There is now widespread interest in the role of neighborhoods in cancer and cancer disparities (Baker, Hoel, Mohr, Lipsitz, & Lackland, 2000; Krieger et al., 2002; Singh, Miller, Hankey, & Edwards, 2003). This new focus on the geography of cancer and its risk factors reflects mounting evidence that both are unequally distributed across neighborhoods that differ in poverty, ethnicity, and rurality (Campbell et al., 2002; Coughlin, Leadbetter, Richards, & Sabatino, 2008; Coughlin et al., 2006; Hsu, Jacobson, & Soto Mas, 2004; Onega et al., 2008). Neighborhoods contribute to disparities in cancer and its risk factors and, hence, understanding the nature of that contribution can highlight new community-level approaches to reducing cancer disparities (Kawachi & Berkman, 2003).
This topic provides a brief overview of the geography of cancer and its risk factors. Due to space limitations, we focus only on Blacks and Whites, and exclude data on other ethnic-minority groups. Likewise, we focus on the role of geographic variables in only three of many cancer risk factors (i.e., smoking, obesity, cancer screening), to devote more attention to the role of such variables in cancer incidence, stage at diagnosis, treatment, mortality, and survival. The geographic (area-level) variables of focus are neighborhood socioeconomic status (SES), residential segregation, and rurality, the three most often examined in research. First, we define these area variables and strategies for measuring them. Next, we examine their association with the three cancer risk factors. This is followed by a summary of their associations with cancer incidence, treatment, stage at diagnosis, mortality, and survival. Finally, we discuss the possible mechanisms of area-level influences.
AREA-LEVEL VARIABLES
Neighborhood-SES
Neighborhood-SES refers to the level of poverty, economic deprivation, or affluence of an area, and can be measured at a variety of area levels, including state, county, zip code, and census tract (Krieger et al., 2002; Krieger, Chen, Waterman, Rehkpof, & Subramanian, 2003, 2005). In general, the smaller the area in which area-SES is assessed, the stronger is the association between area-SES and health. This is because large areas (e.g., states, zip codes) contain the entire range of smaller SES-areas within them—that is, both poor and affluent neighborhoods (Krieger et al., 2002, 2003, 2005). Consequently, significant area-SES differences are more often found within large areas than between them. Hence, measuring area-SES in small areas (i.e., census tract) is widely regarded as superior to measurement at larger area levels, with census-tract (CT) level measurement most predictive of health disparities (Krieger et al., 2005). This is because, unlike zip codes (mean N = 30,000) and Metropolitan Statistical Areas (MSA) such as Los Angeles-Long Beach, CA (N = several million), CTs are small (mean N = 4,000) areas that are largely homogenous in ethnicity, living conditions (e.g., amenities), life circumstances (e.g., owning vs. renting), and area-SES (e.g., property values).
Area-SES can be measured in many ways, including median price of homes, percentage of residents below the poverty line (% BPL), median household income, and the Townsend (deprivation) Index (Krieger et al., 2003, 2005). Clearly, some of these measures are simply aggregates of the individual-level SES of the area’s residents (e.g., median household income), whereas others are more contextual (e.g., median price of homes, average rent paid). Hence, the correlation between individual- and area-level SES can be large when aggregate measures are used. Irrespective of whether aggregate or contextual measures are used, however, the correlation between individual- and area-SES is larger for Whites than for Blacks, because Blacks—irrespective of individual-level SES (e.g., household income)—tend to live in poorer neighborhoods than Whites of matched individual-level SES (Krieger et al., 2005; Krieger, Williams, & Moss, 1997). Nonetheless, as will be shown, individual- and area-level SES have independent associations with cancer and cancer risk factors.
Segregation
Residential segregation refers to the geographic separation of ethnic minorities (Blacks in this topic) from Whites in residential areas (Iceland, Weinberg, & Steinmetz, 2002; Johnston, Poulsen, & Forrest, 2007). Segregation can be measured at any area level, with measurement at smaller (e.g., CT) levels preferred. Again, this is because large areas (e.g., states, counties) contain both highly segregated and integrated smaller areas within them, such that segregation differences are larger within than between large areas. Hence, like area-SES, measuring segregation at the CT level—or at the MSA level, with MSA-segregation calculated from CT data—is psychometrically superior and preferred (Iceland et al., 2002; Johnston et al., 2007).
Segregation can be measured in many ways. These include dissimilarity (the uneven distribution of Blacks and Whites in a residential area), Isolation (the probability that Blacks will encounter only other Blacks in their residential area), concentration (the population density of Black neighborhoods), clustering (the extent to which Black neighborhoods are surrounded by other Black neighborhoods), centralization (the degree to which Black neighborhoods are in a city’s urban center vs. the suburbs), and hypersegregation, the simultaneous occurrence of all of these (Massey & Denton, 1988; Massey, White, & Phua, 1996; Wilkes & Iceland, 2004). Of these well-accepted measures, the Isolation Index exhibits greater psychometric integrity (i.e., validity, interpretability) than the others (Acevedo-Garcia, Lochner, Osypuk, & Subramanian, 2003; Chang, Hillier, & Mehta, 2009). Crude measures with questionable validity (e.g., percentage of Blacks in an area) often are used in research as well (Kramer & Hogue, 2009).
The correlation between segregation and area-SES is small (i.e., r =.10-.20), that is, there are both affluent and poor, mostly Black, neighborhoods. Likewise, the correlation between individual-SES and segregation is small; this is because, as a result of housing discrimination, most Blacks (65%) live in mostly Black (segregated) neighborhoods irrespective of their individual-SES and their preference to live in integrated areas.
Rurality
The definition of rural versus urban areas is constantly changing (Probst, Moore, Glover, & Samuels, 2004), but generally refers to area population density. As of 2000, the U.S. Census Bureau defines urban areas as those with populations of 250,000 to >1 million, rural areas as those with populations of <2,500, and suburban areas as those adjacent to but outside of urban areas, with residents commuting to the urban area for work (http://www.census.gov).
AREA-LEVEL VARIABLES AND CANCER RISK FACTORS
Smoking
Area-SES
Numerous studies have examined the role of neighborhood-SES in cigarette smoking. A few of these are summarized in Table 6.1, where all studies controlled for individual-level SES. As shown, as area-SES decreases, smoking prevalence generally increases among adult women and men alike, even when controlling for age, education, and income. For example, in their study of 41,726 Black women, Datta, Subramanian et al. (2006), found an overall smoking prevalence rate of 16.2%. Smoking prevalence was nearly twice as high in high-poverty CTs (defined as >20% BPL) than in low-poverty CTs (<5% BPL), after controlling for individual-level demographic factors. Likewise, a study of the 39,695 (multiethnic) participants in the Third National Health and Nutrition Examination Survey (NHANES-III) found that the percentage of smokers increased significantly with increasing neighborhood deprivation, the latter defined as the percentage of households without telephones or plumbing, median rent paid, and other variables; this effect held even after controlling for gender, age, marital status, race/ethnicity, and education and income (Stimpson, Ju, Raji, & Eschback, 2007).
Rurality and Segregation
Studies consistently have found a higher prevalence of smoking in rural than in urban areas (among men in particular), with cigarette smoking often combined with use of smokeless tobacco among rural men (e.g., Bell et al., 2009; Doescher, Jackson, Jerant, & Hart, 2006; Nelson et al., 2006). Studies of residential segregation and smoking among African Americans have yielded less consistent results (Bell, Zimmerman, Mayer, Almgren, & Huebner, 2007; Datta, Subramanian et al., 2006; Dell, Whitman, Shah, Silva, & Ansell, 2005; Northridge et al., 1998). One study found no relationship (Datta, Subramanian et al., 2006), two found smoking rates higher among Blacks in segregated-Black than in integrated areas (Dell et al., 2005; Northridge et al., 1998), and one found a U-shaped relationship between segregation and smoking during pregnancy among Black women (Bell et al., 2007). These inconsistencies may reflect sample differences in gender (women only vs. women and men), as well as differences in the measure of segregation used, as noted below.
Obesity
Area-SES
Numerous studies have found strong relationships between area-SES and obesity, that is, body mass index (BMI) ^ 30 (e.g., Black & Macinko, 2008; Boardman, Saint Onge, Rogers, & Denney, 2005; Drewnowski, Rehm, & Solet,2007; Glass, Rasmussen, & Schwartz, 2006; Robert & Reither, 2004; Stimpson et al., 2007; Wang, Soowan, Gonzalez, MacLeod, & Winkleby, 2007). A few of these are summarized in Table 6.1, where all studies controlled for individual-level SES. As shown, as area deprivation or poverty increases, BMI increases among women and men, even when controlling for individual-level SES and other variables.
TABLE 6.1 Selected Studies on the Role of Area Variables in Cancer Risk Factors
|
Cigarette Smoking |
|
|
Author |
Participants |
Area-SES Measures |
Low Area-SES Associated With |
Datta et al. (2006) |
N = 41,726 U.S. Black women |
Census tract (CT) residents, % BPL |
Increased smoking prevalence |
Diez-Roux et al. (1997) |
N = 12,601 U.S. adults |
Census block (CB) median housing price, education, occupation |
Increased smoking prevalence |
Diez-Roux et al. (2003) |
N = 3,472 U.S. women and men. Blacks and Whites, ages 18-30 years |
CT and CB income, education, and occupation |
Increased smoking prevalence among Whites only |
Ross (2000) |
N = 2,482 Illinois adults, 59% women, 84% White |
CT poverty and education |
Increased smoking prevalence among men only |
Tseng et al. (2001) |
N = 648 North Carolina women, 42% Black |
CB poverty, unemployment, home and car ownership |
Increased smoking prevalence |
Obesity/BMI |
|||
Author |
Participants |
Area-level SES Measures |
Low Area-SES Associated With |
Drewnowski et al. (2007) |
N = 8,803 adults from 74 zip codes |
Zip code % BPL, median value of homes, etc. |
Increased prevalence of obesity |
Stimpson et al. (2007) |
N = 39,695 adults in the NHANES III |
CT % BPL, percentage without phones and other measures |
Increased prevalence of obesity |
Wang et al. (2007) |
N = 7,595 California adults |
CT and CB median housing values, percentage unemployed and other indices |
Higher BMI and higher rates of obesity |
TABLE 6.1 (continued)
Author |
Participants |
Segregation Measure |
High Segregation Associated With |
Boardman et al. (2005) |
National sample of N = 402,154 adults |
CB proportion of Blacks |
Increased obesity prevalence |
Chang (2006) |
National sample of N = 46,881 adults |
Metropolitan statistical area (MSA) Black Isolation averaged from CT Isolation |
Increased prevalence of overweight and obesity |
Chang et al. (2009) |
N = 6,608 Philadelphia women |
CT-level Black Isolation Index |
Increased BMI and obesity prevalence |
Mobley et al. (2006) |
N = 2,692 women |
Zip code and county-level Black Isolation Index |
No significant association |
Robert and Reither (2004) |
N = 3,617 adults |
CT-level percent Blacks |
No significant association |
Cancer Screening |
|||
Author |
Participants |
Area-SES Measure |
Low Area-SES Associated With |
Jackson et al. (2009) |
Breast; N = 33,938 California women, ages 40-84 years |
CT-level income |
Lower rates of breast cancer screening |
Harper et al. (2009) |
Breast; women in SEER and NHIS 1987-2005 |
County-level % BPL |
Lower rates of breast cancer screening |
Coughlin et al. (2006) |
Cervical; N = 49,231 women in 2000 and 2002 BRFSS |
County income and education |
Lower rates of PAP testing within the last 3 years |
Author |
Participants |
Area-SES Measure |
Low Area-SES Associated With |
Pruitt et al. (2009) |
Breast, cervical, and colorectal; review of 19 studies |
Varied from CT to MSAs and county level, and from area-income/education to % BPL |
Lower rates of all three types of cancer screening, largely irrespective of area-level and area-SES measures |
Schootman et al. (2006) |
Breast, cervical, and colorectal; N = 118,000 adults in 98 MSAs and 740 counties |
MSA % BPL, calculated from CT data |
Lower rates of all three types of screening, and higher rates of never-screening |
Author |
Participants |
Segregation Measure |
High Segregation Associated With |
Dai (2010) |
Breast; 12,413 women with diagnosed breast cancer |
Zip code Black Isolation Index |
Increased diagnosis of late-(vs. early-) stage breast cancer |
Mobley et al. (2008) |
Breast; N = 224,585 women with breast cancer, from 11 states, SEER data |
Zip code Black Isolation Index |
Lower and higher screening depending on state |
For example, a study of a random sample of 1,140 U.S. adults found that residents of hazardous neighborhoods were nearly 2 times more likely than residents of nonhazardous neighborhoods to be obese (53% vs. 27%, respectively). Neighborhood hazard was defined by the Townsend Index (housing quality, unemployment, crowding, etc.), as well as by the number of liquor stores and violent crimes (Glass et al., 2006). Likewise, in their study of the 39,695 participants in the NHANES III, Stimpson et al. (2007) found that BMI increased with increasing neighborhood deprivation, even when controlling for age, gender, income, physical activity levels, alcohol use, smoking, race/ethnicity, and education. On the whole, such findings are stronger for women than for men, suggesting that gender may mediate the relationships between neighborhood-SES and BMI (Black & Macinko, 2008; Do et al., 2007).
Rurality and Segregation
Only a few studies have examined the role of residential segregation in obesity among Black adults; these are summarized in Table 6.1 (i.e., Boardman et al., 2005; Chang, 2006; Chang et al., 2009; Mobley et al., 2006; Robert & Reither, 2004). As shown, results are mixed, with three studies finding positive relationships and two finding none. For example, Chang (2006) found a strong association between segregation and obesity among Black (but not White) adults, in which each 1 standard-deviation increase in segregation (Black Isolation) was associated with a 0.423 increase in Black BMI, and a 14% increase in Blacks’ odds of being overweight—even when controlling for individual-level SES, physical activity, age, gender, and diet. Boardman et al. (2005) found nearly identical results. Alternatively, Mobley et al. (2006) and Robert and Reither (2004) found no effect. These inconsistencies probably reflect differences in the area-level and segregation measures used. Indeed, the inconsistent findings for segregation and smoking, and for segregation and obesity, may both reflect the null results obtained when using either a crude measure of segregation (percent Black) or a very large (zip code, county) area (Kramer & Hogue, 2009).
Unlike the inconsistent results for segregation and BMI, studies consistently have found that the prevalence of overweight (BMI ^ 25) and obesity (BMI ^ 30) are higher in urban than in rural areas among women and men, children, and adolescents (Bodor, Rice, Farley, Swalm, & Rose, 2010; Drewnowski et al., 2007; Dunton, Kaplan, Wolch, Jerrett, & Reynolds, 2009; Ford & Mokdad, 2008; Mascie-Taylor & Goto, 2007).
Cancer Screening
Area-SES
Numerous studies have found a strong association between area-SES and cancer screening. As shown by the examples in Table 6.1, the majority found that cancer screening increases with increasing area-SES (e.g., Coughlin, King, Richards, & Ekwueme, 2006; Harper et al., 2009; Jackson et al., 2009; Pruitt, Shim, Mullen, Vernon, & Amick, 2009; Schootman, Jeffe, Baker, & Walker, 2006). The few negative findings are likely to reflect differences in the area level used, with null findings generally emerging in studies that used large area levels.
Rurality and Segregation
One of the most clearly established geographic relationships is the strong association between rural residence and lower cancer screening prevalence. Consistently across a variety of types of studies and types of cancer screening, researchers have found that those who reside in rural areas are significantly less likely to have been screened for cancers, and likewise are more likely to be diagnosed at a later cancer stage than nonrural residents (Coughlin, Leadbetter, Richards, & Sabatino, 2008; Coughlin, Uhler, Bobo, & Caplan, 2004; Huang, Dignan, Han, & Johnson, 2009; Jackson et al., 2009).
Only a handful of studies have examined the association between cancer screening and residential segregation, with contradictory findings obtained (Dai, 2010; DeChello, Gregorio, & Samociuk, 2006; Haas et al., 2008; Mobley, Kuo, Driscoll, Clayton, & Ansell, 2008). Moreover, many studies did not assess cancer screening but, instead, examined cancer stage at diagnosis, with late versus early stage interpreted as a proxy for low versus high screening, respectively. For example, Dai (2010) found that as segregation (Black Isolation at the zip code level) increased, the likelihood of being diagnosed with later-stage cancers increased, this being interpreted as implying a lack of early detection. Similarly, Mobley et al. (2008) found that as Black segregation increased, the probability of cancer screening decreased in some states; in other states, increasing Black segregation (Isolation) was associated with increased screening. Such findings suggest that segregation plays a role in cancer screening, but that role varies with the type of screening (e.g., breast vs. cervical) and the region of the United States (Mobley et al., 2008).