Impact of age, rurality and distance in predicting contralateral prophylactic mastectomy for breast cancer in a Midwestern state: a population-based study

Iowa is among several rural Midwestern states with the highest proportions of contralateral prophylactic mastectomy (CPM) in women < 45 years of age. We evaluated the role of rurality and travel distance in these surgical patterns. Women with unilateral breast cancer (2007–2017) were identified using Iowa Cancer Registry records. Patients and treating hospitals were classified as metro, nonmetro, and rural based on Rural–Urban Continuum Codes. Differences in patient, tumor, and treatment characteristics and median travel distance (MTD) were compared. Characteristics associated with CPM were evaluated with multivariate logistic regression. 22,158 women were identified: 57% metro, 26% nonmetro and 18% rural. Young rural women had the highest proportion of CPM (52%, 39% and 40% for rural, metro, nonmetro women < 40 years). Half of all rural women had surgery at metro hospitals; these women had the longest MTD (62 miles). Among all women treated at metro hospitals, rural women had the highest proportion of CPM (17% rural vs 14% metro/nonmetro, p = 0.007). On multivariate analysis, traveling ≥ 50 miles (ORs 1.43–2.34) and rural residence (OR = 1.29) were independently predictive of CPM. Other risk factors were young age (< 40 years: OR = 7.28, 95% CI 5.97–8.88) and surgery at a metro hospital that offers reconstruction (OR = 2.30, 95% CI 1.65-3.21) and is not NCI-designated (OR = 2.34, 95% CI 1.92–2.86). There is an unexpectedly high proportion of CPM in young rural women in Iowa, and travel distance and availability of reconstructive services likely influence decision-making. Improving access to multidisciplinary care in rural states may help optimize decision-making.


Introduction
Contralateral prophylactic mastectomy (CPM) for breast cancer has become increasingly common [1] despite the lack of survival benefit in average risk women [2], and according to multiple guidelines, should be discouraged [3][4][5]. CPM is largely driven by patient demand, with higher income, education and private insurance associated with greater likelihood of CPM [1,[6][7][8][9]. These determinants are typically attributed to women in metropolitan areas [10]. The Midwest was recently noted to have a high proportion of CPM, with Iowa having the second highest proportion nationally (44.9%) among women aged 20-44 during 2010-2012 [9,11]. Like Iowa, most of the Midwest is agricultural with over half of Midwestern counties considered rural [12]. Therefore, the high proportion of Midwest women undergoing CPM is unexpected and deserve further investigation.

3
A study utilizing the National Cancer Database (NCDB) found that women who travel further for surgery are twice as likely to receive CPM [9]. The role of rurality was not examined. Since the NCDB only captures patients treated at Commission on Cancer (CoC)-accredited centers, which are rare in rural communities, it is possible that greater travel distance may be related to rural residence for some patients [13]. We postulate that travel distance is a factor contributing to the high rate of CPM in Iowa.
Rurality is associated with disparities in breast cancer throughout its spectrum of care [13][14][15][16]. Rural women are more likely to receive mastectomy and less likely to receive reconstruction, radiation, or chemotherapy [16][17][18]. The impact of rurality on breast cancer disparities and specifically surgical treatment could reflect a number of factors, including travel distance, socioeconomic and educational status, and access to multidisciplinary cancer care [19][20][21]. A recent study using the NCDB did not find a higher proportion of CPM in rural women [10]. However, only 1.7% of the patients captured by this database were from a rural residence, highlighting the fact that a high proportion of rural women are not treated in CoC-accredited institutions. We hypothesized that rural women are more likely to receive CPM, perhaps because of lack of access to multidisciplinary care and expertise in the treatment of breast cancer.
In the same study showing high rates in Iowa, three adjoining upper Midwestern states (South Dakota, Missouri, Nebraska) and 4 other states (Colorado, Tennessee, Maine, Montana) had proportions of CPM exceeding 40% in the 20-44 age group [11]. The percent population living in rural areas in these states ranged from 14 to 61%, with a median of 35% [22]. We postulated that the treatment patterns identified in Iowa may also be present in other states and could help explain this geographical variation in CPM. Iowa has a high-quality state cancer registry that includes facilities of all types and sizes, unlike the NCDB that only includes CoC-accredited hospitals. We used these data to examine the complex relationship between age, residence, travel distance, treating institution and surgery type.

Data source and study population
A secondary data analysis of Iowa breast cancer patients was performed. Cases were extracted from the Iowa Cancer Registry Surveillance, Epidemiology, and End Results (SEER) Data Management System (SEER*DMS) using the following inclusion criteria: Iowa women with unilateral, first primary, stage 0-III breast cancer (AJCC 6th edition) diagnosed between 2007 and 2017 (n = 23,210). Patients were excluded if they had synchronous breast cancers (n = 699) or lobular carcinoma in situ (ICD-O-3 histology code 8520/2; n = 353). This study was granted human subject exemption status by the University of Iowa Institutional Review Board.

Study variables
Patient demographics in the dataset included age at diagnosis, race, marital status, insurance, year of diagnosis, and patient residence at diagnosis. Patients' county of residence and treating hospital locations were categorized based on the US Department of Agriculture's Rural-Urban Continuum Codes (RUCC), which categorize counties based on metropolitan areas within the county and the degree of urbanization in adjacent counties ( Fig. 1) [23]. Iowa metropolitan (metro) counties included RUCC 1-3, nonmetro 4-6, and rural 7-9, as used elsewhere [24]. Nonmetro counties contain cities with ≥ 20,000 people and/or are adjacent to counties containing metro areas, whereas rural counties are not adjacent to counties containing metro areas and/or only have cities with < 2500 people. A geographic analysis of our rural category revealed that patients in these counties, on average, would have to travel ≥ 50 miles to the closest CoC-accredited hospital (data not shown).
County-level patient demographics of poverty level and percentage of county with bachelor's degree or higher from the 2016 American Community Survey were included [22]. Tumor characteristics included stage, lymph node status, histology, and breast cancer subtype.  [25].
Characteristics of the surgery hospital were linked to patients and included rurality, CoC accreditation status and NCI cancer center designation, bed size, availability of reconstruction services (derived by identifying any facility in the cohort that had provided reconstruction surgery between 2007 and 2017), and travel distance to surgery facility, which was calculated with Great Circle Distance in Arc-GIS utilizing latitudes and longitudes of hospital location and patient residence at the time of cancer diagnosis. There were 1585 patients (877 metro, 437 nonmetro, and 271 rural patients) who did not match to a surgery hospital and hence are missing from the surgery hospital variables. This usually occurred when surgery was performed at a nonhospital facility, such as an ambulatory surgery center. Patients with an Iowa address who were treated outside the state were also captured (10.3% of total).

Statistical analysis
Rates were calculated using SEER*Stat, version 8.3.5 [26]. Differences in patient, tumor, and treatment characteristics were compared between patient and hospital rurality groups using Chi-square tests. A multivariable logistic regression was used to determine the characteristics associated with CPM compared to other surgery types. All variables were included in a stepwise selection model with an exit significance level of 0.2 and the significant variables were included in the final model. All variables listed above were considered for inclusion into the model. ANOVA analysis was performed to compare median travel distance between patient rurality groups. Analyses were conducted in SEER*Stat, SAS 9.4 (SAS Institute, Cary, NC) and ArcGIS 10.5.1 (ESRI, Redlands, CA). p values < 0.05 were considered statistically significant. Patients with missing relevant variables were excluded from the models.

Results
The study cohort consisted of 22,158 women (Table 1). Over half (57%) of the women lived in metro, 26% in nonmetro and 18% in rural counties. Regardless of residence, twothirds were married, 90% had private or Medicare insurance, and nearly two-thirds had early stage disease (stage 0 or 1). Overall, 60% had lumpectomy, 25% had UM, and 13% had CPM.
The overall proportion of CPM in Iowa was consistently higher compared to SEER-18 registries throughout the interval from 2007 to 2014 (Fig. 2). When examined by patient residence, the proportion for nonmetro women follows the national trend. For metro and rural women, however, the proportions were persistently higher between 2007 and 2014, peaking in 2013, with a trend toward convergence with national rates after 2014. The proportion of CPM was significantly higher for both rural (13%) and metro (14%) women compared to nonmetro women (11%). Young rural women had the highest proportion of CPM use, with 52% of rural women age < 40 years undergoing CPM compared to 40% and 39% of metro and nonmetro women, respectively.

Demographic, tumor and treatment characteristics by patient residence
Women residing in metro counties were younger, and their counties were more likely to have low poverty and more highly educated residents than nonmetro and rural counties ( Table 1). There were no clinically significant differences in stage at presentation, receptor subtype, or use of neoadjuvant chemotherapy by patient residence (results not shown).

Distance traveled by patient residence and type of surgery
Rural women traveled farther than metro and nonmetro women for surgery, regardless of the type of operation

Treatment patterns of rural women by location of surgical care
Almost one-third (29%) of rural women underwent breast cancer surgery at a rural hospital from 2007 to 2017, with the proportion decreasing over the study period (Table 2).
In contrast, nearly all metro (99%) women had surgery at metro hospitals, whereas 40% of nonmetro women had surgery at nonmetro hospitals. Rural women who had surgery at rural hospitals were older, more likely to be unmarried and have Medicare as primary insurance compared to other rural women. They were also much more likely to have UM without *p values < 0.05 are in bold reconstruction and traveled the shortest distance for surgery (Table 2). Overall, 50% of rural women were treated at metro hospitals. Their stage distribution was similar to those treated at rural hospitals. In contrast, rural women treated at nonmetro hospitals (21%) tended to be diagnosed at a lower stage than those treated at other hospitals and had the highest rate of lumpectomy and the lowest rate of UM.
The rate of CPM in rural women varied widely by hospital rurality (17% metro, 14% nonmetro, 8% rural) as did the rate of reconstruction post mastectomy (31% metro, 23% nonmetro, 3% rural). The highest rate of reconstruction in rural women (57%) was seen in those undergoing CPM at metro hospitals (Table 2).

Interaction between hospital rurality and patient residence
To evaluate whether differences in surgical treatment for rural patients were related solely to the hospital where they had surgery, patients treated in the same hospital location were stratified by area of residence (Table 3, Appendix 1). Compared with nonrural women treated at metro or nonmetro hospitals, rural women were more likely to be married, have private insurance, and less likely to come from a county with high poverty. They were also more likely to have surgery at a larger, CoC-accredited or NCI-designated institution that offered reconstruction.
The highest rate of CPM in both metro (Table 3) and nonmetro (Appendix 1) hospitals was seen in their rural patients. Rural women also had the highest rate of reconstruction for both UM and CPM at these hospitals.

Multivariate analysis of factors affecting CPM
A multivariate analysis was performed to evaluate the likelihood of CPM based on patient residence, hospital location, and travel distance (Table 4). Using metro women who traveled < 50 miles as a reference, we found that travel distance ≥ 50 miles predicted CPM regardless of residence. The strongest association was seen in metro patients, who were 2.3 times more likely to have CPM if traveling farther. However, rural patients who traveled < 50 miles were also significantly more likely to have CPM. The largest risk factor for CPM was younger age at diagnosis, especially for those less than age 40 vs. 60 +. CPM was also more likely for patients with stage 3 disease and invasive lobular cancer, and less likely for those with Medicare and from counties with higher poverty. Surgery hospital characteristics associated with CPM included metro location, reconstruction capability, and absence of NCI designation. CoC accreditation was associated with a lower likelihood of CPM (OR = 0.91) in univariate analysis, but this was not significant in the multivariate model. Distance traveled for surgery (with/without reconstruction) for stage 0-III breast cancer by patient residence. *Circle, plus, and cross represent mean distance; Whiskers extend to 1.5 (IQR) above/ below Q1/Q3

Discussion
Iowa was reported to have the second highest proportion of CPM among women under 45 years of age between 2004 and 2012, with a high proportion of CPM in older women as well [11]. We found that younger age was the strongest predictor of CPM for women across Iowa. Although the proportion of metro and nonmetro women under 40 years undergoing CPM was high (40 and 39%), the highest proportion was seen in rural women in this age group (52%). Age is consistently associated with CPM in the literature [1,[6][7][8][9]; however, the association of young age and rural residence with CPM is a novel finding. Examining the factors related to this finding may shed some light on the determinants of CPM in Iowa and other rural states. Initially, we had hypothesized that rural women were more likely to receive CPM at rural hospitals with limited expertise in breast cancer treatment, but found instead that many rural women traveled long distances to seek care at large metro hospitals, and that those women were the most likely to undergo CPM. About one-fifth of the population of Iowa reside in areas with a population density of fewer than 20,000 people [22,23]. Rural counties had no CoC-accredited hospitals, and breast reconstruction was rarely performed. Many Iowa rural counties are adjacent to more populous nonmetro areas (Fig. 1); however, most nonmetro hospitals also did not perform breast reconstruction or have CoC accreditation. It is therefore not surprising that only 20% of rural women with breast cancer were treated at rural hospitals from 2015-2017, and that half of nonmetro women had surgery at metro hospitals instead.
Receipt of CPM is strongly associated with white race, private insurance and higher education and socioeconomic status [1,[6][7][8][9]. Although Iowa is a rural state, its rates of poverty, high school noncompletion and uninsured residents are well below national averages [22,27]. Indeed, rural counties in Iowa have better educational achievement, insurance status and lower poverty when compared to rural counties in other states [27]. It is notable that most of the other rural states with a high proportion of CPM also have comparatively better education, income and insurance in their rural counties [22,27]. In this study, we were unable to examine the relationship of education with CPM because it was not collected at the patient level, and the strong collinearity of county-level education with rurality precluded inclusion into the model. Our multivariate analysis confirmed the association of CPM with insurance and socioeconomic status. For metro women, the large proportion of white, college-educated women with private insurance may in part explain the higher-than-average rates in this population. Although a larger proportion of rural women in Iowa have characteristics associated with CPM, it would not explain why the proportion of CPM in rural women exceeds that of the metro group.
The role of travel distance as a determinant of CPM in Iowa is complex and appears to be strongly influenced by the patient's residence. The largest effect size was seen in metro women, with those who traveled ≥ 50 miles twice as likely to receive CPM compared to those who did not. However, this accounted for only 4% of metro women, most of whom had short travel distances with little variation between surgery types. In contrast, rural patients were more likely to undergo CPM regardless of travel distance. This may be because even those rural women who traveled < 50 miles had a longer median travel distance than metro and nonmetro patients in the < 50-mile category, as they were often driving to hospitals in a different community (data not shown).
Rural Iowans have demonstrated a willingness to travel farther to seek a higher level of cancer care [28]. Travel distance likely does not reflect the closest institution providing surgical care for breast cancer, even for rural patients. Rural women tended to seek out larger metro hospitals and those that were CoC-accredited, or NCI-designated, including those in adjacent states. The greatest travel distances were noted in both CPM and UM patients with reconstruction, and the desire for reconstruction may explain some of these patterns [29]. However, only half of all CPM patients ultimately received reconstruction, and the relationship between travel distance and CPM persisted even after adjusting for reconstruction, as it did in the NCDB study [9]. Rural women had a higher proportion of CPM use than other women of similar ages treated at both metro and nonmetro hospitals, implying an intrinsic preference for CPM in this group. CPM is associated with a more active decision-making role for surgery [30] and greater likelihood of seeking a second opinion [31], and thus perhaps a willingness to travel for care. Women facing a high travel burden for their desired level of care may perceive CPM as one way to minimize future need for healthcare utilization [18,32].
Our findings suggest that the high rate of CPM in young rural women may be in part due to patient preference. While this would not be considered a disparity, it highlights how differences in access to care could impact patient preferences and decisions. Both travel distance and low plastic surgeon density have been shown to limit receipt of reconstruction [5,29,33] and may impact access for rural patients in particular. The characteristics of rural women who traveled farther for surgery suggests that they had more resources than those who did not. The emotional and socioeconomic burden of travel varies with patient resources, and rural patients with fewer means are likely to have less access to multidisciplinary care and breast reconstruction. Although young women with CPM had a high rate of reconstruction, the overall rate in Iowa was low compared to other states [33,34]. Regional differences in provider attitudes and practices have also been shown to influence geographic differences in surgical care [35,36]. This likely plays some role in the high proportion of CPM in Iowa. Age was strongly associated with CPM, even though age alone is not considered an indication for CPM [3,5,37]. Both lobular carcinoma and stage 3 disease were associated with CPM, a finding also reported in SEER [1]. Since lobular carcinoma is not an indication for CPM and prophylactic surgery would provide even less benefit than usual in higher stage disease [4], it is unclear whether these patients are being recommended CPM inappropriately or are being incompletely counseled about risks and benefits. These findings support a role for provider education in the state. While rural women who traveled to metro hospitals were more likely to select NCI-designated facilities in Iowa and elsewhere, surgery at an NCI-designated hospital conferred half the risk of receiving CPM compared to other hospitals. We also found CoC accreditation was associated with a lower likelihood of CPM, although this finding did not meet statistical significance on multivariate analysis. In Iowa, CoC accreditation has been shown to correlate with more comprehensive services [28], and such facilities may have greater awareness of consensus statements published in 2016 and 2017 cautioning against overuse of CPM [3][4][5]38]. The proportion of CPM in Iowa has been falling in all populations since 2013 and in 2015 was close to or below national rates, which also appear to be leveling off. These changes may reflect an increasing knowledge of the trends in CPM use and issues surrounding it.
Our study has several limitations. The Iowa Cancer Registry does not capture everything that can impact hospital selection or surgical decision-making, such as genetic mutation, family history, additional testing or number of visits [2,30]. Iowa has a predominantly White population with few rural non-Whites, so we were unable to examine the interaction of race and rurality. In addition, because our analysis was restricted to one state, generalization of our findings to other similar states may not be feasible. However, restricting analysis to one state's cancer registry, combined with our familiarity with local communities, ensures that our definitions of rurality reflect clinically significant differences in populations and resources. This is especially relevant considering the existing variability in how rurality is defined [39].

Conclusions
Our findings suggest that higher travel burden increases the likelihood of CPM and disproportionately impacts rural women with breast cancer. Travel distance may be dictated not just by the distance to the nearest hospital but rather to the nearest hospital offering multidisciplinary care and breast reconstruction. In Iowa, many rural women are willing to travel farther to access this kind of care, with the potential for significant socioeconomic burden and less than optimal care. Counseling for patients desiring CPM should take into consideration rurality, access to local healthcare, expectations regarding subsequent visits for surveillance and completion of reconstruction, and measures to make follow-up care less burdensome. Improving navigation for rural patients, supporting provider education and quality accreditation for hospitals may increase access to multidisciplinary care and reduce travel burden as a determinant of CPM.