Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2021 Dec 1.
Published in final edited form as: Appl Geogr. 2020 Sep 23;125:102324. doi: 10.1016/j.apgeog.2020.102324

Identifying county-level factors for female breast cancer incidence rate through a large-scale population study

Tingting Zhao 1,*, Zihan Cui 2, Mary Grace McClellan 1, Disa Yu 2, Qing-Xiang Amy Sang 3, Jinfeng Zhang 2,*
PMCID: PMC7543978  NIHMSID: NIHMS1631247  PMID: 33041393

Abstract

Female breast cancer (FBC) incidence rate (IR) varies greatly across counties in the United States (U.S.). Factors contributing to these geographic disparities have not been fully understood at the population level. In this study, we investigated the relationships between the county-level FBC IR and a diverse set of variables in demographics, socioeconomics, life style, health care accessibility, and environment. Our study included 1,277 counties in the U.S. where the female population was 10,000 or above for at least one race/ethnicity. After controlling for the racial/ethnic and other significant factors, percent of husband-wife family households (pHWFH) for a racial/ethnic group in a county is negatively associated with FBC IR (p < 0.001). A 10% increase in married family households may lower a county’s IR by 5.2 cases per 100,000 females per year. We also found that PM2.5 (fine inhalable particles with a diameter of 2.5 micrometers or less) is positively associated with FBC IR (p < 0.001). Counties with the highest level of PM2.5 have approximately 4 additional FBC new cases per 100,000 females per year than counties with the lowest level of PM2.5. Furthermore, we found that the county-level factors contributing to FBC IR vary significantly for different racial groups using race-specific models. While confirming most of the previously known patient- and neighborhood-level risk factors (such as race/ethnicity, income, and health care accessibility), our study identified two significant county-level factors contributing to the spatial disparity of FBC IR across the U.S. The newly-identified beneficial factor (marriage) and risk factor (PM2.5), together with the verified known factors, may help provide insights to officials of health departments/organizations for them to make decisions on cancer intervention strategies.

Keywords: cancer epidemiology, geographic disparity, marital status, air pollution, public health, United States

1. Introduction

In the United States (U.S.), female breast cancer (FBC) continues to be the number-one leading site of all diagnosed cancers in women (American Cancer Society, 2018). The incidence rate (IR) of FBC varies greatly by states and counties across the country (Buck, 2016, Scott et al., 2017). The spatial variance of FBC incidence is found to be associated with income inequality in previous studies (Liu et al., 2016). It also relates to urban-rural inequality; for example, the overall late-stage breast cancer risk was the highest in the most urbanized areas in Illinois, U.S. (McLafferty et al., 2011). A more recent study across eight U.S. states reported high incidence of late-stage breast cancer in areas with predominately black population, where education and screening availability were low (Tatalovich et al., 2015). Meanwhile, advanced spatial statistics, such as spatial scan statistics initially developed for cancer count data (Kulldorff, 1997, Kulldorff et al., 2006) and subsequently extended to analyze ordinal and continuous measures (Huang et al., 2007, Jung et al., 2007), regionalization methods (Wang et al., 2012), measures of health care accessibility (Xu et al., 2017, Wan et al., 2012), spatial autocorrelation filtering techniques (Hu et al., 2020), have been developed to identify and analyze the geographic variance of cancer outcomes including FBC incidence.

Despite extensive research in the past, factors associated with FBC spatial disparities have not been fully understood. One of the major challenges in the population-level research of cancer health disparities is to be inclusive of potential variables into a single study. In practice, most studies include only a limited number of factors under investigation. Considering socioeconomic status (SES), scholars found that the overwhelming majority of previous studies used poverty as the single socioeconomic measure for cancer outcomes research (Zahnd and McLafferty, 2017). However, factors excluded from those studies (such as percent of female-headed households, percent of less educated population, and percent of recent immigrants) were found to be significant socioeconomic predictors for breast cancer stage at diagnosis in California (Davidson et al., 2005). Second, while many studies incorporating SES measures, only a few included neighborhood health and environmental variables such as housing, food, and physical activity (Gomez et al., 2015). These factors may play a role in the spatial disparity of cancer outcomes. According to a recent study in Baltimore City, the spatial distribution of FBC incidence was found to be associated with crime and business measures (Torres et al., 2018). Third, the scale of spatial units and analytical approaches varied greatly, adding difficulties to compare across different studies (Gomez et al., 2015).

In this study, we explored a large, diverse set of factors contributing to the county-level FBC IR across the U.S. Many of them are known risk factors for FBC incidence at the patient level. These factors include demographic, socioeconomic (Akinyemiju et al., 2015, Yost et al., 2001), lifestyle (Robert et al., 2004, Seiler et al., 2017), health accessibility (Broeders et al., 2018), and environmental variables (Hiatt and Brody, 2018) compiled at the county scale. All came from various census, health, and other geographic data sources. Additionally, variables characterizing people’s living condition, such as household/family and labor/occupation, are also included in this study. However, genetic/biological factors and reproductive history are not included in the present analysis due to their unavailability at the county scale. As race is a well-known, significant risk factor for FBC IR (Seiler et al., 2017, Joslyn et al., 2005, Li et al., 2017, Shi et al., 2017) and also interacts with many other variables, data for many of the predictor variables in this study were collected separately for different racial/ethnic groups.

As far as we know, this is the largest-scale study focusing on spatial variation of FBC IR in the U.S., in terms of the extent of geographic regions covered, number of risk factors taken into consideration, and the multiple races/ethnicities under investigation. Our research is consistent with National Cancer Institute’s aim of continuously promoting research efforts concerning spatial and environmental aspects of cancer etiology (Schootman et al., 2017). We believe the results of our study provide insight for local governments and health organizations to create intervention strategies (Polite et al., 2017, Forman et al., 2015) for behavioral, healthcare, and environmental changes to reduce breast cancer risks.

2. Data

FBC incidence rates (IR, Table 1) were extracted from the web portal of State Cancer Profiles (National Cancer Institute, 2017). This data portal provides age-adjusted IR measured as new cases per 100,000 females per year. The 5-year incidence from 2009 to 2013 was acquired for four race/ethnicity groups including White Non-Hispanic, Black (includes Hispanic), Asian/Pacific Islander (includes Hispanic), and Hispanic (any race). The tabulation data were linked spatially to the 2010 Census counties (Figure 1a–1d). In the following parts of this paper, FBC IR refers to the FBC incidence rate broken down by racial/ethnic groups. The racial and ethnic groups are broadly defined as Non-Hispanic White (NHW), Black, Asian/Pacific Islander (API), and Hispanic Americans.

Table 1.

List of variables and data sources. Descriptive statistics are reported in Appendix A.

Variable Race/Ethnicity* Description Source of Raw Data
Breast Cancer data
Incidence rate (IR) Individual New cases per 100,000 females per year State Cancer Profiles: Incidence Rates Table1
Demographic and socioeconomic data
Percent of population (pPop) Individual Percent of people who reported a single race and Hispanic/Latino origin Profile of General Population and Housing Characteristics: 20102
Percent of urban population (pUrban) Combined Percent of population in urbanized areas and urban clusters Urban and Rural: 20102
Mean household size (meanHS) Combined Average number of people living in a housing unit Profile of General Population and Housing Characteristics: 20102
Mean family size (meanFS) Combined Average number of people in a family Profile of General Population and Housing Characteristics: 20102
Female median age (FMedAge) Individual Median age of females for each race/ethnicity (years) 2010 Census: Median Age by Sex by Race2
Female population by age groups Individual Percent of female population in each age group (under 20, 20s, 30s, 40s, 50s, 60s, 70 and older, 50 and older, and 60 and older) for reach race/ethnicity 2010 Census: Sex by Age2
Female total (fPop) Individual Total population of females for each race/ethnicity 2010 Census: Population by Sex by Age by Race2
Percent of family households (pFH) Individual Number of family households divided by total households for each race/ethnicity (%) 2010 Census: Household Type by Race2
Percent of husband-wife family households (pHWFH) Individual Number of husband-wife family households divided by total households for each race/ethnicity (%) 2010 Census: Household Type by Race2
Percent of female-householder family households (pFFH) Individual Number of female-householder (with no husband present) family households divided by total households for each race/ethnicity (%) 2010 Census: Household Type by Race2
Percent of husband-wife family with children households (pHWFHWC) Individual Number of husband-wife family with own children under 18 years divided by total households tor each race/ethnicity (%) 2010 Census: Family Type by Presence and Age of Own Children by Race2
Percent of female-householder family with children households (pFFHWC) Individual Number of female-householder (with no husband present) with own children under 18 years divided by total households for each race/ethnicity (%) 2010 Census: Family Type by Presence and Age of Own Children by Race2
Median household income (medlncome) Individual Median household income for each race/ethnicity (dollars) 2009–2013 5-Year American Community Survey: Median Income in the Past 12 Months (in 2013 Inflation-Adjusted Dollars) by Race2
Percent of females with high school diploma or higher education (pFHS) Individual Percent of females with high-school diploma or higher education for females 25 years and over 2009–2013 5-Year American Community Survey: Sex by Educational Attainment for the Population 25 Years and Over by Race2
Percent of females in MBS A or SO occupations (pF MBSASO) Individual Percent of females in Management, Business, Science, and Arts occupations (MBSA) or Sales and Office (SO) occupations among all females 2009–2013 5-Year American Community Survey: Sex by Occupation for the Civilian Employed Population 16 Years and Over by Race2
Percent unemployed (pUnem) Combined Percent of population age 16+ unemployed (2011) County Health Rankings, 2013: Ranked Measures Data - Bureau of Labor Statistics3
Lifestyle and Health Care Data
Percent of smokers (pSm) Combined Percent of adult smoking (2005–2011) County Health Rankings, 2013: Ranked Measures Data - Behavioral Risk Factor Surveillance System3
Percent excessive drinking (pExDr) Combined Percent of adults who report heavy drinking (2005–2011) County Health Rankings, 2013: Ranked Measures Data - Behavioral Risk Factor Surveillance System3
Percent obese (pOb) Combined Percent of adults that report a BMI >=30 (2009) County Health Rankings, 2013: Ranked Measures Data - National Center for Chronic Disease Prevention and Health Promostion3
Percent diabetic (pDia) Combined Percent of adults aged 20 and above with diagnosed diabetes (2009) County Health Rankings, 2013: Additional Measures - National Center for Chronic Disease Prevention and Health Promotion, Division of Diabetes Translation3
Percent diabetic screening (pDiaSc) Combined Percent of diabetics that receive HbAlc screening (2010) County Health Rankings, 2013: Ranked Measures Data - National Center for Chronic Disease Prevention and Health Promotion, Division of Diabetes Translation3
Percent physically inactive (pPhylnact) Combined Percent of adults that report no leisure time physical activity (2009) County Health Rankings, 2013: Ranked Measures Data - National Center for Chronic Disease Prevention and Health Promotion3
Percent uninsured (pUnins) Combined Percent of population without health insurance (2010) County Health Rankings, 2013: Ranked Measures Data - Small Area Health Insurance Estimates3
Primary Care Provider (PCP) rate (PCPrate) Combined Primary care providers per 100,000 population (2011–2012) County Health Rankings, 2013: Additional Measures - HRSA Area Resource File3
Percent mammography screening (pMam) Combined Percent of females that receive screening (2010) County Health Rankings, 2013: Ranked Measures Data - Dartmouth Atlas of Health Care3
Physical and Social Environment
Average daily PM2.5 (PM2.5) Combined Average daily fine particulate matter (ug/mJ), 2008 County Health Rankings, 2013: Ranked Measures Data - CDC WONDER Environmental Data3
Percent population in violation of water safety (pWatVio) Combined Percent of population exposed to water exceeding a violation limit (2012) County Health Rankings, 2013: Ranked Measures Data - Safe Drinking Water Information System3
Recreational facility rate (RecRate) Combined Access to recreational facilities (rate/100,000 population), 2010 County Health Rankings, 2013: Ranked Measures Data - County Business Patterns3
Percent limited access to healthy food (pLAHF) Combined Percent of population who lives in poverty and does not live close to a grocery story (2012) County Health Rankings, 2013: Ranked Measures Data - USDA Food Environment Atlas3
Percent fast food restaurants (pFFR) Combined Percent of all restaurants that are fast food (2010) County Health Rankings, 2013: Ranked Measures Data - County Business Patterns3
Open in a new tab
*

“Individual” indicates statistics were extracted by each single race/ethnic group, while “combined” refers to variable regardless of racial/ethnic differences due to data availability. In our study, race includes White (not Hispanic or Latino), Black, Asian, and Native Hawaiian and Other Pacific Islander. “Asian” and “Native Hawaiian and Other Pacific Islander” were combined to match cancer statistics for the Asian/Pacific Islander group. Ethnicity refers to Hispanic or Latino origin.

Figure 1.

Figure 1.

Open in a new tab

Female breast cancer (FBC) incidence rate (IR; a-d) and percent (%) of population (e-h) for each racial/ethnic group in the U.S. counties. Only the conterminous U.S. is shown in maps. IR: Incidence rate (cases per 100,000 population per year, age-adjusted) over 2009–2013; NHW: Non-Hispanic White; API: Asian/Pacific Islander.

The demographic and socioeconomic data came from the U.S. Census Bureau’s 2010 Census as well as the 2009–2013 5-Year American Community Survey (Table 1). Variables include population by racial/ethnic groups, population by urban/rural areas, household/family size, female age (median age and population counts by age groups), household family characteristics (such as marital status and presence of children), household income, female counts by four levels of educational attainment, and female counts by five types of occupations. Most of the demographic and socioeconomic variables were extracted by each of the four racial/ethnic groups. They were all linked spatially to the 2010 Census counties; for example, percent population by racial/ethnic group (Figure 1e–1h).

Lifestyle, health care, and environmental data came from the 2013 County Health Rankings National Data (Remington et al., 2015). We selected several risk factors such as smoking, drinking, obesity, diabetes, and physical activities to measure life style. Primary Care Provider rate (i.e., number of PCPs per 100,000 people), mammography screening, diabetic screening, and uninsured rate were extracted as indicators of accessibility to health care. We also included environmental variables such as air quality (measured as average daily fine particulate matter, i.e., PM2.5), water quality, recreation facilities, and access to healthy food. Variables from the County Health Rankings dataset are for the population as a whole rather than specific racial/ethnic groups. They were also linked spatially to the 2010 Census counties.

3. Methods

For statistical analyses, only counties with a female population of 10,000 and above for one or more racial/ethnic groups were selected for statistical tests and modeling. The rationale is to exclude counties with too few samples of FBC incidence due to very low female population. Statistical methods adopted in this study included descriptive statistics (Appendix A), hierarchical linear regression (HLR), and linear regression. All analyses were performed in IBM SPSS Statistics version 22.

For HLR, weighted least squares regression was adopted. Model dependent variable was the county-level FBC IR by race/ethnicity. The model was weighted by county’s female population by race/ethnicity. In the initial analysis, predictors for the first block were binary variables of race/ethnicity, while predictors for the second block included all other independent variables in Table 1. We used the forward selection process to identify factors associated with the dependent variable. Then we examined each parameter’s significance as well as its variance inflation factor (VIF, a measure of multicollinearity) in the model outputs. Multicollinearity refers to correlation among the predictor variables in multiple regression. A VIF of 10 and above indicates a multicollinearity problem (Goldberger, 1991). In our analyses, parameters with VIF above 10 were dropped out from the HLR one at a time. In the final step, we created a three-block HLR. The first block was race/ethnicity. The second block included measures of cancer screening accessibility. And the last block included all other variables determined statistically significant based on the previous model forward selection and variable elimination (according to VIF) processes.

In addition, to examine factors contributing to FBC IR for individual racial/ethnic groups, we performed four race-specific linear regression analyses (one for each race/ethnicity separately). Again weighted least squares regression was adopted. Variable forward selection process was used and VIF values were used to reduce multicollinearity.

4. Results

4.1. County-level FBC IR and county’s socioeconomic characteristics by racial/ethnic groups

In total, 1,609 of 3,143 U.S. counties had a female population of 10,000 or more for at least one racial/ethnic group. We excluded counties with no incidence data reported due to state legislation and regulations. Therefore, 1,277 counties were included in our statistical analysis. FBC IR weighted by county’s female population was highest for NHW (129.01), followed by Black (123.99), Hispanic (92.40), and API (87.94).

For these 1,277 counties, the median age of females (Figure 2a) and the median household income (Figure 2b) were significantly higher for NHWs and APIs than they were for the Black or Hispanic group (Appendix B). Among women 25 years or older, a significantly higher proportion within the NHW groups obtained an education of high-school diploma or higher compared to the other racial/ethnic groups (Figure 2c; Appendix B). NHW and API groups had a significantly higher percentage of working women in MBSA (i.e., management, business, science, and arts) and SO (i.e., sales and office) occupations combined compared to the other two racial/ethnic groups (Figure 2d; Appendix B).

Figure 2.

Figure 2.

Open in a new tab

Boxplots of the county-level female median age (a), median household income (b), female educational attainment (c), female occupations (d) and family characteristics (e. husband-wife family, f. husband-wife family with children, g. female-householder family, h. female-householder family with children) by four racial/ethnic groups

In terms of family characteristics, approximately 62% of all API households were husband-wife family, significantly higher than all other racial/ethnic groups (Figure 2e; Appendix B). The proportion of husband-wife families with children present was significantly higher for APIs and Hispanics compared to the other groups (Figure 2f; Appendix B). Approximately 30% of the Black households were female-householder family (Figure 2g) with 18% having at least one child (Figure 2h); both numbers are significantly higher than those of the other racial/ethnic groups (Appendix B).

4.2. Significant variables for FBC IR

HLR with the race/ethnicity variables alone explained 53.8% of the total variance in the county-level FBC IR (Model 1 in Table 2). This indicates that race and ethnicity plays a significant role. API, Hispanic, and black populations have a significantly lower IR compared to NHW. Adding measures of cancer screening accessibility, i.e., rate of PCP (PCPrate) and percent uninsured population (pUnins), the model explained 63.2% variance of the county-level FBC IR (Model 2 in Table 2). Rate of PCP was positively related to IR, while percent uninsured population was negatively related to IR. These indicate counties with a higher number of PCPs and higher percentage of people with health insurance saw an increasing FBC incidence. The model R2 change between Model 1 and 2 was 0.094 (Sig. = 0.000), which indicates these two variables explained 9.4% of the total variance in the county-level FBC IR.

Table 2.

Factors contributing to FBC IR based on hierarchical weighted least squares regression analyses. Dependent variable: IR by four races/ethnicity. Weighted by female population of each racial/ethnic group. Model 1 adjusted R2: 0.538; Model 2 adjusted R2: 0.632; Model 3 adjusted R2: 0.707; Model 2 R2 change: 0.094 (Sig. = 0.000); Model 3 R2 change: 0.076 (Sig. = 0.000). PCPrate: count per 100,000 population; medIncome: thousand dollars; PM2.5: μg/m3.

Model
 Unstandardized Coefficients Standardized Coefficients Beta t Sig. 95.0% Confidence Interval for Collinearity Statistics
B
B Std. Error Lower Bound Upper Bound Tolerance VIF
1 (Constant) 129.015 .403 319.931 .000 128.224 129.806
Black −5.026 .956 −.084 −5.257 .000 −6.901 −3.151 .955 1.047
API −41.074 1.533 −.423 −26.798 .000 −44.080 −38.068 .978 1.023
Hispanic −36.615 .887 −.660 −41.267 .000 −38.355 −34.875 .953 1.049
2 (Constant) 116.913 1.612 72.523 .000 113.751 120.074
Black −5.983 .861 −.100 −6.953 .000 −7.671 −4.295 .940 1.063
API −43.631 1.379 −.449 −31.642 .000 −46.335 −40.927 .963 1.038
Hispanic −34.676 .836 −.625 −41.502 .000 −36.315 −33.037 .857 1.166
PCPrate .220 .012 .282 18.539 .000 .196 .243 .842 1.187
pUnins −.234 .059 −.063 −3.963 .000 −.350 −.118 .776 1.289
3 (Constant) 112.634 4.733 23.799 .000 103.352 121.916
Black −6.696 1.369 −.112 −4.891 .000 −9.380 −4.011 .295 3.386
API −39.608 1.378 −.408 −28.737 .000 −42.311 −36.905 .766 1.305
Hispanic −25.949 1.312 −.468 −19.780 .000 −28.522 −23.376 .276 3.619
PCPrate .103 .012 .132 8.427 .000 .079 .127 .632 1.582
pUnins −.127 .057 −.034 −2.228 .026 −.239 −.015 .659 1.518
medlncome .442 .021 .370 21.434 .000 .401 .482 .517 1.934
pHWFH −.522 .044 −.271 −11.928 .000 −.608 −.436 .300 3.339
PM2.5 .482 .167 .038 2.878 .004 .154 .810 .864 1.158
FMedAge .126 .068 .043 1.861 .063 −.007 .258 .294 3.403
Open in a new tab

The final model with all significant variables explained 70.7% of the total variance in the county-level FBC IR (Model 3 in Table). The model R2 change between Model 2 and 3 was 0.076 (Sig. = 0.000). This means that 7.6% of the IR variance were explained by median household income, percent of married-couple families (pHWFH), PM2.5, and female median age combined. Income is positively related to IR; specifically, every increase of $10,000 in median household income is associated with 4.4 additional cases of FBC per 100,000 females. Our model also indicates a significant relationship between family composition and the county-level FBC IR. This is the first time that this variable has been identified as significantly associated with FBC IR at the population level. As measured by counties and separate racial/ethnic groups, for every 10% increase in husband-wife family households (among all households), FBC IR decreases by 5.2 cases per 100,000 people. We also found that the county-level measure of average daily PM2.5 is positively related to FBC incidence. For an increase of 1 μg/m3 in the fine particulate matter, there are 0.48 additional cases of FBC per 100,000 females.

4.3. Race-specific models

The weighted least squares regression analyses for each race/ethnicity showed a divergent pattern with a few factors contributing differently to the race-specific, county-level FBC IR models (Table 3). However, indicators positively contributing to cancer screening (such as high PCP rate and low uninsured population) were generally positively associated with FBC IR. High median household income and high female education (pFHS) attainment were associated with high FBC IR. Unemployment rate (pUnem) was positively associated with FBC IR for NHW and API population. Family marriage status as measured by pHWFH (or, the reverse, pFFH) was a significant factor for all races except for the Hispanic category. Counties with a large percent of married-couple families (pHWFH for NHW and Black) saw a decreasing FBC IR, while those with a large percent of female-householder family (pFFH for Black) were associated with increasing FBC IR. In addition, counties of high percent of married-couple families with children (pHWFHWC) were associated with declined FBC IR for API. PM2.5 was significantly positively associated with FBC incidence in NHW and Black populations. Percent of obese population (pOb) was positively related to FBC IR for African Americans.

Table 3.

Factors associated with FBC IR for each race/ethnicity based on weighted least squares regression analyses (weighted by female population of each racial/ethnic group). Dependent variable is IR for each of the four racial/ethnic groups. Model adjusted R2 for NHW is 0.558, Black 0.309, API 0.512, and Hispanic 0.282. PCPrate: count per 100,000 population; medIncome: thousand dollars; meanFS: number of people; PM2.5: μg/m3.

NHW Black API Hispanic
Variables B Sig. VIF B Sig. VIF B Sig. VIF B Sig. VIF
(Constant) 26.917 0.016 −64.530 0.007 −16.071 0.393 56.644 0.000
PCPrate 0.093 0.000 1.837 0.120 0.000 1.581 0.087 0.041 1.475
pDiaSc 0.588 0.000 1.292
pUnins −0.280 0.000 1.907 −1.283 0.000 1.462 −0.372 0.022 1.623
medlncome 0.420 0.000 2.683
pUnem 0.436 0.010 1.735 1.899 0.000 1.527
pFHS 1.217 0.000 1.989 0.740 0.001 1.509 0.593 0.000 1.202
pHWFH −0.476 0.000 1.687 −0.317 0.021 2.085
pHWFHWC −0.596 0.000 1.548
pFFH 0.618 0.037 2.926
meanFS 8.200 0.000 2.813
pUrban 0.095 0.000 2.028
PM2.5 0.478 0.007 1.194 0.998 0.025 1.396
FMedAge 0.166 0.042 1.596 0.894 0.000 1.204 1.800 0.000 1.364
pOb 0.845 0.000 2.175
Open in a new tab

5. Discussion

Breast cancer is a very complex disease and many associated risk factors have complex interactions among themselves. Therefore, it is desirable to include as many variables as possible in a study to understand how a particular variable may affect FBC IR while controlling the other variables. However, integrative studies that incorporate SES, lifestyle and environmental factors are challenging at the national scale (Lynch and Rebbeck, 2013, Gomez et al., 2015). Most recently there are a few studies focused on either incorporating as many factors as possible at the local scale (Torres et al., 2018) or exploring spatial disparity across the nation with limited factors (Scott et al., 2017). One study explored a large set of factors contributing to the spatial patterns of a cancer outcome, measured as mortality to incidence ratio, across the U.S. counties (Buck, 2016). The measure was based on cancer incidence and mortality rates for all cancer sites combined. Research focusing on individual cancer types is in need, since each type of cancer has its own risk factors that may vary across space.

Although our study at the national scale tried to include as many factors as possible, the factors under investigation took no direct consideration of genetics, biology, or reproductive history of individuals. Therefore, impacts from those factors and their potential interaction with factors under investigation are missing from the present models. Many of the life style factors are aggregated data, which do not separate races or genders. Therefore, they may have small effects on the modeling of FBC IR, which is an outcome measure for females only and was also separated by racial/ethnic groups in our study. The modeling technique used in this study does not capture non-linear relationships or conditional dependencies of model parameters, which may be enhanced in the future research using advanced statistical models such as non-linear models or Bayesian network models. In addition, our dependent variable was the aggregated incident rate at the county level. In the future, analyses on case-control data and variables at the location of the geocoded addresses (if permitted) may enhance our understanding on factors contributing to FBC spatial disparities. Regardless of these limitations, our analyses captured relationships between the population-level FBC IR and counties’ demographic, socioeconomic, health, and pollution characteristics.

5.1. Convergent and controversial factors

Race/ethnicity still proves to be the most important factor associated with a county’s FBC IR, accounting for 53.8% of the county-level IR variance. We found that income and female education attainment are both positively associated with FBC IR (Table 2 and 3); the latter plays significant roles in all individual racial/ethnic groups except for NHW (Table 3). These are consistent with most previous studies that found breast cancer incidence is higher in affluent classes (Klassen and Smith, 2011, Scott et al., 2017). We also found increased health accessibility, measured by higher PCP rate, lower uninsured population, and (for NHW population only) percent diabetic screening, is positively associated with FBC IR (Table 2 and 3). This converges with an earlier study based on SEER-18 registries (Moss et al., 2017). A higher rate of PCP may indicate a higher chance of detecting and diagnosing breast cancer (Gangnon et al., 2015, Kirby et al., 2017, Mobley et al., 2012, Wang and Onega, 2015).

None of the life style factors, except percent obese (for Black population), stands out in our models of factors contributing to a county’s FBC IR (Table 2 and 3). Obesity was found to be positively associated with breast cancer incidence in our study, consistent with existing literature (Robert et al., 2004, Hiatt et al., 2014, Engmann et al., 2017). Although diabetes (Hardefeldt et al., 2012), smoking (Swift and Lukin, 2008, Rosenberg et al., 2013) and excessive drinking (Key et al., 2006) are both risk factors of breast cancer based on clinical data, we were not able find any significant association from these variables aggregated at the population level. One possible explanation may be that these county-level measures include counts of both men and women; in contrast, the FBC IR is a measure of cancer outcome for women only. Take smoking as an example. There is a high spatial variance in female smoking rates (Osypuk et al., 2006); therefore, the county-level percentage of total population who smoke cannot substitute the proportion of females who smoke.

5.2. Newly identified factor: Family marital status and FBC incidence

Our HLG analyses (Table 2) showed that, at the county level, a 10% increase in pHWFH is equivalent to 5.2 fewer new cases. Given that the national average of FBC IR is 124 new cases per 100,000 females (American Cancer Society, 2018), such a positive impact from family composition on reducing cancer rate is fairly notable.

According to the U.S. Census, the average pHWFH is approximately 30% for the Black population. The average for Hispanics and NHWs is approximately 50% and 53%, respectively. All are significantly lower than the average for API, which is approximately 62% (Figure 2e; Appendix B). Among the four groups, API is associated with the lowest FBC incidence (Table 2). A portion of the FBC IR disadvantage for the racial/ethnic groups other than API may be attributable to family composition characteristics, according to our research findings. In particular, counties with a higher proportion of husband-wife family households saw a lower rate of breast cancer incidence for non-Hispanic White and Black populations (Table 3).

A literature search on family composition, especially marital status, and its impacts on FBC incidence returned very few studies. It is unclear how family characteristics directly affect breast cancer incidence, although perhaps marital status may affect women’s level of stress (Lillberg et al., 2003), timing of girls’ pubertal onset (Deardorff et al., 2011), parity and/or practice of contraception, which is not directly accounted for in our current statistical models. Higher parity has been documented to decrease FBC risk in East-Asian women, especially in BRCA1 carriers (Park et al., 2017). A meta-analysis provided evidence of an increase in breast cancer risk associated with oral contraception use, especially for long-term users (Zhu et al., 2012). A local-scale study in the U.S. based on patient records reported 17.5% of the married women reported no contraceptive use, 4.6 percent higher than single women (Williams et al., 2012). Similarly an earlier study in Europe found that, among women who use oral contraceptives, only 30–45% of them were ever married (Oddens and Lehert, 1997). These studies indicate that married women appear to have a lower rate of oral contraception use, which may have led to a lower rate of breast cancer incidence. However, controversy about the impacts of oral contraceptives on breast cancer incidence still exists (Tirona et al., 2010, Hiatt and Brody, 2018). Too few studies have been focused on marital status and contraception, let alone its broader impacts on cancer incidence. Thus, marital status and its impact on breast cancer incidence requires further research before related recommendations regarding breast cancer intervention and education can be determined.

5.3. PM2.5 and FBC incidence

The impacts of environmental pollutants on breast cancer risk have been documented in a growing number of studies over the past two decades (Jagai et al., 2017, Hiatt and Brody, 2018). The evidence supports the association between breast cancer and polycyclic aromatic hydrocarbons (PAHs), which are organic compounds that may be produced by the incomplete combustion in engines and incinerators or the burning of biomass (Brody et al., 2007, Gray et al., 2009). PAH residues can bind to particulate matters (PM), which are microscopic solid or liquid matter suspended in the atmosphere. Exposure to high concentrations of total suspended particulates at birth was found to increase the chance of breast cancer incidence in Erie and Niagara Counties, State of New York (Bonner et al., 2005).

PM2.5 is a subtype of PM, referring to fine inhalable particles with a diameter of 2.5 micrometers or less. It is the main cause of reduced visibility in parts of the United States. More recently, a U.S. study based in Atlanta, GA found that PM2.5 is significantly positively associated with breast cancer incidence (Parikh and Wei, 2016). A study in 15 cohorts from nine European countries found positive but statistically insignificant association between incidence of postmenopausal breast cancer and exposure to PM2.5 (Andersen et al., 2017).

In our study, after controlling for race/ethnicity, income, and other variables, PM2.5 still appears to be a significant variable positively associated with FBC IR at the county level (Table 2). An increase of the fine particulate matter by 1 μg/m3 implies 0.48 additional cases of FBC per 100,000 females. Among the 1,277 counties with female population of 10,000 or above for at least one race/ethnicity, average daily PM2.5 varied between 6.56 and 14.53 μg/m3. This may cause counties with the highest levels of PM2.5 to observe approximately 4 additional cases of FBC incidence per 100,000 females compared to counties with the lowest levels of PM2.5.

The biological mechanism of how PM2.5 influences breast cancer incidence still remains poorly understood, although a recent study may shed some light. According to a study of Canada’s Quebec regional cohort, air pollution (including PM2.5, NO2, SO2 and O3) interacts with the gene atad2 (Favé et al., 2018). Overexpression of atad2 identifies breast cancer patients with poor prognosis (Kalashnikova et al., 2010). Additionally, our research found that non-Hispanic White and Black populations may respond to PM2.5 concentration differently from the Asian or Hispanic population (Table 3). In the future, large-scale big data analyses of environmental exposure to air pollution and its interactions with genotypes will help cancer researchers to better understand breast cancer risks at the population scale.

6. Conclusions

In this county-level FBC IR analysis for multiple races/ethnicities, we integrated geographical factors (e.g., population, economic, physical environment) and behavioral data for 1,277 counties across the U.S. The dominant literature-documented patient- and neighborhood-level risk factors, such as race/ethnicity, income, and health accessibility, were found associated with FBC incidence disparities at the county scale. Higher household income, smaller uninsured population, and increasing number of primary care providers (PCP) at the county level are all associated with increasing FBC IR for a county. We also identified two new county-level factors contributing to the spatial disparity of FBC IR; and found that FBC IR is lower in counties with a higher proportion of married women and higher in counties with a higher level of air pollution.

Supplementary Material

1
2

Highlights:

  • Spatial disparities of female breast cancer incidence rates were assessed across the U.S.

  • A large number of geographical factors and behavioral variables were integrated

  • Multiple races/ethnicities were included

  • New risk factors were found besides income and health care accessibility

  • 10% increase in married-couple families lowers 5.2 FBC new cases per 100,000 females

  • Elevated PM2.5 may generate up to 4 more FBC new cases per 100,000 females

Acknowledgments

Jinfeng Zhang was supported by the National Institute of General Medical Sciences of the National Institute of Health under award number R01GM126558.

List of Abbreviations

API

Asian/Pacific Islander

FBC

female breast cancer

FMedAge

female median age

HLR

hierarchical linear regression

IR

incident rate

meanFS

mean family size

medIncome

median household income

NHW

Non-Hispanic White

PAHs

polycyclic aromatic hydrocarbons

PCP

primary care provider

PM2.5

particulate matters (PM) with a diameter of 2.5 micrometers or less

pDiaSc

percent diabetic screening

pFFH

percent of female-householder family households

pFHS

percent of females with high school diploma or higher education

pHWFH

percent of husband-wife family households

pHWFHWC

percent of husband-wife family with children households

pOb

percent obese

pUnem

percent unemployed

pUnins

percent of population without health insurance

pUrban

percent of urban population

SES

socioeconomic status

U.S.

United States

VIF

variance inflation factor

Footnotes

Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

References

  1. AKINYEMIJU TF, GENKINGER JM, FARHAT M, WILSON A, GARY-WEBB TL & TEHRANIFAR P 2015. Residential environment and breast cancer incidence and mortality: a systematic review and meta-analysis. Bmc Cancer, 15. [DOI] [PMC free article] [PubMed] [Google Scholar]
  2. AMERICAN CANCER SOCIETY 2018. Cancer Facts and Figures 2018. Atlanta, Ga: American Cancer Society. [Google Scholar]
  3. ANDERSEN ZJ, STAFOGGIA M, WEINMAYR G, PEDERSEN M, GALASSI C, JORGENSEN JT, OUDIN A, FORSBERG B, OLSSON D, OFTEDAL B, AASVANG GM, AAMODT G, PYKO A, PERSHAGEN G, KOREK M, DE FAIRE U, PEDERSEN NL, OSTENSON CG, FRATIGLIONI L, ERIKSEN KT, TJONNELAND A, PEETERS PH, BUENO-DE-MESQUITA B, PLUSQUIN M, KEY TJ, JAENSCH A, NAGEL G, LANG A, WANG M, TSAI MY, FOURNIER A, BOUTRON-RUAULT MC, BAGLIETTO L, GRIONI S, MARCON A, KROGH V, RICCERI F, SACERDOTE C, MIGLIORE E, TAMAYO-URIA I, AMIANO P, DORRONSORO M, VERMEULEN R, SOKHI R, KEUKEN M, DE HOOGH K, BEELEN R, VINEIS P, CESARONI G, BRUNEKREEF B, HOEK G & RAASCHOU-NIELSEN O 2017. Long-Term Exposure to Ambient Air Pollution and Incidence of Postmenopausal Breast Cancer in 15 European Cohorts within the ESCAPE Project. Environmental Health Perspectives, 125. [DOI] [PMC free article] [PubMed] [Google Scholar]
  4. BONNER MR, HAN DW, NIE L, ROGERSON P, VENA JE, MUTI P, TREVISAN M, EDGE SB & FREUDENHEIM JL 2005. Breast cancer risk and exposure in early life to polycyclic aromatic hydrocarbons using total suspended particulates as a proxy measure. Cancer Epidemiology Biomarkers & Prevention, 14, 53–60. [PubMed] [Google Scholar]
  5. BRODY JG, MOYSICH KB, HUMBLET O, ATTFIELD KR, BEEHLER GP & RUDEL RA 2007. Environmental pollutants and breast cancer - Epidemiologic studies. Cancer, 109, 2667–2711. [DOI] [PubMed] [Google Scholar]
  6. BROEDERS MJM, ALLGOOD P, DUFFY SW, HOFVIND S, NAGTEGAAL ID, PACI E, MOSS SM & BUCCHI L 2018. The impact of mammography screening programmes on incidence of advanced breast cancer in Europe: a literature review. BMC Cancer, 18, 860. [DOI] [PMC free article] [PubMed] [Google Scholar]
  7. BUCK KD 2016. Modelling of geographic cancer risk factor disparities in US counties. Applied Geography, 75, 28–35. [Google Scholar]
  8. DAVIDSON PL, BASTANI R, NAKAZONO TT & CARREON DC 2005. Role of community risk factors and resources on breast carcinoma stage at diagnosis. Cancer, 103, 922–930. [DOI] [PubMed] [Google Scholar]
  9. DEARDORFF J, EKWARU JP, KUSHI LH, ELLIS BJ, GREENSPAN LC, MIRABEDI A, LANDAVERDE EG & HIATT RA 2011. Father Absence, Body Mass Index, and Pubertal Timing in Girls: Differential Effects by Family Income and Ethnicity. Journal of Adolescent Health, 48, 441–447. [DOI] [PMC free article] [PubMed] [Google Scholar]
  10. ENGMANN NJ, GOLMAKANI MK, MIGLIORETTI DL, SPRAGUE BL, KERLIKOWSKE K & BREAST CANC SURVEILLANCE C 2017. Population-Attributable Risk Proportion of Clinical Risk Factors for Breast Cancer. Jama Oncology, 3, 1228–1236. [DOI] [PMC free article] [PubMed] [Google Scholar]
  11. FAVÉ M-J, LAMAZE FC, SOAVE D, HODGKINSON A, GAUVIN H, BRUAT V, GRENIER J-C, GBEHA E, SKEAD K, SMARGIASSI A, JOHNSON M, IDAGHDOUR Y & AWADALLA P 2018. Gene-by-environment interactions in urban populations modulate risk phenotypes. Nature Communications, 9, 827. [DOI] [PMC free article] [PubMed] [Google Scholar]
  12. FORMAN MR, WINN DM, COLLMAN GW, RIZZO J & BIRNBAUM LS 2015. Environmental exposures, breast development and cancer risk: Through the looking glass of breast cancer prevention. Reproductive Toxicology, 54, 6–10. [DOI] [PubMed] [Google Scholar]
  13. GANGNON RE, SPRAGUE BL, STOUT NK, ALAGOZ O, WEEDON-FEKJAER H, HOLFORD TR & TRENTHAM-DIETZ A 2015. The Contribution of Mammography Screening to Breast Cancer Incidence Trends in the United States: An Updated Age-Period-Cohort Model. Cancer Epidemiology Biomarkers & Prevention, 24, 905–912. [DOI] [PMC free article] [PubMed] [Google Scholar]
  14. GOLDBERGER AS 1991. A course in econometrics, Harvard University Press. [Google Scholar]
  15. GOMEZ SL, SHARIFF-MARCO S, DEROUEN M, KEEGAN THM, YEN IH, MUJAHID M, SATARIANO WA & GLASER SL 2015. The impact of neighborhood social and built environment factors across the cancer continuum: Current research, methodological considerations, and future directions. Cancer, 121, 2314–2330. [DOI] [PMC free article] [PubMed] [Google Scholar]
  16. GRAY J, EVANS N, TAYLOR B, RIZZO J & WALKER M 2009. State of the Evidence The Connection Between Breast Cancer and the Environment. International Journal of Occupational and Environmental Health, 15, 43–78. [DOI] [PubMed] [Google Scholar]
  17. HARDEFELDT PJ, EDIRIMANNE S & ESLICK GD 2012. Diabetes increases the risk of breast cancer: a meta-analysis. Endocrine-related cancer, 19, 793–803. [DOI] [PubMed] [Google Scholar]
  18. HIATT RA & BRODY JG 2018. Environmental Determinants of Breast Cancer. Annu Rev Public Health, 39, 113–133. [DOI] [PubMed] [Google Scholar]
  19. HIATT RA, PORCO TC, LIU FC, BALKE K, BALMAIN A, BARLOW J, BRAITHWAITE D, DIEZ-ROUX AV, KUSHI LH, MOASSER MM, WERB Z, WINDHAM GC & REHKOPF DH 2014. A Multilevel Model of Postmenopausal Breast Cancer Incidence. Cancer Epidemiology Biomarkers & Prevention, 23, 2078–2092. [DOI] [PubMed] [Google Scholar]
  20. HU L, CHUN YW & GRIFFITH DA 2020. Uncovering a positive and negative spatial autocorrelation mixture pattern: a spatial analysis of breast cancer incidences in Broward County, Florida, 2000–2010. Journal of Geographical Systems, 18. [Google Scholar]
  21. HUANG L, KULLDORFF M & GREGORIO D 2007. A spatial scan statistic for survival data. Biometrics, 63, 109–118. [DOI] [PubMed] [Google Scholar]
  22. JAGAI JS, MESSER LC, RAPPAZZO KM, GRAY CL, GRABICH SC & LOBDELL DT 2017. County-Level Cumulative Environmental Quality Associated With Cancer Incidence. Cancer, 123, 2901–2908. [DOI] [PMC free article] [PubMed] [Google Scholar]
  23. JOSLYN SA, FOOTE ML, NASSERI K, COUGHLIN SS & HOWE HL 2005. Racial and ethnic disparities in breast cancer rates by age: NAACCR Breast Cancer Project. Breast Cancer Research and Treatment, 92, 97–105. [DOI] [PubMed] [Google Scholar]
  24. JUNG I, KULLDORFF M & KLASSEN AC 2007. A spatial scan statistic for ordinal data. Stat Med, 26, 1594–607. [DOI] [PubMed] [Google Scholar]
  25. KALASHNIKOVA EV, REVENKO AS, GEMO AT, ANDREWS NP, TEPPER CG, ZOU JX, CARDIFF RD, BOROWSKY AD & CHEN H-W 2010. ANCCA/ATAD2 Overexpression Identifies Breast Cancer Patients with Poor Prognosis, Acting to Drive Proliferation and Survival of Triple-Negative Cells through Control of B-Myb and EZH2. Cancer Research, 70, 9402–9412. [DOI] [PMC free article] [PubMed] [Google Scholar]
  26. KEY J, HODGSON S, OMAR RZ, JENSEN TK, THOMPSON SG, BOOBIS AR, DAVIES DS & ELLIOTT P 2006. Meta-analysis of studies of alcohol and breast cancer with consideration of the methodological issues. Cancer Causes & Control, 17, 759–770. [DOI] [PubMed] [Google Scholar]
  27. KIRBY RS, DELMELLE E & EBERTH JM 2017. Advances in spatial epidemiology and geographic information systems. Annals of Epidemiology, 27, 1–9. [DOI] [PubMed] [Google Scholar]
  28. KLASSEN AC & SMITH KC 2011. The enduring and evolving relationship between social class and breast cancer burden: A review of the literature. Cancer Epidemiology, 35, 217–234. [DOI] [PubMed] [Google Scholar]
  29. KULLDORFF M 1997. A spatial scan statistic. Communications in Statistics-Theory and Methods, 26, 1481–1496. [DOI] [PMC free article] [PubMed] [Google Scholar]
  30. KULLDORFF M, SONG CH, GREGORIO D, SAMOCIUK H & DECHELLO L 2006. Cancer map patterns - Are they random or not? American Journal of Preventive Medicine, 30, S37–S49. [DOI] [PMC free article] [PubMed] [Google Scholar]
  31. LI Y, STEPPI A, ZHOU Y, MAO F, MILLER PC, HE MM, ZHAO T, SUN Q & ZHANG J 2017. Tumoral expression of drug and xenobiotic metabolizing enzymes in breast cancer patients of different ethnicities with implications to personalized medicine. Scientific Reports, 7, 4747. [DOI] [PMC free article] [PubMed] [Google Scholar]
  32. LILLBERG K, VERKASALO PK, KAPRIO J, TEPPO L, HELENIUS H & KOSKENVUO M 2003. Stressful life events and risk of breast cancer in 10,808 women: A cohort study. American Journal of Epidemiology, 157, 415–423. [DOI] [PubMed] [Google Scholar]
  33. LIU ZY, ZHANG K & DU XL 2016. Risks of developing breast and colorectal cancer in association with incomes and geographic locations in Texas: a retrospective cohort study. Bmc Cancer, 16. [DOI] [PMC free article] [PubMed] [Google Scholar]
  34. LYNCH SM & REBBECK TR 2013. Bridging the Gap between Biologic, Individual, and Macroenvironmental Factors in Cancer: A Multilevel Approach. Cancer Epidemiology Biomarkers & Prevention, 22, 485–495. [DOI] [PMC free article] [PubMed] [Google Scholar]
  35. MCLAFFERTY S, WANG FH, LUO L & BUTLER J 2011. Rural-urban inequalities in late-stage breast cancer: spatial and social dimensions of risk and access. Environment and Planning B-Planning & Design, 38, 726–740. [DOI] [PMC free article] [PubMed] [Google Scholar]
  36. MOBLEY LR, KUO T-M, URATO M, SUBRAMANIAN S, WATSON L & ANSELIN L 2012. Spatial Heterogeneity in Cancer Control Planning and Cancer Screening Behavior. Annals of the Association of American Geographers, 102, 1113–1124. [DOI] [PMC free article] [PubMed] [Google Scholar]
  37. MOSS JL, LIU BM & FEUER EJ 2017. Urban/Rural Differences in Breast and Cervical Cancer Incidence: The Mediating Roles of Socioeconomic Status and Provider Density. Womens Health Issues, 27, 683–691. [DOI] [PMC free article] [PubMed] [Google Scholar]
  38. NATIONAL CANCER INSTITUTE. 2017. State Cancer Profiles: Incidence Rates Table [Online]. Available: https://www.statecancerprofiles.cancer.gov/incidencerates/ [Accessed January 27 2017].
  39. ODDENS BJ & LEHERT P 1997. Determinants of contraceptive use among women of reproductive age in Great Britain and Germany .1. Demographic factors. Journal of Biosocial Science, 29, 415–435. [DOI] [PubMed] [Google Scholar]
  40. OSYPUK TL, KAWACHI I, SUBRAMANIAN SV & ACEVEDO-GARCIA D 2006. Are state patterns of smoking different for different racial/ethnic groups? An application of multilevel analysis. Public Health Reports, 121, 563–577. [DOI] [PMC free article] [PubMed] [Google Scholar]
  41. PARIKH PV & WEI YD 2016. PAHs and PM2.5 emissions and female breast cancer incidence in metro Atlanta and rural Georgia. International Journal of Environmental Health Research, 26, 458–466. [DOI] [PubMed] [Google Scholar]
  42. PARK B, L. HOPPER J, WIN A, G. DOWTY J, SUNG H, AHN C, KIM S-W, LEE M, LEE J, LEE J, KANG E, YU J-H, KIM K, MOON B-I, HAN W, NOH D-Y, K. PARK S & STUDY GROUP, K. 2017. Reproductive factors as risk modifiers of breast cancer in BRCA mutation carriers and high-risk non-carriers. Oncotarget, 8. [DOI] [PMC free article] [PubMed] [Google Scholar]
  43. POLITE BN, ADAMS-CAMPBELL LL, BRAWLEY OW, BICKELL N, CARETHERS JM, FLOWERS CR, FOTI M, GOMEZ SL, GRIGGS JJ, LATHAN CS, LI CI, LICHTENFELD JL, MCCASKILL-STEVENS W & PASKETT ED 2017. Charting the Future of Cancer Health Disparities Research: A Position Statement from the American Association for Cancer Research, the American Cancer Society, the American Society of Clinical Oncology, and the National Cancer Institute. Cancer Research, 77, 4548–4555. [DOI] [PubMed] [Google Scholar]
  44. REMINGTON PL, CATLIN BB & GENNUSO KP 2015. The County Health Rankings: rationale and methods. Population Health Metrics, 13, 11. [DOI] [PMC free article] [PubMed] [Google Scholar]
  45. ROBERT SA, STROMBOM I, TRENTHAM-DIETZ A, HAMPTON JM, MCELROY JA, NEWCOMB PA & REMINGTON PL 2004. Socioeconomic risk factors for breast cancer - Distinguishing individual- and community-level effects. Epidemiology, 15, 442–450. [DOI] [PubMed] [Google Scholar]
  46. ROSENBERG L, BOGGS DA, BETHEA TN, WISE LA, ADAMS-CAMPBELL LL & PALMER JR 2013. A prospective study of smoking and breast cancer risk among African-American women. Cancer Causes & Control, 24, 2207–2215. [DOI] [PMC free article] [PubMed] [Google Scholar]
  47. SCHOOTMAN M, GOMEZ SL, HENRY KA, PASKETT ED, ELLISON GL, OH A, TAPLIN SH, TATALOVICH Z & BERRIGAN DA 2017. Geospatial Approaches to Cancer Control and Population Sciences. Cancer Epidemiology Biomarkers & Prevention, 26, 472–475. [DOI] [PMC free article] [PubMed] [Google Scholar]
  48. SCOTT L, MOBLEY LR & IL’YASOVA D 2017. Geospatial Analysis of Inflammatory Breast Cancer and Associated Community Characteristics in the United States. International Journal of Environmental Research and Public Health, 14. [DOI] [PMC free article] [PubMed] [Google Scholar]
  49. SEILER A, MURDOCK KW, GARCINI LM, CHIRINOS DA, RAMIREZ J, JACKSON B & FAGUNDES CP 2017. Racial/Ethnic Disparities in Breast Cancer Incidence, Risk Factors, Health Care Utilization, and Outcomes in the USA. Current Breast Cancer Reports, 9, 91–99. [Google Scholar]
  50. SHI Y, STEPPI A, CAO Y, WANG J, HE MM, LI L & ZHANG J 2017. Integrative Comparison of mRNA Expression Patterns in Breast Cancers from Caucasian and Asian Americans with Implications for Precision Medicine. Cancer Res, 77, 423–433. [DOI] [PMC free article] [PubMed] [Google Scholar]
  51. SWIFT M & LUKIN JL 2008. Breast Cancer Incidence and the Effect of Cigarette Smoking in Heterozygous Carriers of Mutations in the Ataxia-Telangiectasia Gene. Cancer Epidemiology Biomarkers & Prevention, 17, 3188–3192. [DOI] [PubMed] [Google Scholar]
  52. TATALOVICH Z, ZHU L, ROLIN A, LEWIS DR, HARLAN LC & WINN DM 2015. Geographic disparities in late stage breast cancer incidence: results from eight states in the United States. International Journal of Health Geographics, 14. [DOI] [PMC free article] [PubMed] [Google Scholar]
  53. TIRONA MT, SEHGAL R & BALLESTER O 2010. Prevention of Breast Cancer (Part I): Epidemiology, Risk Factors, and Risk Assessment Tools. Cancer Investigation, 28, 743–750. [DOI] [PubMed] [Google Scholar]
  54. TORRES AZ, PHELAN-EMRICK D & CASTILLO-SALGADO C 2018. Evaluating Neighborhood Correlates and Geospatial Distribution of Breast, Cervical, and Colorectal Cancer Incidence. Frontiers in Oncology, 8. [DOI] [PMC free article] [PubMed] [Google Scholar]
  55. WAN N, ZHAN FB, ZOU B & CHOW E 2012. A relative spatial access assessment approach for analyzing potential spatial access to colorectal cancer services in Texas. Applied Geography, 32, 291–299. [Google Scholar]
  56. WANG F & ONEGA T 2015. Accessibility of cancer care: disparities, outcomes and mitigation. Annals of GIS, 21, 119–125. [Google Scholar]
  57. WANG FH, GUO DS & MCLAFFERTY S 2012. Constructing geographic areas for cancer data analysis: A case study on late-stage breast cancer risk in Illinois. Applied Geography, 35, 1–11. [DOI] [PMC free article] [PubMed] [Google Scholar]
  58. WILLIAMS SL, PARISI SM, HESS R & SCHWARZ EB 2012. Associations between recent contraceptive use and quality of life among women. Contraception, 85, 282–287. [DOI] [PMC free article] [PubMed] [Google Scholar]
  59. XU YQ, FU C, ONEGA T, SHI X & WANG FH 2017. Disparities in Geographic Accessibility of National Cancer Institute Cancer Centers in the United States. Journal of Medical Systems, 41. [DOI] [PMC free article] [PubMed] [Google Scholar]
  60. YOST K, PERKINS C, COHEN R, MORRIS C & WRIGHT W 2001. Socioeconomic status and breast cancer incidence in California for different race/ethnic groups. Cancer Causes & Control, 12, 703–711. [DOI] [PubMed] [Google Scholar]
  61. ZAHND WE & MCLAFFERTY SL 2017. Contextual effects and cancer outcomes in the United States: a systematic review of characteristics in multilevel analyses. Annals of Epidemiology, 27, 739–748. [DOI] [PubMed] [Google Scholar]
  62. ZHU H, LEI X, FENG J & WANG Y 2012. Oral contraceptive use and risk of breast cancer: A meta-analysis of prospective cohort studies. European Journal of Contraception and Reproductive Health Care, 17, 402–414. [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

1
NIHMS1631247-supplement-1.xlsx (14.6KB, xlsx)
2
NIHMS1631247-supplement-2.xlsx (15KB, xlsx)

RESOURCES