, for example, margarine is fortified but milk is not (104). Similarly, fortification of margarine is mandatory in Australia, but fortification of milk and other dairy products is voluntary (104). These differences suggest that studies from countries with different regulations and practices regarding vitamin D fortification are not strictly comparable. If vitamin D is the component of dairy products that influences breast cancer risk, comparisons should take into account not only the specific dairy product but also the level of vitamin D fortification.
Overall, the published studies reviewed herein do not provide consistent evidence for an association between dairy product consumption and breast cancer risk. On one hand, mechanisms have been proposed to suggest that certain types of fat, growth factors, or environmental contaminants found in milk could increase risk of breast cancer. However, the evidence to date regarding these hypotheses remains less than compelling. On the other hand, mechanisms based on the cellular roles of vitamin D, calcium, or CLA have been put forth to suggest that the consumption of dairy products could protect against the development of breast cancer. It is conceivable that the micro- and macronutrients in dairy products influence multiple pathways related to the development of breast cancer, but the net effect is neither an increase nor a decrease in risk. Therefore, additional studies that address the association with certain hormonal and environmental factors and micronutrients may help to shed light on the association between dairy product consumption and breast cancer risk. Alternatively, the limitations of dietary assessment in epidemiologic studies may have attenuated any true effect of dairy product consumption on breast cancer risk. Given somewhat compelling evidence of a protective effect of dairy products from animal studies (31, 38-40), future studies are needed to assess the extent to which measurement error may have obscured any association with dairy products in epidemiologic studies. In conclusion, despite several intriguing hypotheses linking dairy product consumption and breast cancer, the available epidemiologic evidence does not support a strong association between the consumption of milk or other dairy products and breast cancer risk.
ACKNOWLEDGMENTS
We thank Dale Sandler and Beth Ragan for helpful comments on the manuscript. An earlier draft of this review was commissioned by The Dairy Council (Great Britain). The conclusions are solely those of the authors.
REFERENCES