Issues and controversies in California that could be addressed by the CBCRP
a. Supposed cancer clusters
Studies of breast cancer incidence show considerable variation within countries and within smaller geographic units, particularly in the United States. A suspected “cluster” of breast cancer has been identified in the northeastern United States (Kulldorf, 1997). More recent analyses suggest that “pockets” of increased breast cancer incidence are found in the western and mid-western United States, as well as the possibility of smaller “clusters” in Marin county and other regions of California. A major limitation of these studies is a failure to examine incidence rates in comparable geographic units. Historically, breast cancer incidence and mortality rates in California have been compared at the county level. Rates in small counties such as Marin are compared to rates in large counties such as Los Angeles. Residual heterogeneity and possible sub-county aggregation cannot be addressed in this manner. The comparability problem could be addressed by computing breast cancer incidence rates for census tracts and aggregating regions with comparable demographics. Incidence could then be compared in regions with comparable demographics, for example Marin County and Beverly Hills. Areas with high as well as low breast cancer incidence rates need to be studied, since low incidence areas might provide clues to identify protective factors.
Such studies would require applications of novel statistical methods, and need pilot testing and validation. Incidence rates could be aggregated according to the newest cluster methods, such as the “moving window” approach (Kulldorf, 1997). Bayes or semi-Bayes methods would need to be used to compute incidence rates in small geographic units, and would require development of new software applications. There are several reasons why this type of work is important: it would answer important questions for residents of high incidence areas, it would avoid unnecessary expenditure of time and money in regions where clusters do not actually exist, and it would lay an important foundation for monitoring breast cancer rates in the future.
b. Reporting biases
According to the California State Cancer Registry, 40 percent of breast cancer cases are reported from hospitals without tumor boards. The advent of increased SEER support will help to determine how much variation in cancer incidence rates in the state may be due to differences in reporting. Such differences are important to address when making comparisons of rates according to social class, and urban/rural status.
c. Race, social class and breast cancer
Comparisons of breast cancer incidence and mortality rates in different racial and ethnic groups often fail to address differences in social class (Krieger et al., 1997; Faggiano, 1997). Information on residence histories and social class could be gathered from cancer patients in California and included in the Cancer Registry database. Patient residence history could be used to link cancer rates to census tract and group level information on income, employment, social class, and other risk factors for breast cancer.
A dramatic increase has occurred in the number of persons living in poverty in the United States. The US Census Bureau reported that the number of Americans living in poverty increased by 1.7 million in 2002, the second year in a row. The median family income declined for the second year in a row. The gap between the wealthiest and poorest individuals in the US has widened, and in California the gap is now the highest in the nation. It has become increasingly important to understand how income and social class differences influence risk of breast cancer and access to effective treatment.
d. Distributions and determinants of age at menarche, age at menopause, and other reproductive risk factors for breast cancer
Survey methods could be used to gather information at the census tract level on traditional risk factors for breast cancer. This information could be used to determine whether such factors aggregate in specific geographic regions and population subgroups.
The information could also be used to adjust for confounding in ecologic studies of environmental exposures and breast cancer. For example, earlier age at menarche is an accepted risk factor for breast cancer. Average age at menarche appears to be decreasing among women in the US, especially in specific population subgroups defined by geography, race, and social class. Determinants of age at menarche, attained height during adolescence, and obesity are some of the outcomes that need to be studied at the individual as well as the aggregate level. For example, one could study average ages at menarche across geographic regions (ecologic studies) and link this information to environmental databases. One could then determine whether average age at menarche was lower in regions with heavy use of pesticides or other sources of environmental contamination. In areas where correlations were observed at the group level, investigators could conduct investigations that gather individual level information (cross-sectional or case-control studies). Exposure histories in individual women could be further investigated, perhaps by including blood levels of persistent organochlorines and biomarkers of exposure to other environmental chemicals.
e. The effect of mammography on stage at diagnosis of breast cancer
One goal of screening mammography is to decrease the incidence of late stage breast cancer. However, SEER data for the US and screening trials in Canada suggest that an increase in incidence of in situ breast cancer is not followed by a reduction in the incidence of late stage breast cancer. It is possible that such a reduction takes longer than ten to twenty years, but it is also possible that early detection finds many breast cancers that would never develop into life-threatening invasive cancer. If the latter is true, then considerable resources are being spent on mammography that could be more effectively used for other breast cancer-related activities. Analyses of long-term breast cancer incidence rates in California would address this issue. Such studies must take into account the prevalence of mammographic screening, since it is possible that women who are diagnosed with late stage breast cancer are the women who are not being screened. It is important to collect incidence data over longer time periods than recent clinical trials of screening mammography.
Differences in screening prevalence also need to be addressed in comparisons of breast cancer rates by social class and geography. Data on geographic differences present according to stage at diagnosis of breast cancer would be interesting. Specific studies could be funded to investigate potential explanations for such differences, including geographic and economic barriers to receiving health care.
f. Environmental contaminants and breast cancer risk
A recent study showed that levels of bromine-based fire retardants are found at higher levels in breast milk from women in the San Francisco Bay Area than anywhere else in the world. The study was too small to address where in the Bay Area population and for whom these levels might be highest, and no correlations were made with environmental sampling to determine sources of exposure. No attempt was made to address risk of breast cancer from such exposure. Other potential breast carcinogens do not accumulate in the body, and thus cannot be studied at the individual level using blood or fat measurements. Group level data may be the only way to study such exposures. Information on environmental levels of these compounds could be studied using GIS and geo-coding and linked to breast cancer incidence rates or to existing case-control studies of breast cancer.
One example of such an approach is a recent study by Aschengrau et al. (2003). The investigators used residence histories and inspection of home piping systems to estimate exposure to perchloroethlyene (PCE)-contaminated drinking water. Study participants with the highest PCE exposures had a small increase in risk of breast cancer. Ecologic studies of drinking water have linked increased breast cancer rates to higher levels of a variety of environmental pollutants. One strength of the Aschengrau et al. (2003) study was the fact that investigators collected individual level information on traditional breast cancer risk factors and combined this with ecologic level exposure measurement using geographic information system (GIS) mapping software and other techniques.
