Abstract
Background: Although social exposures have complex and dynamic relationships and interactions, the existing literature on the impact of rural–urban residence on stage at breast cancer diagnosis does not examine heterogeneity of effect. We examined the joint effect of social support, social relationship strain, and rural–urban residence on stage at breast cancer diagnosis.
Methods: Using data from the Women's Health Initiative (WHI) (n = 161,808), we describe the distribution of social, behavioral, and clinical factors by rural–urban residence among postmenopausal women with incident breast cancer (n = 7,120). We used rural–urban commuting area (RUCA) codes to categorize baseline residential addresses as urban, large rural city/town, or small rural town, and the surveillance, epidemiology, and end results staging system to categorize breast cancer stage at diagnosis (dichotomized as early or late). We then used univariable and multivariable logistic regression to estimate odds ratios (ORs) and associated 95% confidence intervals (95% CI) for the relationship between rural–urban residence and stage at breast cancer diagnosis. We included separate interaction terms between rural–urban residence and social strain and social support to test for statistical interaction.
Results: Of the social, behavioral, and clinical factors we examined, only younger age at WHI enrollment screening was significantly associated with late stage at breast cancer diagnosis (p = 0.003). Contrary to our hypothesis, rural–urban residence was not significantly associated with stage at breast cancer diagnosis among postmenopausal women ([adjusted OR, 95% CI] for urban compared with small town: 1.08 [0.76–1.53]; large town compared with small town: 1.16 [0.74–1.84]; and urban compared with large town: 0.93 [0.68–1.26]).The associations did not vary by social support or social strain (p for interaction between RUCA and social strain and social support, respectively: 0.99 and 0.17).
Conclusions: Future studies should examine other potential effect modifiers to identify novel factors predictive or protective for late stage at breast cancer diagnosis associated with rural–urban residence.
Keywords: : stage at breast cancer diagnosis, rural–urban, postmenopausal, social strain and support
Introduction
Breast cancer is a serious public health issue, especially for older women. Stage at cancer diagnosis is directly related to cancer screening behaviors and predicts both survival and quality of life after cancer.1,2 Access to screening mammography increases the likelihood of early detection and successful breast cancer treatment. However, factors such as access to screening, various health behaviors, and quality of care are highly patterned by where people live.3,4 Rural residents tend to be older, have lower socioeconomic position, are more likely to smoke and be obese, lack health insurance coverage, and have reduced access and utilization of cancer screening than residents of urban areas.5,6 All of these factors contribute to late stage at breast cancer diagnosis.7 Rural residents comprise 20% of the U.S. population, and cancer disparities in this group can have a significant impact on the overall health of the nation.6
Some studies have documented a “rural penalty,” in which persistently higher cancer incidence and mortality rates occur in rural versus urban areas.5,8,9 Others have found evidence of an urban disadvantage in terms of cancer stage at diagnosis.10 The authors of the latter study suggested that their findings may be a function of the unique geographic distribution of vulnerable residents in urban (Chicago, IL) compared with more rural areas.10 More research is needed to understand the unique context for life and risk of cancer across the rural–urban continuum.11
The rural–urban commuting area (RUCA) taxonomy12,13 is the recommended and most commonly used method for categorizing rural versus urban census tracts, especially for cancer researchers, given that it takes into account population density, travel distance, and commuting flows.14,15 The existing literature on the association between rural–urban residence (using the RUCA taxonomy) and stage at breast cancer diagnosis is small and suggests no significant direct effect.2,16 However, social exposures such as the residential environment likely interact with and have complicated and dynamic relationships with other social factors.17,18 Said differently, the impact of rural–urban residence on stage at breast cancer diagnosis might be exacerbated or buffered by other social factors.
Social support (the useful aspects of relationships)19 and social strain (negative aspects of social relationships or nonsupportive social ties)20 can positively affect cancer screening behaviors such as repeated mammography, as well as clinical- and self-breast examinations.21,22 Furthermore, a sizeable literature has identified social support as an important protective factor for breast cancer development and progression,23–26 and as a predictor of physical activity and health status among rural breast cancer survivors.27 In contrast, strained relationships have been associated with earlier mortality,28,29 and socially isolated women have been shown to have higher breast cancer mortality than their socially integrated counterparts.30 Given this, it is plausible that the association between rural–urban residence and stage at breast cancer diagnosis is buffered by social support and/or exacerbated by social strain, but no studies have tested these hypotheses.
To build upon the existing literature, we examined whether rural–urban residence was associated with stage at breast cancer diagnosis among a large well-defined racially/ethnically diverse cohort of postmenopausal women, with adjudicated incident cancer outcomes, from the Women's Health Initiative (WHI) observational study and clinical trials. We also tested whether social strain and social support modified the relationship between rural–urban residence and stage at breast cancer diagnosis. We hypothesized that the association between rural–urban residence and stage at diagnosis would be exacerbated by social strain and buffered by social support.
Materials and Methods
Study population
The WHI included 161,808 racial/ethnically diverse postmenopausal women, with 68,132 enrolled in dietary modification (DM), calcium /vitamin D, and hormone therapy (HT) clinical trials, and 93,676 in the observational study (OS). Enrollment occurred from 1993 to 1998, and included women from across 40 U.S. clinical centers, from 24 states and the District of Columbia. Details of the WHI have been published.31–33 Women were eligible to participate in the WHI if they were (1) between 50 and 79 years old, (2) postmenopausal, (3) willing to give informed consent, and (4) expected to survive and not relocate for the next 3 years.
From the original cohort of 161,808 patients, our analytic sample included n = 7,120 study participants who were diagnosed with incident breast cancer between 1995 and 2014. Updates on incident breast cancers were reported semiannually for the clinical trials and annually for the observational study, and all cancers were confirmed by trained physician adjudicators after review of the medical records. Study participants' estimated probability of developing breast cancer was quantified with the Gail risk assessment tool that utilized age, history of benign breast disease, age at menarche, age at first live birth, race/ethnicity, and number of relatives with breast cancer.34
Rural–urban classification
RUCA codes are used to classify all U.S. census tracts into 1 of 10 main categories based on the rural–urban core and the extent of commuting, as well as 33 subcategories based on secondary commuting flows.35 WHI participant addresses were geocoded as previously described,36,37 and their corresponding census tracts were assigned RUCA codes based on a reference document provided by the USDA.35 An aggregation scheme was used to categorize codes into one of four classes: (1) urban or “metropolitan” area, (2) large rural city/town or “micropolitan,” (3) small rural town, and (4) isolated small rural town. Given the sparsity of WHI participants living in isolated small rural communities, we combined the small rural (n = 113) and isolated small rural town (n = 99) categories for our analysis. Considering the possibility of multiple RUCA codes (the RUCA taxonomy is not consistent across census years)35, we assigned each participant a RUCA code using the version of RUCA closest to the date at which her home address was collected. For our analysis, we used the RUCA codes based on the year 2000 census, since this time point is closest to the midpoint of WHI data collection (1995–2014) and allows us to include the largest number of WHI participants and thus the largest number of incident breast cancer cases in the analysis.
Stage at breast cancer diagnosis
Stage at breast cancer diagnosis was coded based on the surveillance, epidemiology, and end results (SEER) criteria.38 There were 6,782 (95%) women who had centrally adjudicated incident cancers and SEER-coded data on stage at breast cancer diagnosis, as well as an appropriate address, closest to the time they were diagnosed, for geocoding.
Covariates
A priori, we identified several potential confounding variables based on clinical relevance. Self-reported age at breast cancer screening (50–59, 60–69, 70–79, and 80+), education (less than high school, high school diploma or general equivalency diploma (GED), some college, and college graduate), health insurance (yes/no), and clinical trials participation (HT, dietary modification, and calcium/vitamin D) were included in multivariable models.
Effect modifiers
Social strain was derived from a validated measure of negative aspects of social relationships20 and was measured with four questions at baseline: How many of the people who are important to you (1) get on your nerves, (2) ask too much of you, (3) did not include you, and (4) try to get you to do things you do not want to do. Answers were coded on a Likert scale and responses ranged from 1 (none) to 5 (all). A summary score was estimated by summing individual items, with higher scores representing more social strain (range of scores: 4–20, Cronbach's alpha: 0.71).
Social support was assessed with a nine-item questionnaire from the medical outcomes study.39 Participants reported how often specific types of social support were available to them, including emotional, affection, tangible support, and positive interactions. Responses ranged from 1 to 5 for each item, and a composite score was calculated by summing the nine items, with higher scores representing more social support (range of scores: 9–45, Cronbach's alpha: 0.93).
Analysis
Given that the WHI data have not been described by RUCA, we summarized distributions of several participant sociodemographic and clinical characteristics (not limited to covariates of interest) by RUCA category, for future hypothesis generation. Stage at breast cancer diagnosis was dichotomized as early (in situ and local) versus late (regional and distant). For descriptive purposes, we included dichotomized social strain and social support at the median (low versus high), given that no established cut-points exist,40 and for consistency with previously published research.26 However, we used the continuous versions of social support and strain variables in our logistic regression models. The assumption of linearity in the logit was assessed graphically (using univariable Lowess plots, plots of the estimated logistic regression coefficients vs. approximate quartile midpoints) as well as using the method of fractional polynomials,41 no serious deviations were noted. Unadjusted and adjusted (for all covariates mentioned earlier) logistic regression models estimated odds ratios (ORs) of late versus early stage at breast cancer diagnosis and 95% confidence intervals (95% CI), comparing women across the rural–urban continuum. Separate interaction terms were included in the models to test whether the impact of rural–urban residence on stage at breast cancer diagnosis was modified by social strain or social support; we present stratified results if warranted. Analyses were performed using SAS/STAT software, Version 9.4 of the SAS System for Windows (SAS Institute, Inc., Cary, NC).
Results
Tables 1 and 2 show the distribution of sociodemographic and psychosocial factors, health behaviors and health history, and clinical indicators in postmenopausal women with incident breast cancer from the WHI, stratified by rural–urban residence (urban, large town, and small town). A higher proportion of younger women (aged 50–59 years) and a lower proportion of the oldest women (aged 70–79 years) resided in small rural towns than urban and large rural towns. In comparison to the urban population, greater proportions of the small and large rural study participants were white. In addition, relative to their urban and large rural counterparts, a larger proportion of women residing in small rural areas had not earned at least a high school diploma or GED, and reported having high social strain and high social support.
Table 1.
Characteristics of Women with Incident Breast Cancer by Rural–Urban Residence at the Time of Diagnosis; Women's Health Initiative (n = 7,120), 1993–2014
| Rural–urban category | ||||
|---|---|---|---|---|
| Characteristics | Total sample N = 7,120 (%) | Urban N = 6,643 (%) | Large rural N = 265 (%) | Small rural N = 212 (%) |
| Sociodemographic/psychosocial | ||||
| Age (years) | ||||
| 50–59 | 2,137 (30) | 1,989 (30) | 75 (28) | 73 (34) |
| 60–69 | 3,336 (47) | 3,109 (47) | 128 (48) | 99 (47) |
| 70–79 | 1,647 (23) | 1,545 (23) | 62 (23) | 40 (19) |
| Race/ethnicity | ||||
| White | 6,192 (87) | 5,732 (86) | 257 (97) | 203 (96) |
| Black | 508 (7) | 495 (7) | 4 (2) | 9 (4) |
| Other | 403 (6) | 399 (6) | 4 (2) | 0 |
| Education | ||||
| None/some HS | 266 (4) | 241 (4) | 11 (4) | 14 (7) |
| HS/GED | 1,094 (15) | 988 (15) | 58 (22) | 48 (23) |
| >HS | 5,700 (80) | 5,359 (81) | 192 (72) | 149 (70) |
| Any insurance | ||||
| No | 221 (3) | 201 (3) | 11 (4) | 9 (4) |
| Yes | 6,843 (96) | 6,389 (96) | 252 (95) | 202 (95) |
| Social strain | ||||
| Median (IQR) | 6 (4,8) | 6 (4,8) | 6 (4,8) | 6 (4,8) |
| (min, max) | (4,20) | (4,20) | (4,16) | (4,16) |
| Social strain | ||||
| Low (≤5) | 3,047 (43) | 2,846 (43) | 117 (44) | 84 (40) |
| High (>5) | 3,935 (55) | 3,667 (55) | 145 (55) | 123 (58) |
| Social support | ||||
| Median (IQR) | 37 (32,43) | 37 (32,42) | 37 (34,53) | 38 (32,44) |
| (min, max) | (9,45) | (9,45) | (12,45) | (17,45) |
| Social support | ||||
| Low (≤37) | 3,553 (50) | 3,321 (50) | 135 (51) | 97 (46) |
| High (>37) | 3,404 (48) | 3,163 (48) | 129 (49) | 112 (53) |
| Health behaviors/history | ||||
| Family history of breast cancer | ||||
| No | 5,083 (71) | 4,768 (72) | 171 (65) | 144 (68) |
| Yes | 1,656 (23) | 1,527 (23) | 72 (27) | 57 (27) |
| Gail risk | ||||
| <1.67% | 3,526 (50) | 3,290 (50) | 116 (44) | 120 (57) |
| ≥1.67% | 3,594 (50) | 3,353 (50) | 149 (56) | 92 (43) |
| Mammogram in past 2 years | ||||
| Yes | 6,065 (85) | 5,678 (85) | 215 (81) | 172 (81) |
| No | 861 (12) | 778 (12) | 44 (17) | 39 (18) |
| WHI trial participation | ||||
| No | 2,853 (40) | 2,660 (40) | 107 (40) | 86 (41) |
| Yes | 4,267 (60) | 3,983 (60) | 158 (60) | 126 (59) |
| WHI study type | ||||
| OS | 4,267 (60) | 3,983 (60) | 158 (60) | 126 (59) |
| E-alone | 228 (4) | 205 (3) | 9 (3) | 14 (7) |
| E+P trial | 436 (6) | 391 (6) | 27 (10) | 18 (8) |
| DM | 1,883 (26) | 1782 (27) | 59 (22) | 42 (20) |
| E-alone/E+P trial + DM | 306 (4) | 282 (4) | 12 (5) | 12 (6) |
| Alcohol | ||||
| Nondrinker | 667 (9) | 578 (9) | 46 (17) | 43 (20) |
| Former drinker | 1,208 (17) | 1,125 (17) | 44 (17) | 39 (18) |
| Current drinker | 5,188 (73) | 4,885 (74) | 175 (66) | 128 (60) |
| Smoking | ||||
| Never smoker | 3,432 (48) | 3,156 (48) | 147 (55) | 129 (61) |
| Past smoker | 3,149 (44) | 2,981 (45) | 104 (39) | 64 (30) |
| Current smoker | 443 (6) | 416 (6) | 10 (4) | 17 (8) |
| Self-rated health | ||||
| Excellent/very good | 4,245 (60) | 3,988 (60) | 150 (57) | 107 (50) |
| Good | 2,264 (32) | 2,088 (31) | 93 (35) | 83 (39) |
| Fair/poor | 575 (8) | 533 (8) | 22 (8) | 20 (9) |
| Physical activity (MET/week) | ||||
| None | 1,030 (14) | 940 (14) | 52 (20) | 38 (18) |
| > 0–3.75 | 987 (14) | 922 (14) | 41 (15) | 24 (11) |
| 3.75–8.75 | 1,411 (20) | 1,316 (20) | 53 (20) | 42 (20) |
| 8.75–17.5 | 1,586 (22) | 1,484 (22) | 62 (23) | 40 (19) |
| ≥17.5 | 1,744 (24) | 1,648 (25) | 41 (15) | 55 (26) |
| Hysterectomy | ||||
| No | 4,398 (62) | 4,120 (62) | 159 (60) | 119 (56) |
| Yes | 2,717 (38) | 2,518 (38) | 106 (40) | 93 (44) |
| Bilateral oophorectomy | ||||
| No | 5,700 (80) | 5,320 (80) | 212 (80) | 168 (79) |
| Yes | 1,277 (18) | 1190 (18) | 47 (18) | 40 (19) |
| Hormone therapy use | ||||
| Never | 2,802 (39) | 2,611 (39) | 97 (37) | 94 (44) |
| Past user | 1,077 (15) | 997 (15) | 42 (16) | 38 (18) |
| Current user <5 years | 841 (12) | 788 (12) | 31 (12) | 22 (10) |
| Current user 5 to <10 years | 826 (12) | 783 (12) | 28 (11) | 15 (7) |
| Current user ≥10 years | 1,569 (22) | 1,459 (22) | 67 (25) | 43 (20) |
| Oral contraceptive use | ||||
| No | 4,235 (59) | 3,972 (60) | 146 (55) | 117 (55) |
| Yes | 2,885 (41) | 2,671 (40) | 119 (45) | 95 (45) |
| Aspirin use | ||||
| No | 5,607 (79) | 5,244 (79) | 201 (76) | 162 (76) |
| Yes | 1,513 (21) | 1,399 (21) | 64 (24) | 50 (24) |
| NSAID use | ||||
| No | 4,744 (67) | 4,453 (67) | 159 (60) | 132 (62) |
| Yes | 2,376 (33) | 2,190 (33) | 106 (40) | 80 (38) |
| Current diabetes | ||||
| No | 6,828 (96) | 6,378 (96) | 252 (95) | 198 (93) |
| Yes | 287 (4) | 261 (4) | 13 (5) | 13 (6) |
| Cholesterol medication use | ||||
| No | 5,748 (81) | 5,369 (81) | 208 (78) | 171 (81) |
| Yes | 939 (13) | 874 (13) | 38 (14) | 27 (13) |
| Myocardial infarction | ||||
| No | 6,970 (98) | 6,504 (98) | 259 (98) | 207 (98) |
| Yes | 147 (2) | 136 (2) | 6 (2) | 5 (2) |
| Angina | ||||
| No | 6,684 (94) | 6,237 (94) | 246 (93) | 201 (95) |
| Yes | 376 (5) | 347 (5) | 19 (7) | 10 (5) |
| Charlson comorbidity index (modified) | ||||
| 0 | 4,178 (59) | 3,921 (59) | 150 (57) | 107 (50) |
| 1+ | 2,775 (39) | 2,564 (39) | 110 (42) | 101 (48) |
DM, dietary modification; E+P, estrogen + progesterone; HS, high school; GED, general equivalency diploma; IQR, interquartile range; MET, metabolic equivalent values; NSAID, nonsteroidal anti-inflammatory drugs; OS, observational study.
Table 2.
Clinical Characteristics of Women with Incident Breast Cancer by Rural–Urban Residence at Time of Diagnosis; Women's Health Initiative (n = 7,120), 1993–2014
| Rural–urban category | ||||
|---|---|---|---|---|
| Characteristics | Total N = 7,120 (%) | Urban N = 6,643 (%) | Large rural N = 265 (%) | Small rural N = 212 (%) |
| Approximate age (years) at diagnosis | ||||
| 50 to <60 | 995 (14) | 940 (14) | 27 (10) | 28 (13) |
| 60 to <70 | 3,089 (43) | 2,858 (43) | 134 (51) | 97 (46) |
| 70 to <80 | 2,648 (37) | 2,480 (37) | 90 (34) | 78 (37) |
| ≥80 | 388 (5) | 365 (5) | 14 (5) | 9 (4) |
| SEER stage | ||||
| In situ | 1,210 (17) | 1,122 (17) | 45 (17) | 43 (20) |
| Local | 4,178 (59) | 3,901 (59) | 156 (59) | 121 (57) |
| Regional | 1,331 (19) | 1,240 (19) | 55 (21) | 36 (17) |
| Distant | 63 (1) | 57 (1) | 1 (0) | 5 (2) |
| Morphology—grade | ||||
| 1: Well differentiated | 1,421 (20) | 1,337 (20) | 43 (16) | 41 (19) |
| 2: Moderately differentiated | 2,481 (35) | 2,304 (35) | 95 (36) | 82 (39) |
| 3: Poorly differentiated | 1,583 (22) | 1,477 (22) | 64 (24) | 42 (20) |
| 4: Anaplastic | 476 (7) | 441 (7) | 22 (8) | 13 (6) |
| Unknown/not done/missing | 1,159 (16) | 1,084 (16) | 41 (15) | 34 (16) |
| Hormone receptor status | ||||
| ER/PR+ | 3,605 (51) | 3,369 (51) | 138 (52) | 98 (46) |
| ER+/PR− | 802 (11) | 756 (11) | 20 (8) | 26 (12) |
| ER−/PR+ | 83 (1) | 77 (1) | 3 (1) | 3 (1) |
| ER/PR− | 792 (11) | 721 (11) | 37 (14) | 34 (16) |
| Borderline/unknown/missing ER and/or PR status | 1,838 (26) | 1,720 (26) | 67 (25) | 51 (24) |
| HER-2/Neu | ||||
| Positive | 693 (10) | 638 (10) | 37 (14) | 18 (8) |
| Negative | 2,919 (41) | 2,735 (41) | 97 (37) | 87 (41) |
| Borderline/unknown/missing | 3,508 (49) | 3,270 (49) | 131 (49) | 107 (50) |
ER, estrogen receptor; PR, progesterone receptor; SEER, surveillance, epidemiology, and end results.
Regarding health history and health behaviors, the proportion of women with a family history of breast cancer was lower among urban women than among their small and large rural counterparts. Elevated breast cancer risk as indicated by Gail risk scores >1.67 occurred more frequently among women residing in large rural areas versus others. Greater proportions of small rural town women reported never smoking, being a nondrinker, never using HT, having received a hysterectomy, having medical comorbidities (as indicated by the Charleston comorbidity index), and participating in the estrogen-alone trial. A smaller proportion of large rural town residents reported engaging in high physical activity (≥17 metabolic equivalent values [METS]/week) than residents of urban and small rural towns. A smaller proportion of the women who resided in small towns participated in the diet modification trial, and reported “excellent/very good” self-rated health. Higher proportions of urban women than women in either rural category reported oral contraceptive use and receiving a mammogram in the past 2 years.
There also were rural–urban differences in sociodemographic and clinical characteristics of women with incident breast cancer (Table 2). The distribution of approximate age at incident breast cancer diagnosis was similar across rural–urban areas, except that women who lived in large rural towns were more often between 60 and 69 years old. In addition, a greater proportion of women residing in large rural towns had HER-2/Neu positive cancers and regional stage breast cancer. In contrast, a smaller proportion of women residing in large rural towns had HER-2/Neu negative cancers, and had well-differentiated tumors, compared with those of urban and small rural residents. Among women residing in small rural towns, a greater proportion of tumors were moderately differentiated, a smaller proportion had estrogen receptor/progesterone receptor positive (ER/PR+) tumors and a larger proportion had estrogen receptor/progesterone receptor negative (ER/PR−) tumors relative to their urban and large rural counterparts.
Of the sociodemographic, behavioral, and clinical covariates examined, only age at screening for enrollment into WHI was significantly associated with stage at diagnosis, with older age at screening being protective (OR for age 60–69 years vs. <50–59 years = 0.79; 95% CI: 0.69–0.91; OR for age >70–79 years vs. <50–59 years = 0.82; 95% CI: 0.69–0.96; overall p = 0.003). There was no significant association between rural–urban residence and stage at breast cancer diagnosis (Table 3). Furthermore, there was no evidence that the association between rural–urban residence and stage at breast cancer diagnosis was modified by either social strain or social support.
Table 3.
Relationship Between Rural–Urban Residence and Stage at Breast Cancer Diagnosis, and Interactions with Social Strain and Social Support; Women's Health Initiative (n = 6782), 1993–2014
| Rural–urban comparisons | Unadjusted OR (95% CI) | Adjusted OR (95% CI) |
|---|---|---|
| Urban | 1.03 (0.73–1.46) | 1.08 (0.76–1.53) |
| Large town | 1.11 (0.71–1.75) | 1.16 (0.74–1.84) |
| Small town | 1.00 (reference) | 1.00 (reference) |
| Urban | 0.93 (0.69–1.25) | 0.93 (0.68–1.26) |
| Large town | 1.00 (reference) | 1.00 (reference) |
Social strain X RUCA p for interaction: 0.17; social support X RUCA p for interaction: >0.99.
Adjustments: age at screening, education, insurance, and trial membership.
CI, confidence interval; OR, odds ratio; RUCA, rural–urban commuting area.
Discussion
We describe, for the first time, the distribution of sociodemographic and clinical characteristics of participants of WHI participants, by the most popular and advantageous method used to define rural–urban residence (RUCA taxonomy), in cancer research.14,15 We found little evidence of an association between rural–urban residence and stage at breast cancer diagnosis, among postmenopausal women, regardless of social strain or social support.
Our null main effects are consistent with three studies that have examined the effect of rural–urban residence (using the RUCA classification system) on stage at breast cancer diagnosis. The earliest study by Celaya et al. used data from the predominantly rural state of New Hampshire (n = 5,966), and the authors noted limitations stemming from their use of cancer registry data from 1998 to 2005 and lack of information on mammography use for women with incident cancers.42 Similarly, Henry et al. used a 10-state cancer registry data set covering the period 2004–2006 (n = 161,619) and, similar to the study by Celaya, did not have mammography utilization data.1 Finally, Markossian and Hines used 1992–2007 data from urban (Atlanta) and rural Georgia National Cancer Institute's SEER program registries (n = 23,500); however, their results are not generalizable and are based on aggregate county-level data on rural–urban residence.43
Although the existing literature on the impact of rural–urban residence on stage at breast cancer diagnosis has not examined potential moderation by social support and social strain, these social factors have been shown to impact general as well as cancer-specific health behaviors21–27 and health outcomes.28–30 Given this literature, it is theoretically plausible that these factors might exacerbate (social strain) or buffer against (social support) the effects of rural residence on stage at breast cancer diagnosis. We add to the extant literature on this topic a by testing novel effect modification hypotheses, and using data from a large geographically diverse multiracial/ethnic cohort of postmenopausal women in which mammography utilization data are available. Future studies with more variability in rural–urban residence may be able to confirm our null interaction effects or extend our findings by providing support for the joint effects.
The following limitations of this study should be considered. First, we could not use the Tumor growth, lymph Nodes, and distant Metastasis (TNM) breast cancer staging system44 due to missing data on affected lymph nodes at the time of diagnosis. Given the multidecade follow-up of the WHI, and RUCA codes that are not fully comparable over time, we also had to limit exposure classification to data from Census year 2000 to optimize our sample size. We were, therefore, unable to exploit the longitudinal nature of the WHI or answer research questions about risk of breast cancer associated with urban versus rural residence. Certainly, rural geographies are not monolithic,45 and although the WHI study participants resided in a large number of states in the United States, the entire country was not represented in the sample, which may limit the generalizability of our findings. Given sample size constraints, we were also unable to assess potential effect modification by geographic region or race/ethnicity. Finally, 93% of the study population resided in urban areas, which may have limited our power to detect rural–urban differences in stage at breast cancer diagnosis.
This study also has several strengths. First, we have comprehensive sociodemographic and cancer risk factor assessment, and central adjudication of incident cancers. We also applied a popular rural–urban classification taxonomy to a large multiethnic well-characterized cohort of postmenopausal women from across 40 clinical centers in the United States, which will be useful for future research. Finally, we tested novel effect modification hypotheses, and sought to identify predictive and protective factors for the association between rural–urban residence and stage at breast cancer diagnosis.
In conclusion, the “rural penalty,” in terms of stage at breast cancer diagnosis, was not apparent among WHI participants, and there was little evidence of effect modification by social support or social strain. Future studies should consider effect modification by other modifiable factors to identify putative heterogeneity in the association between rural–urban residence and stage at breast cancer diagnosis. This information could be useful in informing future intervention studies focused on the cancer control continuum.
Acknowledgment
The WHI program is funded by the National Heart, Lung, and Blood Institute, National Institutes of Health, U.S. Department of Health and Human Services through contracts HHSN268201600018 C, HHSN268201600001 C, HHSN268201600002 C, HHSN268201600003 C, and HHSN268201600004 C.
Short List of WHI Investigators
Program Office: (National Heart, Lung, and Blood Institute, Bethesda, Maryland) Jacques Rossouw, Shari Ludlam, Joan McGowan, Leslie Ford, and Nancy Geller.
Clinical Coordinating Center: (Fred Hutchinson Cancer Research Center, Seattle, WA) Garnet Anderson, Ross Prentice, Andrea LaCroix, and Charles Kooperberg.
Investigators and Academic Centers: (Brigham and Women's Hospital, Harvard Medical School, Boston, MA) JoAnn E. Manson; (MedStar Health Research Institute/Howard University, Washington, DC) Barbara V. Howard; (Stanford Prevention Research Center, Stanford, CA) Marcia L. Stefanick; (The Ohio State University, Columbus, OH) Rebecca Jackson; (University of Arizona, Tucson/Phoenix, AZ) Cynthia A. Thomson; (University at Buffalo, Buffalo, NY) Jean Wactawski-Wende; (University of Florida, Gainesville/Jacksonville, FL) Marian Limacher; (University of Iowa, Iowa City/Davenport, IA) Jennifer Robinson; (University of Pittsburgh, Pittsburgh, PA) Lewis Kuller; (Wake Forest University School of Medicine, Winston-Salem, NC) Sally Shumaker; (University of Nevada, Reno, NV) Robert Brunner; and (University of Minnesota, Minneapolis, MN) Karen L. Margolis.
Women's Health Initiative Memory Study: (Wake Forest University School of Medicine, Winston-Salem, NC) Mark Espeland.
Author Disclosure Statement
No competing financial interests exist.
References
- 1. Henry KA, Boscoe FP, Johnson CJ, Goldberg DW, Sherman R, Cockburn M. Breast cancer stage at diagnosis: Is travel time important? J Community Health 2011;36:933–942 [DOI] [PubMed] [Google Scholar]
- 2. Nguyen-Pham S, Leung J, McLaughlin D. Disparities in breast cancer stage at diagnosis in urban and rural adult women: A systematic review and meta-analysis. Ann Epidemiol 2014;24:228–235 [DOI] [PubMed] [Google Scholar]
- 3. Akinyemiju TF, Genkinger JM, Farhat M, Wilson A, Gary-Webb TL, Tehranifar P. Residential environment and breast cancer incidence and mortality: A systematic review and meta-analysis. BMC Cancer 2015;15:191. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4. Smedley BD, Stith AY, Nelson AR. Unequal treatment. Confronting Racial and Ethnic Disparities in Health Care. Washington, DC: The National Academy Press, 2003:100. [PubMed] [Google Scholar]
- 5. Matthews KA. Health-related behaviors by Urban-Rural County Classification—United States, 2013. MMWR Surveill Summ 2017;66:1–8 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6. Zahnd WE, James AS, Jenkins WD, et al. Rural-Urban Differences in Cancer Incidence and Trends in the United States. Cancer Epidemiol Biomarkers Prev 2017. DOI: 10.1158/1055-9965.Epi-17-0430 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7. Tatalovich Z, Zhu L, Rolin A, Lewis DR, Harlan LC, Winn DM. Geographic disparities in late stage breast cancer incidence: Results from eight states in the United States. Int J Health Geogr 2015;14:31. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8. Cossman JS, James WL, Cosby AG, Cossman RE. Underlying causes of the emerging nonmetropolitan mortality penalty. Am J Public Health 2010;100:1417–1419 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9. Cosby AG, Neaves TT, Cossman RE, et al. Preliminary evidence for an emerging nonmetropolitan mortality penalty in the United States. Am J Public Health 2008;98:1470–1472 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10. McLafferty S, Wang F. Rural reversal? Rural-urban disparities in late-stage cancer risk in Illinois. Cancer 2009;115:2755–2764 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11. Fogleman AJ, Mueller GS, Jenkins WD. Does where you live play an important role in cancer incidence in the U.S.? Am J Cancer Res 2015;5:2314–2319 [PMC free article] [PubMed] [Google Scholar]
- 12. Morrill R, Cromartie J, Hart G. Metropolitan, urban, and rural commuting areas: Toward a better depiction of the United States settlement system. Urban Geogr 1999;20:727–748 [Google Scholar]
- 13. Hart LG, Larson EH, Lishner DM. Rural definitions for health policy and research. Am J Public Health 2005;95:1149–1155 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14. Meilleur A, Subramanian SV, Plascak JJ, Fisher JL, Paskett ED, Lamont EB. Rural residence and cancer outcomes in the United States: Issues and challenges. Cancer Epidemiol Biomarkers Prev 2013;22:1657–1667 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15. Pruitt SL, Eberth JM, Morris ES, Grinsfelder DB, Cuate EL. Rural-urban differences in late-stage breast cancer: Do associations differ by rural-urban classification system? Tex Public Health J 2015;67:19–27 [PMC free article] [PubMed] [Google Scholar]
- 16. Williams F, Jeanetta S, James AS. Geographical location and stage of breast cancer diagnosis: A systematic review of the literature. J Health Care Poor Underserved 2016;27:1357–1383 [DOI] [PubMed] [Google Scholar]
- 17. Hertzman C, Boyce T. How experience gets under the skin to create gradients in developmental health. Annu Rev Public Health 2010;31:329–347 323p following 347 [DOI] [PubMed] [Google Scholar]
- 18. VanderWeele TJ, Knol MJ. A tutorial on interaction. Epidemiol Methods 2014;3:33–72 [Google Scholar]
- 19. House JS, Kahn RL. Social support and health. In: Cohen S, S.L. S, eds. Measures and concepts of social support. New York: Academic Press, 1985:83–108 [Google Scholar]
- 20. Antonucci T, Kahn R, Akiyama H. Psychosocial factors and the response to cancer symptoms. In: Yancik R, Yates JW, eds. Cancer in the elderly: Approaches to early detection and treatment. New York: Springer Publishing Co, 1989:40–52 [Google Scholar]
- 21. Messina CR, Lane DS, Glanz K, et al. Relationship of social support and social burden to repeated breast cancer screening in the women's health initiative. Health Psychol 2004;23:582–594 [DOI] [PubMed] [Google Scholar]
- 22. Documet P, Bear TM, Flatt JD, Macia L, Trauth J, Ricci EM. The association of social support and education with breast and cervical cancer screening. Health Educ Behav 2015;42:55–64 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23. Cassileth BR, Walsh WP, Lusk EJ. Psychosocial correlates of cancer survival: A subsequent report 3 to 8 years after cancer diagnosis. J Clin Oncol 1988;6:1753–1759 [DOI] [PubMed] [Google Scholar]
- 24. Pinquart M, Duberstein PR. Associations of social networks with cancer mortality: A meta-analysis. Crit Rev Oncol Hematol 2010;75:122–137 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25. Garssen B. Psychological factors and cancer development: Evidence after 30 years of research. Clin Psychol Rev 2004;24:315–338 [DOI] [PubMed] [Google Scholar]
- 26. Kroenke CH, Michael Y, Tindle H, et al. Social networks, social support and burden in relationships, and mortality after breast cancer diagnosis. Breast Cancer Res Treat 2012;133:375–385 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27. Azuero A, Benz R, McNees P, Meneses K. Co-morbidity and predictors of health status in older rural breast cancer survivors. SpringerPlus 2014;3:102. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28. Birditt K, Antonucci TC. Life sustaining irritations? Relationship quality and mortality in the context of chronic illness. Soc Sci Med (1982) 2008;67:1291–1299 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29. Antonucci TC, Birditt KS, Webster NJ. Social relations and mortality: A more nuanced approach. J Health Psychol 2010;15:649–659 [DOI] [PubMed] [Google Scholar]
- 30. Kroenke CH, Kubzansky LD, Schernhammer ES, Holmes MD, Kawachi I. Social networks, social support, and survival after breast cancer diagnosis. J Clin Oncol 2006;24:1105–1111 [DOI] [PubMed] [Google Scholar]
- 31. Hays J, Hunt JR, Hubbell FA, et al. The Women's Health Initiative recruitment methods and results. Ann Epidemiol 2003;13(9 Suppl):S18–77 [DOI] [PubMed] [Google Scholar]
- 32. Design of the Women's Health Initiative clinical trial and observational study. The Women's Health Initiative Study Group. Control Clin Trials 1998;19:61–109 [DOI] [PubMed] [Google Scholar]
- 33. Matthews KA, Shumaker SA, Bowen DJ, et al. Women's health initiative. Why now? What is it? What's new? Am Psychol 1997;52:101–116 [DOI] [PubMed] [Google Scholar]
- 34. Gail MH, Brinton LA, Byar DP, et al. Projecting individualized probabilities of developing breast cancer for white females who are being examined annually. J Natl Cancer Inst 1989;81:1879–1886 [DOI] [PubMed] [Google Scholar]
- 35. Cromartie J. Rural-Urban Commuting Area Codes: Overview and Documentation. 2012; Available at: www.ers.usda.gov/data-products/rural-urban-commuting-area-codes.aspx Accessed September13, 2017
- 36. Whitsel EA, Quibrera PM, Smith RL, et al. Accuracy of commercial geocoding: Assessment and implications. Epidemiol Perspect Innov 2006;3:8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37. Whitsel EA, Rose KM, Wood JL, Henley AC, Liao D, Heiss G. Accuracy and repeatability of commercial geocoding. Am J Epidemiol 2004;160:1023–1029 [DOI] [PubMed] [Google Scholar]
- 38. Curb JD, McTiernan A, Heckbert SR, et al. Outcomes ascertainment and adjudication methods in the Women's Health Initiative. Ann Epidemiol 2003;13(9 Suppl):S122–S128 [DOI] [PubMed] [Google Scholar]
- 39. Sherbourne CD, Stewart AL. The MOS social support survey. Soc Sci Med (1982) 1991;32:705–714 [DOI] [PubMed] [Google Scholar]
- 40. Hardy SE, Concato J, Gill TM. Resilience of community-dwelling older persons. J Am Geriatr Soc 2004;52:257–262 [DOI] [PubMed] [Google Scholar]
- 41. Royston P, Altman DG. Regression using fractional polynomials of continuous covariates: Parsimonious parametric modelling. Appl Statist 1994:429–467 [Google Scholar]
- 42. Celaya MO, Berke EM, Onega TL, et al. Breast cancer stage at diagnosis and geographic access to mammography screening (New Hampshire, 1998–2004). Rural Remote Health 2010;10:1361. [PMC free article] [PubMed] [Google Scholar]
- 43. Markossian TW, Hines RB. Disparities in late stage diagnosis, treatment, and breast cancer-related death by race, age, and rural residence among women in Georgia. Women Health 2012;52:317–335 [DOI] [PubMed] [Google Scholar]
- 44. Edge SB, Compton CC. The American Joint Committee on Cancer: The 7th edition of the AJCC cancer staging manual and the future of TNM. Ann Surg Oncol 2010;17:1471–1474 [DOI] [PubMed] [Google Scholar]
- 45. James WL. All rural places are not created equal: Revisiting the rural mortality penalty in the United States. Am J Public Health 2014;104:2122–2129 [DOI] [PMC free article] [PubMed] [Google Scholar]