Environmental influences and breast cancer
a. Descriptive epidemiology
It is quite clear that environmental factors play an important role in the etiology of breast cancer. Incidence rates for breast cancer vary tremendously throughout the world (Henderson et al., 1996). The highest incidence rates are found in industrialized countries, including the US, Northern Europe, and Canada, while lower rates are found in Asia and Africa. Differences in incidence could be due to differences in screening and reporting, but comparisons of international breast cancer mortality rates show similar patterns. Mortality rates are based upon death certificates and do not depend upon cancer registries for data collection, and are not influenced as strongly by cancer screening as incidence rates. Some of the most compelling evidence for environmental factors comes from long-term comparisons of breast cancer incidence rates within the United States. Here, a longterm, “background” increase in breast cancer incidence since the 1940s remains largely unexplained, even after accounting for mammographic screening and trends in distributions of known breast cancer risk factors (Feuer, 1992). The increase in breast cancer incidence in the United States of about 1 percent per year for the past sixty years could be due to a variety of factors (Feuer et al., 1993; King & Schottenfeld, 1996), but one thing is quite clear: it cannot be ascribed to genetics alone.
Studies of women who migrate from Asia to the United States, in particular the Bay Area of Northern California, suggest that breast cancer rates rise to US levels soon after women migrate to the US (Thomas & Karagas, 1996). Migrant studies suggest that features of the physical environment, including early life experiences, play an important role in breast cancer risk. Even if genetic factors play a role in breast cancer risk among women of other countries, migrant studies suggest that these genetic factors must interact with environment factors in the host country. Twin studies estimate that the heritability of breast cancer is very low, about 27% (Lichtenstein, 2000). This does not mean that the environment alone is responsible for the remaining 73 percent of breast cancer, since gene-environment interactions could be involved. Nor does it mean that genes cause 27 percent of breast cancer, since twins also share environments. But it does mean that genes most certainly are not the only cause of breast cancer.
b. Clues to environmental influences
The traditional view is that hormones cause breast cancer, except in rare cases where women in high-risk families inherit mutations in genes such as BRCA1 or BRCA2. Recent evidence shows that this view is too narrow. A broader view of breast cancer causation is needed that takes environmental factors into account. But how do we define the environment?
Defining “environment”
A conference on breast cancer and the environment convened by the National Breast Cancer Coalition in Washington DC in 1998 proposed a working definition of the “environment” to include “voluntary exposures as well as involuntary exposures, social class, and urban/ rural differences, and exposures that occur outside the body as well as those that modify the internal milieu.” A summit on breast cancer and the environment held in Santa Cruz, California in 2002 defined the environment as “the totality of living and working conditions as well as the physical, biological, social, and cultural responses to those conditions.” The latter summit emphasized that “environmental exposures are often influenced by social, economic, and cultural factors such as employment, income, and housing” and include exposures related to occupation or residence, as well as industrial emissions, pollution, and hazardous chemicals. For the purposes of this paper, environment will be taken in the broadest sense, as proposed by these two conferences. Another way to categorize environmental factors in this context is “everything except for genes,” recognizing that genes and environment often interact.
Definitions of causality
Epidemiologists define risk factors as exposures that are associated with an increase or decrease in the number of cases of illness in a population. The association may be direct (as in the case of cigarette smoking and lung cancer) or indirect (when an exposure such as age serves as a proxy for a more proximal risk factor). The association may be causal (true) or non-causal (due to chance or bias). Causal associations fall into several categories: the exposure may be a necessary cause (present in all cases of disease), sufficient cause (able to cause disease on its own), or a contributory cause (neither necessary nor sufficient, but present as one of many component causes in some but not all cases of the disease). Associations may also be classified as strong (e.g., risk ratios or odds ratios greater than four) or weak (risk ratios close to the null value of one). Most known risk factors for breast cancer are contributory causes, exposures with weak effects that are neither necessary nor sufficient for disease. Therefore, it should be no surprise that environmental risk factors for breast cancer, including any that are newly identified or under investigation, are likely to have very weak effects and act only in combination with other risk factors.
c. Hormonal risk factors
The most widely recognized risk factors for breast cancer are female gender and increasing age. Women develop breast cancer at over one hundred times the rate as men, and rates increase dramatically as women get older. Classically, age and female gender serve as proxies for cumulative exposure to ovarian hormones. The mitogenic effects of estrogen combine with progesterone to increase proliferation of breast epithelial cells, increase the likelihood of mutation, and thereby lead to tumor formation (Pike et al., 1993). For most epidemiologists, in fact, female hormones are the main cause of breast cancer. Reproductive profiles, including age at menarche, age at menopause, parity, as well as obesity and height, are well-established risk factors and are presumed to be proxies for lifetime estrogen exposure. Epidemiologic studies consistently demonstrate a positive correlation between blood levels of estrogen and related metabolites and breast cancer risk later in life (Pike at al., 1993). For this reason, most epidemiologists would probably state that breast cancer is a “hormonal” cancer, and all or most risk factors for breast cancer operate by modulating levels of estrogen and progesterone, or related metabolites.
Under the hormonal theory of breast carcinogenesis, environmental exposures that occur outside the body contribute to increased breast cancer risk only by raising levels of estrogen-related metabolites, or act to mimic such metabolites within the body. Much research has been devoted to identifying environmental exposures such as xenoestrogens that may act in this manner. Promising work is being conducted in the area. However, a broader view of breast cancer causation is emerging that suggests that even hormonal-related exposures are more complex than we once thought. For example, alcohol consumption and postmenopausal hormone replacement therapy (HRT) show consistent associations with increased breast cancer risk, especially long-term use, and until now both exposures fit well within the hormonal theory of causation. Both alcohol (by raising estrogen levels) and HRT (synthetic hormones) operate through hormone-mediated pathways. But recent evidence suggests that alcohol and HRT may also increase levels of oxidative stress. Estrogen metabolites are involved in redox cycling, and yield byproducts that can cause DNA damage and mutation (Zhu and Conney, 1998; Lin et al., 2003). Metabolites of alcohol (including aldehydes) cause oxidative DNA damage in a variety of organs.
Oxidative DNA damage due to estrogen exposure and estrogen metabolites may underlie many of the established hormonal risk factors for breast cancer (Zhu and Conney, 1998). In the past, the focus for investigation of environmental risk factors for breast cancer has often been on xenoestrogens and environmental pollutants or contaminants that may mimic estrogen. If the underlying mechanism for estrogen-mediated breast carcinogenesis is really oxidative stress, then the search for environmental risk factors should also consider chemicals that alter redox cycling and exposures that modulate levels of oxidative DNA damage, rather than just estrogen or estrogen-like compounds per se.
Limitations of the hormonal theory of breast carcinogenesis
An unfortunate outgrowth of the hormonal theory of breast carcinogenesis is the assumption that all breast cancer risk factors are hormonal. Hormonal risk factors may represent component causes, neither necessary nor sufficient for breast cancer. Several hormonal risk factors for breast cancer, especially age at first birth, actually represent the degree of differentiation of breast tissue. In rats, later age at first full term pregnancy leads to increased susceptibility to a variety of environmental carcinogens. It may be that reproductive history and other hormonal risk factors for breast cancer merely set up a “fertile soil” for the effects of environmental exposures later in life. Thus, hormones may act in combination (or interact) with environmental exposures. In one simple model for carcinogenesis, environmental exposures cause mutations in DNA. These mutations become fixed and perpetuated in daughter cells by hormone-induced cellular proliferation.
Why do these theories matter? If interactions between hormones and environment are important, we will miss these effects if we estimate main effects for environmental factors while adjusting for hormonal risk factors as potential confounders. Countless epidemiologic studies have “ruled out” associations for environmental exposures and breast cancer simply by adjusting away these effects, while ignoring the potential for interactions between the environment and “known” risk factors.
Endogenous hormone levels could represent intermediates, determined in large part by environmental exposures, especially those that act early in life. Onset of menarche and regular menstrual cycling may be influenced by diet, obesity, and possibly a variety of estrogenic exposures (including chemicals) in the environment. Not enough research has been done of how physical activity and chemical exposures influence age at menarche and maturation/differentiation of breast tissue in young women. Some epidemiologic studies estimate effects for environmental exposures after adjusting for hormone levels in blood. This approach is not valid if hormone levels are intervening variables between environmental exposures and disease, and will lead to invalid estimates of effect.
Thus, we can see how, rather than “explaining” breast cancer, hormonal theories for breast cancer may actually be pointing us towards whole new avenues of research. These new research areas include studying interactions with environmental exposures and investigating new biochemical pathways for breast carcinogenesis. Rather than “closing the book” on the etiology of breast cancer, the latest knowledge of how hormones might work opens the door to the investigation of new environmental risk factors for breast cancer. Most importantly, some of these new environmental risk factors may be modifiable.
d. Environmental risk factors
The strongest known environmental risk factor for breast cancer is exposure to ionizing radiation. A strong association has been observed between high dose exposure in atomic bomb survivors and persons undergoing prolonged radiation treatment. But few studies have been conducted of low dose occupational exposures or common medical procedures (Henderson et al., 1996). Women who underwent radiation treatment for Hodgkin's disease and other cancers are at increased risk of breast cancer later in life. So are women who were treated with radiation for scoliosis, underwent repeated cardiac catheterizations, or underwent other diagnostic procedures. But at present it is not possible to identify which women should avoid such procedures or undergo alternative treatments. The risk of breast cancer from nuclear power plants and other low-level sources of ionizing radiation has not been extensively studied, but is thought to be negligible. Nevertheless, considerable concern exists about such an association among many grassroots advocacy groups.
Despite scores of epidemiologic studies and decades of research, the association between tobacco smoke and breast cancer remains controversial (Laden and Hunter, 1998). Difficulties in measuring exposure, particularly passive exposure early in life, and disentangling the effects of the complex mixture of compounds within tobacco smoke are a few of the problems encountered. Cigarette smoke may increase breast cancer risk by raising levels of oxidative DNA damage. Exposure to ionizing radiation also increases levels of oxidative damage, so it is possible that hormones, alcohol, smoking, radiation, and many other environmental factors share oxidative damage and perhaps other biochemical pathways as common mechanisms of action in breast carcinogenesis.
A variety of factors have been identified as suspected environmental risk factors for breast cancer. These include: light at night (disruptions in melatonin secretion), hormone disruptors (including an extensive list of widespread compounds such as phthalates), environmental pollutants (hydrocarbons, organochlorines), and occupational exposures (chemical, radiation). The role of electromagnetic fields has been given less attention recently, with more emphasis on light at night as a source of melatonin disruption. Epidemiologic studies have shown fairly consistent associations between shift work and other sources of exposure to light at night and increased risk of breast cancer. It has been estimated that 20 percent or more of employed women in California may be exposed to light at night through shift work on jobs in health care, manufacturing, janitorial work, and transportation.
Infectious agents have long been suspected to play a role in breast cancer, including Epstein-Barr virus (EBV) and leukemia viruses of animals. Studies of EBV (Dr. Esther John) and Bovine Leukemia Virus (Dr. Gertrude Buehring) have been supported by the California Breast Cancer Research Program, and may yield important new information. Incidence rates of Hodgkin's disease and breast cancer show strong correlations in nationwide SEER data, which is of interest since delayed exposure to EBV is a risk factor for Hodgkin's disease. No studies have been done to examine potential geographic clusters of breast cancer and delayed EBV infection, but such studies would be interesting to study in relation to population density.
A number of lifestyle factors may affect breast cancer risk, including residence history and social class. To date, not enough research has been done in this area. Why do women of higher income and higher social class have higher incidence rates of breast cancer within the United States than women of lower social class? How much of the difference is due to early detection or increased reporting? How much is due to different life histories and environmental exposures? Mortality from breast cancer is higher in women of lower income and social class. How much of this difference is due to differences in access to health care? Studies of social class and breast cancer have been few and far between, and suffer from lack of data. Comprehensive databases are needed that collect information on race, social class, residence history, and access to care at the time of diagnosis (Krieger, 1990). Urban-rural differences are also important to study. Within the Carolina Breast Cancer Study, a population-based case-control study of breast cancer in African American and white women in North Carolina, researchers observed that women who lived or worked on farms and lived in rural areas had half the risk of breast cancer compared to women who lived in urban areas. The association persisted after adjusting for income and known breast cancer risk factors (Duell et al., 2001). Incidence rates of breast cancer are lower in rural areas than urban areas in many areas of the United States. How much of the urban-rural difference is due to lifestyle or other modifiable risk factors for breast cancer? How much is due to differences in reporting? These are important questions that could be readily addressed in California, an area with extensive urban, suburban, and rural populations. The gradient in population density within California makes it the ideal place to study urban-rural differences in breast cancer.
In order to answer these questions, complementary data collection would be needed. Residential histories and social class would need to be obtained with survey instruments that are developed to integrate with the California Cancer Registry database. Routine abstraction of medical records does not include such variables.
e. Why environmental factors for breast cancer are hard to study
There are several reasons why it has been difficult to study breast cancer and the environment. A few possible explanations and potential solutions are listed below.
Lack of biologic knowledge
The chief obstacle in studies of the environment and breast cancer factors has been lack of a firm biologic foundation. At present, for example, we do not know exactly how hormones increase risk of breast cancer, whether by stimulating cell proliferation, increasing levels of oxidative stress, or some other mechanism. This knowledge is important for understanding how hormones and the environment may interact. We do not know what mutations are necessary for breast cancers to develop, or what causes them. And we do not understand what changes in breast epithelial cells may be reversible, particularly later in life after damage has been done.
We do not know how protective factors work to lower the risk of breast cancer. Studies of physical activity and breast cancer need to be set on a firmer biologic foundation by investigating whether hormonal profiles, oxidative stress, the immune system, or other biologic pathways are involved. Do non-steroidal anti-inflammatory drugs lower risk of breast cancer by reducing oxidative stress or by enhancing apoptosis? What signaling cascades are important for breast cancer, especially non-hormonal pathways? Do some environmental exposures stimulate cell-signaling pathways to increase breast cancer risk? Do other exposures interfere with cell signaling to lower breast cancer risk? For example, recent epidemiologic studies suggest that insulin resistance may be a mechanism for breast carcinogenesis. Increased levels of insulin-like growth factor and related mitogens increase proliferative activity in the breast. Diet, physical activity, obesity, and a variety of other environmental exposures act to alter insulin-related biochemical pathways. Further study is needed to determine the extent to which insulin resistance and related biochemical pathways contribute to risk of breast cancer
Misleading arguments on both sides
There are two fallacious arguments surrounding the role of environmental risk factors for breast cancer. The first, a “pro-environment” stance, states that environmental factors must play a strong role in breast cancer etiology, because “the majority of breast cancer patients have no known risk factors.” The second, an “anti-environment” argument, proposes that environmental exposures are not worth studying because the effects are quite weak.
Contrary to the first argument, epidemiologic studies of breast cancer demonstrate that the majority of breast cancer cases do in fact have at least one “known” risk factor for breast cancer, defined by the list of hormonerelated risk factors and family history (the “usual suspects”). The problem is that most unaffected women also have risk factors for breast cancer. For example, in the Carolina Breast Cancer Study (Newman et al., 1995; Millikan et al., 1995), 97 percent of cases and 96 percent of controls had one or more known hormonerelated risk factors for breast cancer. The problem is not that traditional risk factors for breast cancer are rare, it is that they are too common: well-accepted risk factors for breast cancer do not distinguish well who will develop breast cancer and who will not. The effects of known risk factors, even in combination, are quite weak. This shortcoming can be seen in the Gail-model, which is based upon traditional risk factors for breast cancer. The model has only modest discriminatory power at the individual level, and cannot predict with high accuracy among individual women who will develop breast cancer and who will not (Rockhill et al., 2001).
Contrary to the second argument, one should not dismiss environmental risk factors for breast cancer because they are likely to have weak effects. Based on what we already know about traditional risk factors for breast cancer, one would predict that environmental factors will also have very weak effects, and represent contributory causes, neither necessary nor sufficient for breast cancer. Thirteen epidemiologic studies, including the Long Island Breast Cancer Study, have shown that risk of breast cancer is increased by roughly 50 percent in women with high exposure to polycyclic aromatic hydrocarbons. This increase in risk is twice that of hormone replacement therapy (HRT), which increases risk of breast cancer by 26 percent over a ten-year period, a level of risk sufficient to halt the HRT arm of the Women's Health Study in July 2002. Postmenopausal obesity increases risk of breast cancer by 50 percent, and it is likely that intervention on the basis of diet and physical activity to reduce obesity would have a large impact on breast cancer occurrence in the United States. Environmental factors are not going to be the “smoking gun” for breast cancer. On the contrary, these exposures will usually demonstrate weak effects, similar to most known risk factors for breast cancer. Most probably, environmental factors act in aggregate, rather than as independent exposures, in combination with hormonal factors at many different periods of a woman's life over long periods of time.
Disease heterogeneity
cDNA expression array data has shown that breast cancer is not one disease, but many diseases. Only some subtypes of breast cancer, for example, may be caused by exposure to tobacco smoke or other environmental factors. We should begin to think about sub-classifying breast cancer according to histology, patterns of somatic alterations, cDNA arrays and other characteristics in epidemiologic studies. Somatic genetic alterations in breast cancer have been well studied, but very little has been done to link these changes to specific etiologic agents. Aggregated data from large populations as well as data pooling will be needed for such investigations.
Latency and early life exposures
Measuring early life exposure is a fundamental problem in the study of the environment and any type of cancer. Recognizing the need for such research, the National Institute of Child Health and Development recently launched a cohort study to investigate the role of early life exposures in risk of cancer and other health outcomes later in life. Similar studies have recently been funded as part of the NCI/NIEHS combined program on Centers for Breast Cancer and the Environment. Breast cancer is one of the main health outcomes of interest. One way that breast cancer susceptibility can be studied in younger women and girls is to study alternative health outcomes. Rather than breast cancer as the health outcome, intermediate health outcomes can be studied that are already known to increase risk of breast cancer later in life, for example, early age at onset of menarche.
Failure to address susceptibility
Epidemiologists generally recognize that exposure does not cause disease by itself, but disease is caused by a combination of exposure and susceptibility. Susceptibility can come in many forms. Susceptibility to breast cancer is a function of age and stages of breast differentiation and development. Race, social class, and population density appear to influence susceptibility to breast cancer in ways that are not well understood. Many studies of the environment and breast cancer neglect to study exposure within this greater context, or focus on populations of women with similar income and social standing. The latter has decreased power to detect the effects of a variety of environmental exposures.
Currently, increased attention is being devoted to genetic or inherited susceptibility. The advent of new molecular techniques has brought numerous advances and will undoubtedly provide important opportunities for greater understanding of breast carcinogenesis. Genetic variation in susceptibility to the effects of environmental factors such as tobacco smoke may explain why only some women appear to be susceptible to these exposures. The breast itself has metabolic activity, and may activate as well as sequester a variety of environmental contaminants (Morris and Seifter, 1992). While pesticides, including organochlorines, and aromatic hydrocarbons in general do not appear to be strong influences on breast carcinogenesis, the jury is still out on a variety of chemical exposures. Some of these compounds undergo metabolism in the liver, lung, breast, and other tissues. Some act by damaging DNA. Genetic differences in carcinogen metabolism and DNA repair may help to sort out the association between many environmental exposures and breast cancer. Genetic markers need to be incorporated into epidemiologic studies of breast cancer before concluding that suspect chemicals play no role in etiology. For example, three epidemiologic studies have shown that the association of breast cancer and exposure to polychlorinated biphenyls is modified by inherited polymorphisms in cytochrome P450 genes. Most groups that fund breast cancer research list gene-environment interaction as a high priority. However most of the emphasis recently has been on genes, not the environment. Even studies of gene-environment interactions and smoking in breast cancer are often small and include only rudimentary exposure histories.
Need for improved exposure assessment
The summit on breast cancer and the environment in Santa Cruz, California, in 2002 listed improved exposure assessment as its highest priority for future research into breast cancer and the environment. Better biomarkers are needed to identify exposure to chemicals, pollutants, and agents that modify cell signaling within breast tissue. When disease outcomes are measured close to the time of action of environmental exposures, power to detect effects is increased. Thus, biomarkers are needed for early disease within the breast that can be linked to etiology, not just for early detection and clinical intervention but also to better understand the causes of breast cancer. We need to develop improved biomarkers of exposure (e.g. assays for low-levels of chemical pollutants in blood and tissue) and biomarkers of early disease (e.g. alterations in cell signaling pathways, activation of oncogenes, down-regulation of tumor suppressor genes) that can be linked to these exposures
Candidate environmental exposures such as chemical pollutants also need to be studied at the aggregate or group level, not just at the individual level. Databases that include levels of environmental pollution could be linked to breast cancer incidence data from cancer registries. In this manner, regions with higher or lower breast cancer rates could be studied in relation to ground water contamination, pesticide and herbicide use, and levels of endocrine disruptors in water and soil. Studies on Long Island and Cape Cod have addressed some of these issues, and provide promising leads. But much of the environmental exposure information in these studies had to be collected de novo. In California, due to Proposition 65, records of pesticide use and other environmental exposures are routinely captured in publicly accessible databases and could be linked to incidence data from the California Cancer Registry. One issue in such studies is population migration, but obtaining residence histories for persons residing over long periods of time in California would overcome many of these problems. California is the only state with this level of environmental data, and mining it to understand the causes of breast cancer is critical.
