Residential proximity to Superfund sites and breast cancer incidence in a nationwide cohort

Abstract

Background:

The United States Environmental Protection Agency-designated Superfund sites are a source of potential exposure to harmful contaminants, including endocrine-disrupting chemicals and carcinogens, that may be relevant to breast cancer risk. However, there is limited research on the possible association between residential proximity to Superfund sites and the risk of breast cancer.

Methods:

Using data from the prospective US-wide Sister Study cohort (N = 49,594), geocoded enrollment addresses were linked to active United States Environmental Protection Agency Superfund sites to determine residential proximity to sites. Proximity metrics included residing within 3-, 5-, and 10-miles of Superfund sites. Cox proportional hazards models were used to estimate hazard ratios (HRs) and 95% confidence intervals (CIs) for the relationship between residential proximity to Superfund sites and breast cancer incidence. We further evaluated whether associations for residential proximity and breast cancer incidence varied according to individual contaminants and latent classes that grouped Superfund sites based on the mixture of contaminants present at the site.

Results:

Over an average of 12 years of follow-up, 4,565 breast cancer cases were identified. Overall, there was little evidence to support that living within three miles of a single Superfund site was strongly related to breast cancer incidence. However, living within three miles of two or more sites was associated with a 20% higher breast cancer incidence (HR = 1.20; 95% CI = 1.00, 1.45). There was little evidence of an association by contaminant latent classes, although when considering individual contaminants, we observed a possible association with proximity to sites containing organophosphates (HR = 1.33, 95% CI = 0.90, 1.96).

Conclusion:

In this large, US-wide cohort, no overall pattern of association between residential proximity to Superfund sites and breast cancer incidence was found. However, elevated HRs were observed for living near two or more sites or sites that are contaminated with organophosphates.

What this study adds.

Studies investigating the association between residential proximity to Superfund sites and incident breast cancer are limited. This study advances our knowledge by using individual-level proximity metrics and considering the impact of site contaminants to better understand how exposure to hazardous sites may contribute to breast cancer risk.

Introduction

Breast cancer is the most common cancer diagnosed in women in the United States, with incidence rates increasing since the 2000s.¹ While breast cancer-specific mortality rates have been decreasing since 1989, breast cancer remains the second leading cause of cancer-related death in women in the United States.²

Established risk factors for breast cancer are estimated to account for only around 50% of incident cases and include genetics, reproductive history characteristics (e.g., age at menarche, gravidity, age at first birth, and breastfeeding history), postmenopausal obesity, physical activity, and exogenous hormone use.³ Consistent associations with factors related to reproductive history and use of exogenous hormones underscore the importance of hormones in breast cancer etiology and provide support for the hypothesis that exposure to environmental pollutants, particularly those that have endocrine-disrupting capabilities, may also be contributors to breast cancer incidence.⁴

United States Environmental Protection Agency (EPA) Superfund sites are locations where hazardous waste has been improperly managed and may be important sources of human exposure to harmful contaminants.⁵ These sites can range from manufacturing facilities and landfills to dry cleaners and transportation facilities. Superfund sites may contain a single chemical or compound or mixtures of compounds.⁶ The EPA estimates that in the United States, approximately 78 million people live within three miles of a Superfund site and 23 million live within one mile of a site.⁷ Residential proximity to Superfund sites has been associated with adverse health effects ranging from preterm birth to certain chronic diseases.⁸

Currently, there is little information specifically on whether residential proximity to a Superfund site is related to the risk of breast cancer. However, there are studies that have examined the association between exposure to various environmental and anthropogenic chemicals, including those commonly found at Superfund sites, and breast cancer risk. For example, human and animal studies support the hypothesis that polychlorinated biphenyls (PCBs)⁹ exposure is associated with a higher risk of developing breast cancer through multiple mechanisms, including altered cell proliferation, cell death, genotoxicity, epigenetic alterations, and oxidative stress.¹⁰ A 2020 systematic review concluded that there is some epidemiologic evidence supporting an association between exposure to metals, polycyclic aromatic hydrocarbons, per- and polyfluoroalkyl substances, PCBs, phthalates, organochlorine pesticides, dioxins, and polybrominated diphenyl ethers and a higher risk of breast cancer.¹¹ A separate review that focused solely on endocrine-disrupting chemicals indicated that dichlorodiphenyltrichloroethane, dioxins, and bisphenol A may be related to a higher risk of breast cancer.⁴ These reviews included chemicals ranked on the Substance Priority List as the most common and toxic contaminants found at Superfund sites. Moreover, Superfund sites rarely include single chemicals but rather contain mixtures of chemicals. The purpose of our study was to investigate how residential proximity to Superfund sites, overall and when considering site contaminant profiles, may be related to incident breast cancer using data from a prospective, US-wide cohort study.

Methods

Study population

The Sister Study is a prospective study of 50,884 women ages 35–74 years who were recruited from 2003 to 2009 and living in the United States including Puerto Rico, and who had at least one sister diagnosed with breast cancer but had no personal history of breast cancer themselves.¹² At enrollment, Sister Study participants completed extensive questionnaires using computer-assisted telephone interviews and mailed surveys to report on demographics, lifestyle factors, occupational history, residential history, medical history, and personal care product use. Sister Study participants are contacted annually for health updates, including for information on incident breast cancer and other cancer diagnoses, which are confirmed by medical records. Detailed follow-up questionnaires are mailed every 2–3 years to update lifestyle and risk factor information and to report any other health updates. The Sister Study was approved by the Institutional Review Board of the National Institutes of Health.

We excluded women who withdrew their data (n = 3), who were diagnosed with breast cancer before or at unknown timing relative to completion of baseline study components (n = 84), had missing or insufficient geocoding to examine in relation to Superfund sites (n = 704), or were missing covariate information (n = 499), with a final analytic sample of 49,594 women. This data includes women with follow-up through 30 September 2021 (Data Release 11.1).

Outcome

Our primary outcome of interest was breast cancer reported after enrollment (N = 4,565), including both invasive breast cancer and ductal carcinoma in situ (DCIS). Breast cancer estrogen receptor (ER) status was a secondary outcome of interest. We confirmed self-reported diagnoses and obtained information on tumor invasiveness and ER status through pathology reports from medical records when available (>80% of cases). When records were unavailable, we used self-reported diagnosis information based on high agreement between self-reports and medical records in the cohort.¹³

Exposure

We used a dataset from the United States EPA¹⁴ that contained information for all Superfund sites in the United States, including site location, chemicals present at the site, years the site has been active (e.g., known to be releasing chemicals), and the industry associated with the site. For this analysis, we restricted to sites that were defined as active during the time period of Sister Study enrollment (2003–2009). Each participant’s geocoded residential location at enrollment was linked to these Superfund sites using multiple proximity metrics, including living within 3, 5, and 10 miles of a site (yes/no) and the sum of the inverse-distance weighted number of sites (IDW score, sites/mile capped at 20 miles maximum impact distance). Formulae for these metrics can be found in Supplemental Table 1; https://links.lww.com/EE/A404. The IDW score can be interpreted as a weighted number of proximal Superfund sites; for example, a single site one mile away from the residence is assumed to yield equivalent exposures as two sites that are two miles away. The three-mile buffer was selected to be consistent with previous reports from EPA. We also included 5-and 10-mile distances to explore whether any associations might persist beyond this three-mile buffer and to be consistent with our prior research that examined residential proximity to sites and toenail concentrations of metals.¹⁵ We previously used latent class analysis (LCA) to identify subgroups of Superfund sites with similar contaminant profiles.¹⁶ LCA is a technique that creates subgroups (classes) using a probabilistic approach to uncover clusters in data by grouping subjects using prespecified features based on unobservable membership.¹⁷ In this application of LCA, we used the Superfund contaminant indicator variables, which were created using the EPA dataset, as the inputs to determine class membership. Specifically, these defining variables included the following dichotomous chemical classifications used by the Agency for Toxic Substances and Disease Registry for contaminants found at the Superfund sites: metals, benzidines/aromatic amines, dioxins/furans/PCBs, halogenated compounds, inorganic substances, nitrosamines/ethers/alcohols, organophosphates, phenols/phenoxy acids, phthalates, radionuclides, and volatile organic compounds. The LCA resulted in a seven-class model that was deemed to be the best-fitting model; Superfund sites were assigned to the latent class for which they had the highest probability of membership.

The names of the latent classes were derived from the chemical profiles and predominant industrial activity subcategories. A class was determined to be “high diversity exposure” if more than half of the 12 defined chemical indicator categories were present with high probabilities. Class 1 was labeled “Low diversity exposure, lumber and wood industry” due to the presence of fewer chemical categories compared with other classes, with the lumber and wood products industry being the most common (19% of sites). Class 2 was designated “High diversity exposure, industrial and municipal landfills.” Class 3, predominantly wood and lumber (64% of sites), had a broad range of chemical categories and was named “High diversity exposure, lumber and wood industries.” Class 4 included sites across waste, landfill, and chemical industries, earning it the name “High diversity exposure, landfills and industrial waste facilities, chemical industries.” Class 5 also featured a variety of chemical classes and was named “High diversity exposure, landfills and waste, metal and electronic industries.” Class 6 was termed “Manufacturing and processing facilities, electronic/electrical and chemical industries,” while Class 7 was named “Batteries and scrap metal industries, metal processing.” Supplemental Table 2; https://links.lww.com/EE/A404 provides the probability distribution of contaminant indicators in each class.

Statistical analysis

Hazard ratios (HRs) and 95% confidence intervals (CIs) were estimated using Cox proportional hazards regression models for associations between residential proximity to Superfund sites and breast cancer incidence. Age served as the timescale of the Cox model with women entering the risk set at their age at study enrollment and followed until age at breast cancer diagnosis or censoring event (death, loss to follow-up, age at last contact). We considered how associations between living near a Superfund site and breast cancer risk varied by extent of disease (invasive vs. DCIS), by ER status, and menopausal status. For associations by ER status, women entered the Cox model at their age at baseline and were followed until age at diagnosis of ER+ tumor, with tumors that were ER− or had missing subtype information censored at time of diagnosis. For analyses related to menopausal status, women were additionally censored in premenopausal breast cancer models when they became postmenopausal. For postmenopausal breast cancer, women entered the Cox model at either their age at enrollment or their age at menopause, whichever was later.

We considered proximity to a single site, as well as living near two or more sites. We stratified our analyses by Superfund latent class (classes 1–7) and chemical contaminant class (metals, benzidines/aromatic amines, dioxins/furans/PCBs, halogenated compounds, inorganic substances, nitrosamines/ethers/alcohols, organophosphates, phenols/phenoxy acids, phthalates, radionuclides, and volatile organic compounds) for the site closest to the participant’s residence to examine differences among site contaminant profiles.

Models were adjusted for potential confounders that were selected using a directed acyclic graph¹⁸ and included race/ethnicity (non-Hispanic White, non-Hispanic Black, Hispanic, and other), education (no high school diploma, high school diploma or General Educational Development, some college or associate’s degree, bachelor’s degree or higher), income (<$50,000, between $50,000 and $99,999, above $100,000), area deprivation index (ADI, ≤29, >29, where 29 was the median), smoking status (never smoked, former smoker, current smoker), and primary source of drinking water (bottled water, community or private well, city or town water, other). The ADI is a neighborhood-level measure that includes factors such as employment, housing quality, income, and education and indicates community disadvantage. A higher ADI reflects higher community disadvantage.¹⁹

We performed two sensitivity analyses with this data. In the first, we restricted the analysis to women who lived in their enrollment residences for at least 10 years, which was the median time reported living at the home. The other analysis included additional covariate adjustment for Census geographic region (East, West, South, and North).

All analyses were conducted using SAS statistical software version 9.4 (SAS Institute, Cary, NC).

Results

Over 12 years of follow-up, 4,510 breast cancer cases were identified. During the enrollment period (2003–2009), 1,493 Superfund sites were active. Overall, the cohort is predominately non-Hispanic White (85%), has a bachelor’s degree or higher (51%), and are never smokers (56%). Demographic and lifestyle distributions were largely similar between participants living within three miles of a Superfund site and those living further away. However, participants living within three miles of a site were more likely to use bottled water and city or town water and were less likely to primarily drink water from a community or private well than those living further away (Table 1).

Table 1.

Characteristics of Sister Study participants, enrolled 2003–2009

	Overall	Lives within 3 miles of a site
		Yes N (%)	No N (%)
Total	48,939	4,849	44,090
Race
Non-Hispanic White	41,774	4,088 (84%)	37,686 (85%)
Non-Hispanic Black	4,322	444 (9%)	3,878 (9%)
Hispanic	1,565	198 (4%)	1,367 (3%)
Other	1,278	119 (2%)	1,159 (3%)
Educational attainment
No high school diploma	469	53 (1%)	416 (1%)
High school diploma or GED	6,876	658 (14%)	6,218 (14%)
Some college or associate’s degree	16,525	1,552 (32%)	14,973 (34%)
Bachelor’s degree or higher	25,069	2,586 (53%)	22,483 (51%)
Household income
Income <$50,000	12,038	1,189 (25%)	10,849 (25%)
Income $50,000 to $99,999	20,227	1,966 (41%)	18,261 (41%)
Income $100,000 or more	16,674	1,694 (35%)	14,980 (34%)
Smoking status
Never smoked	27,340	2,563 (53%)	24,777 (56%)
Past smoker	17,602	1,856 (38%)	15,746 (36%)
Current smoker	3,997	430 (9%)	3,567 (8%)
Primary source of drinking water
Bottled water	4,611	554 (11%)	4,057 (9%)
Community or private well	9,210	656 (14%)	8,554 (19%)
City or town water	34,588	3,604 (74%)	30,984 (70%)
Other	530	35 (1%)	495 (1%)
Urbanicity
Urban	9,295	1,300 (27%)	7,995 (18%)
Suburban	19,159	2,182 (45%)	16,977 (39%)
Small town	10,506	965 (20%)	9,541 (22%)
Rural or other	9,979	402 (8%)	9,577 (22%)
Area deprivation index
ADI ≤ 29	24,674	2,543 (52%)	22,131 (50%)
ADI > 29	24,265	2,306 (48%)	21,959 (50%)
Body mass index
BMI < 25	19,024	1,894 (39%)	17,130 (39%)
BMI ≥ 25	29,915	2,955 (61%)	26,960 (61%)
Breast cancer
Breast cancer diagnosis	4,510	433	4,077
DCIS	926	94 (22%)	832 (20%)
Invasive	3,574	338 (78%)	3,236 (80%)
Estrogen receptor status
ER+	3,240	320 (86%)	2,920 (85%)
ER−	568	53 (14%)	515 (15%)

GED indicates General Educational Development.

Overall, we observed little evidence of an association between the presence of a single Superfund site and breast cancer incidence (Table 2). Living within a 3-, 5-, or 10-mile buffer of any one Superfund site was not associated with a higher incidence of breast cancer (e.g., <3 vs. ≥3 miles, HR = 0.96; 95% CI = 0.87, 1.06). However, participants who lived near a higher density of sites, with two or more Superfund sites within three miles, had a 20% higher incidence of breast cancer (HR = 1.21; 95% CI = 1.00, 1.46). This association was attenuated when evaluating larger distance buffers (5 miles; HR = 1.04; 95% CI = 0.93, 1.15) and (10 miles; HR = 1.04; 95% CI = 0.97, 1.11). The pattern of associations for ER+ breast cancer was similar to the patterns for overall breast cancer. We observed little evidence of an association for proximity to site in relation to either premenopausal or postmenopausal breast cancer.

Table 2.

Superfund site exposure and breast cancer risk overall and by ER status in the Sister Study

	3 miles of site		5 miles of site		10 miles of a site
	Case N	Adjusted HRa (95% CI)	Case N	Adjusted HRa (95% CI)	Case N	Adjusted HRa (95% CI)
Overall breast cancer
No site	4,077	1	3,533	1	2,351	1
Any one site (1,493 sites)	433	0.96 (0.87, 1.06)	977	0.98 (0.91, 1.05)	2,159	1.02 (0.97, 1.09)
Two or more sites	111	1.21 (1.0, 1.46)	358	1.04 (0.93, 1.15)	1,274	1.04 (0.97, 1.11)
Class 1 (103 sites)	30	0.88 (0.62, 1.27)	79	1.06 (0.85, 1.33)	164	1.13 (0.96, 1.32)
Class 2 (212 sites)	92	1.13 (0.92, 1.39)	217	1.15 (1.00, 1.32)	425	1.09 (0.98, 1.21)
Class 3 (50 sites)	8	0.60 (0.30, 1.20)	24	0.72 (0.48, 1.07)	77	0.97 (0.77, 1.22)
Class 4 (274 sites)	76	0.89 (0.71, 1.12)	192	0.96 (0.83, 1.10)	442	1.02 (0.92, 1.13)
Class 5 (346 sites)	82	0.87 (0.70, 1.08)	186	0.93 (0.80, 1.08)	446	1.01 (0.91, 1.12)
Class 6 (217 sites)	61	0.93 (0.72, 1.20)	129	0.89 (0.74, 1.06)	295	1.01 (0.89, 1.14)
Class 7 (291 sites)	84	1.12 (0.90, 1.39)	150	0.95 (0.81, 1.12)	310	0.96 (0.85, 1.08)
ER+ breast cancer
No site	2,920	1	2,514	1	1,663	1
Any one site (1,493 sites)	320	0.99 (0.88, 1.11)	726	1.02 (0.94, 1.10)	1,577	1.05 (0.98, 1.13)
Two or more sites	80	1.21 (0.97, 1.51)	275	1.10 (0.98, 1.25)	921	1.04 (0.96, 1.12)
Class 1 (103 sites)	20	0.81 (0.52, 1.26)	58	1.07 (0.82, 1.39)	110	1.05 (0.87, 1.28)
Class 2 (212 sites)	64	1.11 (0.86, 1.42)	158	1.18 (1.00, 1.38)	316	1.14 (1.01, 1.29)
Class 3 (50 sites)	7	0.73 (0.35, 1.54)	19	0.80 (0.51, 1.26)	58	1.01 (0.78, 1.32)
Class 4 (274 sites)	53	0.87 (0.66, 1.14)	146	1.02 (0.86, 1.21)	330	1.07 (0.95, 1.21)
Class 5 (346 sites)	66	0.97 (0.76, 1.24)	144	1.01 (0.85, 1.19)	329	1.04 (0.93, 1.17)
Class 6 (217 sites)	46	0.97 (0.73, 1.31)	92	0.88 (0.72, 1.09)	209	0.99 (0.86, 1.15)
Class 7 (291 sites)	64	1.19 (0.93, 1.53)	109	0.98 (0.81, 1.18)	225	0.98 (0.85, 1.12)
ER− breast cancer
No site	515	1	453	1	319	1
Any one site (1,493 sites)	53	0.95 (0.72, 1.26)	115	0.92 (0.75, 1.12)	249	0.90 (0.76, 1.06)
Two or more sites	15	1.33 (0.80, 2.22)	44	1.05 (0.77, 1.43)	144	0.93 (0.77, 1.13)
Class 1 (103 sites)	4	0.99 (0.37, 2.64)	8	0.89 (0.44, 1.79)	24	1.28 (0.84, 1.93)
Class 2 (212 sites)	13	1.27 (0.73, 2.21)	27	1.12 (0.76, 1.66)	46	0.88 (0.65, 1.20)
Class 3 (50 sites)	0	–	2	–	5	0.49 (0.20, 1.18)
Class 4 (274 sites)	10	0.92 (0.49, 1.71)	22	0.85 (0.56, 1.31)	52	0.89 (0.66, 1.19)
Class 5 (346 sites)	5	0.43 (0.18, 1.04)	19	0.76 (0.48, 1.21)	46	0.80 (0.59, 1.09)
Class 6 (217 sites)	9	1.13 (0.59, 2.19)	15	0.84 (0.50, 1.40)	35	0.92 (0.65, 1.31)
Class 7 (291 sites)	12	1.29 (0.72, 2.29)	22	1.11 (0.72, 1.70)	41	0.96 (0.69, 1.33)

^aAdjusted for race, income, education, smoking status, and primary source of drinking water.

When stratifying by the latent class contaminant profiles of the nearest site, we observed little evidence of elevated HRs for breast cancer across the seven classes (Table 2). We observed some evidence of elevated HRs for participants living within three miles of Superfund sites in Class 2 (high diversity exposure, industrial and municipal landfills) (HR = 1.13; 95% CI = 0.92, 1.39) and Class 7 (batteries and scrap metal industries, metal processing) (HR = 1.12; 95% CI = 0.90, 1.39). Class-specific associations showed similar patterns by ER status.

When evaluating proximity to Superfund sites by the presence of 12 individual contaminant categories, we observed increased HRs for the organophosphate site category (Table 3). The incidence of breast cancer associated with exposure to organophosphate sites was strongest for sites within three miles and trended downward with increasing distance from the Superfund sites (3-mile HR = 1.33; 95% CI = 0.90, 1.97; 5-mile HR = 1.22; 95% CI = 0.97, 1.54; 10-mile HR = 1.10; 95% CI = 0.97, 1.25). There was little evidence of strong associations with either ER+ or ER− tumors, with the exception of a suggestive higher incidence for organophosphate-containing sites (e.g., ER+ 3-mile HR = 1.33, 95% CI = 0.82, 2.13 and 5-mile HR = 1.27, 95% CI = 0.97, 1.66; ER− 3-mile HR = 1.99, 95% CI = 0.84, 4.76 and 5-mile HR = 1.45, 95% CI = 0.80, 2.62).

Table 3.

Superfund site exposure by contaminant category and breast cancer risk overall and by ER status in the Sister Study

	Breast cancer diagnosis			ER+ breast cancer			ER− breast cancer
	Case N	Noncase N	Hazard ratio (95% CI)	Case N	Noncase N	Hazard ratio (95% CI)	Case N	Noncase N	Hazard ratio (95% CI)
Benzidine and aromatic amine sites
Site within 3 miles	8	124	0.71 (0.36, 1.39)	4	124	–	1	124	–
Site within 5 miles	39	427	1.01 (0.73, 1.38)	27	427	0.98 (0.67, 1.43)	6	427	1.24 (0.56, 2.78)
Site within 10 miles	160	1,966	0.86 (0.73, 1.01)	113	1,966	0.84 (0.69, 1.01)	21	1,966	0.95 (0.61, 1.47)
Dioxin-like compound sites
Site within 3 miles	217	2,397	0.98 (0.86, 1.13)	162	2,397	1.03 (0.88, 1.20)	24	2,397	0.86 (0.57, 1.29)
Site within 5 miles	576	6,183	1.01 (0.93, 1.10)	439	6,183	1.08 (0.98, 1.20)	60	6,183	0.83 (0.63, 1.08)
Site within 10 miles	1,504	15,690	1.05 (0.99, 1.12)	1,095	15,690	1.07 (0.99, 1.15)	168	15,690	0.91 (0.75, 1.09)
Hydrocarbon sites
Site within 3 miles	316	3,473	0.98 (0.88, 1.10)	232	3,473	1.01 (0.88, 1.15)	37	3,473	0.92 (0.66, 1.28)
Site within 5 miles	746	8,189	0.98 (0.91, 1.06)	563	8,189	1.04 (0.95, 1.14)	82	8,189	0.85 (0.67, 1.07)
Site within 10 miles	1,808	19,186	1.03 (0.97, 1.10)	1,319	19,186	1.05 (0.98, 1.13)	213	19,186	0.95 (0.80, 1.13)
Halogenated hydrocarbon and halogenated VOC sites
Site within 3 miles	284	3,229	0.95 (0.84, 1.07)	207	3,229	0.97 (0.84, 1.11)	35	3,229	0.94 (0.67, 1.33)
Site within 5 miles	688	7,603	0.97 (0.90, 1.06)	513	7,603	1.01 (0.92, 1.12)	81	7,603	0.92 (0.72, 1.16)
Site within 10 miles	1,775	18,724	1.04 (0.98, 1.11)	1,293	18,724	1.06 (0.99, 1.14)	204	18,724	0.93 (0.78, 1.10)
Heavy metals sites
Site within 3 miles	349	3,831	0.99 (0.89, 1.10)	263	3,831	1.04 (0.92, 1.18)	39	3,831	0.88 (0.63, 1.21)
Site within 5 miles	807	8,853	0.98 (0.91, 1.06)	606	8,853	1.04 (0.95, 1.13)	94	8,853	0.91 (0.73, 1.13)
Site within 10 miles	1,912	20,471	1.02 (0.96, 1.08)	1,405	20,471	1.05 (0.98, 1.13)	218	20,471	0.88 (0.74, 1.05)
Nitrosamines, ethers, and alcohols sites
Site within 3 miles	132	1,472	0.97 (0.81, 1.15)	97	1,472	0.99 (0.81, 1.21)	17	1,472	1.01 (0.62, 1.63)
Site within 5 miles	377	3,998	1.02 (0.92, 1.13)	282	3,998	1.06 (0.94, 1.20)	42	3,998	0.92 (0.67, 1.25)
Site within 10 miles	1,104	11,491	1.05 (0.98, 1.12)	800	11,491	1.05 (0.97, 1.14)	127	11,491	0.96 (0.79, 1.17)
Inorganic sites
Site within 3 miles	357	3,954	0.98 (0.88, 1.09)	268	3,954	1.03 (0.90, 1.16)	41	3,954	0.89 (0.65, 1.23)
Site within 5 miles	825	9,062	0.98 (0.91, 1.06)	618	9,062	1.03 (0.94, 1.13)	97	9,062	0.92 (0.74, 1.14)
Site within 10 miles	1,928	20,690	1.01 (0.95, 1.08)	1,415	20,690	1.05 (0.97, 1.12)	221	20,690	0.89 (0.75, 1.05)
Organophosphate sites
Site within 3 miles	25	209	1.33 (0.90, 1.97)	17	209	1.33 (0.82, 2.13)	5	209	1.99 (0.84, 4.76)
Site within 5 miles	73	666	1.22 (0.97, 1.54)	53	666	1.27 (0.97, 1.66)	11	666	1.45 (0.80, 2.62)
Site within 10 miles	251	2,492	1.10 (0.97, 1.25)	184	2,492	1.13 (0.97, 1.31)	28	2,492	1.00 (0.69, 1.47)
Phenols and phenoxy acids sites
Site within 3 miles	168	1,809	1.00 (0.85, 1.16)	121	1,809	0.99 (0.83, 1.19)	23	1,809	1.11 (0.73, 1.68)
Site within 5 miles	436	4,603	1.03 (0.93, 1.13)	333	4,603	1.09 (0.97, 1.22)	49	4,603	0.92 (0.69, 1.24)
Site within 10 miles	1,171	12,239	1.05 (0.98, 1.12)	860	12,239	1.07 (0.98, 1.15)	124	12,239	0.86 (0.70, 1.04)
Phthalates sites
Site within 3 miles	165	1,816	0.98 (0.84, 1.14)	121	1,816	1.00 (0.84, 1.20)	19	1,816	0.90 (0.57, 1.42)
Site within 5 miles	434	4,732	0.99 (0.90, 1.09)	327	4,732	1.04 (0.93, 1.17)	49	4,732	0.89 (0.66, 1.19)
Site within 10 miles	1,240	12,947	1.05 (0.98, 1.12)	904	12,947	1.06 (0.98, 1.15)	148	12,947	1.01 (0.83, 1.21)
Radionuclide sites
Site within 3 miles	87	930	1.02 (0.82, 1.26)	63	930	1.02 (0.80, 1.31)	14	930	1.32 (0.78, 2.25)
Site within 5 miles	241	2,667	0.96 (0.85, 1.10)	173	2,667	0.96 (0.82, 1.11)	35	2,667	1.15 (0.82, 1.62)
Site within 10 miles	741	7,905	1.00 (0.93, 1.09)	527	7,905	0.98 (0.89, 1.07)	93	7,905	1.04 (0.83, 1.31)
VOC sites
Site within 3 miles	348	3,916	0.96 (0.86, 1.07)	261	3,916	1.00 (0.88, 1.13)	39	3,916	0.86 (0.62, 1.18)
Site within 5 miles	823	9,149	0.96 (0.89, 1.04)	618	9,149	1.02 (0.93, 1.11)	93	9,149	0.86 (0.69, 1.08)
Site within 10 miles	1,938	20,640	1.03 (0.97, 1.09)	1,415	20,640	1.05 (0.98, 1.13)	220	20,640	0.88 (0.75, 1.05)

Results for residential proximity to Superfund sites using inverse distance weight estimates were consistent with results based on fixed buffers (Supplemental Table 3; https://links.lww.com/EE/A404). We observed similar patterns of associations when considering invasive breast cancer, DCIS, and premenopausal and postmenopausal breast cancer (Supplemental Tables 4 and 5; https://links.lww.com/EE/A404).

In the sensitivity analyses, our results remained similar when restricted to women residing in their homes for 10 years or more and when additionally adjusting for census geographic region (Supplemental Tables 6 and 7; https://links.lww.com/EE/A404).

Discussion

In this large prospective cohort of women living across the United States, we observed only limited evidence of a higher breast cancer incidence associated with residential proximity to Superfund sites overall or for sites based on individual contaminants or contaminant profiles. There was some evidence that living within three miles of two or more sites or sites contaminated with organophosphates was associated with a higher incidence of breast cancer. This is the first study to comprehensively evaluate Superfund sites nationwide in relation to breast cancer incidence.

Although there was limited evidence that living near a single Superfund site contributed to breast cancer incidence, we did observe elevated HRs for women who lived near two or more active Superfund sites. This finding suggests that these women may experience a higher burden of chemical exposure due to their proximity to multiple Superfund sites. This finding may also reflect underlying differences in the areas with multiple Superfund sites compared with those with no or only one Superfund site, such as the overall industrialization, which may result in higher exposure to other potential breast carcinogens. A limitation of this study is that we had limited statistical power to consider distances shorter than three miles, for which associations might be more evident.

We also observed elevated, but imprecise, associations for women who lived near Superfund sites contaminated with organophosphates. Notably, for participants who lived closest to Superfund sites containing organophosphates, the highest HR was observed for the three-mile buffer size, with associations decreasing with increasing buffer sizes. These findings are in line with previous studies that suggest that organophosphates may contribute to the risk of developing breast cancer. For example, breast cancer risk was elevated among women who reported ever using organophosphate pesticides in the Agricultural Health Study.²⁰ In a separate study within the same population, female spouses of pesticide applicators have been found to have a 20% higher incidence of breast cancer.²¹ These epidemiologic findings are supported by experimental studies that have demonstrated that chlorpyrifos, an organophosphate pesticide, promotes breast tumor progression, stimulating cell migration and invasion in vitro as well as tumor growth.^22,23

While there are limited studies that have evaluated proximity to industrial emissions and/or Superfund sites and cancer, a recent study that utilized the Florida Cancer Data system found that living near a single Superfund site in Florida was associated with 27% higher likelihood of developing metastatic breast cancer.²⁴ This study evaluated how exposure to Superfund sites within a census-designed place of their residence was associated with the development of metastatic disease among women with breast cancer. Additionally, the Nurses’ Health Study II found an increased risk of invasive breast cancer in women who resided within 10 km (6.2 miles) of any municipal solid waste incinerator, with positive associations observed for higher dioxin emissions from municipal solid waste incinerators within 3, 5, and 10 km.²⁵ This is consistent with findings from the Sister Study that found an increased risk of breast cancer among women residing within 3 km of industrial sites that had the highest quartile of emissions of polychlorinated dibenzo-p-dioxins and dibenzofurans.²⁶ While we did not observe an association with residential proximity to dioxin-containing Superfund sites in this study or evaluate associations with metastatic breast cancer, these findings highlight the need for more comprehensive studies to better capture residential exposure to chemicals from a variety of potentially relevant point sources.

The Sister Study is a large nationwide prospective cohort of women that has extensive individual-level data on lifestyle and residential characteristics. The study design allowed for adjustment of important demographic and socioeconomic confounders to more rigorously assess the incidence of breast cancer in women that live in close proximity to Superfund sites. As we had geocoded measures of both the participant’s residence and the Superfund site location, we considered multiple indicators of proximity-based exposure including buffers of varying distances and an IDW score. This is an important strength of this study compared with the larger literature on the health impacts of Superfund sites that has largely relied on exposure characterization at a much broader geographic scale, such as the county or zip-code level, which could introduce more measurement error.

An additional strength of this study is our use of contaminant data at each Superfund site to better characterize the different types of chemicals individuals living near sites may be exposed to. The latent class approach was used to identify unique contaminant mixtures that better represent real-life exposure patterns; the utility of these classes was reflected in the fact that these classes appeared to correspond well with industry categories, suggesting that we were capturing important differences in the types of Superfund sites. Additionally, we considered individual contaminant categories, where we were able to focus on more specific exposure groups. However, while sites often contained mixtures of contaminants, the rate of release and rate of migration of different contaminants may vary, so what is present at the site may or may not be what a nearby resident experiences, and the relative proportion of contaminants they are exposed to may differ.

Women with lower socioeconomic status and racial and ethnic minoritized groups often experience disproportionately higher exposure to industrial waste and emissions due to historical and continuing practices of placement of polluting sources in lower-income neighborhoods.^27–30 Additionally, we have previously shown that individuals living closest to the most hazardous waste sites were more likely to be from disadvantaged communities.³¹ We had limited statistical power to explore associations within those subgroups in the present study population, reinforcing the need for additional research that can evaluate the impacts of exposure in these vulnerable populations.

In this study, we assessed exposure to Superfund sites using only the residence at the time of cohort enrollment, which is a proxy for adult-level exposure and may not capture the critical windows of exposure for breast cancer etiology. These critical exposure windows in a woman’s life—such as prenatal development, puberty, pregnancy, and menopause—are marked by changes in mammary tissue that can make breast cells more vulnerable to damage from endocrine-disrupting chemicals, potentially leading to breast cancer later in life.³² Future research should consider how residential exposure to environmental contaminants during these windows may impact later breast cancer incidence.

We focused on residential exposures at baseline and did not account for exposures that may have occurred at the workplace or over follow-up. However, the Sister Study cohort is relatively stable (e.g., 10 years is the median time spent at baseline address), and Superfund sites are not remediated quickly. Although we collect updated residential addresses over follow-up, it is likely that past exposure may be more relevant, considering the potential for a long latency period for breast cancer. Additionally, most Superfund sites are not quickly remediated and thus, we expect exposure to be relatively stable over time at a given residential address. We did not capture whether women had occupations that involved interaction with the Superfund sites. This would be expected to not be very likely as this is a higher socioeconomic cohort, and few women experience any industrial exposure.³³

In this large prospective US-wide cohort, we observed some evidence suggesting that women living near two or more Superfund sites and those living in close proximity to sites contaminated with organophosphate pesticides had a higher incidence of breast cancer. These findings highlight the need for further investigation, particularly in studies that may better capture specific populations that experience a higher burden of exposure. Many individuals are unaware of their proximity to Superfund sites and the possibility that these chemical exposures that could impact their health.

Conflicts of interest statement

The authors declare that they have no conflicts of interest with regard to the content of this report.