How do Geographical/Regional Factors Influence Breast Cancer Survival and Incidence Among US Women

Background

Breast Cancer is the second leading cancer among United States (US) women creating a burden of disease that demands research into etiology that can inform prevention and control (NIH National Cancer Institute, 2018). Large, population-based studies surveilling mortality and incidence data are able to identify trends and risk factors that exist among breast cancer cases in the US, and current studies have been providing greater resolution into a variety of these independent variables influencing breast cancer outcomes which makes them invaluable sources of information. While many of these studies have revealed genetic factors that are known risk factors for breast cancer, genetics often do not account for all of the observed variance in breast cancer incidence and survival. This role of geography within the United States is especially evident given the incredible variation in the rate of new Breast Cancer cases observed in each state (CDC, 2015). In fact, research has identified many sociodemographic, environmental, and health access related risks which point to the importance of place in the story of breast cancer etiology (REFERENCES). Given that many of these place related risks, such as health access, mammographic screening, environment, etc. can be modified, understanding what factors are indicative of lower survival or increased incidence is a unique opportunity to paint a complete picture of breast cancer etiology and potentially identify opportunities for targeted evidenced-based interventions. For these reasons, considering geography and its relation to breast cancer outcomes is salient. This study seeks to summarize the existing literature on geography and its relation to breast cancer and what mediates that relationship by addressing the following question: Among US women, how do geographical/regional factors influence breast cancer survival and incidence?

Methods

A thorough search of the literature was performed in PubMed, search parameters were set only to include articles published within the last 5 years. Articles were included if they were: 1. Based on United States populations, 2. Directly related to geography/place and factors related to area of living, and 3. Based on incidence and mortality data. Articles were excluded if they were: 1. Based on a population in another country, 2. Were not clearly linked to geography, 3. Were exclusively studies looking at insurance and uptake (particularly among Medicare populations), and 3. Were Interventional studies as the focus of this paper is on epidemiological cross-sectional studies of population level data. The first search combined the terms “geographic*”, “breast cancer”, and “survival.” This search yielded 183 articles, of which 4 met the inclusion criteria. The next search combined the terms “geographic*”, “breast cancer”, “mortality” which only yielded 82 articles, this provided 1 additional article that had not already been identified. The next search combined terms the following geographical terms one at a time “regional”, “county”, and “state” with “breast cancer”, “mortality”, “survival” and “united states” yielding 145 articles, 3 additional articles were added to the study.

Multi-State Studies

States: Atlanta, GA; Connecticut; Detroit, Michigan; Hawaii; Iowa; New Mexico; San Francisco-Oakland, California; Seattle, Washington; Utah; Los Angeles, California; San Jose-Monterey, California; Rural Georgia; Greater California; Kentucky; New Jersey.

The study performed by Akinyemiju, Moore, Ojesina, Waterbor, and Altekruse (2016) “Racial disparities in individual breast cancer outcomes by hormone-receptor subtype, area-level socio-economic status and healthcare resources” explores the effect of race/ethnicity, healthcare resources, socioeconomic status, and hormone-receptor subtype on breast cancer survival. Each broad category was an aggregate of proxy variables listed in Table 1. These covariates were used to perform survival analysis, for consecutive multilevel regression modeling, and to calculate odds ratios and hazards ratios for the following outcomes: 1. Stage at Diagnosis; 2. Surgical Treatment; 3. Radiation Treatment; and 4. Breast Cancer Survival. The study used data from all Surveillance, Epidemiology, and End Results database (SEER) reporting registries. Non-Hispanic Black (NH-Black) and Hispanic women tended to live in areas with lower SES as measured by greater proportion unemployed and under the federal poverty level and also were less likely to live in rural areas. Interestingly, NH-Black women lived in areas with greater healthcare access, on average. Despite this, racial disparities were clear and apparent—NH-Black Women had 42% higher hazards of breast cancer mortality and both NH-Black and Hispanic women were more likely to have late-stage diagnosis. While the study found that Hormone-Receptor subtype explained the greatest amount of the variance in late stage diagnosis, survival, and treatment, even when this was controlled for geographical, socio-demographic, and socio-economic covariates did have a significant influence.

The strength of this study was its thorough and rigorous models that allowed for the adjustment for multiple geographical, socioeconomic, and biological covariates in order to quantify the contribution of each covariate to the observed incidence and survival. However, the results found for healthcare access lacked clear rationale given that NH-Black women had the greatest access yet the worst outcomes. This is likely because healthcare access was defined by proxies that do not explain utilization. Additionally, county-level socio-economic covariates cannot be directly linked to the individual cancer cases. SEER also lacked data on Her2 status.

States: Baltimore, MD; Chicago, IL; Dallas, TX; Detroit, MI; Houston, TX; Los Angeles, CA; Memphis, TN; New York City, NY; Philadelphia, PA; Washington D.C.

Sighoko, Hunt, Irizarry, Watson, Ansell, & Murphy (2018) “Disparity in breast cancer mortality by age and geography in 10 racially diverse US cities” explored the disparity in breast cancer mortality utilizing age-stratified Non-Hispanic Black to Non-Hispanic White rate ratio (RR) and Mortality Risk Differences (RD). In this descriptive analysis, the authors used National Center for Health Statistics mortality data from 1999-2013. The study found an interesting distribution of disparity, revealing that though the lowest mortality burden was among the younger age groups (under 40 and 40-49), the highest disparity exists in this age group while among the older age group, specifically the 65+ age group, had the lowest disparity and the highest burden of breast cancer mortality. Between cities, the same pattern of disparity was sustained however the magnitude of disparity differed. For instance, cities in the Eastern US tended to have lower disparity in breast cancer mortality. While this study is a useful source of descriptive evidence, it only utilizes descriptive epidemiological methods to observe the Non-Hispanic Black breast cancer disparities as compared to Non-Hispanic White in large metropolitan US cities with the top 10 largest black populations. While this study reveals the disparities that exist and how they differ from city to city, the study does not use methods or control for variables in order to describe why the disparities exist. In order to understand what drives that disparity, there is a need for more in-depth analysis to see what mediates the variation seen by geographic location between breast cancer mortality and race.

States: California, Georgia, Iowa, Louisiana, Kentucky, New Jersey, New Mexico, and Utah

Tatalovich, Zhu, Rolin, Lewis, Harlan, & Winn (2015) “Geographic disparities in late stage breast cancer incidence: results from eight states in the United States” used descriptive analysis, ANOVA with Bonferroni correction, backward stepwise linear regression, and geospatial modeling to study geographical variation in age-adjusted late-stage breast cancer incidence based on SEER data from 2006-2010. The covariates studied included proxy variables which illustrated socio-demographic and economic characteristics, accessibility to health care, and availability of screening services in the given “Health Service Areas” (HSAs) defined by the National Center for Health Statistics and modified by the National Cancer Institute.

New Jersey had the highest incidence of late stage breast cancer diagnosis while New Mexico had the lowest. Analysis of variance revealed statistically significant differences in the mean incidence rates of late stage breast cancer diagnosis between states, however further study with Bonferroni correction revealed that New Mexico had a significantly lower rate than NJ, GA, KY, and CA which explained that variance. Interestingly, the proportion of high, medium, and low incidence Health Service Areas varied dramatically. New Jersey had a staggering 80% of its HSAs in the “high” incidence category while New Mexico had 80% of its HSAs in the “low” incidence category. The other 6 states had varied proportions falling between these two extremes. The regression analysis revealed that of all the covariates tested, four in particular had significant relationships with late stage incidence. the number of mammography facilities per person, the percent of the population with bachelor’s degree or greater, and percent with English literacy were associated with lower incidence of late stage diagnosis. The percentage of Black population in a given area was associated with greater incidence of late stage diagnosis. The study effectively illustrated inter and intrastate differences in late stage breast cancer incidence. The geospatial mapping was particularly useful for visually representing the overlap between the independent geographic variables and the health service areas with high incidence. However, the conclusion that there is a significant relationship between college education and late stage breast cancer incidence is weak given that the p-value was 0.010. Additionally, the study acknowledges the difficulty in quantifying and representing these geographic independent covariates.

Single State Studies

State: Wisconsin, Southeastern

Beyer, Zhou, Matthews, Hoormann, Bemanian, Laud, & Nattinger (2016) “Breast and Colorectal Cancer Survival Disparities in Southeastern Wisconsin” focused on Southeastern Wisconsin counties (Milwaukee, Jefferson, Kenosha, Ozaukee, Racine, Walworth, Washington, & Waukesha) to better understand the distribution of survival disparities. The study utilized Cox Proportional Hazards Model, Kaplan-Meier analysis, and Adaptive Spatial Filtering (mapping) in order to study cause-specific breast cancer mortality, all-cause breast cancer mortality. The study found significant survival disparities for race and ethnicity. Specifically, both Hispanic/Latino and Black/African American women had significantly poorer survival for both all cause and cause specific breast cancer than white women. Survival was also poorer for those with Late-Stage diagnosis and older age. Geography was analyzed using Adaptive Spatial Filtering which is essentially a univariate analysis of 5-year survival. These results found that the city of Milwaukee and several rural areas had lower survival rates.

The methods of this study are intriguing; however, the spatial analysis does not allow for a thorough understanding of what factors are contributing to the observed survival disparities. This modeling method does not allow for the adjustment of covariates; thus, it is hard to tell even with the provided maps representing race/ethnicity and poverty what is truly driving the lower regional survivals. The survival analysis on race/ethnicity, late-stage diagnosis, marital status, and age is unique only because of the data set. Currently, more thorough analyses with greater resolution that observes intra-ethnic diversity exists. The state also has a relatively low number of minority populations represented in this registry. If anything, this study reveals a greater need for a thorough survival analysis with a model that allows for the adjustment of covariates.

State: Nevada

Callahan, Pinheiro, Cvijetic, Kelly, Ponce, & Kovetz (2017) “Worse Breast Cancer Outcomes for Southern Nevadans, Filipina, and Black Women” observed Nevada by three regions: Northwestern, Southern, and Rural to analyze cause-specific breast cancer mortality and stage-specific survival and how it varied by region. The study utilized 5-year adjusted survival analysis, log rank test, and Cox proportional hazards regression modelling to describe the observed differences in survival between regions with data obtained from the Nevada Central Cancer Registry (NCCR) for the years 2003-2010. Three models were created, the first adjusted only for age; the second adjusted for age, race/ethnicity, insurance status, SES, and NV region; the third adjusted for age, race/ethnicity, insurance status, SES, NV Region, Stage, Estrogen-Receptor Status, and Grade of Tumor.

Nevada as a state had a significantly lower survival rate (84.4%) than the US as a whole (89.2%). 68% of breast cancer mortality cases were localized in Southern NV. Survival was lowest in Southern and Rural Nevada regions. This observed elevation in risk of death (16%) observed in the Southern region remained even after adjusting for demographic, social, and pathological covariates. In Nevada, Black and Filipina women had higher hazards of cause specific mortality than white women. The pathological factor, stage at diagnosis, was the biggest factor for cause specific mortality. This study utilizes three Cox proportional hazards regression models which include many covariates to rigorously study what impacts breast cancer survival outcomes. Modelling that controlled for all demographic, social, and pathological factors specifically revealed the role of region in breast cancer survival outcomes. Despite variance being most significantly described by stage at diagnosis, the disparity in the Southern Region of Nevada remained significant demonstrating the significant and important role of geography in breast cancer survival. This study did not include any covariates that served as a proxy for healthcare access. There may have been geographic disparities in access, thus it may describe the regional differences in survival that exist.

State: Louisiana

Carrol, Lawson, Jackson, & Zhao (2017) “Assessment of spatial variation in breast cancer-specific mortality using Louisiana SEER data” studied cause-specific breast cancer mortality by Louisiana parish using SEER data from 2000-2013. The study used an accelerated failure time model with spatial frailty estimates, a complex model enabling a high-resolution analysis of Louisiana by parish. The study considered a multitude of covariates ranging from socio-demographic composition of parishes to environmental, industrial, and proximity to key geographical features of parishes. Overall, it was found that parishes with shorter survival time were lower income and positioned alongside either Red or Mississippi Rivers. There was heterogeneity between parishes, the best survival was in Orleans parish and the worst survival was in Terrebonne—those in Orleans Parish survived 1.5 times longer. Additional factors contributing to low survival in parishes included access and quality of care, food availability (fresh vs food desert), socioeconomic status, percent urban, percent farmland, and percent fishing mining, forestry, and agriculture. The study also included emissions as potential environmental risk factors and found that agriculture associated emissions such as ammonia and particulate matter were associated with shorter survival in parishes.

The study paints a complete picture of "place" and its role in breast cancer survival. The model utilized is complicated but its scope and the number of covariates assessed provides a detailed evaluation. The study took a holistic approach in determining its risk factors and covered sociodemographic, socioeconomic, environmental, and occupational variables. The study reveals interesting and modifiable characteristics of areas that can be targeted and also environmental and occupational exposures that could be mitigated.

State: Texas

Pruitt, Lee, Tiro, Xuan, Ruiz, & Inrig (2015) “Residential racial segregation and mortality among black, white, and Hispanic urban breast cancer patients in Texas, 1995 to 2009” sought to determine the role of segregation, as calculated by an LQ distribution that compares non-Hispanic Blacks to non-Hispanic whites in a given area, on cause-specific breast cancer mortality and all-cause breast cancer mortality. The data was derived from the Texas Cancer Registry for the years 1995-2009. The study utilized descriptive analyses, Chi-Square, ANOVA, Spearman Correlation Coefficients, and Cox Proportional Hazards to determine the significance of the following covariates with breast cancer mortality: 1. Age, 2. Summary Stage, 3. Diagnosis Year, 4. Tumor Grade, 5. Histology, 6. Neighborhood poverty (% of redisdents living in poverty by census tract), 7. # of Mammography Machines per 10,000 women aged 50 or older in county of residence, 8. Segregation: LQ distribution comparing NH Blacks to NH Whites and Hispanics to NH Whites.

The study found that Non-Hispanic Blacks live in more segregated neighborhoods of Texas and Hispanics do as well, but to a lower extent. This study is unique amongst the other studies because it looks at neighborhood composition in the context of segregation for NH Blacks and Hispanics. The findings reveal the consistent finding of racial disparities in breast cancer survival. While there was weak univariate evidence to suggest that segregation is associated with poorer survival, segregation does not explain the disparities in survival for NH blacks. It is correlated; however, it does not explain it. The covariates included in the model allowed for appropriate adjustments to determine the relationships underlying the neighborhood differences between metropolitan areas. It is possible, as stated by the authors, that one of the other covariates is the mediator of segregation, such as poverty, and thus controlling for that covariate removed the relationship between segregation and breast cancer mortality.

Cite this page