Abstract
Objectives. We sought to determine whether social class modifies the effect of BMI on breast cancer incidence.
Methods. Participants included 5642 postmenopausal White women recruited in 1989 to CLUE II, a prospective cohort study in Washington County, Maryland. We obtained exposure data from CLUE II and the 1990 US Census. We used survival and random-effects Cox proportional hazards analyses to determine the association of social class and BMI with breast cancer incidence.
Results. Education was independently associated with increased risk of breast cancer incidence (hazard ratio [HR] = 1.06; 95% confidence interval [CI] = 1.01, 1.11; P < .05); contextual measures of social class were not. Education modified the effect of BMI at age 21 years (HR = 0.98; 95% CI = 0.97, 0.99); area-level social class modified the effect of BMI at baseline (HR = 0.97; 95% CI = 0.94, 0.99) and BMI change (HR = 0.98; 95% CI = 0.95, 1.00). Subpopulation analyses that were adjusted for hormone use, parity, and breast-feeding found similar effects.
Conclusions. Social class moderates the influence of body size on breast cancer incidence. Public health efforts, therefore, should advocate for policies that improve social conditions to decrease the burden of breast cancer.
Higher social class and heavier body size are established independent risk factors for postmenopausal breast cancer. Their interaction, however, has not been well studied in relation to breast cancer incidence. Breast cancer incidence is higher among women of higher social class, in part because of class-based differences in reproductive behavior.1–3 Women of higher social class are more likely to be older at first birth, have fewer total births, and to be older at menopause (all risk factors for breast cancer1–3). Heavier body size, measured by body mass index (BMI), and weight gain increase postmenopausal breast cancer risk because high levels of adiposity result in the production of excess levels of circulating endogenous estrogens, which in turn promote mammary carcinogenesis.4,5 Because social class in general has been found to be inversely associated with body size, there is an expectation that this association would lead to higher rates of breast cancer incidence among women of low social class. No previous studies have examined the dual effects of social class and body size on breast cancer incidence. Furthermore, only a few studies4,6–8 have examined both static measures and measures of change over time in body size.
To date, breast cancer research has not fully explored both individual and contextual indicators of social class. Individual-level indicators alone may not fully capture the sociostructural and economic conditions that influence the distribution of disease. Utilization of both area- and individual-level indicators of social class may therefore contribute to a better understanding of the effects of social class on breast cancer incidence. We examined the joint association of social class and body size with breast cancer incidence among a cohort of postmenopausal women and tested whether area-based measures of social class provided distinct social effects on breast cancer incidence apart from individual-based measures of social class. Our analyses focused on postmenopausal women, for whom heavier body size increases breast cancer risk (rather than premenopausal obesity, which has a protective effect).
METHODS
This study is nested within CLUE II, an ongoing prospective cohort study of residents of Washington County, Maryland, begun in 1989. The name “CLUE” comes from the campaign slogan that was used for recruitment: “Give us a CLUE to cancer and heart disease.” At baseline, participants provided a blood specimen and completed a food frequency questionnaire and a brief questionnaire on demographic characteristics, physiological factors, and behavioral factors.9 Similar to the county population, 98% of the participants were White. Women who were more educated and older had higher participation rates. Of the total adult participants, 6285 were postmenopausal White women. All participants provided written consent at the beginning of the study.
We identified incident breast cancer cases (International Classification of Diseases, 10th Edition,10 code 174; n = 272) occurring between 1990 through 2005 by linkage of the study cohort to the Washington County and Maryland cancer registries. The county registry identifies cases from death certificates and pathology reports at the county's single hospital9; the Maryland Cancer Registry receives mandatory reporting of all cancers from hospitals, pathology laboratories, and physicians throughout the state and through data sharing from other state registries.11 Additionally, CLUE II cohort follow-up questionnaires periodically ask respondents about new cancer diagnoses, which are then verified through registries. We identified deceased participants through death certificates and daily reviews of obituaries. Thus, only participants who both moved out of state and refused follow-up were missing from incident counts.
Analyses included postmenopausal White women residing in Washington County at baseline (1989) who had complete data for age, education, self-reported BMI at baseline and at age 21 years, breast cancer incidence status, and geocodable residential addresses. Based on self-reports and registry information, women with known breast cancer at baseline were excluded as representing prevalent cases (n = 299).
Residential address information was obtained for all participants of CLUE II at baseline and geocoded with ArcGIS 9.0 (ESRI, Redlands, CA). We assigned participants by geocoded address to census block groups to generate area-level variables. Area-level sociodemographic data on Washington County came from the 1990 US Census (US Census Bureau Summary Tape File 3A).12 Ninety-five percent of addresses geocoded to block groups with complete census data; participants with nongeocodable addresses (n = 310) showed no trends across individual-level variables and were excluded.
In addition, women missing data on education (n = 4), self-reported BMI at baseline (n = 9), or self-reported BMI at age 21 years (n = 26) were excluded. We excluded 643 respondents for 1 or more reason, leaving 5642 participants for the analyses.
A 1996 follow-up questionnaire collected information on reproductive risk factors. Of the 5642 women, 2231 did not participate in the 1996 follow-up and 199 were excluded for missing hormone use information, resulting in 3212 women available for this subanalysis.
Measures
Social class indicators.
We used both individual- and area-level social class indicators. Education was the only available individual-level social class measure and was assessed by the question, “How many grades of school, including college, have you completed?” Area-level social class indicators from the 1990 census block-group data included per capita income and percentage of individuals 25 years and older with a high school diploma; unemployed individuals 16 years and older; working individuals 16 years and older with white-collar occupations; individuals in poverty; households without a car; households without a telephone; and owner-occupied housing units. We classified 5 of the 13 census-defined occupational groups as white collar: executive; administrative and managerial; professional specialty, technician, and related support; sales; and administrative support and clerical. Throughout this article, social class refers to both individual and contextual measures.
Body size indicators.
We calculated BMI, defined as weight in kilograms divided by height in meters squared,2,13 from self-reported weight and height at baseline. We assessed the change in BMI by the difference between BMI at baseline and BMI at age 21 years.
Cancer outcome.
The cancer outcome was time to breast cancer incidence. We counted time as years between baseline and diagnosis of breast cancer cases or, among noncases, between baseline and the last date they were known to be alive.
Covariates.
Covariates included age at baseline and, for the subanalysis, age at first birth, number of live births, breast-feeding (ever or never), and hormone use (ever or never).
Statistical Analyses
We used descriptive statistics to characterize the study population and determine frequency distributions, measures of central tendency, and dispersion. We used exploratory factor analysis to reduce the 8 area-level social class indicators to a more parsimonious set of factors with an assumed underlying dimensional structure of social class. Factor analysis reduced the 8 area-level indicators to 2 factors. Factor 1 included measures of material resources (percentage of individuals below poverty, percentage of households without a telephone, percentage of households without a car, and percentage of individuals living in owner-occupied housing). Factor 2 included measures of social class (per capita income, percentage of high school graduates, percentage of white-collar workers, and percentage of employed individuals). Based on the results from the exploratory factor analysis, we created separate indices for material resources and social class. We reverse-scored deprivation measures (e.g., percentage of individuals without a car) and divided annual income by $1000. For the 2 indices, we summarized and standardized block-group values and assigned block-group index values to respondents.
We used Cox proportional hazards and multilevel Cox proportional hazards models to examine the association of social class, self-reported body size, and breast cancer incidence for the individual- and area-level regression analyses, respectively. Survival analysis was appropriate in this older cohort given potential variation in obesity-related and overall mortality risk and, thus, variation in the potential time period for developing breast cancer. Continuous variables were centered at their mean to aid in interpretation of interaction terms.
We built models in the following manner: to investigate the effect of body size within social classes, we first modeled the risk of breast cancer with education and each of the 3 self-reported body size measures. Interactions between body size measures and education were then tested to determine whether the risks associated with body size differed by social class. We then added area-level material resources and area-level social class to the individual-level predictor models to investigate the additional explanatory power of area-level social influences on breast cancer risk. For all models, we report age-adjusted estimates. For subanalyses models, we report estimates adjusted for age, age at first birth, breast-feeding, and hormone use.
RESULTS
Table 1 shows the characteristics of the study population. In the full cohort, mean years of schooling was 11.7 years. Respondents averaged a BMI of 26.2 kg/m2 at baseline, a BMI of 21.2 kg/m2 at age 21 years, and a BMI change of 5.0 kg/m2. Average age was 63.3 years. The subanalysis population had similar sociodemographic characteristics and reported an average of 2.7 live births, an average age of first birth of 22.6 years, with 15.8% reporting hormone use, and 41.1% ever breast-feeding. We identified approximately 5% of both full and subanalysis populations as incident breast cancer cases.
TABLE 1.
Characteristics of Postmenopausal White Women: CLUE II, Washington County, MD, 1989–2005
| Full Samplea |
Subgroup With Reproductive Datab |
|||
| Mean (SD) | Range | Mean (SD) | Range | |
| Individual-level predictors | ||||
| Education, y | 11.7 (2.6) | 2–22 | 12.0 (2.5) | 3–21 |
| BMI, kg/m2 | ||||
| At baseline | 26.2 (4.9) | 13.7–57.6 | 26.3 (4.8) | 13.7–57.6 |
| At 21 years | 21.2 (3.2) | 12.3–51.5 | 21.1 (3.1) | 12.9–41.6 |
| Change | 5.0 (4.6) | −13.9 to 31.0 | 5.2 (4.4) | −13 to 31 |
| Age, y | 63.3 (8.8) | 50–96 | 61.6 (7.7) | 50–88 |
| Age at first birth, y | … | … | 22.6 (4.4) | 14–42 |
| Number of live births | … | … | 2.7 (1.6) | 0–11 |
| Hormone use, % | … | … | 15.8 (…) | … |
| Breast-fed, % | … | … | 41.1 (…) | … |
| Block group measures | ||||
| % poverty | 8.9 (8.7) | 0.0–46.8 | 8.2 (7.6) | 0.0–46.8 |
| % without phone | 3.9 (5.1) | 0.0–41.0 | 3.6 (4.8) | 0.0–41.0 |
| % without car | 9.4 (10.7) | 0.0–70.8 | 8.2 (8.7) | 0.0–70.8 |
| % owner-occupied housing | 66.9 (21.3) | 0.0–97.2 | 69.0 (19.3) | 0.0–97.2 |
| Material resources | 344.7 (40.5) | 178.9–390.9 | 349.1 (34.7) | 178.9–390.9 |
| % high school graduate | 71.3 (12.1) | 25.4–92.7 | 72.1 (11.3) | 25.4–92.7 |
| % white collar | 52.5 (12.8) | 7.9–81.1 | 52.7 (12.7) | 7.9–81.1 |
| Per capita income, in $1000 | 14.0 (3.6) | 5.6–23.3 | 14.2 (3.5) | 5.6–23.3 |
| % unemployed | 3.8 (2.7) | 0–24 | 3.8 (2.6) | 0–24 |
| Social class | 359.6 (57.3) | 180.2–502.8 | 362.7 (56.2) | 180.2–502.8 |
Note. BMI = body mass index. For the full sample, N = 5642; for the subgroup with reproductive data, n = 3212. There were 98 block groups in Washington, MD. Ellipses indicate that data is not available.
In the full sample, 5370 (95.2%) did not have breast cancer, and 272 (4.8%) did.
In the subgroup, 3040 (94.6%) did not have breast cancer, and 172 (5.4%) did.
Respondent block groups had an average of 71.3% high school graduates, 52.5% white-collar workers, 66.9% owner-occupied residences, and $14 000 per capita income. There was relatively little material deprivation, with an average of 8.9% of residents living below poverty level, 3.8% unemployment, 3.9% of households without phones, and 9.4% without cars.
Table 2 shows the results of the Cox proportional hazards model for the association between individual-level education and body size with breast cancer incidence. In age-adjusted analyses, education and 2 of the 3 weight measures (current BMI and BMI change) were statistically significant predictors of increased breast cancer risk. We found only 1 significant interaction, that of education and BMI at age 21 years, indicating that the risk associated with education was reduced among women who were heavier at age 21. Subgroup analyses, which were adjusted for reproductive factors, resulted in similar risk estimates as the age-adjusted models. However, perhaps because of reduced sample size, only BMI change remained statistically significant (P < .05).
TABLE 2.
Cox Proportional Hazards Ratios (HRs) for the Association of Education, Body Size, and Breast Cancer Incidence Among Postmenopausal White Women: CLUE II, Washington County, MD, 1989–2005
| Model 1 |
Model 2 |
|||
| Full Sample, HR (95% CI) | Subgroup, HR (95% CI) | Full Sample, HR (95% CI) | Subgroup, HR (95% CI) | |
| Education and BMI | ||||
| Education,a y | 1.06* (1.01, 1.11) | 1.06 (0.99, 1.13) | 1.06* (1.01, 1.11) | 1.06 (0.99, 1.13) |
| BMI,b kg/m2 | 1.03* (1.01, 1.05) | 1.03 (1.00, 1.06) | 1.03* (1.00, 1.05) | 1.03 (1.00, 1.06) |
| Education × BMI at baseline | 0.99 (0.99, 1.00) | 1.00 (0.99, 1.01) | ||
| Education and BMI at age 21 y | ||||
| Education,a y | 1.05* (1.00, 1.10) | 1.05 (0.98, 1.12) | 1.05* (1.00, 1.10) | 1.05 (0.99, 1.12) |
| BMI at age 21 y,c kg/m2 | 0.98 (0.94, 1.02) | 0.99 (0.94, 1.04) | 0.97 (0.93, 1.01) | 0.98 (0.93, 1.03) |
| Education × BMI at age 21 y | 0.98* (0.97, 0.99) | 0.98* (0.96, 1.00) | ||
| Education and BMI change | ||||
| Education,a y | 1.06* (1.01, 1.11) | 1.06 (0.99, 1.13) | 1.06* (1.019, 1.11) | 1.06 (0.99, 1.13) |
| BMI change,d kg/m2 | 1.04* (1.02, 1.07) | 1.04* (1.00, 1.07) | 1.04* (1.02, 1.07) | 1.04* (1.00, 1.07) |
| Education × BMI change | 1.00 (0.99, 1.01) | 1.01 (0.99, 1.02) | ||
Note. CI = confidence interval; BMI = body mass index. We adjusted full group models for age. We adjusted subgroup models for age, hormone use, age at first birth, number of live births, and breastfeeding. Analyses for Model 1 included education and body size measure; analyses for Model 2 included education, body size measure, and the interaction term.
Mean years of education = 11.7 years.
Mean BMI = 26.2 kg/m2.
Mean BMI at age 21 years = 21.2 kg/m2.
BMI change is the difference between BMI at baseline and BMI at age 21 years.
P < .05, 2 sided.
Table 3 shows the additional effect of block group–level material resources on breast cancer incidence, after we controlled for individual-level education, body size, and age. Overall, material resources were not significantly associated with incidence, nor did including this measure in models significantly change the relationships of individual-level education and body size to incidence.
TABLE 3.
Association of Material Resources, Education, Body Size, and Breast Cancer Incidence for Postmenopausal White Women: CLUE II, Washington County, MD, 1989–2005
| Model 1 |
Model 2 |
|||
| Full Sample, HR (95% CI) | Subgroup, HR (95% CI) | Full Sample, HR (95% CI) | Subgroup, HR (95% CI) | |
| Material resources, education, and BMI | ||||
| Material resources | 0.95 (0.84, 1.08) | 0.92 (0.78, 1.09) | 0.98 (0.86, 1.12) | 0.94 (0.79, 1.13) |
| Education,a y | 1.06* (1.01, 1.11) | 1.06 (1.00, 1.13) | 1.06* (1.01, 1.11) | 1.06 (0.99, 1.13) |
| BMI,b kg/m2 | 1.03* (1.01, 1.05) | 1.03 (1.00, 1.06) | 1.03* (1.00, 1.05) | 1.03 (1.00, 1.06) |
| Material resources × BMI | 0.98 (0.96, 1.00) | 0.98 (0.95, 1.01) | ||
| Material resources, education, and BMI at age 21 y | ||||
| Material resources | 0.95 (0.84, 1.07) | 0.92 (0.78, 1.09) | 0.95 (0.84, 1.08) | 0.94 (0.78, 1.13) |
| Education, y | 1.05* (1.00, 1.10) | 1.05 (0.99, 1.12) | 1.05* (1.00, 1.10) | 1.05 (0.99, 1.12) |
| BMI at age 21 y, kg/m2c | 0.98 (0.94, 1.02) | 0.99 (0.94, 1.04) | 0.98 (0.94, 1.02) | 0.99 (0.94, 1.04) |
| Material resources × BMI at age 21 yc | 0.99 (0.96, 1.02) | 0.97 (0.93, 1.01) | ||
| Material resources, education, and BMI change | ||||
| Material resources | 0.95 (0.84, 1.08) | 0.92 (0.78, 1.10) | 0.97 (0.85, 1.11) | 0.93 (0.78, 1.10) |
| Education, y | 1.06* (1.01, 1.11) | 1.06 (0.99, 1.13) | 1.06* (1.01, 1.11) | 1.06 (0.99, 1.13) |
| BMI change,d kg/m2 | 1.04* (1.02, 1.07) | 1.04* (1.00, 1.07) | 1.04* (1.02, 1.07) | 1.04* (1.00, 1.07) |
| Material resources × BMI change | 0.99 (0.97, 1.01) | 0.99 (0.95, 1.03) | ||
Note. HR = hazards ratio; CI = confidence interval; BMI = body mass index. We adjusted full group models for age. We adjusted subgroup models for age, hormone use, age at first birth, number of live births, and breastfeeding. Analyses for model 1 included material resources, education, and body size measure; analyses for model 2 included material resources, education, body size measure, and the interaction term.
Mean years of education = 11.7 years.
Mean BMI = 26.2 kg/m2.
Mean BMI at age 21 years = 21.2 kg/m2.
BMI change is the difference between BMI at baseline and BMI at age 21 years.
P < .05, 2 sided.
Table 4 shows that the results for area-level social class were similar to those for material resources. Area-level social class did not have a significant independent relationship to breast cancer risk when included in models with individual-level education. However, individual-level education, BMI, and BMI change continued to convey excess risk for breast cancer.
TABLE 4.
Association of Area-Level Social Class, Education, Body Size, and Breast Cancer Incidence for Postmenopausal White Women: CLUE II, Washington County, MD, 1989–2005
| Model 1 |
Model 2 |
|||
| Full Sample, HR (95% CI) | Subgroup, HR (95% CI) | Full Sample, HR (95% CI) | Subgroup, HR (95% CI) | |
| Area-level social class, education, and BMI | ||||
| Area-level Social Class | 0.98 (0.86, 1.12) | 1.03 (0.87, 1.22) | 0.98 (0.86, 1.12) | 1.03 (0.87, 1.22) |
| Education,a y | 1.06* (1.01, 1.11) | 1.06 (0.99, 1.13) | 1.06* (1.01, 1.11) | 1.05 (0.99, 1.13) |
| BMI,b kg/m2 | 1.03* (1.01, 1.05) | 1.03 (1.00, 1.06) | 1.02 (1.00, 1.05) | 1.02 (0.99, 1.06) |
| Area-level social class × BMI | 0.97* (0.94, 0.99) | 0.97* (0.94, 1.00) | ||
| Area-level social class, education, and BMI at age 21 y | ||||
| Area-level social class | 0.97 (0.85, 1.11) | 1.02 (0.87, 1.21) | 0.95 (0.83, 1.09) | 1.00 (0.85, 1.18) |
| Education,a y | 1.05* (1.00, 1.10) | 1.05 (0.98, 1.12) | 1.05* (1.00, 1.10) | 1.05 (0.98, 1.12) |
| BMI at age 21 y,c kg/m2 | 0.98 (0.94, 1.02) | 0.99 (0.94, 1.04) | 0.97 (0.93, 1.01) | 0.98 (0.93, 1.04) |
| Area-level social class × BMI at age 21 y | 0.96 (0.92, 1.00) | 0.94 (0.89, 1.00) | ||
| Area-level social class, education, and BMI change | ||||
| Area-level social class | 0.98 (0.86, 1.12) | 1.03 (0.87, 1.22) | 1.00 (0.87, 1.14) | 1.04 (0.88, 1.22) |
| Education,a y | 1.06* (1.01, 1.11) | 1.05 (0.99, 1.13) | 1.06* (1.01, 1.11) | 1.05 (0.98, 1.13) |
| BMI change,d kg/m2 | 1.04* (1.02, 1.07) | 1.04* (1.00, 1.07) | 1.04* (1.01, 1.07) | 1.04* (1.00, 1.07) |
| Area-level social class × BMI change | 0.98* (0.95, 1.00) | 0.98 (0.95, 1.02) | ||
Note. HR = hazards ratio; CI = confidence interval; BMI = body mass index. We adjusted full group models for age. We adjusted subgroup models for age, hormone use, age at first birth, number of live births, and breastfeeding. Analyses for Model 1 included area-level social class, education, and body size measure; analyses for Model 2 included area-level social class, education, body size measure, and the interaction term.
Mean education = 11.7 years.
Mean BMI = 26.2 kg/m2.
Mean BMI at age 21 years = 21.2 kg/m2.
BMI change is the difference between BMI at baseline and BMI at age 21 years.
P < .05, 2 sided.
With the addition of the interaction terms between body size measures and area-level social class, education remained a statistically significant predictor of breast cancer incidence in the models. Although not significant as a main effect, area-level social class attenuated the association between current BMI and breast cancer incidence (hazard ratio [HR] = 0.97; 95% confidence interval [CI] = 0.94, 0.99; P < .05), and BMI change (HR = 0.98; 95% CI = 0.95, 1.00; P < .05). The interaction of area-level social class and BMI at age 21 years was not statistically significant. (Additional multilevel models, which were adjusted for rurality, produced similar results and are not shown.)
DISCUSSION
Our results provide additional evidence of links between both body size and social class, measured in several ways, and breast cancer incidence in postmenopausal women. In addition, the analyses found that these relationships were primarily independent of each other, suggesting that the mechanisms through which women of greater body size increased their risk for breast cancer were similar across social class and that risk associated with social class did not operate through body size. However, there is evidence that social class may to some degree moderate the influence of body size on breast cancer risk.
Body Size and Breast Cancer
Consistent with previous studies (e.g., Ahn et al.14), BMI change was a consistently strong predictor of breast cancer incidence, after adjustment for age, individual and contextual social class indicators, and rural residence. BMI change continues to be a predictor of breast cancer incidence even after adjustment for reproductive factors. Current BMI was also a statistically significant predictor of breast cancer incidence, although it was a slightly weaker predictor than was BMI change.4,6–8,14–18 However, after adjustment for reproductive factors, we found that current BMI became nonstatistically significant. BMI, a static measure, does not capture lifelong weight patterns and their metabolic influence on the promotion of mammary carcinogenesis. BMI change, however, is a more dynamic measure and, therefore, may be a more sensitive marker of the metabolic role of obesity in altering ovarian hormone, glucose metabolism, and growth factors that may promote mammary carcinogenesis.19 In models that included both BMI and BMI change (absolute and percentage), only BMI change was significant, further supporting this conclusion (results not shown).
BMI at age 21 years had no statistically significant association with breast cancer incidence. This is consistent with an endocrine mechanism model for the effect of body size on postmenopausal breast cancer. Premenopausal overweight and obese women have reduced levels of estradiol during the anovulatory cycles, conferring decreased breast cancer risk.
Social Class and Breast Cancer
With respect to social class indicators, we found education to be a statistically significant predictor of breast cancer incidence after adjustment for age and body size measures. This finding was consistent both with previous literature, which shows an increased risk of breast cancer among women of higher educational levels,4 and with the theory of social stratification.
The results of the subanalyses do not fully answer whether the effect of social class on breast cancer risk works exclusively through reproductive behaviors. As previously described, risk factors for breast cancer incidence such as earlier age at menarche, later age at first birth, fewer total births, later age at menopause, and not having breast-fed have historically been more prevalent among women of higher social class.1–3 In our analyses, risk estimates remained consistent in models before and after adjustment for reproductive factors. Differences in statistical significance may reflect differences in sample size and statistical power. This suggests that reproductive factors, although important, may not explain the entire relationship between social class and breast cancer.
Individual Versus Contextual Measures of Social Class
Among the women in our study, a single individual-level measure, years of formal schooling, sufficiently captured the relationship between social class and breast cancer risk. Our 2 measures of social resources at the neighborhood level were not significant, independent predictors of breast cancer risk when examined in models without individual-level education (data not shown), nor did they add significantly when included in models with individual-level education. This lack of association differs from previous findings.20–22
There are several possible explanations for this. One is that current social resources, reflected by one's area of residence, may have less relevance for breast cancer etiology than the level of education attained, which shapes the social resources, norms, and behaviors of women across the life course, including the critical time periods for breast cancer development. However, it is possible that a single county, although showing some degree of social resource variation, was not a large or diverse enough geographic area to reveal the additional influence of area-level resources.
Our results suggest that, under certain circumstances, the influence of body size on breast cancer incidence does vary by social class. Education and area-level social class reduced the effect of BMI at age 21 years (although itself not a statistically significant predictor) on breast cancer incidence, and area-level social class reduced the effect of BMI and BMI change on risk only in the unadjusted and age-adjusted models. In the model adjusting for reproductive factors, we found education to modify the effect of BMI at age 21 years. One interpretation of these results--that social resources in some way work to influence the risk associated with larger body size--merits exploration in additional research.
Conclusions
Although our study used both constant and changing measures of body size, future research should examine a more accurate measure of body size by taking into consideration weight during periods of hormonal change and changes in weight gain or weight loss over the lifespan, because they may have different biological effects on the endocrine mechanism and because they may more precisely identify women at risk for breast cancer incidence.
To further understand the causal effects of social factors or body size on breast cancer incidence, theoretically informed hypotheses are required. These hypotheses would improve upon our current knowledge of the effects of social class and body size on breast cancer incidence, an improvement that would contribute to an enhanced understanding of the inequalities contributed by both factors.
Because of the independent and shared effects of social class and body size on breast cancer incidence, these factors need to be considered in the development and implementation of targeted interventions to further promote behaviors that reduce the risk and burden of breast cancer in the population.
Acknowledgments
C. M. Torio was partially supported by the Ruth L. Kirschtein National Research Service Institutional Training Grant (CA 09314-23), sponsored by the National Institutes of Health. Other funding sources came from the Breast Specialized Programs of Research Excellence project (grant CA 94-027) sponsored by the National Cancer Institute (NCI); the Department of Defense (grant CA ES-93-0); and CLUE II (grant CA 47503), sponsored by the NCI.
Note. Funding sources had no involvement in study design, collection, analysis, data interpretation, writing of the article, or in the decision to submit the article for publication.
Human Participant Protection
This research received institutional review board approval from the Johns Hopkins Committee on Human Research.
References
- 1.International Agency for Research on Cancer (IARC)/World Health Organization Social Inequalities and Cancer Lyon, France: IARC Scientific Publications; 1997 [Google Scholar]
- 2.Kelsey JL, Horn-Ross PL. Breast cancer: magnitude of the problem and descriptive epidemiology. Epidemiol Rev 1993;15(1):7–16 [DOI] [PubMed] [Google Scholar]
- 3.Krieger N. Exposure, susceptibility, and breast cancer risk: a hypothesis regarding exogenous carcinogens, breast tissue development, and social gradients, including Black/White differences, in breast cancer incidence. Breast Cancer Res Treat 1989;13(3):205–223 [DOI] [PubMed] [Google Scholar]
- 4.Paffenbarger RS, Jr, Kampert JB, Chang HG. Characteristics that predict risk of breast cancer before and after the menopause. Am J Epidemiol 1980;112(2):258–268 [DOI] [PubMed] [Google Scholar]
- 5.Stoll BA. Obesity, social class and Western diet: a link to breast cancer prognosis. Eur J Cancer 1996;32A(8):1293–1295 [DOI] [PubMed] [Google Scholar]
- 6.Barnes-Josiah D, Potter JD, Sellers TA, et al. Early body size and subsequent weight gain as predictors of breast cancer incidence (Iowa, United States). Cancer Causes Control 1995;6(2):112–118 [DOI] [PubMed] [Google Scholar]
- 7.Brinton LA, Swanson CA. Height and weight at various ages and risk of breast cancer. Ann Epidemiol 1992;2(5):597–609 [DOI] [PubMed] [Google Scholar]
- 8.Huang Z, Hankinson SE, Colditz GA, et al. Dual effects of weight and weight gain on breast cancer risk. JAMA 1997;278(17):1407–1411 [PubMed] [Google Scholar]
- 9.Helzlsouer KJ, Selmin O, Huang HY, et al. Association between glutathione S-transferase M1, P1, and T1 genetic polymorphisms and development of breast cancer. J Natl Cancer Inst 1998;90(7):512–518 [DOI] [PubMed] [Google Scholar]
- 10.International Classification of Disease, 10th edition. Geneva, Switzerland: World Health Organization; 2006 [Google Scholar]
- 11.Tri-Analytics I. Final Maryland Cancer Registry Data Report. Bel Air, MD: Maryland Cancer Registry; 1993 [Google Scholar]
- 12.US Census of Population and Housing, 1990: Summary Population and Housing Characteristics. Washington: Government Printing Office, 1991 [Google Scholar]
- 13.World Health Organization Physical Status: The Use and Interpretation of Anthropometry. Geneva, Switzerland: World Health Organization; 1995. Report No. 854 [Google Scholar]
- 14.Ahn J, Schatzkin A, Lacey JV, et al. Adiposity, adult weight change, and postmenopausal breast cancer risk. Arch Intern Med 2007;167(19):2091–2102 [DOI] [PubMed] [Google Scholar]
- 15.Feigelson HS, Jonas CR, Teras LR, et al. Weight gain, body mass index, hormone replacement therapy, and postmenopausal breast cancer in a large prospective study. Cancer Epidemiol Biomarkers Prev 2004;13(2):220–224 [DOI] [PubMed] [Google Scholar]
- 16.La Vecchia C, Negri E, Franceschi S, et al. Body mass index and post-menopausal breast cancer: an age-specific analysis. Br J Cancer 1997;75(3):441–444 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Le Marchand L, Kolonel LN, Earle ME, et al. Body size at different periods of life and breast cancer risk. Am J Epidemiol 1988;128(1):137–152 [DOI] [PubMed] [Google Scholar]
- 18.Sonnenschein E, Toniolo P, Terry MB, et al. Body fat distribution and obesity in pre- and postmenopausal breast cancer. Int J Epidemiol 1999;28(6):1026–1031 [DOI] [PubMed] [Google Scholar]
- 19.Ballard-Barbash R. Anthropometry and breast cancer. Body size—a moving target. Cancer 1994;74(suppl 3):1090–1100 [DOI] [PubMed] [Google Scholar]
- 20.Krieger N. Social class and the Black/White crossover in the age-specific incidence of breast cancer: a study linking census-derived data to population-based registry records. Am J Epidemiol 1990;131(5):804–814 [DOI] [PubMed] [Google Scholar]
- 21.Liu L, Deapen D, Bernstein L. Socioeconomic status and cancers of the female breast and reproductive organs: a comparison across racial/ethnic populations in Los Angeles County, California (United States). Cancer Causes Control 1998;9(4):369–380 [DOI] [PubMed] [Google Scholar]
- 22.Yost K, Perkins C, Cohen R, et al. Socioeconomic status and breast cancer incidence in California for different race/ethnic groups. Cancer Causes Control 2001;12(8):703–711 [DOI] [PubMed] [Google Scholar]