Living in a Disadvantaged NYC Neighborhood Is Associated With Poorly Differentiated Breast Cancer and Shorter Breast Cancer-specific Survival

Abstract

Objective:

To evaluate the association between living in disadvantaged neighborhoods in New York City (NYC) with tumor grade, a clinical proxy for proliferation and tumor aggressiveness, and breast cancer-specific survival (BCSS).

Background:

Neighborhood disadvantage (ND) is associated with shorter BCSS, independent of individual-level, tumor, and treatment characteristics, highlighting unmeasured factors associated with this survival disparity.

Methods:

Women with stage I to III breast cancer (BCa) living in NYC treated at Memorial Sloan Kettering Cancer Center from 2013 to 2024 were included. ND was stratified using the Area Deprivation Index (ADI). The median ADI for the cohort was 3 and was used as the cutoff between neighborhood advantage (NA, ADI 1–3) and ND (ADI 4–10). Multivariable logistic regression and Cox proportional hazards modeling, controlling for individual, tumor, and treatment factors, were used to determine the association between ND and tumor grade and BCSS, respectively.

Results:

Five thousand four hundred fifty-two women with BCa were included. Three thousand four hundred seventy-nine (64%) lived in NA and 1973 (36%) in ND. On multivariable analysis, ND had higher odds of poorly versus well/moderately differentiated tumors (adjusted odds ratio: 1.23, CI: 1.03–1.48), independent of age, race/ethnicity, insurance, body mass index, smoking/alcohol, stage, and subtype. ND was also associated with shorter BCSS (adjusted hazard ratio: 1.56, CI: 1.05–2.38).

Conclusions:

Women living in ND in NYC were more likely to present with poorly differentiated tumors and have shorter BCSS. These findings merit further inquiry and lay the foundation for future translational studies to externally validate the mechanisms by which ND “gets under the skin” to impact aggressive BCa tumor biology, and ultimately survival.

Despite significant advances in breast cancer screening, diagnosis, and treatment over the last 20 years, disparities in breast cancer-specific survival persist across the United States. Most studies have attributed these disparities to differences in access to care barriers leading to later stage at diagnosis and lower rates of guideline-concordant treatment completion.^1–3 In addition, race-based disparities in breast cancer subtype, such as higher odds of triple-negative breast cancer, a more aggressive subtype, are an important source of breast cancer survival disparities.⁴

Moreover, recent studies have identified that place-based factors, specifically neighborhood disadvantage, are also associated with shorter breast cancer-specific survival. A recent study among a majority-minority South Florida population identified that independent of individual-level socioeconomic variables (eg, race and ethnicity, insurance status), health habits (eg, smoking, alcohol), comorbidities (eg, diabetes), tumor characteristics (eg, stage and subtype), and receipt of NCCN guideline-concordant care, patients living in disadvantaged neighborhoods had shorter breast cancer-specific survival when compared with their counterparts in more advantaged neighborhoods.⁵ Similarly, data from the Surveillance, Epidemiology, and End Results (SEER) program further showed that disparities in breast cancer-specific survival by neighborhood disadvantage persist on a national scale.⁶ These persistent survival disparities even after controlling for multilevel factors, including access to care barriers, suggest that neighborhood disadvantage “gets under the skin” to influence more aggressive breast cancer biology.

Recent translational studies in a South Florida population have identified that patients from disadvantaged neighborhoods are subjected to social adversity-related stress, which activates the “fight or flight” sympathetic nervous system (SNS).⁷ This stress-related SNS response from living in neighborhood disadvantage is independently associated with upregulation of transcription factors associated with inflammation and cell proliferation, and downregulation of transcription factors related to protective cellular responses and programmed cell death.^7–10 These alterations in tumor biology are associated with more aggressive disease, reflected by higher OncotypeDX scores and shorter breast cancer recurrence-free survival.¹¹

Given that tumor grade is a proxy for cell proliferation and tumor aggressiveness, the objective of this study was to evaluate the association between disadvantaged and advantaged New York City (NYC) neighborhoods, tumor grade, and breast cancer-specific survival. By relating tumor biology to neighborhood conditions, we aim to expand the generalizability of a growing body of research showing a relationship between neighborhood disadvantage, more aggressive tumor biology, and worse breast cancer-specific survival.

METHODS

Study Population

Memorial Sloan Kettering Cancer Center (MSKCC), a National Cancer Institute-designated Comprehensive Cancer Center provides breast cancer care to a diverse patient population in NYC. After institutional review board approval, patients who underwent surgery at MSKCC from 2013 to 2024 were identified in a prospectively maintained database. Neighborhood disadvantage was stratified using the Area Deprivation Index, a widely used measure that has been validated to the census block-group level and incorporates 17 income, education, employment, and housing quality metrics to give a score from 1 to 10, with 10 being the most disadvantaged neighborhoods.¹² In our study, Area Deprivation Index (ADI) was determined based on address at time of cancer diagnosis and calculated using the UW-Madison Neighborhood Atlas. (Fig. 1) As there is no standard definition of disadvantaged versus advantaged neighborhood by ADI, we used the median ADI for our cohort, which was 3. This stratification is consistent with our prior work where ADI 1 to 3 is defined as advantaged neighborhoods (NA, ADI 1–3), and ADI 4 to 10 as disadvantaged neighborhoods (ND, ADI 4–10).

FIGURE 1.

Neighborhood area deprivation index and tumor differentiation by block group in New York City. A, Manhattan. B, Queens. C, Brooklyn. D, Staten Island. E, The Bronx. Census block groups categorized as disadvantaged (ADI 1–3) are shaded in blue and those with poorly differentiated tumors are shared in maroon. The overlap between disadvantaged neighborhoods and poorly differentiated tumors are shaded in brown.

Covariates

Covariates were chosen based on subject matter and clinical knowledge. They included sociodemographic factors (age at diagnosis <50, 50–69, >70), race (White, Black, Asian, Unknown), Hispanic ethnicity, body mass index (normal, overweight, obese), insurance status (private or Medicare, Medicaid), alcohol use (never, socially, frequently), and smoking status (yes/no). Captured tumor characteristics included clinical stage (I, II, III), tumor grade (well, moderately, or poorly differentiated), tumor subtype [estrogen receptor (ER) and human epidermal growth factor receptor 2 (HER2) status], and receipt of adjuvant chemotherapy, radiation, and endocrine therapy.

Outcomes

The primary outcome was tumor grade. Secondary outcomes included breast cancer-specific survival defined as time from initial diagnosis to death from locoregional or distant disease. Cause of death was determined through review of the medical record.

Statistical Analysis

All statistical analysis was done using R 4.5. Categorical variables were compared using Chi-squared and continuous variables were compared using Wilcoxon rank sum test. In order to assess the factors associated with higher tumor grade and with breast cancer specific survival, a logistic regression model and a Cox regression model were used respectively. All covariates that were significant in the univariate analyses at a permissible type I error rate of 0.05 were then included in the multivariable analyses.

RESULTS

Patient Sociodemographic, Comorbidity, Health Behavior, and Access to Care Measures by Neighborhood Disadvantage

Five thousand four hundred fifty-two women with stage I to III breast cancer were included in the study, with 3479 (64%) living in advantaged neighborhoods (ADI 1–3), and 1310 (36%) living in disadvantaged neighborhoods (ADI 4–10). Sociodemographic and clinical characteristics of the population stratified by ADI are featured in Table 1. Women living in disadvantaged neighborhoods in NYC were more likely to be younger than 70 years old, with 53.6% aged 50 to 69 compared with 47.5% in advantaged areas (P<0.001). A higher proportion were Black (30.8% vs 11.4%) and they were more likely to identify as Hispanic (15.6% vs 9.3%, P<0.001). Significantly more women in disadvantaged neighborhoods had Medicaid insurance (23.0% vs 14.4%), and higher percentages of obesity (40.8% vs 28.9%, P<0.001). Alcohol use was less common in disadvantaged neighborhoods, with a higher proportion reporting never drinking (65.8% vs 57.9%, P<0.001).

TABLE 1.

Sociodemographic Characteristics for Women Living in Advantaged Versus Disadvantaged New York City Neighborhoods

Characteristic	Total	Advantaged neighborhood (ADI 1–3), N (%)	Disadvantaged neighborhood (ADI 4–10), N (%)	P
N	5452	3479 (63.8)	1973 (36.2)
Age
<49	1656	1094 (31.4)	562 (28.5)	<0.001
50–69	2708	1651 (47.5)	1057 (53.6)
>70	1088	734 (21.1)	354 (17.9)
Race
White	3128	2279 (65.5)	849 (43.0)	<0.001
Black	1004	397 (11.4)	607 (30.8)
Asian	670	444 (12.7)	226 (11.5)
Other	651	360 (10.4)	291 (14.8)
Hispanic ethnicity
No	4822	3157 (90.7)	1665 (84.4)	<0.001
Yes	630	322 (9.3)	308 (15.6)
Insurance
Private or Medicare	4495	2977 (85.6)	1518 (77.0)	<0.001
Medicaid	956	502 (14.4)	454 (23.0)
BMI
Normal	1950	1440 (41.4)	510 (25.9)	<0.001
Overweight	1693	1035 (29.8)	658 (33.4)
Obese	1809	1004 (28.9)	805 (40.8)
Alcohol use
Never	3313	2015 (57.9)	1298 (65.8)	<0.001
Socially	1968	1327 (38.1)	641 (32.5)
Frequently	171	137 (3.9)	34 (1.7)
Smoking
Yes	1633	1082 (31.1)	551 (27.9)	0.014
No	3819	2397 (68.9)	1422 (72.1)

Tumor Characteristics by Neighborhood Disadvantage

Women living in disadvantaged neighborhoods had a higher percentage of stage II/III breast cancer (42.0% vs 37.4%, P 0.002) and more poorly differentiated tumors (43.3% vs 35.4%, P<0.001) (Table 2). As shown in Fig. 1, we can visualize disadvantaged neighborhoods, particularly in Harlem and areas in Queens and the Bronx, that have higher percentages of poorly differentiated tumors. Tumor subtype distribution also varied by neighborhood disadvantage, with a greater percentage of triple-negative breast cancer (TNBC, ER−/HER2−) at presentation in patients living in disadvantaged neighborhoods (17.4% vs 12.9%, P<0.001). Although chemotherapy and radiation rates were similar across both groups, disadvantaged women had lower rates of endocrine therapy use (76.4% vs 81.9%, P<0.001), consistent with subtype distribution.

TABLE 2.

Tumor and Treatment Characteristics for Women Living in Advantaged Versus Disadvantaged Neighborhoods in New York City

Characteristic	Total	Advantaged neighborhood (ADI 1–3), N (%)	Disadvantaged neighborhood (ADI 4–10), N (%)	P
N	5452	3479 (63.8)	1973 (36.2)
Clinical stage
1	3164	2801 (62.6)	1083 (58.0)	0.002
2	1709	1036 (31.2)	673 (36.0)
3	319	206 (6.2)	113 (6.1)
Tumor differentiation
Well or moderately	2297	1568 (64.6)	729 (56.7)	<0.001
Poorly	1415	858 (35.4)	557 (43.3)
Tumor Subtype
ER+/HER2−	3411	2189 (73.3)	1213 (69.5)	<0.001
ER+/HER2+	402	263 (8.8)	139 (8.0)
ER−/HER2+	241	150 (5.0)	91 (5.2)
ER−/HER2−	689	386 (12.9)	303 (17.4)
Chemotherapy
No	2906	1885 (54.2)	1021 (51.8)	0.083
Yes	2546	1594 (45.8)	952 (48.3)
Radiation
No	1815	1182 (34.0)	633 (32.1)	0.15
Yes	3637	2297 (66.0)	1340 (67.9)
Endocrine therapy
No	1095	630 (18.1)	465 (23.6)	<0.001
Yes	4357	2849 (81.9)	1508 (76.4)

Multivariable Analysis of Factors Associated With Tumor Grade

On univariable analysis by tumor grade (well and moderately differentiated vs poorly differentiated disease), we found that patients with poorly differentiated disease were more likely to live in disadvantaged neighborhoods [odds ratio (OR): 1.40, 95% CI: 1.22–1.60, P <0.001], identify as non-white (P<0.001), have Medicaid insurance (OR: 1.27, 95% CI: 1.06–1.51, P=0.008), and be obese (OR: 1.21, 95% CI: 1.03–1.42, P=0.03) Higher clinical stage at presentation (Stage 3; OR: 3.44, 95% CI: 2.53–4.69, P<0.001) and tumor subtypes other than ER+/HER2− (TNBC; OR: 20.0, 95% CI: 15.0–27.2, P<0.0010) were also associated with poorly differentiated disease (Table 3).

TABLE 3.

UVA/MVA of Sociodemographic, Comorbidity, and Clinical Characteristics Associated With Poorly Versus Well/Moderately Differentiated Tumors

Characteristic	UVA—OR (95% CI)	P	MVA—OR (95% CI)	P
ADI
1_3	Reference	<0.001	Reference	0.022
4–10	1.40 (1.22–1.60)		1.23 (1.03–1.48)
Age
<49	Reference	<0.001	Reference	0.10
50–69	0.75 (0.64–0.87)		0.80 (0.66–0.98)
>70	0.67 (0.56–0.81)		0.87 (0.68–1.11)
Race
White	Reference	<0.001	Reference	0.027
Black	2.26 (1.89–2.71)		1.46 (1.14–1.87)
Asian	1.34 (1.09–1.65)		1.03 (0.79–1.35)
Other	1.43 (1.16–1.77)		1.07 (0.81–1.40)
Hispanic ethnicity
No	Reference	0.3
Yes	1.11 (0.89–1.38)
Insurance
Private or Medicare	Reference	0.008	Reference	0.6
Medicaid	1.27 (1.06–1.51)		1.07 (0.85–1.34)
BMI
Normal	Reference	0.03	Reference	0.5
Overweight	1.01 (0.86–1.19)		0.92 (0.74–1.13)
Obese	1.21 (1.03–1.42)		1.03 (0.83–1.28)
Alcohol use
Never	Reference	0.015	Reference	0.6
Socially	0.99 (0.86–1.13)		0.99 (0.82–1.18)
Frequently	0.58 (0.39–0.84)		0.80 (0.49–1.35)
Smoking
No	Reference	<0.001	Reference	0.014
Yes	1.32 (1.13–1.51)		1.26 (1.02–1.54)
Clinical stage
1	Reference	<0.001	Reference	<0.001
2	2.32 (2.01–2.69)		1.76 (1.47–2.12)
3	3.44 (2.53–4.69)		2.82 (1.9–4.20)
Tumor subtype
ER+/HER2−	Reference	<0.001	Reference	<0.001
ER+/HER2+	4.04 (3.13–5.24)		3.49 (2.65–4.61)
ER−/HER2+	14.2 (9.05–23.1)		12.5 (7.78–21.1)
ER−/HER2−	20.0 (15.0–27.2)		17.6 (13.0–24.4)

BMI indicates body mass index; MVA, multivariable; UVA, univariable.

These differences by tumor grade persisted on multivariable analysis after controlling for age, race, insurance, body mass index, alcohol use, smoking history, stage at presentation, and tumor subtype. Women with poorly differentiated disease were more likely to live in disadvantaged neighborhoods (OR: 1.23, 95% CI: 1.03–1.48, P=0.022), identify as Black (OR: 1.46, 95% CI: 1.14–1.87, P=0.027), present with more advanced disease (Stage 3; OR: 2.82, 95% CI: 1.90–4.20, P<0.001), and have TNBC (OR: 17.6, 95% CI: 13.0–24.4, P<0.001) (Table 3).

Cox Proportional Hazards Modeling for Breast Cancer-specific Survival

The median follow-up time for our study was 4.7 years. On multilevel Cox proportional hazards modeling for breast cancer-specific survival, we found that women in disadvantaged neighborhoods had a worse breast cancer-specific survival when compared with their more advantaged counterparts [hazard ratio (HR): 1.56, 95% CI: 1.02–2.38, P=0.04], independent of race, insurance, clinical stage, tumor differentiation, tumor subtype, and receipt of chemotherapy, radiation, and endocrine therapy (Table 4). In addition, we identified that having Medicaid insurance (HR: 1.91, 95% CI: 1.21–3.01, P=0.01), stage II or III disease at diagnosis (II—HR: 2.16, 95% CI: 1.36–3.44, P<0.001; III—HR: 5.88, 95% CI: 3.24–10.7, P<0.001), poorly differentiated disease (HR: 2.03, 95% CI: 1.22–3.37, P=0.006), and triple-negative disease (HR: 2.53, 95% CI: 1.26–5.08, P<0.001) were all associated with increased breast cancer-specific mortality.

TABLE 4.

Sociodemographic, Comorbidity, and Clinical Characteristics Associated With Breast Cancer-specific Survival

Characteristic	UVA—HR (95% CI)	P	MVA—HR (95% CI)	P
ADI
1–3	Reference	0.033	Reference	0.040
4–10	1.43 (1.03–1.99)		1.56 (1.02–2.38)
Age
<49	Reference	0.3
50–69	1.24 (0.84–1.82)
>70	1.42 (0.89–2.28)
Race
White	Reference	0.025	Reference	0.5
Black	1.80 (1.22–2.65)		1.07 (0.63–1.82)
Asian	1.05 (0.6–1.83)		0.98 (0.48–2.02)
Other	1.45 (0.89–2.36)		1.56 (0.87–2.79)
Hispanic ethnicity
No	Reference	0.14
Yes	1.43 (0.90–2.27)
Insurance
Private or Medicare	Reference	<0.001	Reference	0.008
Medicaid	2.10 (1.47–3.00)		1.91 (1.21–3.01)
BMI
Normal	Reference	0.5
Overweight	1.07 (0.71–1.61)
Obese	1.26 (0.86–1.86)
Alcohol use
Never	Reference	0.3
Socially	0.78 (0.55–1.10)
Frequently	1.03 (0.45–2.35)
Smoking
No	Reference	0.2
Yes	0.8 (0.57–1.12)
Clinical stage
1	Reference	<0.001	Reference	<0.001
2	2.96 (2.01–4.38)		2.16 (1.36–3.44)
3	9.51 (6.07–14.9)		5.88 (3.24–10.7)
Tumor differentiation
Well or moderately	Reference	<0.001	Reference	0.006
Poorly	3.98 (2.60–6.07)		2.03 (1.22–3.37)
Tumor subtype
ER+/HER2−	Reference	<0.001	Reference	<0.001
ER+/HER2+	0.90 (0.41–1.97)		0.59 (0.23–1.51)
ER−/HER2+	0.88 (0.32–2.43)		0.20 (0.003–1.56)
ER−/HER2−	5.17 (3.62–7.38)		2.53 (1.26–5.08)
Chemotherapy
No	Reference	<0.001	Reference	0.113
Yes	1.95 (1.39–2.75)		0.67 (0.41–1.10)
Radiation
No	Reference	<0.001	Reference	0.5
Yes	2.23 (1.46–3.40)		1.16 (0.71–1.90)
Endocrine therapy
No	Reference	<0.001	Reference	0.8
Yes	0.26 (0.19–0.36)		0.93 (0.47–1.83)

BMI indicates body mass index; MVA, multivariable; UVA, univariable.

DISCUSSION

Our study found that living in a disadvantaged NYC neighborhood is associated with higher odds of poorly differentiated tumors and shorter breast cancer-specific survival, independent of individual-level, tumor, and treatment characteristics. We also visualized our findings, which revealed overlap between disadvantaged areas in Manhattan (eg, Harlem), the Bronx, and Queens, that overlap with historically redlined areas, leading disinvestment over the years and downstream factors associated with social adversity such as higher crime rates. These findings support and expand the generalizability of prior studies, which suggest an association between neighborhood disadvantage, aggressive tumor biology, and shorter breast cancer-specific survival.^5,6 In addition, we identified important sociodemographic, comorbidity, and access to care barriers by neighborhood disadvantage.

Neighborhood Disadvantage and Poorly Differentiated Tumors

The “social genomic determinants of health” provide one explanation for our findings between living in a disadvantaged neighborhood and poorly differentiated tumors. Disadvantaged neighborhoods are chronic epicenters of social adversity and distress shaped by economic and/or residential segregation.¹³ They are often under-resourced and characterized by increased rates of poverty, violent crimes, limited social networks, and poor housing conditions, leading to higher levels of social adversity associated stress.¹⁰ Recent social genomic studies by Goel and colleagues have mapped the “social genomic determinants of health,” or the biological pathways by which social adversity, particularly from one’s built neighborhood environment and one’s perceptions of their neighborhood, activate fight-or-flight SNS neural signaling.¹⁴ Specifically, the authors performed transcriptional analysis of 148 breast cancer blood and tissue samples from a South Florida population and identified that social adversity from both objective measures of neighborhood disadvantage and subjective perceptions of one’s neighborhood disadvantage, particularly higher crime and a sense of “threat to safety” are independent predictors of higher levels of myeloid lineage immune cell transcription factor activity consistent with upregulation of NF-kB (reflective of proinflammatory and proliferative pathways), downregulation of IRF (protective cellular immune response pathways), and upregulation of CREB (a marker of SNS activation).^7,10,14 Combined, these transcription factor alterations have been shown to cultivate a prometastatic niche and were associated with shorter recurrence-free survival, reflective of more aggressive tumor biology.^7,10 Since NF-kB, is also a marker of cell proliferation, our current findings that neighborhood disadvantage, particularly areas with higher levels of crime and poverty, is an independent predictor of poorly differentiated tumors (a marker of cell proliferation), reinforces an association between chronic SNS activation and aggressive tumor biology. These findings highlight the importance of future social genomic studies in this NYC population.

Neighborhood Disadvantage, Sociodemographic, and Tumor Subtype Differences

Tumor Subtype Differences by Neighborhood Disadvantage

In line with previous studies, we also identified significant sociodemographic, comorbidity, tumor subtype, and BCSS differences by neighborhood disadvantage, with more patients of Black race, Hispanic ethnicity, obesity, and TNBC residing in disadvantaged neighborhoods.¹³ This racial and ethnic distribution is a downstream consequence of historical redlining, a government-sanctioned discriminatory practice that flagged areas with primarily minority populations as high risk, resulting in denial of mortgages and blockage of the path to homeownership.^15–17 The term “redlining” came from the color-coding scheme used by the Home Owners Loan Corporation to identify neighborhoods that they identified as “hazardous”. The residential racial segregation that followed prompted a lack of investment in neighborhood infrastructure and services, sparking the access inequities that have persisted long after the ban of redlining in 1968. In New York City, the effects of redlining are far-reaching and continue to impact health outcomes today. Historically redlined neighborhoods in the Bronx, Harlem, and Brooklyn that showed higher rates of more aggressive disease and worse breast cancer-specific outcomes in our study have also been found to have higher prevalence of strokes and preterm births.^18,19 Nationally, similar disparities have been seen in men with prostate cancer, with a recent study showing that Black men with prostate cancer were significantly more likely to live in high redlined neighborhoods compared with their white counterparts, and that men in high redlined areas had lower prostate cancer-specific survival and higher mortality.²⁰

These neighborhoods also overlap with food deserts, areas where there are few supermarkets, and with food swamps, areas where food options are primarily from convivence or fast-food stores.²¹ Combined with lack access to green space and safe walking areas, this leads to higher odds of obesity, a risk factor in and of itself for breast cancer, particularly TNBC.^22,23

We similarly saw that women living in disadvantaged neighborhoods were more likely to have TNBC. It has been well-established that Black women are at 2-fold higher risk for development of TNBC, and that women with TNBC present with more aggressive disease and worse survival.^4,24–26 Western sub-Saharan ancestry has been shown to confer an increased risk of TNBC, related to the DARC/ACKR1 allele; however, this study was unable to control for neighborhood-level or socioeconomic factors.²⁷ More recent studies taking a translational epidemiological approach by merging ancestry and neighborhood-level factors identified that living in a disadvantaged neighborhood is associated with increased odds of TNBC, independent of West African ancestry and BRCA1 mutational status.²⁸ These findings raise the possibility that factors associated with one’s built environment, including social adversity, through social epigenomic mechanisms, might be associated with development of TNBC.^4,24–26,28

Neighborhood Disadvantage and Shorter Breast Cancer-specific Survival

Turning our attention to breast cancer-specific survival outcomes, we identified that along with established clinical predictors of shorter survival (eg, later stage, higher grade, and TNBC), that living in a disadvantaged neighborhood and having Medicaid insurance were independent predictors of shorter breast cancer-specific survival.^29–31 These findings highlight access to care barriers associated with Medicaid, along with unaccounted for aggressive tumor biology, not completely captured by grade, stage, or subtype, further laying the foundation for future social genomic studies in our NYC population to validate prior genomic signatures of adversity.

Patients with Medicaid insurance also presented with later-stage disease. This is consistent with previous studies that have shown diagnosis at more advanced stages in women with Medicaid insurance, and higher odds of TNBC.²⁹ However, to our knowledge there have been no studies evaluating the impact of insurance status on grade at presentation. Medicaid insurance has also been found to be independently associated with delays in care, even in women being treated at a comprehensive breast center.³⁰ These delays in care secondary to insurance type can be compounded in women who live in disadvantaged neighborhoods, where lack of access to health care resources and appropriate referrals can further delay diagnosis and lead to later stage at presentation, resulting in shorter breast cancer-specific survival.³¹ Lack of transportation and poor social support may affect the completion of appropriate treatment, which is associated with shorter breast cancer-specific survival.³² This combination of more aggressive tumor biology and social barriers highlights a “double hit” phenomenon, with women from disadvantaged neighborhoods developing more aggressive disease and subsequently facing barriers that delay timely and appropriate care, ultimately resulting in shorter breast cancer-specific survival.

Next Steps and Interventions

Future directions include validating the mechanisms by which social adversity influences tumor biology. To do this, we aim to develop interventions that can reverse social genomic TFs associated with more aggressive tumor biology (eg, NF-kB, AP-1, CREB). Pharmaceutical interventions, like the use of beta-blockade to mitigate the stress response seen by the SNS, have shown promise in managing stress related to breast cancer diagnosis. Preclinical studies and a recent phase II randomized controlled trial (RCT, n=30 per arm) in Australia showed that breast cancer patients on 7 days of propranolol presurgery had downregulated TF NF-kB and reduced self-reported patient anxiety.³³ Other strategies for stress reduction, such as cognitive behavioral stress management (CBSM), have been previously shown to lower distress during breast cancer treatment, reverse CTRA biology, and improve survival.^34,35 In addition, policy interventions to improve infrastructure in disadvantaged neighborhoods, particularly areas where we see poorly differentiated tumors, such as regions in Harlem, Queens, and the Bronx, are critical to help improve survival outcomes for all patients living in NYC (Fig. 1). Further evaluation of NYC neighborhoods for a potential protective effect of ethnic enclaves, previously observed in Hispanic communities in South Florida, could provide additional insight into the impact of social cohesion in shaping outcomes in disadvantaged neighborhoods.³⁶

This study was limited by its retrospective nature. Although we were able to account for neighborhood disadvantage, using ADI, a validated, multidimensional tool that captures data down to the block-group level, allowing for more precise evaluation of neighborhood deprivation, we were unable to account for additional exposures such as pollution, which might influence tumor biology.³⁷ In addition, although we have data on whether patients received adjuvant chemotherapy, radiation and endocrine therapy, we do not know if patients completed treatment as recommended. Most of our patients were non-Hispanic White and lived in advantaged neighborhoods, which can limit the generalizability of our study, though this ADI distribution is consistent with previous work. Another limitation is that ADI was determined by the patient’s address at diagnosis, but we do not have additional information about the length of time lived at that address, or any prior addresses. However, previous research linking mesothelioma to non-asbestos toxic air exposure looked at residential histories in NYC and found that 43% of their patients lived in NYC and on average lived at 3 addresses with a mean residential duration of 10.4 years, suggesting stability.³⁸ Despite these limitations, we were able to control for granular-level risk factors associated with tumor grade such as obesity, which are not available in national databases.³⁹

To our knowledge, this is the first study to show an association between neighborhood disadvantage and poorly differentiated breast cancer in a diverse NYC population. Our findings expand generalizability of prior studies showing that women living in disadvantaged neighborhoods are more likely to have aggressive disease and worse breast cancer-specific survival, even when controlling for individual, tumor and treatment characteristics. These findings merit further inquiry and lay the foundation for future translational studies to externally validate the mechanisms by which neighborhood disadvantage “gets under the skin” to impact aggressive breast cancer tumor biology, and ultimately survival, in this large, diverse urban population treated in a tertiary care setting, paving the way for intervention strategies to achieve health equity.

DISCUSSANT

Dr. Caprice C. Greenberg (Chapel Hill, NC)

What a fantastic presentation. It’s quite amazing to think about how far we’ve come in our understanding of the complex relationship between socioeconomic disadvantage and cancer outcomes. After decades of just describing disparities by factors such as insurance status, race, or education, we began to explore concepts such as differential access, health literacy, and financial toxicity. And now you’re taking it to the next step to identify the unique influence of the external built environment and neighborhood-level concepts that can impact patient outcomes.

We also have become much more sophisticated in our understanding of the mechanisms that underlie these observed associations. The systemic stress in disadvantaged neighborhoods leads to physiological changes. The upregulation of factors associated with inflammation and cell proliferation and downregulation of transcription factors that are related to protective cellular responses and programmed cell death lead to the development of higher grade, more aggressive cancers. Once cancer develops, access to care barriers can lead to delayed diagnosis and delayed treatment. And together, these more aggressive cancer biologies and delays in care lead to poor breast cancer-specific survival.

In your study, you build on this work to assess the relationship between neighborhood disadvantage and tumor characteristics and survival for women with stage I through III breast cancer. Disadvantage in your study is measured by the Area Deprivation Index, which is a census-based index that combines measures such as household income, education, employment, and housing conditions into a single estimate of disadvantage. You also can look at important individual health behaviors, like alcohol and tobacco use, that aren’t available in most studies.

As you considered both these individual and neighborhood-level effects into your multivariable modeling, one thing that really struck me was the fact that the effect size of ADI in your multivariable model predicting poorly differentiated tumor histology was about half of the univariate effect size when you considered things like race, insurance, and health behaviors. In contrast, when you looked at your model for breast cancer-specific survival, the effect of ADI increased when you added these other characteristics.

And so I have 2 questions here for you. I’m curious how you interpret these findings. And given the complexity of these factors, I’m wondering if you included interaction terms and/or a stratified analysis by characteristics such as race or histologic grade.

I’m also curious about your future directions. In your manuscript, you talked about the potential to develop pharmaceutical interventions that target these biological pathways that are impacted by social adversity. There seems to be convincing data that chronic stress from environmental factors leads to more aggressive tumor biology through these epigenetic pathways. Once cancer is diagnosed, differential treatments such as poor access to care and differences in quality, compliance, treatment, nutrition, exercise, and all these factors are also important.

You referenced an important paper that I want people in this audience to be aware of, which was a study by Rehm and colleagues that looked at cognitive behavioral stress management as a way to intervene for breast cancer patients and looked at the effect according to ADI. In that study at baseline, cortisol levels were 35% higher in women from higher ADI, which would be expected reflecting the chronic environmental stress that we’ve been talking about. Interestingly, although the cognitive behavioral therapy decreased the cancer stress in both advantaged and disadvantaged women, the cortisol levels actually were only lowered in women who were from less disadvantaged areas, suggesting the ability to intervene for the acute stress of a cancer diagnosis but not the ability to intervene on the chronic stress that comes with living in these disadvantaged environments.

So is there data that suggest that post-diagnosis interventions targeting these epigenetic pathways can improve survival? Or do we risk increasing the survival gap if interventions are more effective in patients from lower disadvantaged areas than from higher disadvantaged areas?

Thank you so much for your work advancing this really exciting area.

Response from Neha Goel

Thank you so much for your thoughtful discussion and these very important questions.

Regarding your point about Medicaid and why it was seen to have an impact on overall survival, including breast cancer-specific survival, we actually went back, and in our population found no significant difference among those who had Medicaid or Medicare or private insurance with respect to receipt of guideline-concordant treatment, suggesting these effects of neighborhood disadvantage are beyond access to care barriers.

So, this really brings up the question, is Medicaid a surrogate for individual-level stress? It could be because individual-level income, which is very difficult to capture from any registry data, is a potential marker of individual-level stress, and it’s often used as a surrogate. So that could be one reason Medicaid is still showing up as a clinically significant factor.

And thank you so much for bringing up the second comment regarding interventions. It’s absolutely critical. Moving patients from a disadvantaged neighborhood to an advantaged neighborhood is not practical. So that really brings up the question of what we can do and what interventions we can develop.

And our team is really focused on 2. One, similar to what you mentioned, is cognitive behavioral stress management, or really helping patients learn how to control their stressors through active coping mechanisms.

And to your point of that paper, and that paper was really to serve as showing that there is a knowledge gap—we need CBSM interventions that focus on the other sources of stress such as from one’s neighborhood. The fact that CBSM did not have effects on cortisol outcomes in high ADI women can be explained by (1) the original randomized control trial was focused on reducing mainly breast cancer-specific stress and anxiety, not neighborhood and other sources of stress and (2) cortisol not reflecting SNS activity as directly as norepinephrine. Thus, the goal of that paper was really to show that, okay, this gap exists, and the science needs to really develop interventions that target one’s sympathetic nervous system activity.

Thus, we need larger trials with biological and behavioral data and CBSM would need to be tailored to address the stress of living in high ADI environments to optimize its relevance to women living there. Tailoring could focus on identifying the chief sources of stress that drive tumor biology and immune/inflammatory signaling vis SNS activation, which may be perceptions of safety/threat and lower social cohesion.

And similarly, in our basic science work, we’ve shown that patients—and this is to our point of the second intervention of potentially beta-blockade, who were incidentally on nonselective beta blockers had tumors that had lower proinflammatory transcription factor activity.

So, we need to target the right mechanism to develop beneficial interventions.

Thank you, and I’ll take any additional questions.

Dr. Christobel Saunders (Melbourne, AUS)

Thank you very much for that important paper. It was great to hear.

I wanted to explore with you 2 potential confounders, both of which have gotten potentially targetable for better outcomes.

The first is around breast cancer screening. We’re aware that screened populations tend to have more slower-growing ER-positive tumors. So, I was wondering whether one potential effect of that is that these populations are not very well-screened and therefore have more aggressive cancers.

And the second is around your survival outcomes. We know that these populations often have pretty poor adherence to treatment, whether that’s because they can’t afford it or they don’t understand it or whatever other reasons. So, I was also wondering if you would manage to collect data on treatment adherence. Thank you.

Response from Neha Goel

Yes, thank you so much. These are very important questions, and our “social genomic drivers of health” research field is integrating these socioeconomic factors with tumor biology. It’s not one or the other. So, it’s really inclusive of all of these, as our translational epidemiological framework suggests.

With respect to your first question, in larger cohorts we have stratified by subtype, and even among ER-positive early stage 1 and 2 tumors, we do see these differences persist. But absolutely screening is critical and is an important source of disparities. And we, from our current findings, have found that even in ER-positive early-stage disease, which is most likely to be found on screening, we see these differences.

And to your next question, we are capturing adherence data. In terms of receipt, we do see that, independent of where people live, in our population and probably because they’re being treated at a comprehensive cancer center, they are receiving guideline-concordant treatment.

Dr. Ronda Henry-Tillman (Little Rock, AR)

Thank you for the report and the data, similar to ongoing information we know so well. I’m so excited to see a new term from social determinants now to social genomics. And hopefully that will give us a place where we can target this.

How do you plan to target to change these poorly differentiated cancers and get different outcomes as you defined in your paper? Thank you, guys, for all of the work.

Response from Neha Goel

Thank you so much for that important question. And I think the main areas to target are really through clinical trials that we are actively starting and have ongoing.

So currently there is pilot data from small, international phase 2 randomized controlled trials showing that patients who get even a short dose of a nonselective beta blocker 7 days before and 7 days after surgery, you can see changes in their tumors that align with reduction of aggressive social genomic transcription factors. So, we are working on developing a clinical trial in the United States.

And similarly, we have ongoing prospective cohort studies. The Miami Breast Cancer Study, that I founded and lead, has enrolled 700 women, and we are asking them very detailed questions to delineate sources of stress—whether it’s breast cancer-specific stress, general stress, or stress from their environment, or all these combined—that we need to target.

All of these factors really need to be considered. And one of our NIH R37-funded studies is looking and teasing out how each of these distinct sources of stress influence biology, to help tailor interventions to improve outcomes for all our patients.