Environmental exposures and breast cancer risk in the context of underlying susceptibility: A systematic review of the epidemiological literature

Abstract

Background:

The evidence evaluating environmental chemical exposures (ECE) and breast cancer (BC) risk is heterogeneous which may stem in part as few studies measure ECE during key BC windows of susceptibility (WOS). Another possibility may be that most BC studies are skewed towards individuals at average risk, which may limit the ability to detect signals from ECE.

Objectives:

We reviewed the literature on ECE and BC focusing on three types of studies or subgroup analyses based on higher absolute BC risk: BC family history (Type 1); early onset BC (Type 2); and/or genetic susceptibility (Type 3).

Methods:

We systematically searched the PubMed database to identify epidemiologic studies examining ECE and BC risk published through June 1, 2019.

Results:

We identified 100 publications in 56 unique epidemiologic studies. Of these 56 studies, only 2 (3.6%) were enriched with BC family history and only 11% of studies (6/56) were specifically enriched with early onset cases. 80% of the publications from these 8 enriched studies (Type 1: 8/10 publications; Type 2: 8/10 publications) supported a statistically significant association between ECE and BC risk including studies of PAH, indoor cooking, NO₂, DDT; PCBs, PFOSA; metals; personal care products; and occupational exposure to industrial dyes. 74% of Type 3 publications (20/27) supported statistically significant associations for PAHs, traffic-related air pollution, PCBs, phthalates, and PFOSAs in subgroups of women with greater genetic susceptibility due to variants in carcinogen metabolism, DNA repair, oxidative stress, cellular apoptosis and tumor suppressor genes.

Discussion:

Studies enriched for women at higher BC risk through family history, younger age of onset and/or genetic susceptibility consistently support an association between an ECE and BC risk. In addition to measuring exposures during WOS, designing studies that are enriched with women at higher absolute risk are necessary to robustly measure the role of ECE on BC risk.

Introduction

The increase in breast cancer (BC) incidence rates over time points to a potential role of the environment in underlying BC etiology^1,2. While many studies have examined the role of environmental chemical exposures (ECE) in BC risk, including over 150 publications in the past decade³, only a fraction of human BC studies have specifically examined ECE during a window of susceptibility (WOS)^3,4, including the prenatal, pubertal, pregnancy, and menopausal periods. In addition to limited information about WOS, most epidemiological studies examining ECE and BC risk have focused on populations of average BC risk. A prior family history of BC increases risk from 1.8 fold for women with one first degree relative to over 4-fold for women with three or more first-degree relatives with BC⁵. As such, studies that are unselected for BC family history or higher absolute risk in general, may lack sufficient statistical power to adequately examine associations between ECE and BC risk, particularly for women at intermediate or high absolute BC risk. Women with higher absolute risk have more mutations and/or genetic variants in DNA repair genes, which can result in poorer capacity to repair DNA damage from carcinogens. For example, we have previously demonstrated how using a family-based cohort can be an efficient way to examine an ECE within the context of underlying familial risk and overcome methodological challenges such as insufficient statistical power to examine interactions with predicted absolute BC risk⁶.

For a comprehensive review of ECE and BC risk, we refer the reader to prior systematic and comprehensive reviews, including a recent review by Rodgers et al. that includes a summary of the epidemiologic and biological evidence^3,7,8. These reviews, however, did not consider the body of evidence as it relates to women at higher absolute risk. In the present systematic review, we specifically focus on studies that have examined ECE and BC risk by higher absolute BC risk through either the study design or analysis. Specifically, we focus on studies or subgroup analyses conducted in women at higher absolute BC risk based on three types: 1) BC family history, 2) early onset BC defined as below age 50 or premenopausal, and/or 3) underlying genetic susceptibility. For each type, we examine whether the publication was based on study design enrichment and/or subgroup analyses based on higher BC risk.

Materials and Methods

We conducted a review of scientific publications on ECE and BC risk to identify studies that considered high-risk populations or examined differences in risk by subsets of their population that may be at increased BC risk. We performed a search of the PubMed database to identify all primary epidemiologic studies published through June 1, 2019 in the English language using search terms from Rodgers et al ³ for the ECE coupled with search terms related to the inclusion criteria. Additionally, we searched references of included publications to identify additional eligible publications. A list of all search terms can be found in Supplemental table 1. For this review, we defined family history as any assessment of BC family history including studies that considered any family history, first-degree family history, or used a continuous measure based on pedigree-based algorithms. We defined early onset as premenopausal women or women ≤50 years of age. We included studies that evaluated interaction or stratified by family history, early onset BC, or genetic susceptibility; or examined the association of BC with an ECE in a study enriched for women with underlying familial risk of BC, early onset BC cases, or premenopausal women. Additionally, the study must have evaluated BC as an outcome and had a control group (we excluded case-only analyses). The ECE included in this review are air pollution exposures including vehicular traffic related air pollution, indoor heating and cooking, polycyclic aromatic hydrocarbons (PAH), aromatic DNA adducts, ambient fine-particulate matter (PM_2.5, PM₁₀) and nitrogen dioxide (NO₂); pesticides and fungicides including dichlorodiphenyldichloroethylene (DDE)/dichlorodiphenyltrichloroethane (DDT), organophosphate pesticides, hexachlorobenzene (HCB), and β-hexachlorocyclohexane (β-HCH); organic solvents; per- and polyfluoroalkyl substances (PFAS); phthalates; metals; polychlorinated biphenyls (PCBs); polybrominated diphenyl ether (PBDE); occupational exposures such as exposure to metalworking fluids; and personal care products such as chemical hair straighteners and hair dyes. We grouped chemical exposures into three broad categories: 1) air pollution exposures; 2) persistent endocrine disrupting chemicals (EDCs) (as defined by Rodgers et al ³) including dioxins, PCBs, organochlorine pesticides such as DDT, and a mixture of EDCs measured as total effective xenoestrogens burden (TEXB); and 3) other ECEs including other pesticides, fungicides, PFAS, solvents, phthalates, PBDEs, metals, and personal care products.

Figure 1 is a PRISMA diagram of publications included and illustrates the screening process and workflow used in this systematic review ⁹. We used the Covidence (www.covidence.org) tool to import studies, screen titles and abstracts, and conduct a full text review. A first reviewer screened all publication abstracts for appropriateness, and second and third reviewers read two random subsamples of the publications (20% of all abstracts; N=355 abstracts) following the same screening processes. There was 99% agreement across reviewers. The initial search returned 2201 publications, of which 1765 abstracts and titles were screened after duplicates were removed (n=436). After an abstract review, 1639 publications were excluded because they were not relevant, leaving 126 publications to review as full-text. Of these, 26 publications were excluded for the following: the publication did not meet inclusion criteria of assessing underlying familial risk, early onset BC, or gene-environment interaction (n=10); the publication did not include an environmental exposure (n=5); the publication had the wrong study design (n=5); the publication was an animal study (n=2); the publication’s full text was not in the English language (n=2) the publication was an editorial and not a primary research study (n=1); and the publication did not include BC as an outcome (n=1). The remaining 100 publications, which came from 56 unique epidemiologic studies, are included in the present systematic review (Tables 1a and 1b and Figure 1). For included publications, we extracted data on study population, sample size, exposure assessment, confounding assessment, study design or analysis feature that met the inclusion criteria, and relevant effect estimates and corresponding 95% confidence intervals (CIs) that were related to our inclusion criteria.

Figure 1:

PRISMA diagram of publications included in the study

Table 1a:

Summary of Studies Included in Systematic Review

Study	Inclusion criteria satisfied					Study Design a	Exposure(s)
	Type 1: FH		Type 2: EO		Type 3: GS
	Design	Analysis	Design	Analysis	Analyses
Agricultural Health Study 77–79				✕		Cohort	Organophosphate pesticides, pesticides, and insecticides
Black Women’s Health Study 93				✕		Cohort	Chemical hair straighteners
Breast Cancer Environment and Employment Study (BCEES)30,88				✕		CC	Diesel and gasoline engine exhausts, solvents including benzene, aromatic, aliphatic, chlorinated, and alcohol based solvents
Breast Cancer Family Registry (BCFR)6	✕			✕		NCC	Polycyclic aromatic hydrocarbons (PAH)
California Farm Workers UFW 80				✕		NCC	Pesticides
California Teachers Study (CTS)33,34,44,85,113		✕		✕		Cohort	Pesticides, per- and polyfluoroalkyl substances (PFAS), ambient air pollutants, polybrominated diphenyl ether (PBDE)
Campaign against Cancer and Stroke (CLUE I) 53				✕		NCC	Dichlorodiphenyldichloroethylene (DDE), polychlorinated biphenyls (PCBs)
Campaign against Cancer and Heart Disease (CLUE II) 53				✕	✕	NCC	DDE, PCBs
Carolina Breast Cancer Study 74					✕	CC	PCBs
Case-control study in Brazil60				✕		CC	DDE
Case-control study in Colombia 51		✕		✕		CC	DDE
Case-control study in Long Island, NY61				✕		CC	DDE, PCBs, organochlorine pesticides
Case-control study in New York City62				✕		CC	DDT, DDE, trans-nonachlor, higher (HPCB) and lower (LPCB) chlorinated biphenyls
Case-control study in Toronto and Kingston, Ontario 50		✕		✕			PCBs, DDE, DDT, hexachlorobenzene (HCB), Mirex, β-hexachlorocyclohexane (β-HCH), oxychlordane, trans-nonachlor, cis-nonachlor, aroclor
Case-control study in Sichuan Province, China 39				✕	✕	CC	Dietary exposure to heterocyclic amines (HCAs) and PAHs
Case-control study in Western New York 66					✕	CC	PCBs
CECILE Study 55				✕		CC	Organochlorines
Child Health and Development Studies (CHDS) 52,67,111,112			✕			NCC	PCBs, dichlorodiphenyltrichloroethane (DDT)
Danish Nationwide Cohort Study 90				✕		Cohort	Phthalates
Danish National Birth Cohort 82,84			✕		✕	NCC	PFAS
European Prospective Investigation into Cancer and Nutrition (EPIC) – Italy 25				✕		NCC	Aromatic DNA adducts (include HCA, other bulky adducts, and PAH)
EPIC - Spain 24				✕		Case-cohort	Aromatic DNA adducts
Diet, Cancer, and Health cohort 72					✕	NCC	PCBs
Four hospitals in Nagano Prefecture, Japan 56				✕		CC	PCBs, DDT
French E3N cohort 45,99				✕		Cohort	Cadmium air pollution, airborne dioxin
German GENICA study 41					✕	CC	Aromatic amines and HCA from diet and occupation
Greenlandic Inuit Women 83,86					✕	CC	PFAS, PCBs
Japan Public Health Center-based Prospective Study (JPHC Study) 54				✕		NCC	Organochlorines including DDT, DDE, hexachlorobenzene (HCB), and β-hexachlorocyclohexane (β-HCH)
Long Island Breast Cancer Study 103				✕		CC	Proximity to industrial facilities and traffic
Long Island Breast Cancer Study Project (LIBCSP)10–14,17–22,27,29,32,40,42,47,71				✕	✕	CC	PAH, DDT/DDE (self-report of seeing a fogger truck and biomarkers) PCBs, dietary PAH and HCA
Mexico City 17 tertiary hospitals 92					✕	CC	Phthalates
Mexico City Three Hospitals 57				✕		CC	DDT
Mexico del Saguaro Social Hospitals 69				✕		CC	PCBs
Multicase-Control on Cancer (MCC-Spain) Spain 49		✕		✕		CC	total effective xenoestrogen burden of α (TEXB-α) and β (TEXB-β) fraction
National Enhanced Cancer Surveillance System (NECSS) 38,104			✕			CC	nitrogen dioxide (NO2), residential proximity to industrial plants
New Haven Connecticut Hospitals 64,65,70				✕	✕	CC	PCBs, DDT/DDE
New York University Women’s Health Study63				✕		NCC	DDE, PCBs
Northern California Region Kaiser Permanente Medical Care Program 59				✕		NCC	DDE, PCBs
Northern Mexico case-control100,101				✕	✕		Arsenic
Nurses Health Study (NHS) 73					✕	NCC	PCBs
Nurses Health Study (NHS) II 35,36			✕			Cohort	PAH
Population-based case-control study in Finland 96				✕		CC	Hair dye
Population-based case-control study in Northern states of Mexico 91				✕		CC	Phthalates
Population-based case-control study in Ontario, Canada105				✕		CC	Occupational exposure to carcinogens and endocrine disruptors
Population-based case-control study in Poland 89		✕		✕		CC	Organic solvents
Population-based case-control study in Vancouver and Kingston, Ontario (Canada) 15		✕		✕		CC	PAH
Population-based case-control study in three counties of western Washington state 95			✕			CC	Hair product use
Population-based case-control study of premenopausal BC in western NY23			✕			CC	Occupational PAH and benzene
Population-based case-control study in Wisconsin48		✕		✕		CC	Sport-caught fish consumption as a proxy for PCB, DDT, PBDE, and
Prince Edward Island Hospital 81					✕	CC	Pesticides
Shanghai Women’s Health Study 16,97				✕		NCC, Cohort	PAH, hair dye use
Sister Study26,31,37,46,76,87,94,98,102	✕			✕		Cohort	ambient air pollution, vehicular traffic related air pollution, workplace chemical exposures, pesticides, solvents, metals
United Autoworker-General Motors (UAW-GM) cohort106,107				✕		Cohort	Metalworking fluid
Western New York Exposures and Breast Cancer (WEB) study28,43				✕		CC	traffic emissions, total suspended particles
Women’s Contraceptive and Reproductive Experiences (CARE) Study - Los Angeles Site 58				✕		CC	DDE, PCBs

FH= breast cancer (BC) family history; EO= early onset BC; GS= genetic susceptibility

^a.NCC= Nested case-control study, CC= case-control study

Table 1b.

Number of publications for each exposure, by type based on inclusion criteria satisfied: family history (Type 1), early onset breast cancer (Type 2), or genetic susceptibility (Type 3).^a

Results

We identified 100 publications from 56 distinct epidemiologic studies that met our inclusion criteria by considering family history (Type 1), early onset BC (Type 2), or genetic susceptibility (Type 3) through either the study design or analysis. Figure 2 shows the distribution of these publications by inclusion criteria type. Of the 100 total included publications, 55% (n=55) considered early onset BC only, 23% considered genetic susceptibility only (n=23), and 16% (n=16) considered both family history and early onset BC (see Figure 2). Over 50% of publications (53/100) measured ECE using biospecimens, 17% (17/100) used geographic location to estimate exposure (e.g., detailed residential histories), 12% (12/100) used employment history records or self-report of occupational exposures, 17% (17/100) used self-report for other exposures including exposure to air pollution, pesticides, and personal care products, and 1 study used linkage of a Danish nationwide cohort to a national Prescription Registry to assess longitudinal phthalate exposure. Tables 2a, 3a and 4a summarize the overall findings by inclusion criteria type (family history, early onset BC and genetic susceptibility) for each category of environmental exposure. Detailed information on individual study findings by each category of ECE, with information about study design or analysis feature that met our inclusion criteria, study population, exposure assessment, and confounding adjustment are presented in Tables 2b, 3b, and 4b.

Figure 2:

Distribution of included publications (n=100 publications from 56 unique epidemiological studies) by type based on inclusion criteria satisfied: family history (Type 1), early onset (EO) breast cancer (Type 2), or genetic susceptibility (GS) (Type 3).

Table 2a:

Summary of findings from epidemiologic studies of air pollution including polycyclic aromatic hydrocarbons (PAH), ambient fine-particulate matter (PM2.5, PM10) and nitrogen dioxide (NO2), indoor heating and cooking, vehicular exhaust and breast cancer (BC), grouped by type based on inclusion criteria satisfied: family history (Type 1), early onset BC (Type 2), or genetic susceptibility (Type 3).

Table 3a:

Summary of findings from epidemiologic studies of persistent endocrine disrupting chemicals including organochlorine pesticides, Dichlorodiphenyltrichloroethane (DDT)/dichlorodiphenyltrichloroethane (DDE), and polychlorinated bisphenol (PCB), xenoestrogen burden, and breast cancer (BC), grouped by type based on inclusion criteria satisfied: BC family history (Type 1), early onset BC (Type 2), or genetic susceptibility (Type 3).

Table 4a:

Summary of findings from epidemiologic studies of other pesticides, fungicides, per- and polyfluoroalkyl substances (PFAS), solvents, phthalates, polybrominated diphenyl ethers (PBDE), metals, and personal care products and breast cancer (BC), grouped by type based on inclusion criteria satisfied: family history (Type 1), early onset BC (Type 2), or genetic susceptibility (Type 3).

Table 2b:

Epidemiologic studies of air pollution including polycyclic aromatic hydrocarbons (PAH), ambient fine-particulate matter (PM_2.5, PM₁₀) and nitrogen dioxide (NO₂), indoor heating and cooking, vehicular exhaust and breast cancer (BC), grouped by whether the study considered BC family history, early onset BC, or genetic susceptibility

Cohort Author, year Analysis	Effect Size a	(95% CI)b	Study design or analysis feature that meets inclusion criteria	Study population	Exposure assessment	Confounding	Notes
Family history
Breast Cancer Family Registry (BCFR)
Shen, 2017 6			Family History: Study design enriched for women with underlying familial risk of breast cancer (BC); analysis assessed underlying familial risk of BC as estimated by a 10-year absolute risk score of BC (BOADICEA) that uses pedigree information. Early onset: Analysis was stratified by menopausal status, see main results.	Prospective nested case-control in the New York site of the BCFR. Included 80 prospectively ascertained BC cases and 156 age- and ethnicity matched controls. Over 66% of both cases and controls were premenopausal (66.7% of cases, 71.2% of controls).	Plasma PAH-albumin adducts	Adjusted for age at blood draw, body mass index (BMI) and smoking status	The higher the absolute risk of BC, the higher the association of PAH. Did not report findings for post-menopausal women. P-value for multiplicative interaction = 0.09 (for PAH*BOADICEA)
PAH-albumin adducts (fmol mg−1) (ref = non-detectable adducts)
All women, ≥ median detectable	OR 2.89	(1.25, 6.69)
Premenopausal, ≥ median detectable	OR 2.87	(1.01, 8.17)
Interaction models with 10 year BOADICEA risk score (ref = Non-detect or Detect <median, <3.4%)
Detect ≥ Median, <3.4%	OR 1.81	(0.55, 5.94)
Non-detect or detect <median, ≥3.4%	OR 1.54	(0.66, 4.04)
Detect ≥ Median, ≥3.4%	OR 4.09	(1.38, 12.13)
Sister Study
Shmuel, 201726			Family History: Study design enriched for women with underlying familial risk of BC, specifically having at least one sister diagnosed with BC. Analysis also stratified by family history (1 first-degree relative or 2+ first-degree relatives). Early onset: Analysis was stratified by menopausal status at diagnosis, see main results.	Sister Study: Prospective cohort study which included 42,934 women of the original 50,884 Sister Study participants with complete information on exposure and covariates. 2,028 BC cases diagnosed after 6.3 years mean follow-up. 23% and 28% of cases and controls were <50 years of age at baseline, respectively	Self-reported childhood residential exposure to traffic related air pollution before age 14.	Adjusted for age, race/ethnicity, and highest level of education attained in the household at age 13.	Analysis also included a variable for traffic during rush hour, and a combined measure of distance to road, presence of a median/barrier, and traffic volume- no significant associations found for these constructs in overall or stratified analyses.
Characteristics of the Residential Road at Childhood Residence
Number of Lanes (ref =1-2 lanes)
3+ Lanes, overall association	HR 0.8	(0.6, 1.1)
3+ Lanes, pre-menopausal BC	HR 1.3	(0.8, 2.2)
3+ Lanes, post-menopausal BC	HR 0.8	(0.6, 1.0)
3+ Lanes, 1 first degree relative	HR 0.9	(0.6, 1.2)
3+ Lanes, 2+ first degree relative	HR 0.7	(0.4, 1.2)
Presence of Median/Barrier (ref = without median or barrier)
With Median or Barrier of Any Kind, overall association	HR 1.2	(0.9, 1.7)
With Median or Barrier of Any Kind, pre-menopausal BC	HR 1.5	(0.8, 2.7)
With Median or Barrier of Any Kind, post-menopausal BC	HR 1.1	(0.8, 1.7)
With Median or Barrier of Any Kind, 1 first degree relative	HR 1.3	(0.9, 1.9)
With Median or Barrier of Any Kind, 2 + first degree relatives	HR 1.1	(0.6, 1.9)
White, 2017 31			Family History: Study design enriched for women with underlying familial risk of BC, specifically having at least one sister diagnosed with BC. Also tested for interaction and stratified by family history (1 first degree relative or 2+ first degree relatives) Early onset: Analysis was stratified by menopausal status at diagnosis, see main results	Sister Study: Prospective cohort study, which included 47,512 Sister Study participants. 2,416 BC cases diagnosed after 6.4 years mean follow-up ~32 % of participants were pre-menopausal at baseline, and 22% (n=495) cases were premenopausal at diagnosis	Self-reported indoor heating and cooking practices in adulthood	Premenopausal models adjusted for age, race, education, income, marital status, parity, use of hormonal birth control, use of postmenopausal hormones (HRT), and BMI. Overall models are additionally adjusted for age at menopause and menopausal status.	Stratified analysis by menopausal status and number of first degree relatives were only reported for the binary exposure construct : indoor wood burning stove/fireplace P-value for multiplicative interaction = 0.8 (family history and Indoor wood burning stove/fireplace)
Indoor wood burning stove/fireplace (ref = no stove/fireplace)
Yes, overall association	HR 1.11	(1.01, 1.22)
Yes, pre-menopausal cancers	HR 1.09	(0.90, 1.33)
Yes, post-menopausal cancers	HR 1.1	(0.99, 1.23)
Yes, 1 first degree relative	HR 1.08	(0.96, 1.22)
Yes, 2+ first degree relatives	HR 1.16	(0.99, 1.36)
Frequency of use: at least once a week, overall association	HR 1.17	(1.02, 1.34)
Fuel source for indoor wood burning stove/fireplace (ref = no stove/fireplace)
Wood, overall association	HR 1.09	(0.98, 1.21)
Gas, overall association	HR 1.15	(1.00, 1.32)
Reding, 2015 37			Family History: Study design enriched for women with underlying familial risk of BC, specifically having at least one sister diagnosed with BC.	Sister Study: Prospective cohort study, which included 50,884 Sister Study participants. 1,749 cases BC cases diagnosed after 4.95 years mean follow-up	Annual average concentrations of air pollution at baseline home addresses, derived using regulatory monitoring data for ambient fine-particulate matter (PM2.5, PM10) and nitrogen dioxide (NO2)	Models adjusted for age at diagnosis, race, educational attainment, smoking status, and HRT	Although everyone in the study had a family history of BC, did not test an interaction by degree of family history. Units represent an increase in the IQR difference: PM2.5 = 3.6 μg/m3; PM10 = 5.8 μg/m3; NO2 =5.8 parts per billion (ppb).
Overall
NO2	HR 1.02	(0.97-1.07)
PM2.5	HR 1.03	(0.96-1.11)
PM10	HR 0.99	(0.98-1.00)
ER+/PR+ BC subtype
NO2	RR 1.10	(1.02-1.19)
PM2.5	RR 1.00	(0.91-1.09)
PM10	RR 1.02	(0.96-1.09)
ER−/PR− BC subtype
NO2	RR 0.92	(0.77-1.09)
PM2.5	RR 0.99	(0.81-1.20)
PM10	RR 0.96	(0.83-1.1)
Population-based case-control study in Vancouver, British Columbia and Kingston, Ontario
Lee, 2019 15			Family History: Assessed multiplicative interaction with first-degree family history, see main results Early Onset: Analysis stratified by menopausal status, see main results.	Multicenter, population-based case-control study in Vancouver, British Columbia (BCo) and Kingston, Ontario, between 2005 and 2010. Greater Vancouver: BC cases aged 40-80 years, with in-situ or invasive BC, recruited from province-wide, population-based BCo Cancer Registry. Controls recruited from the BCo Cancer Breast Screening Programme. Kingston: Cases and controls under age 80 recruited from the Hotel Dieu Breast Assessment Programme in Kingston, Ontario. Cases had a diagnosis of in situ or invasive BC, and controls had either normal mammography results or a diagnosis of benign breast disease. 1130 cases and 1169 controls. 14% of controls and 19.8% of cases (n=224) had a family history of BC. 40.5% of controls and 38.4% of cases (n=434) were premenopausal	Lifetime work history, industry, occupation and tasks performed for any job held for at least 6 months, was used to infer PAH exposure using a job exposure matrix (JEM) based on a statistical model of coal tar pitch volatiles (CTPV), a common PAH surrogate. Examined ever-never, duration of exposure, weighted duration, and average probability	Models adjusted for age (continuous), centre (Kingston vs Vancouver), education, ethnicity, and smoking (pack-years)	No difference in overall PAH-BC association by menopausal status (P-interaction<0.2) or any other exposure construct. P-interaction for family history = 0.03 for ever-never at maximum level and 0.03 for duration of exposure at the high level. Results were similar for the association of duration of exposure at the medium and high level (p-interaction = 0.01), weighted duration (p-interaction =0.06), and average probability (p-interaction =0.03), BC risk, and family history).
Ever-never at maximum level (maximum exposure level across all occupations, regardless of duration) (ref=never)
Maximum level at high, first-degree family history	OR 2.27	(1.34-3.86)
Maximum level at high, no family history	OR 1.31	(1.04-1.64)
Maximum level at high, premenopausal	OR 1.53	(1.11-2.11)
Maximum level at high, postmenopausal	OR 1.33	(1.02-1.74)
Long duration (7.5 -74.1 years) of exposure at the high level (ref=none)
first-degree family history	OR 2.79	(1.25-6.24)
no family history	OR 1.29	(0.95-1.75)
premenopausal	OR 1.74	(1.10-2.74)
postmenopausal	OR 1.29	(0.90-1.84)
Early Onset or menopausal status
Breast Cancer Environment and Employment Study (BCEES)
Rai, 2016 30			Early onset: Analysis was stratified by menopausal status at diagnosis, see main results	BCEES: Population-based case-control study of women residing in Western Australia (WA). 1,202 cases aged 18-80 diagnosed between 2009-2011. 1,785 frequency age-matched controls identified from the WA Electoral Roll. 23.5% of controls and 30.8% (n=370) of cases were premenopausal at recruitment	Exposure divided into three categories: diesel exhausts, gasoline exhausts, and other exhausts. Assessment based on self-reported occupational history, using algorithms within OccIDEAS job-specific modules that linked tasks with exposures.	Considered confounding by BC risk factors, final models only included age.	The number of cases for “other exhausts” exposure category includes 5 or less cases, cannot make any conclusions.
Ever exposed to engine exhaust (vs never occupationally exposed)
Diesel exhaust
Overall	OR 1.07	(0.81-1.41)
Premenopausal	OR 1.29	(0.77-2.18)
Postmenopausal	OR 0.99	(0.72-1.38)
Gasoline exhaust
Overall	OR 0.98	(0.74-1.28)
Premenopausal	OR 1.39	(0.80-2.43)
Postmenopausal	OR 0.87	(0.63-1.20)
California Teachers Study (CTS)
Liu, 2015 34			Early onset: Analysis was stratified by menopausal status; however, stratified results are not shown in the study. Authors note they are similar to overall results.	112,379 women enrolled in CTS, unaffected with BC, & living in CA in 1995/1996. 5,361 BC cases diagnosed between 1995-2010 27% of cases (n=1,433) and 43% of controls were pre- and perimenopausal at baseline	Annual ambient air concentrations based on EPA National Air Toxics Assessment (NATA) 2002 estimates for 11 hazardous air pollutants (HAPs) identified as estrogen disruptors at the census tract level were assigned to each CTS participant’s address.	Adjusted for race/birthplace, BC family history, age at menarche, age at first full-term pregnancy (AAFP), menopausal status, HRT , alcohol consumption, smoking, physical activity, and BMI	EPI consists of 9 estrogenic HAPs examined except for diesel engine emissions and selenium compounds.
Exposure potential index (EPI) of nine estrogenic HAPs (vs Quantile 1 (Q1))
Quantile 5	HR 1.04	(0.95-1.13)
Diesel engine emissions (vs Q1)
Quantile 5	HR 1.04	(0.95-1.13)
Selenium compounds (vs Q1)
Quantile 5	HR 1.00	(0.92-1.09)
Garcia, 2015 33			Early onset: Analysis was stratified by menopausal status, however most stratified results are not shown in the study, except for propylene oxide - please see main results.	112,379 women enrolled in CTS, unaffected with BC, & living in CA in 1995/1996. 5,676 BC cases diagnosed between 1995-2010 41.4% of non-cases and 24.8% (n=1,405) of BC cases are premenopausal.	Same method as Liu et al (2015), except estimates were for 24 mammary gland carcinogens (MGCs).	Models stratified by age and adjusted for race	Differences by menopausal status only found for 3 of 24 MCGs: propylene oxide (pre/perimenopausal) and carbon tetrachloride, vinyl chloride (postmenopausal-results not shown for premenopausal)
Summary variable (24 MGCs, vs Q1)
Overall - Quantile 5	HR 1.05	(0.96-1.14)
Propylene oxide (vs. Q1)
Overall - Quantile 5	HR 1.01	(0.93-1.10)
Pre/perimenopausal - Quantile 5	HR 1.15	(0.97-1.35)
Case-control study in Western New York
Petralia, 199923			Early onset: study of premenopausal BC	Case-control study in western New York State of histologically confirmed premenopausal BC identified from all major hospitals in Erie and Niagara counties between 1986-1991. Controls were identified from New York State Department of Motor Vehicles and were frequency-matched to cases by age and county of residence. 301 cases and 316 controls	Job-exposure matrices and lifetime occupational histories for PAH and benzene were used to determine which occupations and occupation and industry combinations involved potential exposure	Adjusted for age, years of education, age at first birth, age at menarche, history of benign breast disease, first degree family history, Quetelet index, and months of lactation.	Authors note that the results need to be interpreted with caution given the small sample sizes, The number of exposed cases ranged from 6 (exclusive PAH) to 56 (Benzene) Also that the PAH results should be interpreted with caution given the inability to examine risk independently of benzene. Study also examined duration, average probability, intensity, cumulative exposure, and latency.
Ever being occupationally exposed to : (ref= never)
PAH	OR 1.82	(1.02, 3.16)
Benzene	OR 1.91	(1.18, 3.08)
Exclusively PAH	OR 1.01	(0.55, 3.45)
PAH and benzene	OR 2.01	(1.08, 3.75)
Exclusively benzene	OR 1.70	(1.17, 2.92)
European Prospective Evaluation into Cancer and Nutrition (EPIC) - Spain
Agudo, 2017 24			Early onset: Analysis was stratified by menopausal status , please see main results Also examined interaction by menopausal status (p-interaction = 0.94)	EPIC-Spain: case-cohort including 41.438 subjects (25, 806 women) ages 29-69 years recruited between 1992-96 in five Spanish provinces This analysis included 305 cases and 149 women in subcohort. 151 cases (49.5%) and 78 in sub-cohort (52.3%) were premenopausal	Aromatic DNA adducts measured in white blood cells by P-post labelling method	Adjusted for age, centre, season of blood extraction, education, physical activity, BMI (interaction with menopausal status), waist circumference, height, age at menopause, age at menarche, age at FFTP, lactation, use of OC, alcohol consumption, total fat intake and energy intake; and HRT (for postmenopausal women) .	P-post-labelling method, while more sensitive than other methods, is unable to specify the exact composition of the detected adducts – and includes a wide range of bulky and hydrophobic adducts in addition to PAH adducts. This analysis did not adjust for smoking status
Aromatic DNA adducts (continuous (log))
Premenopausal BC	RR 1.74	(1.25 – 2.45)
Postmenopausal BC	RR 2.04	(1.36 – 3.05)
EPIC - Italy
Saieva, 2011 25			Early onset: Analysis was stratified by menopausal status, results not shown (see notes)	Nested case-control EPIC-Italy: Italian section of EPIC including on 47,749 Subjects (32,578 women), aged 35–64 years, enrolled in five centres across Italy between 1993-1998 This analysis included 292 breast cancer cases and 292 matched controls	Bulky DNA adducts measured by the P-postlabelling method in peripheral leukocytes	Adjusted for BMI, smoking history, education level, age at menarche, age at first delivery and alcohol consumption.	Authors note that there were no significant results found by models carried out separately in pre- and postmenopausal women,
DNA adduct levels (association for overall BC)
Detectable vs undetectable	OR 1.14	(0.67–1.93)
Continuous scale	OR 1.00	(0.98–1.01)
Tertile 3 of detectable vs undetectable	OR 0.81	(0.45–1.46)
Long Island Breast Cancer Study Project (LIBCSP)
Gammon, 2002 13			Early onset: Analysis was stratified by menopausal status , please see main results	LIBCSP: population-based, case-control study in Nassau & Suffolk counties in NY. 575 BC cases diagnosed between 1996-1997. 424 controls frequency-matched by age, identified either by random digit dialing (<65 years) or Health Care Financing Administration (HCFA) rosters (>=65 years). 33.5% of controls 31.3% of BC cases (n=180) are premenopausal	PAH-DNA adducts using an ELISA	Adjusted for age, race, history of infertility problems, season of blood donation, parity, total months of lactation, BMI at age 20, BC family history, AAFB	Although not statistically significant, there appears to be a difference when stratifying by menopausal status. Gammon, 2004 (below) includes these cases and controls in a pooled analysis and finds differences by menopausal status.
Detectable PAH-DNA adduct levels
Overall	OR 1.35	(1.01–1.81)
Premenopausal	OR 1.58	(0.94–2.66)
Postmenopausal	OR 1.19	(0.82–1.72)
Gammon, 2004 12			Early onset: Analysis was stratified by menopausal status, see main results	LIBCSP: Same as Gammon 2002. This analyses included 873 BC /941 controls 34% of controls 32% of BC cases (n=279) are premenopausal	PAH-DNA adducts using an ELISA	Adjusted for age.	Extension of Gammon, 2002 study- pooled analyses includes all bloods analyzed in both rounds. Interaction term with menopausal status not statistically significant.
Detectable PAH-DNA adduct levels (all subjects)
Overall	OR 1.29	(1.05, 1.58)
Premenopausal	OR 1.56	(1.09, 2.23)
Postmenopausal	OR 1.14	(0.88, 1.47)
Steck, 2007 42			Early onset: Analysis was stratified by menopausal status, see main results	LIBCSP: Same as Gammon 2002, this analysis included 1508 cases and 1556 controls.	Lifetime intakes of grilled or barbecued and smoked meats were derived from the interviewer-administered questionnaire data. Dietary intakes of PAH and heterocyclic amines (HCA) were derived from the self-administered modified Block food frequency questionnaire of intake 1 year before reference date	Adjusted for age, energy intake, fruit and vegetable intake, and multivitamin supplement use.	Authors note that although menopausal status was not an effect modifier on a multiplicative scale, they present all analyses stratified given previous studies that have found this association to vary by menopausal status Also examined smoked and grilled meat separately but there were no associations for premenopausal BC
Total grilled/barbecued and smoked meats
Average over lifetime (highest vs lowest)
Premenopausal	OR 1.13	(0.76–1.68)
Postmenopausal	OR 1.35	(1.02–1.79)
Total overall lifetime (highest vs lowest)
Premenopausal	OR 1.03	(0.68–1.54)
Postmenopausal	OR 1.47	(1.12–1.92)
Total BaPs from food
Premenopausal	OR 1.15	(0.68–1.94)
Postmenopausal	OR 1.01	(0.68–1.50)
HCA – PhIP (highest vs lowest)
Premenopausal	OR 0.83	(0.57–1.21)
Postmenopausal	OR 0.92	(0.70–1.22)
HCA – MeIQx (highest vs lowest)
Premenopausal	OR 0.60	(0.40–0.91)
Postmenopausal	OR 0.94	(0.71–1.25)
HCA -Di MeIQx (highest vs lowest)
Premenopausal	OR 0.59	(0.38–0.91)
Postmenopausal	OR 0.91	(0.66–1.26)
Mordukhovich, 2016a 27			Early onset: Examined effect modification by menopausal status, see main results Gene-environment interaction: examined effect modification by tumor TP53 mutation status	LIBCSP: See Gammon 2002. 1995 analysis: 1274 cases/1334 controls. 34% of controls 32% of BC cases (n=397) were premenopausal. 1960-1990 analyses (combined over 30 imputations): 520-551 cases/566-597 controls. Up to ~23% of controls and ~22% of BC cases (n=124) were premenopausal	Linked participant’s residential histories with measures of historical traffic emissions and estimated PAH exposure using benzo[a]pyrene (BaP) as a surrogate for total PAH exposure. Used the same method as the WEB study	Adjusted for age	Interaction for menopausal status and exposure for ≥ 95th percentile was significant for 1995 exposure (p-value = 0.02) but not 1960-1990 exposure (p-value = 0.50). The authors note that they observed no heterogeneity of the effect estimates for tumor TP53-mutation status subtypes. They also note though that the number of case women with TP53-mutation-positive tumor tissue and high traffic exposure estimates was low.
Traffic PAH exposure classification
1995 (ref= <50th percentile)
75th to < 95th percentile
Premenopausal	OR 1.64	(1.13, 2.38)
Postmenopausal	OR 0.80	(0.62, 1.02)
≥ 95th percentile
Premenopausal	OR 1.20	(0.58, 2.47)
Postmenopausal	OR 1.06	(0.69, 1.63)
1960–1990 (ref=<50th percentile)
50th to < 75th percentile
Premenopausal	OR 0.84	(0.42, 1.66)
Postmenopausal	OR 1.00	(0.70, 1.42)
≥ 75th percentile
Premenopausal	OR 1.31	(0.63, 2.71)
Postmenopausal	OR 0.91	(0.64, 1.29)
National Enhanced Cancer Surveillance System (NECSS)
Hystad, 2015 38			Early onset: Premenopausal women with BC were over-sampled to optimize the statistical power needed to adequately characterize the relationships between several different risk factors and BC in younger women. Results stratified by menopausal status, see main results	NECSS: population-based case–control study conducted in 8 Canadian provinces from 1994-1997. 1569 BC cases and 1872 frequency matched by age population controls. 32.6% of controls and 35.2% of cases (n=619) were premenopausal.	Assigned annual mean exposures (1975-1994) to NO2 to each participant using self-reported residential history and used three approaches: 1. NO2 estimates from 2005-2011 satellite NO2 levels and a chemical transport model 2. Annual rescaling of the NO2 satellite surface using monitoring data, and 3. national land use regression (LUR)	Age, study province, age at menarche, parity, AAFB, breastfed, bilateral oophorectomy, BMI, smoking, years since smoking cessation, alcohol consumption, income, education, second hand smoke, meat & vegetable consumption, physical activity, mammography, neighborhood SES deprivation index, years residing in an urban area	Premenopausal women were defined as those women < 55 years of age who also reported that they were still menstruating one year before interview
Unadjusted satellite NO2 (per 2 ppb)
Overall	OR 1.11	(0.93–1.32)
Premenopausal	OR 1.26	(1.05–1.67)
Postmenopausal	OR 1.10	(0.88–1.36)
Scaled satellite NO2 (per 10 ppb)
Overall	OR 1.13	(0.99–1.27)
Premenopausal	OR 1.32	(0.92–1.74)
Postmenopausal	OR 1.10	(0.94–1.28)
National land use regression (LUR) model (per 10 ppb)
Overall	OR 1.08	(0.90–1.28)
Premenopausal	OR 1.28	(0.92–1.79)
Postmenopausal	OR 1.07	(0.86–1.32)
Nurses’ Health Study (NHS) II
Hart, 2016 36			Early onset: Cohort initially recruited younger women aged 25-42 years. Results stratified by menopausal status, see main results	NHSII: Cohort of 115,921 female nurses with no previous history of cancer enrolled in 1989 when they were aged 25 to 42 years. 3,416 BC cases diagnosed between 1993 – 2011 ~60% of BC cases were premenopausal (n=1,966)	Used geocoded biennial residential addresses to estimate particulate matter (PM) based on monthly spatiotemporal prediction models and measures of distance to roadway	Adjusted for age, race, calendar period, history of benign breast disease (BBD), family history, age at menarche, parity, AAFB, height, BMI (current & age 18), alcohol consumption, overall diet quality, oral contraceptive use, HRT, smoking status, physical activity, socioeconomic status, and region of residence.	Examined distance from: all three road types together, two largest road types (A1, A2), and for the largest road type (A1). Very few cases lived close to A1 and A2 roads, so those results should be interpreted with caution.
Proximity to A1 roads (m, ref= ≥ 200m)
Overall, 0-49 m	HR 1.60	(0.80–3.21)
Premenopausal, 0-49 m	HR 1.74	(0.72–4.21)
Postmenopausal, 0-49 m	HR 1.48	(0.47–4.62)
PM10 (per 10 μg/m3), 48 month average
Overall	HR 1.00	(0.93–1.07)
Premenopausal	HR 1.03	(0.93–1.13)
Postmenopausal	HR 0.97	(0.86–1.09)
PM2.5 (per 10 μg/m3), 48 month average
Overall	HR 0.90	(0.79–1.03)
Premenopausal	HR 0.99	(0.83–1.18)
Postmenopausal	HR 0.76	(0.61–0.95)
Hart, 2018 35			Early onset: Cohort initially recruited younger women aged 25-42 years. Results stratified by menopausal status, see main results. Overall results for some HAPs not reported	NHSII: Cohort of 109,239 female nurses with no previous history of cancer enrolled in 1989 when they were aged 25 to 42 years. 3321 BC cases diagnosed between 1993 – 2011 72% of entire NHSII study participants and 62% of BC cases (n=2,059) were premenopausal.	Annual ambient air concentrations based on EPA NATA 2002 estimates for hazardous air pollutants (HAPs) identified as estrogen disruptors (9 HAPs) and mammary carcinogens (23 HAPs) at the census tract level were assigned to each participant’s geocoded address.	Adjusted for age, calendar period, race, family history of BC, BBD, age at menarche, parity, AAFB, menopausal status, HRT, oral contraception use, recent mammogram, height, BMI (age 18), current BMI-BMI age 18, smoking status, physical activity, overall diet quality, alcohol consumption, shift work, individual-level SES, area-level SES, and Census region of residence	Chose HAPs in similar method as Liu, 2015 analyses in California Teachers’ Study
Hazardous Air Pollutant, Quartile 4 vs Q1
1,2-Dibromo-3-Chloropropane
Premenopausal	HR 1.16	(0.97, 1.39)
Benzene (Including Benzene From Gasoline)
Premenopausal	HR 1.07	(0.93, 1.22)
Diesel Engine Emissions
Overall	HR 1.10	(0.99, 1.22)
Premenopausal
Methylene Chloride (Dichloromethane)
Premenopausal	HR 1.08	(0.93, 1.25)
Dimethyl Formamide
Overall	HR 1.08	(0.97, 1.20)
Premenopausal	HR 1.13	(0.99, 1.29)
4-Nitrophenol
Overall	HR 1.07	(0.96, 1.19)
Premenopausal	HR 1.11	(0.97, 1.27)
Shanghai Women’s Health Study (SWHS)
Lee, 2010 16			Early onset: Analysis was stratified by menopausal status; however, stratified results are not shown in the study. Authors note they found no significant associations, similar to overall findings	Nested case-control in the SWHS. 327 cases identified through participant follow-up and linkage with tumor registry from 1997-2004. 654 controls matched by age, sample collection date, collection time of day, antibiotic use in the past week, and menopausal status. 51% of controls & cases (n=182) are premenopausal.	Measured urinary 1-hydroxypyrene and 2-naphthol as PAH metabolites using reverse-phase high-performance liquid chromatography (HPLC)	Adjusted for age of baseline, sample collection date, antibiotic use, previous cancer history, and menopausal status.
1-hydroxypyrene (μmol/mol creatinine)
Quartile 4 vs Quartile 1	OR 0.91	(0.63-1.32)
2-naphthol (μmol/mol creatinine)
Quartile 4 vs Quartile 1	OR 0.83	(0.58-1.21)
Western New York Exposures and Breast Cancer (WEB) study
Nie, 2007 28			Early onset: Analysis was stratified by menopausal status. No postmenopausal results were reported for exposure at menarche.	WEB study: population-based case-control of women aged 35–79 residing in Erie and Niagara Counties in NY. Total number of cases and controls vary by the time period of exposure. Premenopausal: Cases: 181-258, controls: 347-501. Postmenopausal: Cases: 52-717, controls: 76-1265. Numbers are in ranges since the study was analyzed in different time periods.	Linked participant’s residential histories with measures of historical traffic emissions and estimated exposures using a geographic traffic model for time periods of potential breast tissue sensitivity. The model estimated PAH exposure using benzo[a]pyrene (BaP) as a surrogate for total PAH exposure	Adjusted for age, education, race, BMI, age at menarche, age at menopause (for post-menopausal women only), age at first birth, number of births, family history of BC, BBD, and year at interview. Stratified models by smoking status adjusted for age, race, education, AAFB, and year at interview	The observed increased risks were limited to nonsmokers in stratified analyses by smoking status.
Traffic-related PAH exposure (ref = Quartile 1)
At menarche
Premenopausal, Q4	OR 2.07	(0.91–4.72)
Premenopausal, non-smokers, Q4	OR 6.67	(1.74–25.67)
Premenopausal, ever smokers, Q4	OR 0.80	(0.27–2.36)
At first birth
Premenopausal, Q4	OR 1.22	(0.44–3.36)
Premenopausal, non-smokers, Q4	OR 2.06	(0.44–9.73)
Premenopausal, ever smokers, Q4	OR 0.82	(0.23–2.92)
Postmenopausal, Q4	OR 2.58	(1.15–5.83)
Postmenopausal, non-smokers, Q4	OR 6.23	(1.70–22.82)
Postmenopausal, ever smokers, Q4	OR 1.35	(0.47–3.83)
20 years prior
Premenopausal, Q4	OR 1.29	(0.59–2.82)
Postmenopausal, Q4	OR 0.82	(0.58–1.18)
10 years prior
Premenopausal, Q4	OR 1.49	(0.65–3.43)
Postmenopausal, Q4	OR 0.80	(0.55–1.17)
Bonner, 200543			Early onset: Analysis was stratified by menopausal status, see main results	WEB study, see Nie 2007. Premenopausal: Cases: 325; controls: 610 Postmenopausal: Cases: 841; controls 1,495	Total suspended particulates (TSP) was used as a proxy for PAHs exposure. Annual average TSP concentrations (1959- 1997) were obtained from NY state monitors, and prediction maps of TSP concentrations were generated using geospatial methods, and were used to determine TSP exposure at each participant’s address for the relevant time period	Adjusted for age, education, and parity.
TSP (μ/m3) (ref= Quartile 1)
TSP concentrations at birth address
Premenopausal, quartile 4	OR 1.78	(0.62-5.10)
Postmenopausal, quartile 4	OR 2.42	(0.97-6.09)
TSP concentrations at menarche address
Premenopausal, quartile 4	OR 0.66	(0.38-1.16)
Postmenopausal, quartile 4	OR 1.45	(0.74-2.87)
TSP concentrations at first birth address
Premenopausal, quartile 4	OR 0.52	(0.22-1.20)
Postmenopausal, quartile 4	OR 1.33	(0.87-2.06)
Genetic susceptibility
Case-control study in Sichuan Province, China
Lee, 201239			Gene-environment interaction: evaluated the SULT1A1 gene polymorphism, dietary exposure to HCAs and PAHs, and BC risk Early onset: Analysis was stratified by menopausal status, see main results	Case control study of women of Han ethnicity, aged 30–70 years old, and had been living in Sichuan for at least 20 years. Between May 2007 and July 2009, 400 women newly diagnosed with BC via pathology in The Second People’s Hospital of Sichuan Province were recruited. 400 community-based women undergoing routine physical examinations at the Chengdu Municipal Center for Disease Control and Prevention were randomly selected as control subjects. Each control was matched to one patient by age and menopausal status.	All participants were asked about their long-term (≥5 years) dietary habits. Since there is no standardized FFQ for Chinese populations, a semi-quantitative dietary questionnaire was developed, which Included questions about daily dietary intake frequency and quantity, and was designed according to the Chinese Nutrition Association Dietary Guidelines.	Adjusted for BMI, smoking, no. of abortions, oral contraceptive use, total protein intake, and total fat intake.	Developed questionnaire includes six types of food and some special dietary components, such as smoked meat, which are frequently consumed in Sichuan. Smoked meat intake interacted positively with the His variant allele (all g > 1).
Energy-adjusted daily intake of smoked meat (g/day) (ref= low)
Overall, high intake	OR 2.58	(1.82–3.65)
_ Premenopausal, high intake	OR 2.31	(1.46–3.66)
_ Postmenopausal, high intake	OR 3.13	(1.89–5.17)
SULT1A1 Polymorphism & energy-adjusted daily intake of smoked meat (g/day) (ref= low intake and Arg/Arg genotype)
Premenopausal, high intake
Arg/Arg	OR 2.01	(1.19–3.38)
His/His + Arg/His	OR 3.31	(1.66–6.62)
Postmenopausal, high intake
Arg/Arg	OR 2.45	(1.40–4.31)
His/His + Arg/His	OR 3.81	(1.79–8.10)
Overall, high intake
Arg/Arg	OR 2.10	(1.43–3.09)
His/His + Arg/His	OR 3.39	(2.02–5.67)
German GENICA Study
Rabstein, 201041			Gene-environment interaction: explored the risks of potential sources of aromatic and heterocyclic amines (AHA) exposure and deduced acetylation status, based on NAT2 polymorphisms.	GENICA is a population-based case-control study conducted in the Greater Region of Bonn, Germany of Caucasian women aged > 80 years and includes 1155 incident BC cases and 1143 population controls enrolled between August 2000 and September 2004. Controls were frequency matched to cases by year of birth in 5-year classes. Cases were women with histopathologically confirmed BC with diagnosis within 6 months before enrolment	Smoking, diet, and occupation were considered as sources of AHA. Women were classified as current smokers if they were regularly smoking one or more cigarettes per day at interview or in the year before. Consumption of meat and grilled food was based on the food-frequency information An expert rating was applied to assess possible occupational exposure based on self-assessed tasks.	Logistic regression conditional on age, adjusted for family history of BC, hormonal therapy, breast feeding, physical activity, and number of mammograms until 2 years before interview.	Job tasks with a supposed exposure to aromatic amines comprised the developing of films, preparing and mixing of materials in the rubber industry, using dyes for coloring of hair, leather, textiles, paper, painting, and working with tar or tar products.
Consumption of red meat, (ref = rare consumption and slow acetylators)
Slow acetylators, regular consumption	OR 1.71	(1.15–2.55)
Fast acetylators, regular consumption	OR 1.73	(1.15–2.61)
Consumption of grilled food, (ref = rare consumption and slow acetylators)
Slow acetylators, regular consumption	OR 1.70	(1.01–2.88)
Fast acetylators, regular consumption	OR 1.64	(0.83–3.21)
Occupational exposure to aromatic and heterocyclic amines (years) (ref = None or <1 and slow acetylators)
Slow acetylators, ≥ 1 years	OR 0.88	(0.51–1.55)
Fast acetylators, ≥ 1 years	OR 1.32	(0.74–2.37)
Long Island Breast Cancer Study Project (LIBCSP)
Terry, 2004 21			Gene-environment interaction: Evaluated exon 23 polymorphism in nucleotide excision repair gene, XPD.	LIBCSP: population-based, case-control study in Nassau & Suffolk counties in NY. 1053 BC cases diagnosed between 1996-1997. 1102 controls frequency-matched by age, identified either by random digit dialing (<65 years) or Health Care Financing Administration (HCFA) rosters (>=65 years).	PAH-DNA adducts using an ELISA	adjusted for age, race, menopausal status, educational level, and first-degree family history of BC	Multiplicative interaction term for PAH adducts and XPD genotypes was not statistically significant (P = 0.48). There was significant additive interaction between XPD genotype and PAH-DNA adducts
PAH-DNA adducts & XPD genotypes
Multiplicative interaction
Nondetectable (vs nondetectable, AA genotype)
AC	OR 1.25	(0.83-1.86)
CC	OR 0.91	(0.52-1.62)
Detectable, median and above (vs detectable, AA genotype)
AC	OR 1.22	(0.85-1.76)
CC	OR 1.61	(0.99-2.63)
Additive interaction
Detectable, median and above (vs nondetectable adducts, AA genotype)	OR 1.90	(1.15-3.15)
Shen, 2005 19			Gene-environment interaction: Evaluated 2 polymorphisms in base excision repair gene, XRCC1	LIBCSP: See Terry 2004. 866 BC/938 controls.	PAH-DNA adducts using an ELISA	Adjusted for age at reference.	Also reported findings for ever smoking women - no significant increased BC risk in that group. No significant interaction was observed between codon 194Trp or 399Gln and PAHDNA adducts. Non-smokers had substantially elevated risk associated with a 399Gln allele and detectable adducts.
PAH-DNA adducts & XRCC1 genotypes
Codon 194 (Arg/Arg, nondetectable (ND) PAH-DNA adducts = ref) Overall
Arg/Arg, detectable adducts	OR 1.27	(1.02-1.59)
Arg/Trp or Trp/Trp, ND adducts	OR 0.90	(0.55-1.49)
Arg/Trp or Trp/Trp, detectable adducts	OR 1.28	(0.89-1.83)
Never smoking women
Arg/Arg, detectable adducts	OR 1.53	(1.10-2.11)
Arg/Trp or Trp/Trp, ND adducts	OR 1.05	(0.51-2.17)
Arg/Trp or Trp/Trp, detectable adducts	OR 1.30	(0.76-2.22)
Codon 399 (Arg/Arg, ND PAH-DNA adducts = ref) Overall
Arg/Arg, detectable adducts	OR 1.23	(0.88-1.73)
Arg/Gln or Gln/Gln, ND adducts	OR 1.00	(0.69-1.43)
Arg/Gln or Gln/Gln, detectable adducts	OR 1.33	(0.97-1.84)
Never smoking women
Arg/Arg, detectable adducts	OR 1.36	(0.84-2.19)
Arg/Gln or Gln/Gln, ND adducts	OR 1.18	(0.70-1.98)
Arg/Gln or Gln/Gln, detectable adducts	OR 1.92	(1.21-3.07)
Shen, 2006 20			Gene-environment interaction: Evaluated 1 polymorphism in, IGHMBP2, a gene involved in DNA repair, replication and recombination	LIBCSP: See Terry 2004. 866 BC/938 controls.	PAH-DNA adducts using an ELISA	Adjusted for age at reference.	Multiplicative interaction term was not significant (p-value= 0.80)
PAH-DNA adducts & IGHMBP2 Thr671Ala polymorphisms (Non detectable (ND) PAH-DNA adducts, AA genotype = ref)
AA genotype, detectable adducts	OR 1.2	(0.9–1.6)
AG +GG genotypes, ND adducts	OR 1.0	(0.7–1.4)
AG +GG genotypes, detectable adducts	OR 1.4	(1.0–1.8)
Crew, 2007a 10			Gene-environment interaction: Evaluated polymorphisms in, FAS and FASL, apoptosis associated genes	LIBCSP: See Terry 2004. 873 BC cases /941 controls.	PAH-DNA adducts using an ELISA	Adjusted for age at reference.	Although the presence of at least one variant allele in FAS1377 was associated with an increased BC risk in those with detectable DNA adducts, the interaction term was not statistically significant (P-value = 0.20). No association found between FASL genotype, PAH-DNA adducts and BC risk (data not shown).
PAH-DNA adducts & FAS genotypes
FAS1377 (GG genotype, nondetectable PAH-DNA adducts = ref)
GA+AA, Non-detectable adducts	OR 0.81	(0.53–1.24)
GG+AA, detectable adducts	OR 1.36	(1.01–1.83)
GG, detectable adducts	OR 1.21	(0.96–1.53)
FAS670 (GG genotype, nondetectable PAH-DNA adducts = ref)
GA, Non-detectable adducts	OR 0.99	(0.64–1.52)
AA Non-detectable	OR 0.85	(0.51–1.41)
GG Detectable	OR 1.18	(0.78–1.77)
GA Detectable	OR 1.21	(0.83–1.78)
AA Detectable	OR 1.36	(0.89–2.07)
Crew, 2007b 11			Gene-environment interaction: Evaluated polymorphisms in nucleotide excision repair genes: ERCC1, XPA, XPD, XPF, XPG	LIBCSP: See Terry 2004. 873 BC cases /941 controls.	PAH-DNA adducts using an ELISA	Adjusted for age at reference.	The association between BC and the homozygous ERCC1 or XPD Asp312Asn variant genotypes was increased in women with detectable PAH-DNA adducts, but the interaction term was only significant for XPD Asp312Asn (p-value: 0.02) but not ERCC1 (P-value: 0.43). Found no association between XPA, XPF, and XPG genotypes, PAH-DNA adducts, and BC risk.
PAH-DNA adducts & Genotypes
ERCC1 8092 (CC genotype, nondetectable PAH-DNA adducts = ref)
CC, detectable adducts	OR 1.21	(0.91-1.61)
CA, detectable adducts	OR 1.24	(0.92-1.67)
AA, detectable adducts	OR 1.92	(1.14-3.25)
XPD Asp312Asn (GG genotype, nondetectable PAH-DNA adducts = ref)
GG, detectable adducts	OR 1.45	(1.05-2.00)
GA, detectable adducts	OR 1.56	(1.13-2.15)
AA, detectable adducts	OR 1.83	(1.22-2.76)
Gaudet, 2008 14			Gene-environment interaction: Evaluated 3 polymorphisms in TP53, a tumor suppressor gene	LIBCSP: See Terry 2004. 492 BC cases/ 345 controls	PAH-DNA adducts using an ELISA	Adjusted for age at reference.	Also reported ORs for non-detectable adducts. ORs for genotypes of all loci were similar for women with and without detectable PAH–DNA adduct levels.
Detectable PAH-DNA adducts
rs1042522 (ref= CC genotype, detectable adducts)
CG	OR 1.25	(0.88, 1.77)
GG	OR 1.28	(0.68, 2.40)
rs17878362 (ref = DEL/DEL, detectable adducts)
DEL/INS	OR 0.97	(0.67, 1.42)
INS/INS	OR 1.73	(0.53, 5.64)
rs1625895 (ref= AA genotype, detectable adducts)
AG	OR 0.97	(0.66, 1.43)
GG	OR 2.02	(0.54, 7.63)
Shen, 2008 18			Gene-environment interaction: Evaluated 2 polymorphisms in XPC, a nucleotide excision repair gene	LIBCSP: See Terry 2004. 873 BC cases /941 controls.	PAH-DNA adducts using an ELISA	Adjusted for age at reference.	No statistically significant interaction was observed for any of the genotypes or diplotypes examined.
Detectable PAH-DNA adducts
XPC Ala499Val (ref= CC genotype, non-detectable adducts)
CC	OR 1.3	(1.0–1.8)
CT/TT	OR 1.3	(0.9–1.7)
XPCLys939Gln (ref= AA genotype, non-detectable adducts)
AA	OR 1.5	(1.0–2.1)
AC/CC	OR 1.5	(1.1–2.0)
XPC diplotypes (ref=all other diplotypes combined, non-detectable adducts)
All other diplotypes combined, detectable adducts	OR 1.3	(1.0–1.6)
CC-CC diplotype, detectable adducts	OR 1.6	(1.1–2.2)
McCarty, 2009 17			Gene-environment interaction: Evaluated multiple polymorphisms in GST gene, glutathione S-transferases. GST genes code for enzymes critical in the metabolism of toxins, including PAH.	LIBCSP: See Terry 2004. 873 BC cases /941 controls.	PAH-DNA adducts using an ELISA	Adjusted for age at reference.	No statistically significant interactions were observed for any of the genotypes when examined individually. While the OR for women with 3 variants and detectable adducts is elevated, the interaction was not statistically significant (p=0.43)
Number of variants & PAH-DNA adducts (ref=4 common variants, non-detectable adducts)
One variant, detectable adducts	OR 0.97	(0.78, 1.20)
One variant, non-detectable adducts	OR 0.56	(0.40, 0.77)
Two variants, detectable adducts	OR 0.82	(0.66, 1.03)
Two variants, non-detectable adducts	OR 0.78	(0.56, 1.07)
Three variants, detectable adducts	OR 1.56	(1.13, 2.16)
Three variants, non-detectable adducts	OR 0.93	(0.56, 1.56)
Four variants, detectable adducts	OR 1.01	(0.71, 1.43)
White, 2014 32			Gene-environment interaction: Evaluated multiple polymorphisms in GST gene, glutathione S-transferases. GST genes code for enzymes critical in the metabolism of toxins, including PAH. Early onset: tested for interaction with menopausal status (results not shown)	LIBCSP: See Terry 2005. 1508 BC cases /1556 controls	Indoor stove/fireplace use assessed using a structured questionnaire.	Adjusted for age, age at menarche, history of breastfeeding, hormone therapy use, family history of BC, parity, age at first birth, BMI, education, smoking history, alcohol intake, physical activity, race, religion, marital status.	Authors note that wood burning excludes synthetic logs. Interaction term was not statistically significant.
Indoor stove/fireplace use & number of GST variants (ref=no stove/fireplace, <2 variants)
Ever any stove fireplace, < 2 variants	OR 1.04	(0.93, 1.16)
Ever any stove fireplace, ≥ 2 variants	OR 1.13	(0.81, 1.57)
Ever wood burning, < 2 variants	OR 1.02	(0.91, 1.15)
Ever wood burning ≥ 2 variants	OR 1.07	(0.76, 1.51)
Ever synthetic log burning, < 2 variants	OR 1.19	(1.03, 1.39)
Ever synthetic log burning ≥ 2 variants	OR 1.71	(1.09, 2.68)
White, 2015 22			Gene-environment interaction: Evaluated interaction with gene-specific promoter methylation status in tumor tissue for 13 BC related genes (ESR1, PGR, BRCA1, APC, CCND2, CDH1, DAPK1, GSTP1, H1N1, CDKN2A, RARϐ, RASSF1A, and TWIST1) which include both steroid hormone genes and tumor suppressor genes	LIBCSP: See Terry 2004. 873 BC cases /941 controls.	PAH-DNA adducts using an ELISA	Adjusted for age	Multiplicative interaction p-value for RARb = 0.03 and APC = 0.09 indicates that the odds of having an ER+PR+ tumor, given PAH-DNA adduct level, is statistically different across strata of methylated breast tumor. No findings for PAH-DNA adducts and BC by methylation status of other genes. Also examined global methylation, and interaction with PAH-DNA adducts and report no significant findings.
PAH-DNA detectable adducts (vs nondetectable adducts)
RARB : ER+PR+ cases verses controls
Methylated breast tumor	OR 2.15	(1.03, 4.47)
Unmethylated breast tumor	OR 1.05	(0.73, 1.51)
RARB : ER+PR+ cases verses all other (ER+PR−, ER−PR+, ER−PR−) cases
Methylated breast tumor	ROR 2.69	(1.02, 7.12)
Unmethylated breast tumor	ROR 0.79	(0.42, 1.46)
APC: ER+PR+ cases verses controls
Methylated breast tumor	OR 1.59	(0.98, 2.58)
Unmethylated breast tumor	OR 1.14	(0.74, 1.76)
APC: ER+PR+ cases verses all other (ER+PR−, ER−PR+, ER−PR−) cases
Methylated breast tumor	ROR 1.76	(0.87, 3.58)
Unmethylated breast tumor	ROR 0.73	(0.35, 1.53)
Mordukhovich, 2016b 29			Gene-environment interaction: Evaluated interaction with polymorphisms in nucleotide excision repair genes and base excision repair genes: ERCC1, XRCC1, OGG1, ERCC2, XPA, ERCC4, and ERCC5.	LIBCSP: See Terry 2004. For 1995 traffic estimates, 842–905 cases and 911–958 controls; 1960–1990 traffic estimates, 332–429 cases and 368–474 controls	Linked participant’s residential histories with measures of historical traffic emissions and estimated exposures using a geographic traffic model for time periods of potential breast tissue sensitivity. The model estimated PAH exposure using benzo[a]pyrene (BaP) as a surrogate for total PAH exposure. Used the same method as the WEB study	Adjusted for age	In analyses examining each polymorphism separately, only one was significant for 1995 analyses: homozygous variant genotype for the ERCC2 Lys751Gln polymorphism: OR: 2.09 (95% CI: 1.13, 3.90). For 1960-1990 analysis: the homozygous major genotype for XRCC1 Arg399Gln (OR:1.88, 95% CI: 1.04, 3.41); and the homozygous major genotype for OGG1 Ser326Cys (OR:1.77, 95% CI: 1.09, 2.88)
Tertiles of exposure by number of high risk alleles in ERCC2, XRCC1, & OGG1 (vs tertile 1)
1995 analysis
0-1 high risk alleles, tertile 3	OR 0.8	(0.49, 1.57)
2-3 high risk alleles, tertile 3	OR 0.92	(0.70, 1.20)
4-6 high risk alleles, tertile 3	OR 2.32	(1.22, 4.49)
1960-1990 analysis
0-1 high risk alleles, tertile 3	OR 0.89	(0.36, 2.22)
2-3 high risk alleles, tertile 3	OR 1.25	(0.80, 1.95)
4-6 high risk alleles, tertile 3	OR 2.96	(1.06, 8.21)
Parada, 201740			Gene-environment interaction: examined association of 22 polymorphisms in four genes of phase I metabolizing CYP enzymes and interaction with grilled-smoked meat intake	LIBCSP: See Terry 2004. 988 cases/ 1021 controls	LIBCSP participants completed an interviewer-administered main questionnaire and a self-administered food frequency questionnaire.	Adjusted for age at diagnosis, energy intake, fruit and vegetable intake, and multivitamin supplement use.	observed multiplicative and additive interactions (P <.05) between grilled /smoked meat intake (low vs. high) with CYP1A1 rs1048943 and CYP1B1 rs10175338 SNPs
CYP1A1 (rs1048943) and Lifetime servings of grilled and smoked meat
AA , high lifetime servings (ref= AA, low)_	OR 1.21	(0.94, 1.55)
AG + GG, high lifetime servings (ref= AG +GG, low)	OR 0.64	(0.23, 1.80)
CYP1B1 (rs10175338) and Lifetime servings of grilled and smoked meat
GG , high lifetime servings (ref= GG, low)_	OR 1.59	(1.15, 2.20)
GT+ TT, high lifetime servings (ref= GT +TT, low)	OR 0.71	(0.50, 1.02)
CYP3A4 (rs2242480) and Lifetime servings of grilled and smoked meat
CC , high lifetime servings (ref= CC, low)_	OR 1.01	(0.78, 1.30)
CT+ TT, high lifetime servings (ref= CT +TT, low)	OR 1.59	(0.91, 2.77)

^a.OR = Odds Ratio, HR= Hazard Ratio, RR= relative risk, ROR= ratio of the odds ratio

^b.95% CI = 95% Confidence interval

Table 3b:

Epidemiologic studies of persistent endocrine disrupting chemicals including organochlorine pesticides, Dichlorodiphenyltrichloroethane (DDT)/dichlorodiphenyltrichloroethane (DDE), and polychlorinated bisphenol (PCB), xenoestrogen burden, and breast cancer (BC), grouped by whether the study considered BC family history, early onset BC, or genetic susceptibility

Cohort Author, year Analysis	Effect Sizea	(95% CI)b	Study design or analysis feature that meets inclusion criteria	Study population	Exposure assessment	Confounding	Notes
Family history
California Teachers Study (CTS)
Reynolds, 200444			Family history: Analysis was stratified by family history*, however stratified results are not shown in the study. Authors note they are similar to overall results. *Family history is in the first-degree relatives Early onset: Analysis was stratified by menopausal status, however stratified results are not shown in the study. Authors note they are similar to overall results.	Prospective cohort study of California teachers. Included 114,835 study participants who were successfully geocoded, lived in California from 1996 to 1999, and had no prior history of BC. 1552 BC cases diagnosed between 1996-1999 38.3% of cohort was premenopausal and 11.8% had a family history.	Linked participant’s residential histories with California’s Department of Pesticide Regulation PUR database on agricultural pesticide application within the state from 1993-1995 using ArcView GIS software. Exposure assessment grouped by percentiles of lb/mi2 of usage: <1 lb/mi2, 1st-49th percentile, 50th-74th percentile, and ≥ percentile.	Adjusted for age, race/ethnicity, socioeconomic status and urbanization.	Pesticides were grouped by similar toxicity endpoints relevant to BC into the following six groups: probable or likely human carcinogens, possible or suggestive human carcinogens, mammary carcinogens, endocrine disruptors, anticholinesterases, and organo- chlorines. Simazine was individually analyzed based on established toxicological data implicating its role in mammary tumorigenesis and endocrine disruption. Also individually examined diuron, oryzalin, propargite, and methyl bromide and reported no association.
Toxicological groupings
Probable or likely human carcinogens (ref=<1lb/mi2)
≥75th percentile (≥175 lb/mi2)	HR 1.07	(0.86, 1.32)
Probable or suggestive human carcinogens (ref=<1lb/mi2)
>75th% (>99 lb/mi2)	HR 1.06	(0.87, 1.29)
Mammary carcinogens (ref=<1lb/mi2)
≥75th percentile, (≥58 lb/mi2)	HR 1.15	(0.90, 1.40)
Endocrine disruptors (ref=<1lb/mi2)
≥75th percentile (≥324 lb/mi2)	HR 1.03	(0.86, 1.25)
Anticholinesterases (ref=<1lb/mi2)
≥75th percentile (≥111 lb/mi2)	HR 1.09	(0.89, 1.33)
Organochlorines (ref=<1lb/mi2)
≥75th percentile (≥18 lb/mi2)	HR 0.99	(0.63, 1.55)
Individual pesticide groupings
Simazine (ref=<1lb/mi2)
≥75th percentile(≥42 lb/mi2)	HR 1.11	(0.81, 1.50)
Sister Study
Niehoff, 201646			Family history: Cohort enriched for women with a family history of BC. Early onset: Analysis was stratified by menopausal status, see main results	Sister Study: Prospective cohort study which included 50,756 women of the original 50,884 participants from the US and Puerto Rico who had a family history of BC and complete information on exposure and covariates. 2134 BC cases diagnosed after 5.3 mean years of follow-up. ~23% of cases were premenopausal.	Participants were asked to provide residential history prior to age 14 details including: years of residence, property type, pesticide treatment, and proximity to several land use types. Women also asked to report exposure to fog or spray of chemicals.	Adjusted for age, race, age at menarche, parity and breastfeeding	Also evaluated association between other measures of pesticide exposure during childhood and adolescence (residential type, use of a residence as a farm or orchard, and proximity to facilities that often use pesticides such as orchards, commercial nurseries, golf courses)and BC risk, and reported null HRs. The authors note that peak use of DDT was in in 1959, so they examined women specifically born in 1941-1958 (who would have been ages 0-18 in 1959). In restricted analysis of women born in 1944-49 (pubertal at time of peak DDT use), HRs were the same, but there were too few cases to examine premenopausal BC.
Farm exposures ages 0-18
Lived on cotton or tobacco farm vs never lived on farm
overall association	HR 1.3	(1.0., 1.6)
ER+/PR+ BC subtype	HR 1.3	(0.93, 1.9)
premenopausal	HR 1.0	(0.55, 2.0)
postmenopausal	HR 1.3	(1.0, 1.6)
Fogger truck or airplane exposure
Ever exposed to fog or spray and born between 1941-1958
overall association	HR 1.1	(0.99, 1.3)
ER+/PR+ BC subtype	HR 1.1	(0.88, 1.3)
premenopausal	HR 1.3	(0.92, 1.7)
postmenopausal	HR 1.0	(0.90, 1.2)
Spanish Multicase-Control Study on Cancer (MCC-Spain)
Pastor-Barriuso, 201649			Family history: stratified analyses by family history of BC (in both first and second degree relatives). Early onset: stratified by menopausal status, see main results	Population based case-control study between 2008 and 2013 in 12 Spanish provinces. 186 cases and 196 controls frequency matched by province, 5-year age interval, and BMI. 15.4% of controls and 21% of cases (n=39) have a family history of BC. 14.3% of controls and 17.2% of cases (n=32) were premenopausal at baseline.	Measured the total effective xenoestrogen burden (TEXB) in serum samples using a standardized bioassay for the combined estrogenic effect of mixtures of xenoestrogens. Serum TEXB-α, TEXB-β, PCB-138, PCB-153, PCB-180, HCB, p,p′-DDE	Adjusted for province, BMI, Education, serum total lipid levels, smoking, number of births, age at first birth, menopausal status, HRT use, BC family history,	Multiplicative interaction of TEXB-α tertile 3 with family history (p-value: 0.49) and menopausal status (p-value=0.72) not significant; multiplicative interaction of TEXB-β tertile 3 with family history (p-value = 0.98) and menopausal status (p-value=0.49) not significant. Also examined overall association of specific organohalogenated compounds, but not by menopausal status or family history.
TEXB-α, tertile 3 (ref=tertile 1)
overall	OR 3.45	(1.50, 7.97)
women with a family history of BC	OR 5.78	(1.00, 33.3)
premenopausal	OR 4.06	(0.41, 40.7)
postmenopausal	OR 2.62	(1.07, 6.46)
TEXB-β, tertile 3 (ref=tertile 1)
overall	OR 4.01	(1.88, 8.56)
women with a family history of BC	OR 3.94	(0.69, 22.7)
premenopausal	OR 1.92	(0.19, 19.1)
postmenopausal	OR 4.53	(2.00, 10.3)
Case-control study in Bogota, Columbia
Olaya-Contreras, 1998 51			Family history: Assessed multiplicative interaction by family history*, however data not shown. Authors reported interaction term was not statistically significant. Early onset: Analysis was stratified by menopausal status, see main results. 39% of study participants are premenopausal.	Hospital-based case-control study in Bogota, Colombia, recruitment 1995-1996 of women aged 26 -75, included 153 histologically confirmed incident BC cases and 153 age-matched controls from a similar hospital that provides care for noncancer patients	Serum levels of DDE	Adjusted for breastfeeding at first child, family history of breast cancer, parity, Quetelet index, and menopausal status.	*Did not specify if it is first-degree family history or any family history
DDE, tertile 3 (ref=T1)
Overall	OR 1.95	(1.10, 3.52)
Premenopausal	OR 2.46	(0.96, 6.30)
Postmenopausal	OR 1.85	(0.84, 4.05)
Case-control study in Ontario, Canada
Aronson, 2000 50			Family history: Assessed multiplicative interaction by family history*, however data not shown. Authors reported interaction term was not statistically significant. *Includes a first or second degree relative with breast cancer Early onset: Analysis was stratified by menopausal status, see main results. 34% of cases and 45% of controls are premenopausal	Hospital-based case-control study, recruitment 1995-1997, which included 217 breast cancer cases and 213 benign controls frequency matched by study site and age in 5-year groups. Women under the age of 80 were enrolled at time of being scheduled for excision biopsy of suspected breast cancer. Cases were subjects diagnosed with in situ or invasive breast cancer. Controls were subjects with biopsies negative for malignancy, but most were diagnosed with some form of benign breast disease.	Biopsy breast tissue was analyzed for 14 PCB congeners, total PCBs, and 10 other organochlorines	Adjusted for age, study site, menopausal status, ever pregnant, lactation, age last breast fed, present use of HRT, ethnicity, family history, BMI, fat intake, alcohol intake, present smoking, and cumulative smoking.	Authors also examined association of PCB 187, cis-nonachlor, trans-nonachlor, oxychlordane, HCB, and β-HCBH and found no elevated risk of pre-menopausal BC
Overall associations for each chemical are for ≥85th percentile (ref = <28th percentile); stratified associations are for ≥85th percentile (ref = <57th percentile)
PCB 99
Overall	OR 1.92	(0.95, 3.86)
Premenopausal	OR 1.63	(0.71, 3.72)
Postmenopausal	OR 1.70	(0.74, 3.91)
PCB 105
Overall	OR 3.17	(1.51, 6.68)
Premenopausal	OR 3.91	(1.73, 8.86)
Postmenopausal	OR 1.49	(0.70, 3.16)
PCB 118
Overall	OR 2.31	(1.11, 4.78)
Premenopausal	OR 2.85	(1.24, 6.52)
Postmenopausal	OR 1.58	(0.70, 3.58)
PCB 138
Overall	OR 1.56	(0.80, 3.06)
Premenopausal	OR 1.52	(0.69, 3.35)
Postmenopausal	OR 1.69	(0.79, 3.60)
PCB 153
Overall	OR 1.04	(0.51, 2.11)
Premenopausal	OR 1.06	(0.48, 2.34)
Postmenopausal	OR 1.61	(0.72, 3.63)
PCB 156
Overall	OR 1.35	(0.68, 2.69)
Premenopausal	OR 1.35	(0.61, 2.98)
Postmenopausal	OR 1.41	(0.65, 3.06)
PCB 170
Overall	OR 1.15	(0.60, 2.22)
Premenopausal	OR 0.89	(0.41, 1.91)
Postmenopausal	OR 1.63	(0.77, 3.45)
PCB 180
Overall	OR 1.27	(0.66, 2.46)
Premenopausal	OR 0.89	(0.42, 1.91)
Postmenopausal	OR 1.77	(0.85, 3.69)
PCB 183
Overall	OR 1.27	(0.66, 2.45)
Premenopausal	OR 1.37	(0.63, 2.96)
Postmenopausal	OR 1.16	(0.58, 2.33)
Aroclor 1260
Overall	OR 1.15	(0.58, 2.25)
Premenopausal	OR 1.24	(0.58, 2.66)
Postmenopausal	OR 1.53	(0.71, 3.30)
p,p′-DDE
Overall	OR 1.62	(0.84, 3.11)
Premenopausal	OR 1.52	(0.70, 3.33)
Postmenopausal	OR 1.05	(0.50, 2.19)
p,p′-DDT
Overall	OR 1.18	(0.61, 2.29)
Premenopausal	OR 1.09	(0.49, 2.40)
Postmenopausal	OR 1.05	(0.53, 2.06)
MIrex
Overall	OR 1.18	(0.59, 2.38)
Premenopausal	OR 1.72	(0.78, 3.76)
Postmenopausal	OR 1.13	(0.60, 2.13)
Population-based case-control study in Wisconsin
McElroy, 200448			Family history: Assessed multiplicative interaction by family history*, however data not shown. Authors reported interaction term was not statistically significant. *First-degree family history (mother, sister, or daughter) Early onset: Analysis was stratified by menopausal status, see main results.	Population-based case-control study in WI of 1,481 BC cases and 1,301 controls. Cases were female WI residents aged 20-69 years with a new BC diagnosis in 1996-2000 identified by WI Cancer Reporting System. Community controls were selected randomly by lists of licensed drivers and Medicare beneficiary files.	Self-reported recent sport-caught (Great Lakes and other lakes) fish consumption assessed as a potential source of exposure to PCBs, DDT, PBDEs, and other halogenated hydrocarbons.	Models adjusted for age, family history of breast cancer, recent alcohol consumption, parity, age at first full-term pregnancy, lactation, age at menarche, weight at age 18, weight gain since age 18, education, and age at menopause (postmenopausal women only).	Authors also reported RRs by menopausal status further stratified by 5-year age groups (premenopausal age groups: <40, 40-44, 45-49, ≥50 years; postmenopausal age groups: <55, 55-59, 60-64, ≥65 years), with some subgroups having significant associations (see Table 3 in manuscript for details).
Recent sport-caught fish consumption, any (ref = none)
Overall	RR 1.00	(0.86, 1.17)
Premenopausal	RR 1.24	(0.96, 1.59)
Postmenopausal	RR 0.91	(0.74, 1.11)
Recent Great Lakes fish consumption, any (ref= none)
Overall	RR 1.06	(0.84, 1.33)
Premenopausal	RR 1.70	(1.16, 2.50)
Postmenopausal	RR 0.78	(0.57, 1.07)
Early Onset or menopausal status
Child Health and Development Studies (CHDS)
Cohn, 2007112			Early onset: Analysis involved early onset cases.	Nested case-control study of 129 cases and 129 controls from the original Child Health and Development Studies (CHDS). Controls matched using DMV records based on birth year. All cases were <50 years old, and mean age at diagnosis was 44 years.	Serum DDT-related compounds	Adjusted for year of blood draw and o,p ′-DDT	p,p′DDT and age in 1945 interaction p-value =0.02 Also assessed confounding by other BC risk factors (parity, BMI, AAFB, Menarche, race, or breast feeding) and there was little evidence of substantial confounding. Wide confidence intervals reflect small number of case-control pairs (n=29-34) for each age quartile strata
p,p′-DDT level, tertile 3 (ref = tertile 1)
All ages	OR 2.8	(1.2, 6.7)
Age quartile 1, <4 years old in 1945	OR 11.5	(1.0, 138.9)
Age quartile 2, 4-7 years old in 1945	OR 9.6	(0.7, 137.2)
Age quartile 3, 8-13 years old in 1945	OR 3.9	(0.8, 19.2)
Age quartile 4, ≥14 years old in 1945	OR 0.6	(0.1, 3.2)
Age quartiles 1-3, <14 years old in 1945	OR 5.4	(1.7, 17.1)
Cohn, 201267			Early onset: Analysis involved early onset cases.	Nested case-control study of 112 cases and 112 controls from CHDS. Controls matched using DMV records based on birth year. All cases were <50 years old, and mean age at diagnosis was 43 years.	Serum PCBs drawn during the early postpartum period, within 1–3 days of delivery. Net effect of PCB exposure: constructed a post-hoc score that consisted of the ratio of the sum of PCB congener(s) associated with higher BC risk to the sum PCB congeners associated with lower risk as in a previous report on health effects of PCB exposure in this population	Adjusted for cholesterol, triglycerides, race, parity, lactation, BMI, and year of blood sampling. Only net pcb analysis is adjusted.	Group 1 (PCB 101, 187, 201), group 2 (potentially antiestrogenic, immunotoxic, dioxin-like, Group 2a (PCB 66, 74, 105, 118, 156, 167), group 2b (PCB 138 and 170)I group 3 (PCB 99, 153, 180, 183, 203).
PCB 167 (ref=quartile 1)
PCB 167, overall association quartile 2	OR 1.09	(0.48, 2.47)
PCB 167, overall association quartile 3	OR 0.70	(0.27, 1.78)
PCB 167, overall association quartile 4	OR 0.24	(0.07, 0.79)
PCB 187 (ref=quartile 1)
PCB 187, overall association quartile 2	OR 0.94	(0.41, 2.17)
PCB 187, overall association quartile 3	OR 0.92	(0.36, 2.38)
PCB 187, overall association quartile 4	OR 0.35	(0.11, 1.14)
PCB 203 (ref=quartile 1)
PCB 203, overall association quartile 2	OR 1.21	(0.46, 3.18)
PCB 203, overall association quartile 3	OR 2.89	(0.98, 8.55)
PCB 203, overall association quartile 4	OR 6.34	(1.85, 21.73)
net PCB (ref=quartile 1)
net PCB, overall association quartile 2	OR 1.36	(0.53, 3.52)
net PCB, overall association quartile 3	OR 1.78	(0.70, 4.55)
net PCB, overall association quartile 4	OR 2.81	(1.11, 7.09)
Cohn, 2015111			Early onset: Analysis involved early onset cases.	Nested case-control study of 118 BC cases and 354 controls who were daughters from the original CHDS prospective 54-year follow-up of 9300 daughters. Controls matched on birth year and trimester of maternal blood draw. All BC cases were diagnosed in daughters by age 52 years old.	Non-fasting maternal perinatal serum samples collected from 1959 through 1967. Mean age of blood drawn was 26.9 years	Adjusted for cholesterol, triglycerides, maternal age, race, overweight in early pregnancy, parity, maternal history of BC, breastfeeding of daughter.	This study examines in utero exposure to DDT
Mother’s o,p′-DDT levels (ref=quartile 1)
o,p′-DDT, overall association quartile 2	OR 2.0	(0.9, 4.3)
o,p′-DDT, overall association quartile 3	OR 1.8	(0.8, 4.0)
o,p′-DDT, overall association quartile 4	OR 3.7	(1.5, 9.0)
Cohn 201952			Early onset: examined interaction with age at diagnosis and stratified by menopausal status, see main results	Nested case control within the CHDS Conducted a pooled analysis including cases and controls from Cohn 2007 that examined the joint effects of age at exposure and DDT differ according to age at BC diagnosis. Pooled analysis had 282 cases diagnosed through age 54 and 561 controls matched on year of birth.	Serum DDT-related compounds measured using serum collected during 1959-1967 from pregnant CHDS mothers. Made inferences about the timing of the first exposure to DDT by using birth year and the year DDT came into use in the United States, 1945.	Adjusted for o′,p′-DDT, year of blood draw, and parity	All post-menopausal women were early menopausal BC cases diagnosed between ages 50-54 3-way multiplicative p-interaction between age at exposure less than 3, menopausal status, and DDT p=0.03. 3-way p-interaction between age at exposure 3-13, menopausal status, and DDT p=0.049
p,p′-DDT first exposure in 1945, pooled analysis (continuous variable exposure)
premenopausal BC, exposure age < 3	OR 3.7	(1.22, 11.26)
premenopausal BC, exposure age 3-13	OR 5.16	(1.92, 13.82)
premenopausal BC, exposure age ≤ 14	OR 0.98	(0.51, 1.88)
postmenopausal BC, exposure age < 3	OR 0.92	(0.52, 1.63)
postmenopausal BC, exposure age 3-13	OR 1.88	(1.37, 2.59)
postmenopausal BC, exposure age ≥ 14	OR 2.26	(1.22, 4.20)
French E3N Cohort
Danjou 201945			Early onset: Stratified by menopausal status, see main results	Nested case-control study within the French E3N Cohort study (1990-2008) Cases: 429 women with primary invasive BC. Controls: 716 controls matched on age, department of residence, menopausal status, and date at blood collection or at baseline At baseline, 60.8% of cases and 59.5% of controls were premenopausal.	Using Geographic Information System (GIS), atmospheric dioxin exposure (mg/m2) was estimated by linking study subject residential histories, meteorological data, and a retrospective inventory of industrial sources and emissions between 1990 and 2008.	Adjusted for age, residence, menopausal status, date at blood collection or at baseline, existence of a biological sample, index date, family history of BC, age at first full-term pregnancy, physical activity, status of birthplace (urban, rural), and breastfeeding.	No multiplicative interaction by menopausal status (p=0.45).
Cumulative airborne dioxin exposure (ref=Quintile 1)
overall, Quintile 2	OR 1.61	(1.04, 2.49)
overall, Quintile 3	OR 1.40	(0.89, 2.21)
overall, Quintile 4	OR 1.51	(0.95, 2.38)
overall, Quintile 5	OR 1.12	(0.69, 1.82)
premenopausal, Quintile 2	OR 1.36	(0.53, 3.48)
premenopausal, Quintile 3	OR 1.35	(0.45, 4.04)
premenopausal, Quintile 4	OR 0.88	(0.29, 2.70)
premenopausal, Quintile 5	OR 1.41	(0.41, 4.87)
postmenopausal, Quintile 2	OR 1.59	(1.08, 2.36)
postmenopausal, Quintile 3	OR 1.35	(0.90, 2.05)
postmenopausal, Quintile 4	OR 1.34	(0.89, 2.03)
postmenopausal, Quintile 5	OR 1.03	(0.67, 1.60)
CECILE Study
Bachelet, 201955			Early onset: Stratified analyses by age group, used as a proxy for menopausal status (<50 years; 50 years), see main results	Population case-control within the Cote d’Or and Ille-et-Vilaine departments, which recruited from April 2005 to March 2007. Cases: 695 women with primary invasive BC or in situ breast carcinoma. Controls: 1055 controls frequency-matched by 10-year age group who were random digit dialed. At baseline, 23% of cases and 33.8% of controls were less than 50 years old	Plasma organochlorines measured using gas chromatograph coupled to an ion trap mass spectrometer detector. Blood was drawn before the first chemotherapy treatment for cases.	Adjusted for age, study area, level of education, age at menarche, parity, age at first full-term pregnancy, body mass index, hormone replacement therapy, familial history of BC, history of benign breast disease, breastfeeding, and reference date.	Aimed to obtain a distribution by SES category in the control group identical to the ses distribution in the general population of women, conditionally to age.
DDE (ng/g lipids), tertile 3 (ref = < LOD)
Overall	OR 0.93	(0.73, 1.18)
age <50 years old	OR 1.48	(0.90, 2.41)
age ≥50 years old	OR 0.81	(0.61, 1.07)
PCB153 (ng/g lipids), tertile 3 (ref = < LOD)
Overall	OR 0.75	(0.57, 0.97)
age <50 years old	OR 1.46	(0.85, 2.49)
age ≥50 years old	OR 0.65	(0.48, 0.89)
Four hospitals in Nagano Prefecture, Japan
Itoh, 2009 56			Early onset: Analysis stratified by menopausal status, see main results	Matched case–control study of BC with 403 eligible matched pairs from 2001 to 2005 at four hospitals in Nagano Prefecture, Japan. 45% of cases (n=183) and 35% of controls are premenopausal	Serum samples were measured for PCBs and nine pesticide-related organochlorines, including dichlorodiphenyltrichloroet hane (DDT). Blood specimens were collected from all cancer patients prior to surgery	Adjusted for age, residential area , total lipid concentration, BMI, smoking status, fish Consumption vegetable consumption family history of BC in a first-degree relative, age at first childbirth, parity, history of BC screening, and breast feeding	Also examined other pestidcides including trans-Nonachlor, cis-Nonachlor, Oxychlordane, HCB, and β-HCBs and report no association overall, and in strata of menopausal status.
o,p′-DDT, Quartile 4 (ref= Q1)
Premenopausal	OR 0.46	(0.17, 1.26)
Postmenopausal	OR 1.03	(0.44, 2.42)
p,p′-DDT, Quartile 4 (ref= Q1)
Premenopausal	OR 0.45	(0.17, 1.17)
Postmenopausal	OR 1.55	(0.68, 3.52)
p,p′-DDE, Quartile 4 (ref= Q1)
Premenopausal	OR 0.92	(0.32, 2.63)
Postmenopausal	OR 0.89	(0.38, 2.08)
Mirex, Quartile 4 (ref = Q1)
Premenopausal	OR 0.28	(0.10, 0.75)
Postmenopausal	OR 0.36	(0.16, 0.85)
Total PCBs, Quartile 4 (ref=Q1)
Premenopausal	OR 0.31	(0.08, 1.16)
Postmenopausal	OR 0.30	(0.12, 0.75)
Mexico City three hospitals
Lopez-Carrillo, 199757			Early onset: Analysis was stratified by menopausal status, see main results	Case-control study in the three referral public hospitals of the Secretariat of Health in Mexico City between March 1994 and April 1996 of women who resided in the Mexico City metropolitan area for at least 20 years. Cases: women aged 20 -79 years with histologically confirmed BC, Controls: had no history of cancer or breast disease . 50.4% of cases and 48.9% of controls are premenopausal at baseline.	Serum levels of DDE, p′p-DDT, and o’p-DDT on the basis of ppb of lipid weight (ng/g) Blood was drawn before any treatment	Adjusted for age, Quetelet index (kg/m2), breast feeding with first birth, parity, family history, and time elapsed since first birth (years)
DDE (ng/g) tertile 3 (ref = tertile 1)
Overall	OR 0.76	(0.41, 1.42)
premenopausal	OR 0.64	(0.22, 1.90)
postmenopausal	OR 0.79	(0.27, 2.28)
New Haven Connecticut hospitals
Zheng, 1999 65			Early onset: Analysis was stratified by menopausal status, however stratified results are not shown in the study. Authors note they are similar to overall results.	Hospital-based case-control study in New Haven, Connecticut, recruitment 1994-1997, included 304 incident BC cases and 186 benign breast disease controls. Cases and controls were women aged 40-79 years, who had breast-related surgery at the Yale-New Haven Hospital and from whose surgical specimen at least 0.4 g of breast adipose tissue was obtained for chemical analysis.	Lipid-adjusted concentrations of DDE and DDT in breast tissue.	Adjusted for age, body mass index, lifetime months of lactation, age at menarche, and age at first full-term pregnancy, menopausal status, race, and income 10 years before the diagnosis or interview.
DDE (ppb), quartile 4 (ref = Q1)	OR 0.90	(0.50, 1.50)
DDT (ppb), quartile 4 (ref = Q1)	OR 0.80	(0.50, 1.50)
Zheng, 200064
DDE (ppb), tertile 3 (ref = Tertile 1 (T1))	OR 0.96	(0.67, 1.36)	Early onset: Analysis was stratified by menopausal status, however stratified results are not shown in the study. Authors note they are similar to overall results.	Case-control study in New Haven, Connecticut, recruitment 1995-1997, which included 475 incident breast cancer patients and 502 controls. Cases were either residents of Tolland County or who had a breast-related surgery at the Yale-New Haven Hospital and Controls were randomly selected from Tolland County residents or from patients who had newly diagnosed benign breast diseases or normal tissue at Yale-New Haven Hospital. Cases and controls were 30-80 years; efforts were made to frequency match by age within 5-year intervals.	Lipid-adjusted serum levels of PCB and DDE.	Adjusted for age, body mass index, lifetime months of lactation, age at menarche, age at first full-term pregnancy, menopausal status, race, and income 10 years before the diagnosis or interview.
PCB (ppb), tertile 3 (ref = T1)	OR 0.95	(0.68, 1.32)
PCB (ppb) congener group 1: potentially estrogenic, tertile 3 (ref = T1)	OR 1.45	(0.99, 2.11)
PCB (ppb) congener group 2: potentially antiestrogenic, tertile 3 (ref = T1)	OR 1.35	(0.94, 1.93)
PCB (ppb) congener group 3: phenobarbital, CYP1A, and CYP2B inducers, tertile 3 (ref = T1)	OR 0.78	(0.54, 1.15)
Zhang, 200470			Early onset: Analysis was stratified by menopausal status, see main results Gene-environment: evaluated interaction between CYP1A1 gene and serum PCB levels	Case-control study which included 374 cases and 406 controls frequency matched on age who had information on serum PCB level and cyp1a1 genotype. Includes white women aged 30-80 with no previous diagnosis of cancer. 29.4% of cases and 55.6% of controls were premenopausal at baseline.	Serum PCB congeners 74, 118, 138, 153, 156, 170, 180, 183, and 187 lipid-adjusted ng/g	Adjusted for age, BMI, lifetime duration of breastfeeding, BC family history, menopausal status, annual household income, and genotype	A borderline-significant increased risk of BC was found for women with the CYP1A1 m1 variant genotype and higher serum levels of PCBs, but no increased risk was found for the CYP1A1 m4 variant genotype, either alone or in combination with PCB exposure.
Total PCB > median (ref = ≤ median)
Overall	OR 1.2	(0.9, 1.6)
premenopausal	OR 1.3	(0.7, 2.3)
postmenopausal	OR 1.2	(0.9, 1.7)
CYP1A1 M2 (M2 WT/WT genotype, ≤median total PCBs = ref)
M2 variants, total PCB >median	OR 1.6	(0.7, 3.5)
M2 WT/WT, total PCB ≤median	OR 1.2	(0.9, 1.6)
M2 variants , total PCB <median	OR 3.6	(1.5, 8.2)
Hospital-based case-control study in New York, NY
Wolff, 2000 62			Early onset: Analysis was stratified by menopausal status, however stratified results are not shown in the study. Authors note they are similar to overall results.	Hospital-based case control study in New York City: included 175 incident BC cases, 181 controls with benign breast disease (BBD), and 175 controls without BBD	Lipid-adjusted serum levels of DDE, DDT, and PCBs.	Adjusted for age, age2, menopausal status, and race.	. frequency matched on age and race/ethnicity.
p,p′-DDE, tertile 3 (ref = T1)	OR 0.93	(0.56, 1.50)
p,p′-DDT, tertile 3 (ref = T1)	OR 1.34	(0.82, 2.20)
trans-nonachlor, tertile 3 (ref = T1)	OR 0.73	(0.43, 1.20)
HPCB, tertile 3 (ref = T1)	OR 0.78	(0.45, 1.30)
LPCB, tertile 3 (ref = T1)	OR 0.96	(0.53, 1.70)
New York University (NYU) Women’s Health Study
Wolff, 200063			Early onset: Analysis was stratified by menopausal status, however stratified results are not shown in the study. Authors note they are similar to overall results.	Nested case-control in the NYU Women’s Health Study; included 148 incident BC cases diagnosed ≥ 5 months after study enrollment and 295 individually matched controls, Blood samples collected 1987-1992.	Lipid-adjusted serum levels of DDE and PCBs.	Adjusted for age at menarche, number of full-term pregnancies, and age at first full-term pregnancy, family history of breast cancer, lifetime history of lactation, height, BMI, and BMI-menopausal status at blood donation interaction.	Matched on: menopausal status and age at enrollment, number and dates of blood donations, and day of the menstrual cycle for premenopausal women.
DDE, quartile 4 (ref = Q1)	OR 1.30	(0.51, 3.35)
PCB, quartile 4 (ref = Q1)	OR 2.02	(0.76, 5.37)
Northern California Region Kaiser Permanente Medical Care Program
Krieger, 1994 59			Early onset: Analysis was stratified by menopausal status, however stratified results are not shown in the study. Authors note they are similar to overall results.	Nested case-control study conducted in the Northern CA Region Kaiser Permanente Medical Care Program, recruitment 1964-1969 with follow-up through 1990. Study included 150 case patients and 150 matched controls.	Serum levels of DDE and PCBs	Adjusted for body mass index, age at menarche, ever versus never pregnant, and menopausal status at time of case patient’s diagnosis of breast cancer.	Individually matched on race/ethnicity, date of joining Kaiser Permanente, year of multiphasic examination, age at examination, and length of follow-up after examination.
DDT, tertile 3 (ref = T1)	OR 1.33	(0.68, 2.62)
PCB, tertile 3 (ref = T1)	OR 0.94	(0.48, 1.84)
Hospital-based case-control study in Rio de Janeiro City, Brazil
Mendonca, 1999 60			Early onset: Analysis was stratified by menopausal status, however stratified results are not shown in the study. Authors note they are similar to overall results.	Hospital-based case-control study in Rio de Janeiro City, Brazil, recruitment 1995-1996, included 177 invasive BC cases and 350 controls of female visitors at the same hospital. Frequency matched on 5-year age groups.	Serum levels of DDE	Adjusted for age, educational level, parity, lactation, tobacco smoking, family history of breast cancer, and breast size.	Fifteen cases and 19 controls did not have blood specimens available. excluded women >75 years
DDE, quintile 5 (ref = Q5)	OR 0.83	(0.40, 1.60)
Hospital-based case-control study in Long Island, NY
Stellman, 200061			Early onset: Analysis was stratified by menopausal status, however stratified results are not shown in the study. Authors note they are similar to overall results.	Hospital-based case-control study in Long Island, New York, recruitment 1994-1996, which included 232 incident breast cancer cases (malignant breast cancer or carcinoma in situ) and 323 hospital controls admitted to surgery for benign breast disease or nonbreast-related conditions.	Level of seven organochlorine pesticides and 14 PCB congeners in adipose tissue were assayed via a supercritical fluid extraction method followed by gas chromatography with electron capture detection.	Adjusted for age, BMI, hospital, and race.	OCP is the sum of seven organochlorine pesticide species; PCB is the sum of 14 congeners, details are provided in the text of the manuscript.
DDE, tertile 3 (ref = T1)	OR 0.74	(0.44, 1.25)
OCP	OR 0.66	(0.38, 1.17)
PCB	OR 1.01	(0.60, 1.69)
Mexico del Seguro Social hospitals
Recio-Vega, 201169			Early onset: analysis stratified by menopausal status, see main results	Hospital based case-control study with 56 cases and 52 controls who were aged 25-80 residing in Comarca Lagunera, Mexico. Cases had no history of chemotherapy or radiotherapy. 40% of cases and 78.6% of controls were premenopausal at baseline.	Serum PCB congeners 8, 18, 28, 44, 52, 66, 77, 101, 105, 118, 126, 138, 148, 153, 170, 180, 187, 195, 206, 209. Grouped PCB congeners according ot their structure-activity relationship	Adjusted for age, age at menarche, lactation, menopause status (in overall analysis only), BMI, and BC family history	The analysis looked at geometric mean (ppb) of five PCB groups integrated according to their structure-activity relationship as well as total PCB as the sum of all congeners. Only PCB groups 2b and 4 had significantly increased risks for premenopausal BC. Only had 41 postmenopausal individuals.
Group 2b (di-ortho substituted, limited dioxin-like, and persistent)
Overall	OR 1.90	(1.25, 2.88)
premenopausal	OR 1.62	(0.99, 2.64)
postmenopausal	OR 3.66	(1.20, 11.19)
Group 4 (environmental relevance)
Overall	OR 1.57	(1.20, 2.07)
premenopausal	OR 1.49	(1.07, 2.06)
postmenopausal	OR 1.83	(1.08, 3.12)
Total PCB (geometric mean (ppb))
Overall	OR 1.09	(1.02, 1.16)
premenopausal	OR 1.08	(0.99, 1.17)
postmenopausal	OR 1.13	(1.01, 1.25)
Campaign against Cancer and Stroke (CLUE I) & Campaign against Cancer and Heart Disease (CLUE II)
Helzlsouer, 199953			Early onset: analysis stratified by menopausal status. See main results for ORs, did not report 95% Cis for stratified models	CLUE I: Nested case-control study from the original CLUE I prospective cohort of 25,802 persons. 235 cases and 235 controls matched by sex, race, age, menopausal status, date of blood donation and day of menstrual cycle at the time of blood donations. CLUE II: Nested case-control study from the original CLUE II prospective cohort of 32,892 persons. 105 cases and 105 controls matched by on same factors as CLUEI.	Sera DDE and PCBs lipid-adjusted ng/g	Unadjusted because no change when adjusting for BC family history, BMI at age 20 or current, age at menarche, age at first birth, and duration of lactation. Adjusted for age and menopausal status at baseline in gene stratified analyses	Did not provide confidence intervals for stratified analysis by genotypes or menopausal status. CLUEI: ~45 % of cases and controls less than 50 years of age. CLUEII: ~25 % of cases and controls less than 50 years of age.
CLUE I
DDE (lipid-adjusted) Q5 vs. Q1, overall	OR 0.73	(0.40, 1.32)
DDE Q3 vs. Q1
premenopausal association	OR 0.86	n/a
postmenopausal association	OR 0.52	n/a
PCB (lipid adjusted) Q5 vs. Q1, overall	OR 1.12	(0.59, 2.15)
DDE Q3 vs. Q1
premenopausal association	OR 2.21	n/a
postmenopausal association	OR 0.62	n/a
CLUE II
DDE Q3 vs. Q1
Overall association	OR 0.58	(0.29, 1.17)
premenopausal association	OR 1.42	n/a
postmenopausal association	OR 0.50	n/a
PCB Q3 vs. Q1
Overall association	OR 0.76	(0.38, 1.51)
premenopausal association	OR 2.12	n/a
postmenopausal association	OR 0.74	n/a
Long Island Breast Cancer Study Project (LIBCSP)
Gammon, 2002 71			Early onset: Analysis was stratified by menopausal status, however stratified results were only shown for DDE, Peak-4 PCBs, and Chlordane.	LIBCSP: Population-based case-control study in Long Island, NY, recruitment 1996-1997, included 646 cases and 429 controls with blood samples. Cases identified through pathology laboratories of all of the hospitals in Long Island. Controls were female residents of the same two counties and were frequency matched by 5-year age group to the expected age distribution of the cases. Participants ranged in age from 24 to 96 years.	Lipid-adjusted serum levels of DDE and PCBs.	DDE and Peak-4 PCB analyses were adjusted for age, race, history of fertility problems, and history of benign breast disease. Chlordane analyses were adjusted for age, race, history of fertility problems, and gravidity.	Results shown were based on quintiles for the main effects of organochlorines, and results shown were based on tertiles for exploration of possible effect modification.
DDE
Overall, quartile 4 (ref = Q1)	OR 1.20	(0.76, 1.90)
Premenopausal, tertile 3 (ref = T1)	OR 0.86	(0.44, 1.71)
Postmenopausal, tertile 3 (ref = T1)	OR 1.10	(0.70, 1.74)
Peak-4 PCBs
Overall, quartile 4 (ref = Q1)	OR 0.83	(0.54, 1.29)
Premenopausal, tertile 3 (ref = T1)	OR 0.94	(0.51, 1.74)
Postmenopausal, tertile 3 (ref = T1)	OR 0.73	(0.47, 1.12)
Chlordane
Overall, quartile 4 (ref = Q1)	OR 0.98	(0.62, 1.55)
Premenopausal, tertile 3 (ref = T1)	OR 1.11	(0.56, 2.17)
Postmenopausal, tertile 3 (ref = T1)	OR 1.09	(0.70, 1.69)
White, 2013 47			Early onset: Stratified by menopausal status, see main results	LIBCSP: Population case-control study with 1,508 cases and 1,556 controls frequency matched by 5-year age groups. Cases ascertained through pathology departments of 28 hospitals on Long Island, and controls ascertained through random digit dialing.	Exposure to fogger trucks was assessed by questionnaire.	Adjusted for frequency matching factor, 5-year age group.	Seeing a fogger truck= proxy measure for acute DDT exposure
Self-reported fogger truck at residence (ref=never)
overall, ever seen a fogger truck	OR 1.14	(0.98, 1.32)
premenopausal, ever seen a fogger truck	OR 0.96	(0.73, 1.26)
postmenopausal, ever seen a fogger truck	OR 1.25	(1.05, 1.50)
Fogger truck seen ≤ 1972 (ref=never)
overall, ever seen a fogger truck	OR 1.16	(0.98, 1.37)
premenopausal, ever seen a fogger truck	OR 0.97	(0.70, 1.35)
postmenopausal, ever seen a fogger truck	OR 1.24	(1.02, 1.51)
Women’s Contraceptive and Reproductive Experiences (CARE) Study - Los Angeles Site
Gatto, 2007 58			Early onset: analysis stratified by menopausal status. However, stratified results are not shown in the study. Authors note they did not observe effect modification by menopausal status.	Population based case-control study of 354 cases and 326 controls from the original Women’s CARE study of 1,255 cases and 1242 controls. 45.6% of cases and 46.5% of control were premenopausal at baseline.	Serum DDE and PCBs	Adjusted for age, BMI, breastfeeding, lipids
PCB (ref=below detection limit)
overall association, quintile 5	OR 1.01	(0.63, 1.63)
DDE (ref=quintile 1)
overall association, quintile 5	OR 1.02	(0.61, 1.72)
Japan Public Health Center-based Prospective Study (JPHC Study)
Iwasaki, 2008 54			Early onset: analysis stratified by menopausal status, see main results	Nested case-control study within the original Japan Publich Leaht Center-based Prospective Study from 1990-1993 with 67,426 subjects. 139 cases and 278 controls matched by age, PHC area, area, date of blood collection, time of day of blood collection, fasting time at blood collection, and menopausal status. 41.0% of cases and 41.0% of controls were premenopausal at baseline.	Plasma p,p-DDT, p,p′-DDE, hexachlorobenzene (HCB), and B-hexachlorocyclohexane (B-HCH)	Adjusted for age at menarche, menopausal status at baseline, age at menopause, number of births, age at first birth, height, BMI, alcohol consumption
p,p′-DDTQ4 vs. Q1
Overall association	OR 0.99	(0.47, 2.08)
premenopausal	OR 2.45	(0.70, 8.63)
postmenopausal	OR 0.53	(0.18, 1.61)
p,p′-DDEQ4 vs. Q1
Overall association	OR 1.48	(0.70, 3.13)
premenopausal	OR 2.30	(0.73, 7.23)
postmenopausal	OR 1.07	(0.31, 3.63)
HCB Q4 vs. Q1
Overall association	OR 0.82	(0.38, 1.76)
premenopausal	OR 0.48	(0.14, 1.63)
postmenopausal	OR 1.09	(0.33, 3.54)
B-HCH Q4 vs. Q1
Overall association	OR 0.74	(0.39, 1.39)
premenopausal	OR 0.68	(0.24, 1.94)
postmenopausal	OR 0.51	(0.19, 1.38)
Genetic susceptibility
Case-control study in Western New York
Moysich, 1999 75			Gene-environment: evaluated interaction of CYP1A1 genotypes and PCB exposure	154 women with postmenopausal BC and 191 controls that were part of a larger case-control study in Western NY of 933 postmenopausal Caucasian women. Cases frequency-matched by age and county of residence with controls, which were randomly selected from the NY State Motor Vehicle lists (<65 years) and the Health Care Finance Administration rolls (≥65 years).	Serum levels of 56 PCB peaks were determined by high-resolution gas chromatography with electron capture.	Adjusted for age, education, serum lipids, age at menopause, family history of breast cancer, age at first birth, parity, duration of lactation, body mass index, and smoking status homozygous (Ile:Ile) for the CYP1A1 wild-type alleles.	Study limited to postmenopausal women.
CYP1A1 polymorphism and PCB body burden (ref= PCB low-Ile:Ile)
PCB low-Ile:Val/Val:Val	OR 0.88	(0.29, 2.70)
PCB high-Ile:Ile	OR 1.08	(0.62, 1.89)
PCB high-Ile:Val/Val:Val	OR 2.9	(1.18, 7.45)
Nurses’ Health Study (NHS)
Laden, 2002 73			Gene-environment: analysis stratified by detoxification genes CYP1A1-exon 7 and CYP1A1-MspI	Nested case-control study within the NHS original cohort of 32,826 women of which 367 cases and 367 controls matched on year of birth, menopausal status, month and time of blood collection, fasting status at blood draw, and postmenopausal hormone use.	Plasma PCBs. “variants” are all women who are either heterozygous or homozygous for the variant allele.	Adjusted for BC family history, history of benign breast disease, age at menarche, number of children, age at birth of first child, and duration of lactation	PCB no overall association (data not shown). P-interaction CYPA1-exon 7 and PCB : All women=0.19; post-menopausal women 0.05. P-interaction CYP1A1-MspI and PCB : All women=0.21; post-menopausal women 0.22.
CYP1A1-exon 7 (WT/WT genotype, PCB tertile 1 = ref)
Postmenopausal, variants, PCB tertile 3	RR 2.78	(0.99, 7.82)
All women, variants, PCB tertile 3	RR 1.36	(0.60, 3.12)
CYP1A1-MspI (WT/WT genotype, PCB tertile 1 = ref)
Postmenopausal, variants, PBC tertile 3	RR 1.08	(0.47, 2.48)
All women, variants, PCB tertile 3	RR 0.94	(0.44, 2.01)
Carolina Breast Cancer Study
Li, 2004 74			Gene-environment: evaluated interaction of CYP1A1 genotypes and PCB exposure	Carolina Breast Cancer Study population based case-control study of women age 20-74. White women: 370 cases and 357 controls. African American women: 242 cases and 242 controls. All cases and controls frequency-matched on race and age	Plasma levels of PCBs lipid adjusted (ng/ml)	Adjusted for age and sampling fractions	Significant multiplicative interaction for White women with CYP1A1 M2-containing genotypes and elevated PCB levels (p-value: 0.02) ICR=interaction contrast ratio, assessing interaction on additive scale. White women: CYP1A1 M1 interaction with PCB ICR=0.4 (−0.2, 0.9). CYP1A1 M2 interaction with PCB ICR=0.8 (0.1, 1.6). CYP1A1 M2 interaction with PCB ICR=0.8 (0.1, 1.6). AA women: CYP1A1 M1 interaction with PCB ICR=0.0 (-0.9, 0.9). CYP1A1 M3 interaction with PCB ICR=0.8 (−0.3, 1.9)
White women
CYP1A1 M1 (Non-M1 genotype, <median total PCBs = ref)
Non-M1, ≥median total pcb	OR 0.7	(0.5, 1.0)
Any M1, <median total pcb	OR 0.8	(0.5, 1.2)
Any M1, ≥median total pcb	OR 0.8	(0.4, 1.4)
CYP1A1 M2 (Non-M2 genotype, <median total PCBs = ref)
Non-M2, ≥median total pcb	OR 0.7	(0.5, 1.0)
Any M2, <median total pcb	OR 0.4	(0.2, 0.8)
Any M2, ≥median total pcb	OR 0.9	(0.4, 1.9)
CYP1A1 M4 (Non-M4 genotype, <median total PCBs = ref)
Non-M4, ≥median total pcb	OR 0.7	(0.5, 1.0
Any M4, <median total pcb	OR 0.8	(0.4, 1.6)
Any M4, ≥median total pcb	OR 0.7	(0.3, 1.6)
African-American women
CYP1A1 M1 (Non-M1 genotype, <median total PCBs = ref)
Non-M1, ≥median total pcb	OR 1.5	(0.9, 2.5)
Any M1, <median total pcb	OR 1.0	(0.6, 1.7)
Any M1, ≥median total pcb	OR 1.4	(0.8, 2.5)
CYP1A1 M3 (Non-M3 genotype, <median total PCBs = ref)
Non-M3, ≥median total pcb	OR 0.7	(0.5, 1.0)
Any M3, <median total pcb	OR 0.8	(0.4, 1.6)
Any M3, ≥median total pcb	OR 0.7	(0.3, 1.6)
Diet, Cancer, and Health
Brauner, 2014 72			Gene-environment: analysis stratified by CYP1B1 and COMT estrogen metabolism and PCB metabolism genes	Nested case-control study within the original 24,697 Diet, Cancer, and Health Cohort. Participants were born in Denmark and lived in Copenhagen or Aarhus. 409 cases and 409 controls matched on age, postmenopausal status, use of HRT at baseline, age at baseline	Adipose tissue analyses of 18 PCB congeners	Adjusted for education, BMI, alcohol consumption, number of childbirths, age at first delivery, duration of lactation, years of HRT use, and history of benign tumor.	RRs are per doubling in PCB concentration. Found no statistically significant interactions between any of the PCB groups and CYP1B1 or COMT polymorphisms.
PCB group 1b
CYP1B1 CC	RR 0.65	(0.35, 1.20)
CYP1B1 CG	RR 1.01	(0.68, 1.51)
CYP1B1 GG	RR 0.91	(0.44, 1.89)
COMT GG	RR 0.79	(0.36, 1.72)
COMT AG	RR 0.85	(0.53, 1.37)
COMT GG	RR 0.95	(0.59, 1.51)
PCB group 2a
CYP1B1 CC	RR 0.60	(0.32, 1.13)
CYP1B1 CG	RR 1.05	(0.69, 1.59)
CYP1B1 GG	RR 0.74	(0.35, 1.56)
COMT GG	RR 0.89	(0.44, 1.82)
COMT AG	RR 0.78	(0.48, 1.27)
COMT GG	RR 0.88	(0.54, 1.45)
PCB group 2b
CYP1B1 CC	RR 0.65	(0.36, 1.17)
CYP1B1 CG	RR 1.08	(0.74, 1.57)
CYP1B1 GG	RR 0.73	(0.37, 1.48)
COMT GG	RR 0.68	(0.32, 1.47)
COMT AG	RR 0.86	(0.55, 1.35)
COMT GG	RR 0.96	(0.62, 1.48)
PCB group 3
CYP1B1 CC	RR 0.59	(0.32, 1.11)
CYP1B1 CG	RR 1.05	(0.70, 1.57)
CYP1B1 GG	RR 0.71	(0.33, 1.50)
COMT GG	RR 0.73	(0.33, 1.62)
COMT AG	RR 0.81	(0.49, 1.31)
COMT GG	RR 0.88	(0.55, 1.41)

^a.OR = Odds Ratio, HR= Hazard Ratio, RR= relative risk, ROR= ratio of the odds ratio

^b.95% CI = 95% Confidence interval

Table 4b:

Epidemiologic studies of other pesticides, fungicides, per- and polyfluoroalkyl substances (PFAS), solvents, phthalates, polybrominated diphenyl ethers (PBDE), metals, other occupational exposures, and personal care products and breast cancer (BC), grouped by whether the study considered BC family history, early onset BC, or genetic susceptibility

Cohort Author, year Analysis	Effect Size	(95% CI)	Study design or analysis feature that meets inclusion criteria	Study population	Exposure assessment	Confounding	Notes
Family History
Sister Study
Ekenga, 2014 87			Family History: Study design enriched for women with underlying familial risk of BC, specifically having at least one sister diagnosed with BC	Sister Study participants were initially BC-free sisters of women who had a sister diagnosed with BC enrolled 2003-2009 and followed for an average of 4.7 years. 1,798 self-reported incident BC cases 46,381 non-cases	Self-reported lifetime occupational exposure to “solvents, degreasers or cleaning agents” on the job and the first and last date of employment for each job, collected at baseline interview Reference group: participants who were ever employed and never exposed to solvents, degreasers or cleaning agents	Adjusted for race/ethnicity, education, income, parity and age at first birth	No association of elevated BC risk seen for timing of first solvent job after first birth, or for nulliparous women.
Solvent job (ref = Never)
Ever	HR 1.04	(0.88, 1.24)
Ever, ER-positive tumors	HR 1.15	(0.95, 1.39)
Total duration of solvent jobs exposure, years (ref = never)
10+	HR 1.14	(0.90, 1.45)
10+, ER-positive tumors	HR 1.20	(0.92, 1.57)
Time period of first solvent job (ref = never)
Before 1980	HR 1.16	(0.93, 1.44)
Before 1980, ER-positive tumors	HR 1.28	(1.01, 1.62)
Timing of first solvent job (ref = never)
Before first birth	HR 1.24	(0.95, 1.63)
Before first birth, ER-positive tumors	HR 1.39	(1.03, 1.86)
Ekenga, 2015 76			Family History: Study design enriched for women with underlying familial risk of BC, specifically having at least one sister diagnosed with BC; Early onset: Analysis was stratified by menopausal status at diagnosis, see main results	Sister Study: Same as Ekenga, 2014 but followed for an average of 5.2 years. 1,966 self-reported incident BC cases 45,674 non-cases Number of exposed premenopausal BC cases: 7-34, depending on exposure	Self-reported occupational information and workplace exposure to acids, dyes or inks, gasoline or other petroleum products, glues or adhesives, lubricating oils, metals, paints, pesticides, soldering materials, solvents and stains or varnishes. Cumulative lifetime exposure for each agent, quartile cut-points were used to assign exposed study participants to categories.	Adjusted for race/ethnicity, education, income, parity and age at first birth	No significant associations between ever use of the 11 agents evaluated and overall BC risk. No significant association for gasoline or other petroleum products, lubricating oils, and paints and premenopausal BC. Women with cumulative exposure to gasoline or petroleum products at or above the highest quartile cutoff had an increased BC risk (HR: 2.3, 95%CI: 1.1–4.9) compared with women in the lowest quartile group
Exposure, ever use (ref=never use)
Acids, overall BC	HR 1.1	(0.9, 1.4)
Acids, Premenopausal BC	HR 1.3	(0.9, 2.1)
Dyes or inks, Overall BC	HR 1.1	(0.9, 1.3)
Dyes or inks, Premenopausal BC	HR 1.4	(0.9, 2.1)
Glues or adhesives, premenopausal BC	HR 1.1	(0.8, 1.6)
Metals, premenopausal BC	HR 1.2	(0.8, 2.0)
Paints, invasive BC	HR 1.1	(0.8, 1.3)
Pesticides, Overall BC	HR 1.1	(0.7, 1.5)
Pesticides, Premenopausal BC	HR 1.5	(0.8, 2.7)
Soldering materials, Overall BC	HR 1.1	(0.8, 1.4)
Soldering materials, Premenopausal BC	HR 1.8	(1.1, 3.0)
Solvents, Overall BC	HR 1.1	(0.9, 1.3)
Solvents, Premenopausal BC	HR 1.3	(0.9, 1.9)
Stains ever, premenopausal BC	HR 1.1	(0.6, 2.2)
Total number of agents, 2+ (ref = never)
Invasive BC	HR 1.1	(0.9, 1.3)
Premenopausal BC	HR 1.2	(0.9, 1.6)
Taylor, 2018 94			Family History: Study design enriched for women with underlying familial risk of BC, specifically having at least one sister diagnosed with BC Early onset: Analysis was stratified by menopausal status at diagnosis, see main results	Sister Study: See Ekenga 2014 Non-Hispanic white (n=42,447) and non-Hispanic blacks (n=4,450). White women : 32% of cases (n=691) premenopausal. Black women: 37% of cases (n=67) were premenopausal	Self-reported ascertainment of 48 personal care product use during the enrollment phase of the study. Usage patterns of personal care products were assessed using latent class analysis -products categorized into beauty, haircare, or skincare.	Adjusted by race, menopausal status at time of diagnosis or follow-up, age at menarche, age at first birth, parity, duration of breastfeeding, oral contraceptive use, hormone therapy, education, alcohol consumption, adult BMI, family history of BC, smoking status, and region of residence.	Study unable to assess risk associated with specific chemicals, but many personal care products include chemicals that can act as endocrine disruptors. Was only able to assess stratified analyses by menopausal status for White women, underpowered for Black women. Authors note: none of the latent class HRs were elevated or significant in Black women, but there were < 100 cases in any category.
Product category/class from latent class analysis (ref= infrequent user)
Beauty, frequent user
White women
overall	HR 1.15	(1.02, 1.30)
premenopausal	HR 1.01	(0.76, 1.33)
postmenopausal	HR 1.18	(1.14, 1.21)
Black women
overall	HR 0.86	(0.53, 1.39)
Skincare, frequent user
White women
overall	HR 1.13	(1.00, 1.29)
premenopausal	HR 1.06	(0.79, 1.42)
postmenopausal	HR 1.12	(1.09, 1.16)
Black women
overall	HR 0.79	(0.47, 1.34)
Hair, frequent user of hair spray and hair gel; White women
overall	HR 1.02	(0.93, 1.11)
premenopausal	HR 1.04	(0.86, 1.26)
postmenopausal	HR 1.03	(1.01, 1.06)
O’Brien, 201998			Family history: Study design enriched for women with underlying familial risk of BC, specifically having at least one sister diagnosed with BC Early onset: Analysis limited to young-onset BC (diagnosis age <50 years)	Nested case control within the Sister Study (2003-2009) and Two Sister Study (2008-2010) Cases: 1,217 women with young-onset BC (age <50). Controls: 1,217 sister-matched controls 100% of cases were diagnosed before age 50.	10 toxic metals examined in toenail clippings using inductively coupled plasma mass spectrometry.	Adjusted for age and highest achieved education.	No statistically significant association s between any of the 10 metals and young-onset BC.
Toenail Metal Exposure, Quartile 4 (ref= Quartile 1)
Arsenic	OR 1.07	(0.72, 1.57)
Cadmium	OR 1.15	(0.82, 1.60)
Cobalt	OR 1.31	(0.82, 2.08)
Chromium	OR 1.08	(0.69, 1.69)
Copper	OR 1.09	(0.73, 1.62)
Mercury	OR 0.99	(0.68, 1.45)
Molybdenum	OR 1.33	(0.84, 2.12)
Lead	OR 1.11	(0.76, 1.61)
Tin	OR 1.22	(0.78, 1.90)
Vanadium	OR 1.36	(0.84, 2.21)
White, 2019 102			Family History: Study design enriched for women with underlying familial risk of BC, specifically having at least one sister diagnosed with BC Early onset: Analysis was stratified by menopausal status at diagnosis, see main results	Sister Study includes 50,884 participants were initially BC-free sisters of women who had a sister diagnosed with BC enrolled 2003-2009 and followed for an average of 7.4 years. 2,587 self-reported incident BC cases diagnosed through July 2015. 35% were premenopausal at baseline (n=16,749) and ~20% of cases were pre-menopausal BCs (n=536)	Used EPA National Air Toxics Assessment (NATA) 2005 concentrations for the metals antimony (Sb), arsenic, cadmium (Cd), chromium, cobalt (Co), lead (Pb), manganese, mercury (Hg), nickel and selenium (Se) at the census tract level and linked to each participant’s address.	Adjusted for race, education, annual household income, marital status, parity, census-tract level median income, and geographic region	Considered metals individually and as a mixture. Noted that the association for the weighted quantile sum analysis was driven by Cd, Pb, and Hg. No statistically significant elevated overall BC or premenopausal BC risk associated with all metals examined, except for Hg (only for overall BC risk). Cd, Pb, and Se may be associated with elevated overall BC risk and Co with pre-menopausal BC risk, but these associations were not statistically significant.
Antimony, Quantile 5 (ref = Q1)
Overall	OR 0.95	(0.83, 1.1)
Pre-menopausal BC	OR 0.69	(0.51, 0.94)
Arsenic, Quantile 5 (ref= Q1)
Overall	OR 1.0	(0.9, 1.2)
Pre-menopausal BC	OR 0.97	(0.71, 1.3)
Post-menopausal BC	OR 1.1	(0.90, 1.2)
Cadmium, Quantile 5 (ref= Q1), Overall	OR 1.1	(0.96, 1.3)
Cobalt, Quantile 5, (ref=Q1)
Overall	OR 1.0	(0.91, 1.2)
Pre-menopausal BC	OR 1.1	(0.83, 1.5)
Lead, Quantile 5 (ref=Q1), Overall	OR 1.1	(0.94, 1.2)
Mercury, Quantile 5 (ref=Q1), Overall	OR 1.2	(1.0, 1.4)
Selenium, Quantile 5, (ref =Q1), Overall	OR 1.1	(0.93, 1.2)
Weighted quantile sum index
Overall	OR 1.0	(0.98, 1.1)
Post-menopausal BC	OR 1.1	(1.0, 1.1)
Agricultural Health Study (AHS) cohort
Engel, 2005 78			Family history: stratified analyses by first-degree family history of BC, see main results. Early onset: Stratified by menopausal status, see main results	AHS: Original cohort of 57,310 private and commercial pesticide applicators in Iowa and North Carolina; of which, 30,003 female spouses of private applicators were used in this analysis. Study captures 10 - 11 years of follow up.	Self-reported assessment of pesticide use throughout the lifetime.	Adjusted for age, race, and state of residence	Found significant overall association with husbands’ exposure to specific insecticides. Limited power to examine associations by family history (74 cases in cohort)
First-degree family history of BC
Diazinon exposure	RR 1.7	(0.9, 3.2)
Husbands’ use of parathion	RR 4.2	(1.6, 10.6)
Husbands’ use of paraquat	RR 3.9	(1.7, 8.9 )
No family history of BC
Diazinon exposure	RR 0.8	(0.5, 1.2 )
Husbands’ use of parathion	RR 0.9	(0.5, 1.8)
Husbands’ use of paraquat	RR 0.9	(0.5, 1.6 )
Exposure to pesticides
Overall, dichlorvosin	RR 1.2	(0.7, 2.1)
Premenopausal, dichlorvosin	RR 2.3	(1.0, 5.3)
Postmenopausal, dichlorvosin	RR 1.0	(0.4, 2.0)
Exposure to insecticides
Overall, terbufos	RR 1.1	(0.6, 2.1)
Premenopausal, terbufos	RR 2.6	(1.1, 5.9)
Postmenopausal, terbufos	RR 0.5	(0.2, 1.6)
Polish Breast Cancer Study
Peplonska, 2010 89			Family history: Analysis was stratified by BC in first degree relative, see main results Early onset: Analysis was stratified by menopausal status at diagnosis and age at diagnosis, see main results	Population-based case-control study in Poland of 2,383 incident invasive primary BCs diagnosed between 2000-2003 in Warsaw and Lodz. Cases identified by participating hospitals and regional population cancer registries. 2,502 controls randomly selected from Polish Electronic System of Population Evidence, matched to cases on city of residence and 5-year age groups In exposed women: 8.6%-11.6 of cases (n=49 or 10) and 6-8% of controls had a first degree family history.	Occupational exposure collected by in-person questionnaire interviews and examined by industrial hygienists who specifically assessed occupational exposure to organic solvents and to benzene. Cumulative exposure is a product of frequency, intensity, and duration of exposure summed across all jobs, the cut point is the median in exposed controls. Stratified results by menopausal status for cumulative exposure were only presented for organic solvents, and not benzene.	Adjusted for age, study site, education, body mass index (BMI), age at menarche, menopausal status, age at menopause (in postmenopausal women), number of full-term births, age at first full-term birth, breastfeeding, family history of BC and previous screening mammography.	The authors note that estimated intensity of exposure among women in their study was low. Study examined multiplicative interaction by age and menopausal status for both exposures and report no statistically significant interaction. Stratified analyses by menopausal status and family history limited to ever exposed to benzene or organic solvents.
Ever exposed to organic solvents (ref = never)
Ever, overall	OR 1.16	(0.99, 1.4)
Ever, premenopausal	OR 1.21	(0.9, 1.6)
Ever, postmenopausal	OR 1.15	(0.96, 1.4)
Ever, age < 50 years	OR 1.13	(0.8, 1.5)
Ever, age ≥ 50 years	OR 1.20	(0.99, 1.5)
Ever, yes family history	OR 1.08	(0.6, 1.9)
Ever, no family history	OR 1.18	(1.0, 1.4)
Cumulative exposure to organic solvents > 39 (ref = never exposed)
Overall	OR 1.14	(0.9, 1.4)
Premenopausal BC	OR 1.57	(0.99, 2.5)
Postmenopausal BC	OR 1.06	(0.8, 1.4)
Ever exposed to benzene (y), (ref = never)
Ever, overall	OR 1.00	(0.8, 1.3)
Ever, premenopausal	OR 0.78	(0.4, 1.4)
Ever, postmenopausal	OR 1.06	(0.8, 1.4)
Ever, age < 50 years	OR 0.90	(0.5, 1.6)
Ever, age ≥ 50 years	OR 1.04	(0.8, 1.4)
Ever, yes family history	OR 0.49	(0.2, 1.3)
Ever, no family history	OR 1.07	(0.8, 1.4)
Early Onset
French E3N Cohort
Amadou, 2019 99			Early onset: Stratified by menopausal status , see main results	Nested case-control within the French E3N cohort study (1990-2011) Cases: 4,529 women with primary invasive BC. Controls: 4,529 controls matched on residential area, age, date, and menopausal status at blood collection or baseline. At baseline, 59.2% of cases and 59.4% of controls were premenopausal.	Using Geographic Information System (GIS), atmospheric cumulative cadmium exposure (mg/m2) was estimated by linking study subject residential histories, meteorological data, and a retrospective inventory of industrial sources and emissions. 2,700 cadmium sources inventoried throughout France	Adjusted for physical activity, smoking status, level of education, BMI, family history of BC, history of personal benign breast disease, age at menarche, age at first full-term pregnancy, parity, breastfeeding, oral contraceptive use, menopausal hormone replacement therapy use, and status of birthplace	Models were conditioned on the following matching factors: Date of blood collection or return of the first questionnaire, age, department of residence, menopausal status at blood collection or at baseline, and existence of a biological sample. No interaction by menopausal status (p=0.07).
Cumulative airborne cadmium exposure (ref=Quintile 1)
Overall, Quintile 5	OR 0.98	(0.84, 1.14)
premenopausal, Quintile 5	OR 0.72	(0.45, 1.15)
postmenopausal, Quintile 5	OR 1.06	(0.89, 1.27)
Danish Nationwide Cohort Study
Ahern, 2019 90			Early onset: Stratified by menopausal status. Data not shown, but presented in figure 1b of paper. Distribution of menopausal status not reported	Danish Nationwide Cohort Study: 1,122,042 women at risk for BC using several independent registries linked to Danish legal residents. 9.99 million person-years of follow-up (median:10 years) from Jan 1, 2005 to December 31, 2015. 27,111 cases of invasive BC. Source population: all alive Danish females who had no cancer history, unexposed to phthalate containing medications between 1995-2005	Cumulative phthalate exposure from prescription medication estimated by linking the Danish Medicines Agency and Danish National Prescription Registry.	Main models adjusted for age, menopausal status, other phthalate exposures, use of potentially confounding comedications, drug substance exposures, and comorbidity index Menopausal models adjusted by age and drug substance exposures.	The pre- & postmenopausal models were displayed using a figure (Fig1B). This figure did not give exact numbers for Hazard Ratios & Confidence Intervals. The authors report that the overall DBP association was modestly stronger among premenopausal women (Fig1b) menopausal status was imputed as pre- or postmenopausal on the basis of <55 years or ≥55 respectively;
Dibutyl phthalate (ref = unexposed 0mg)
overall, ≥10,000 mg	HR 2.0	(1.1, 3.6)
premenopausal, ≥10,000 mg	Check figure 1b
postmenopausal, ≥10,000 mg	Check figure 1b
Danish National Birth Cohort
Bonefeld-Jørgensen, 2014 82			Early onset: Study of premenopausal BC, age at diagnosis ranged from 30-53 and Mean age at diagnosis was 40.8 and 40.3 years for cases and controls	Nested case-control within the Danish National Birth Cohort (1996-2002) Women recruited during pregnancy and followed for 10-15 years post-partum. Cases: 250 nulliparous women diagnosed BC after recruitment Controls: 233 frequency matched on age and parity; taken at random from the entire cohort at baseline	Blood serum samples collected during the 1st and 2nd pregnancy trimester, analyzed for levels of ten perfluorocarboxylated acids, five perfluorosulfonated acids, and sulfonamide (perflurooctane-sulfonamide, PFOSA) were determined by liquid chromatography-tandem mass spectrometry with electrospray ionization.	Adjusted for age at blood sampling, BMI before pregnancy, gravidity, OC use, menarche age, smoking during pregnancy, alcohol intake, maternal education and physical activity
PFHxS (ref=Quintile 1)
overall association, Q5	RR 0.61	(0.33, 1.12)
≤ 40 years old association, Q5	RR 0.41	(0.17, 0.96)
> 40 years old association, Q5	RR 1.01	(0.40, 2.54)
PFOSA (ref=Quintile 1)
overall association, Q5	RR 1.89	(1.01, 3.54)
≤ 40 years old association, Q5	RR 2.45	(1.00, 6.00)
> 40 years old association, Q5	RR 1.62	(0.61, 4.29)
California Farm Workers UFW
Mills, 2005 80			Early onset: Analysis stratified by age at BC diagnosis (≤54 years and ≥ 55 years)	Registry-based case-control study in farm labor union members (U FW). 128 Cases of newly diagnosed BCs in 1987-2001. 640 controls: UFW members not diagnosed with BC, matched based on ethnicity and year of birth. 50% of cases were less than 50 years of age at diagnosis (n=65).	UFW histories linked with department of pesticide regulation (DPR) pesticide use reports (PUR) to create a surrogate measure of pesticide exposure	Adjusted for age, date of first union affiliation, duration of union affiliation, fertility, and socioeconomic level.	Analysis restricted to Hispanics.
Use of all chemicals combined (ref= low chemical level)
Highest chemical level, ≤ 54 years	OR 1.44	(0.55, 3.75)
Highest chemical level, ≥ 55 years	OR 1.38	(0.34, 5.54)
California Teachers Study (CTS)
Hurley, 2018 85			Early onset: Stratified by menopausal status, see main results	California Teachers Study (CTS) an on-going prospective cohort study initiated in 1995-1996 of 133,479 female California public school professionals. 902 incident invasive BC cases 858 controls who did not develop cancer in the CTS cohort 5% of cases and 7% of controls were premenopausal.	Per- and poly- fluoroalkyl substances (PFASs) were measured in sera. Blood specimen and completed an interview-administered questionnaire at blood draw between October 2011 and August 2015. solid phase extraction method coupled to liquid chromatography and tandem mass spectrometry	Adjusted for age, race/ethnicity, region of residence, blood draw date*, blood draw date*, season of blood draw, total pack-years smoking*, family history of BC*, age at first full-term pregnancy*, total red meat* or pork consumption *postmenopausal women	No significant associations for any of the six PFASs examined. For cases- blood specimens collected on average 35 months after diagnosis
PFNA (ref = low)
Overall association, High	OR 1.04	(0.80, 1.35)
pre/perimenopausal association, High	OR 3.12	(0.98, 9.96)
postmenopausal association, High	OR 0.95	(0.72, 1.25)
Hurley 2019 113			Early onset: Stratified by menopausal status, see main results	Nested case-control study in CTS. 902 cases diagnosed with invasive BC; 936 controls frequency-matched by age, race/ethnicity, and geographic region. 5% of cases and 11% of controls were premenopausal.	Serum samples analyzed for 19 polybrominated diphenyl ether (PBDE) congeners using automated solid phase extraction and gas chromatography/high resolution mass spectrometry.	Adjusted for age, race/ethnicity, study site, total serum lipid content, date of blood draw, season of blood draw, BMI, physical activity, family history of BC, parity/age at first full term pregnancy, menopausal status/HT use, and pork consumption at baseline.	Analysis stratified by menopausal status and tumor hormone responsiveness yielded similar null results. For cases- blood specimens collected on average 35 months after diagnosis
BDE-47 (ref = Q1)
overall association, Q4	OR 0.88	(0.67, 1.17)
premenopausal association, Q4	OR 1.00	(0.34, 2.90)
postmenopausal association, Q4	OR 0.94	(0.71, 1.23)
BDE-100 (ref = Q1)
overall association, Q4	OR 0.89	(0.67, 1.17)
premenopausal association, Q4	OR 2.00	(0.61, 6.55)
postmenopausal association, Q4	OR 0.92	(0.70, 1.21)
BDE-153 (ref = Q1)
overall association, Q4	OR 0.89	(0.67, 1.18)
premenopausal association, Q4	OR 1.50	(0.45, 4.98)
postmenopausal association, Q4	OR 0.88	(0.67, 1.17)
Agricultural Health Study (AHS) cohort
Lerro, 2015 77			Early onset: Stratified by menopausal status, see main results	Original cohort of 57,310 private and commercial pesticide applicators; of which, 30,003 female spouses of private applicators were used in this analysis.	Self-reported assessment of pesticide use throughout the lifetime.	adjusted for age, state of residence, pack-years, race, alcohol use, education, BMI, family history, menopausal status, parity, ever use of OC, being the person who treats the home or lawn for pests, ever use of specific pesticides
Any organophosphate insecticide, ever use vs. never use
overall association	RR 1.20	(1.01, 1.43)
premenopausal association	RR 1.02	(0.77, 1.36)
postmenopausal association	RR 1.27	(1.00, 1.62)
Chlorpyrifos, ever use vs. never use
overall association	RR 1.41	(1.00, 1.99)
premenopausal association	RR 1.36	(0.79, 2.34)
postmenopausal association	RR 1.53	(0.96, 2.44)
Terbufos, ever use vs. never use
overall association	RR 1.52	(0.97, 2.36)
premenopausal association	RR 1.25	(0.61, 2.54)
postmenopausal association	RR 1.73	(0.93, 3.21)
Engel, 2017 79			Early onset: Stratified by menopausal status, see main results Nearly half (46.3%) premenopausal	Original cohort of 57,310 private and commercial pesticide applicators in Iowa and North Carolina; of which, 30,003 female spouses of private applicators were used in this analysis. Study captures 10 - 11 years of follow up.	Self-reported assessment of pesticide use throughout the lifetime.	Adjusted for race, state, and combined parity/age at first birth and use of benomyl, metribuzin, butylate, and toxaphene	No significant overall associations. Weak associations in premenopausal women with phorate and chlorpyrifos. Limited power for exposed premenopausal women (n range 5- 39)
Overall, organophosphates (terbufos)	HR 1.5	(1.0, 2.1)
Premenopausal, organophosphates (terbufos)	HR 2.6	(1.3, 5.4)
Postmenopausal, organophosphates (terbufos)	HR 1.2	(0.8, 1.9)
Canadian National Enhanced Cancer Surveillance System (NECSS)
Pan, 2011 104			Early onset: stratified by menopausal status , see main results	Case-control population based data from the NECSS across Canada (1994-1997) Cases: 2343 women with primary invasive BC. Controls: 2467 controls frequency matched Premenopausal: cases: N= 863; Controls: N= 835 Mean age for premenopausal cases = 44.8 years and 62 years for postmenopausal cases	Database of chemical pollutants based on geographic industrial sources operating from 1960 to early 1990. Linked historical residential zip codes to chemical exposure database	Adjusted for age, province of residence, education, smoking pack years, alcohol consumption, number of live births, age at menarche, total energy intake, and employment in the industry under consideration.	Found overall associations for residential proximity to petroleum refinery, pulp mills, and steel mills as well as 1-9 years duration of residential proximity to steel mills
Residential proximity to thermal power plants, <.0.8 km vs. > 3.2 km or no plant
Overall	OR 1.56	(1.16, 2.10)
Premenopausal	OR 1.73	(1.06, 2.83)
Postmenopausal	OR 1.36	(0.92, 2.00)
Residential proximity to any above plants, <.0.8 km vs. > 3.2 km or no plant
Overall	OR 1.17	(0.97, 1.41)
Premenopausal	OR 1.10	(0.81, 1.49)
Postmenopausal	OR 1.08	(0.85, 1.37)
Duration of close residential proximity (<0.8 km) to thermal power plants
Overall, ≥ 10 years	OR 2.18	(1.21, 3.94
Premenopausal, ≥ 10 years	OR 6.42	(1.42, 29.0)
Postmenopausal, ≥ 10 years	OR 1.53	(0.77, 3.05)
Essex and Kent counties of Southern Ontario, Canada
Brophy, 2012105			Early onset, stratified by menopausal status , see main results	Case-control population-based study. Included1005 cases between regional cancer center recruited from 2002 to 2008 and 146 randomly-selected from the same geographic study area (frequency-matched). Mean year of diagnosis for cases = 2005 and mean year at interview for controls = 2006	Occupational history questionnaires. All jobs were industry- and occupation- coded for the construction of cumulative exposure metrics representing likely exposure to carcinogens and endocrine disruptors.	ORs (matched on age in 3 year- intervals, education, income, pregnancy history, smoking status). Adjusted for BMI in stratified models	Cumulative exposure evaluated at 10 year in high-exposed jobs (lagged 5 year). Study also examined windows of exposure in relation to reproductive events. No overall associations with other occupational exposures like agriculture, textile and dry cleaning
Exposure to auto plastics manufacturing
Overall	OR 1.09	(1.03, 1.15)
Premenopausal	OR 4.76	(1.58, 14.4)
Postmenopausal	OR 2.25	(1.09, 4.66)
Exposure to canning
Premenopausal	OR 5.70	(1.03, 31.5)
Postmenopausal	OR 1.47	(0.55, 3.97)
Northern states of Mexico
Lopez-Carrillo, 2010 91			Early onset: Stratified by menopausal status, see main results	233 cases from 25 hospital units in the northern states of Mexico, who were aged ≥18 years, had no other cancer history, and lived in study area > 1 year. 221 controls identified from housing lists used by the Health Department national surveys, matched to cases by age and residency. 37% of cases (n=87) and 33.5% of controls were premenopausal	Urinary phthalate metabolites measured in first morning void sample collected from 2007-2008 before treatment for cases. solid-phase extraction coupled with high-performance liquid chromatography/isotope dilution/tandem mass spectrometry	Adjusted for current age, age of menarche, parity, and menopausal status plus phthalate metabolites: DEHP metabolites were adjusted for non-DEHP metabolites; MEP, MBP, MiBP, BBzP, and MCPP were adjusted for themselves plus the sum of DEHP metabolites.
DEP/MEP, tertile 3 (ref = tertile 1)
Overall	OR 2.20	(1.33, 3.63)
Premenopausal	OR 4.13	(1.60, 10.7)
Postmenopausal	OR 1.84	(0.99, 3.42)
BBzP/MBzP, tertile 3 (ref = tertile 1)
Overall	OR 0.46	(0.27, 0.79)
Premenopausal	OR 0.22	(0.08, 0.61)
Postmenopausal	OR 0.61	(0.31, 1.19)
DOP and other phthalates /MCPP, tertile 3 (ref= tertile 1)
Overall	OR 0.44	(0.24, 0.80)
Premenopausal	OR 0.18	(0.05, 0.59)
Postmenopausal	OR 0.52	(0.25, 1.08)
DEHP/MECPP, tertile 3 (ref=tertile 1)
Overall	OR 1.68	(1.01, 2.78)
Black Women’s Health Study
Rosenberg, 2007 93			Early onset: Stratified by ages <45 & ≥45, see main results	From the original Black Women’s Health Study cohort, 48,167 women followed for 266, 298 person-years of follow-up with 574 incident cases of BC. Women were ages 21 to 69 and from across the USA.	Self-reported exposure to hair-relaxers	Adjusted for age and questionnaire cycle.	No association between BC risk and any categories of duration of hair relaxer use, frequency of use, age at first use, number of burns experienced during use, or type of relaxer used
Type of hair relaxer used (ref=nonuse)
overall association, Lye	IRR 1.16	(0.82, 1.65)
overall association, No-lye	IRR 1.07	(0.80, 1.43)
overall association, Unknown	IRR 1.05	(0.74, 1.49)
Age <45 year association, Lye	IRR 1.03	(0.53, 2.02)
Age <45 year association, No-lye	IRR 1.01	(0.56, 1.82)
Age <45 year association, Unknown	IRR 1.21	(0.58, 2.53)
Age ≥45 year association, Lye	IRR 1.23	(0.82, 1.87)
Age ≥45 year association, No-lye	IRR 1.10	(0.79, 1.52)
Age ≥45 year association, Unknown	IRR 1.01	(0.68, 1.51)
Breast Cancer Environment and Employment Study (BCEES)
Glass, 2015 88			Early onset: Analysis was stratified by menopausal status at diagnosis, see main results	Population based case-control study of 1,202 first incident invasive primary BC diagnosed between 2009-2011, identified from the Western Australian Cancer Registry. 1,785 controls randomly selected from the Western Australian electoral roll, matched to cases on age. 23.5% controls and 30.4% of cases are pre-menopausal.	Self-administered questionnaire was used to collect lifetime occupational history. The web- based application OccIDEAS was used to obtain details about job characteristics by telephone interview asking questions tailored to assess solvent exposure.	Adjusted for age, BMI, age at menarche, smoking, alcohol consumption, hormonal replacement therapy, age at first birth, parity, and family history of BC had minimal effect (<3%) so these variables were not included as adjustments
All participants, (ref = no)
Benzene	OR 1.08	(0.08, 1.47)
Other aromatic	OR 1.21	(0.97, 1.52)
Aliphatic	OR 1.21	(0.99, 1.48)
Chlorinated	OR 1.05	(0.69, 1.61)
Alcohol	OR 1.15	(0.96, 1.37)
Any Solvent	OR 1.15	(0.98, 1.35)
Premenopausal (ref = no)
Benzene	OR 1.53	(0.84, 2.80)
Other aromatic	OR 1.43	(0.92, 2.21)
Aliphatic	OR 1.33	(0.89, 2.0)
Chlorinated	OR 1.47	(0.62, 3.45)
Alcohol	OR 1.05	(0.74, 1.49)
Any Solvent	OR 1.14	(0.84, 1.56)
Post-menopausal (ref = no)
Benzene	OR 0.96	(0.67, 1.38)
Other aromatic	OR 1.15	(0.88, 1.49)
Aliphatic	OR 1.16	(0.92, 1.46)
Chlorinated	OR 0.94	(0.57, 1.54)
Alcohol	OR 1.16	(0.95, 1.43)
Any Solvent	OR 1.14	(0.95, 1.37)
Population-based case-control study in three counties of western Washington state
Cook, 1999 95			Early onset: Study of young women, only included women aged 45 years or less.	Population based case-control study in Washington state of 844 white women diagnosed with invasive (n = 747) or in situ (n = 97) BC. Cases were born after 1944, and diagnosed between January 1, 1983 and April 30, 1990, and resided in three counties of western Washington (King, Pierce, and Snohomish counties) 960 control women identified through random digit dial and frequency matched on 5-year age groups	Exposure information collected via in-person interview where women were asked about use of hair rinses and dyes, their natural hair color, the types of hair coloring used, the desired color results, the frequency of application, and the amount of time the product remained on their hair.	Adjusted for age in years, parity, weight in kilograms, and history of BC in first degree relative.	Also examined total lifetime episodes, total lifetime exposure, age at first hair color use, and time since last use, in all women and stratified non-exclusive and exclusive use of hair coloring. Also examined hairspray use.
Any use of hair coloring (ref = none)	OR 1.3	(1.0, 1.6)
Types of hair coloring used (ref = none)
Any rinse	OR 1.7	(1.2, 2.5)
Any semi-permanent	OR 1.4	(1.0, 1.8)
Any permanent	OR 1.2	(1.0, 1.6)
Any bleach then dye	OR 2.5	(1.6, 3.9)
Any frosting/tipping	OR 1.5	(1.2, 2.0)
Combination of types (ref = no hair coloring use)
Only one type	OR 1.1	(0.9, 1.3)
Any combination of two	OR 1.6	(1.3, 2.4)
Any combination of three	OR 2.7	(1.7, 5.5)
Population-based case-control study in Finland
Heikkinen, 2015 96			Early onset: Analysis was stratified by menopausal status at diagnosis, see main results	Population-based case-control study in Finland, Of 6,567 BC cases, aged 22–60 years and diagnosed in 2000–2007, and 21,598 matched controls Median age of BC diagnosis : 52 years	Self-administered questionnaire where women were asked to estimate the cumulative number of hair dye episodes during life, age at first use, and the types of dyes used.	Adjusted for birth year, parity, age at first birth, family history of BC, menarche age, use of hormonal contraceptives, physical activity, alcohol use, BMI, and level of education	The authors only report ever use of hair dyes findings for premenopausal women.
Ever use of hair dyes (ref=never)
Overall	OR 1.23	(1.11, 1.36)
Pre-menopausal BC	OR 1.19	(0.81, 1.75)
Post-menopausal BC	OR 1.25	(1.10, 1.43)
Shanghai Women’s Health Study
Mendelsohn, 200997			Early onset: Analysis was stratified by menopausal status at diagnosis, data not shown but authors note it is similar to overall finding of no association	Shanghai Women’s health study is a prospective cohort of women aged 40-70 years of age residing in Shanghai, China. This includes 73,366 women, and 592 BC cases. 47-51% of the cohort is postmenopausal	Self-administered questionnaire where women were asked about hair dye use, frequency of use, and duration of use. Inperson interviews were conducted as a follow-up to ensure completeness.	Adjusted for age, education, and smoking duration in pack/years.
Ever use of hair dyes (ref=never)
Overall BC	RR 0.93	(0.78, 1.09)
United Autoworker-General Motors (UAW-GM) cohort
Garcia, 2018106			Early onset, stratified by menopausal status , see main results	Cohort of hourly workers from 1 of 3 automobile manufacturing plants in Michigan who were exposed to metalworking fluid primarily via inhalation. 4,503 female autoworkers with 221 incident BCs (follow-up, 1985–2013). Mean age at dx = 62.1 years	Cumulative exposure estimates were calculated for each woman from employment records available from hire through 1994 and of a time-varying job-exposure matrix.	Adjusted for year of hire, calendar year, race, and manufacturing plant	Examined premenopausal breast cancer using a series of 4 age cut points, 55, 54, 53, and 52 years.
Exposure to synthetic metalworking fluid
Overall, ≥ 0.07 vs. 0	HR 0.77	(0.47, 1.26)
Premenopausal, ≥ 0.90 vs. 0	HR 3.76	(1.04, 13.51)
Exposure to continuous synthetic metalworking fluid
Overall, ≥ 0.07 vs. 0	HR 0.98	( 0.90, 1.07)
Premenopausal, ≥ 0.90 vs. 0	HR 1.17	(1.01, 1.35)
Thompson, 2005 107			Early onset, stratified by menopausal status , see main results	Nested case-control study in a cohort of 4,680 female hourly automobile production workers employed between January 1, 1941 and January 1, 1985 in 3 large manufacturing plants. Study included 99 BC cases and 626 matched controls	Developed scale factors using all sources of information, (e.g., historical estimates of exposure based on company record) and created exposure matrix. The job exposure matrix was then combined with individual employment histories to estimate cumulative exposure	Controls matching on date of birth, race, and being alive at the date of diagnosis of the case. Models also controlled for exposure during the other time periods	Stratified cases on age of diagnosis before or after age 51 to approximate menopausal status limitation of this study was the lack of data on some potential confounders, particularly reproductive history.
Soluble Metalworking Fluid Exposure, 1-10 years before risk
Dx before 51	OR 1.24	(0.82, 1.89)
Dx after 51	OR 0.89	(0.63, 1.24)
Long Island Breast Cancer Study
Lewis-Michl, 1996103			Early onset, stratified by menopausal status , see main results	Case-control study. Original cohort included 2 840 cases diagnosed with breast cancer between 1984 and 1986, and 1 420 matched controls identified through driver’s license records. Present study restricted to women who resided in Long Island for at least 20 consecutive years prior to diagnosis or interview. Study included 321 (22%) premenopausal women. Premenopausal: cases: N= 166; Controls: N= 155	A geographic information system was used to assign industry and traffic counts to 1-km2 grid cells (5-kM2 grid cells for traffic) and to assign potential exposure values to study subjects, based on 20-y residential histories.	Controls matched individually by age and county of residence; OR estimates were unadjusted	Standard Industrial Classification (SIC) major groups
Nassau County, NY
Premenopausal
Proximity to ≥ 1 industrial facilities in grid cell of residence
Chemical facilities included in SIC or other	OR 1.12	(0.60, 2.08)
Chemical	OR 0.82	(0.34, 1.96)
High-density traffic	OR 1.24	(0.41, 3.73)
Postmenopausal
Proximity to ≥ 1 industrial facilities in grid cell of residence
Chemical facilities included in SIC or other	OR 1.11	(0.84, 1.48)
Chemical	OR 1.62	(1.08, 2.43)
High-density traffic	OR 1.33	(0.80, 2.21)
Suffolk County, NY
Premenopausal
Proximity to ≥ 1 industrial facilities in grid cell of residence
Chemical facilities included in SIC or other	OR 0.95	(0.42, 2.12)
Chemical	OR 0.56	(0.13, 2.43)
High-density traffic	OR 0.96	(0.13, 7.00)
Postmenopausal
Proximity to > 1 industrial facilities in grid cell of residence
Chemical facilities included in SIC or other	OR 1.12	(0.72, 1.74)
Chemical	OR 1.61	(0.73, 3.54)
High-density traffic	OR 0.92	(0.42, 2.03)
Genetic susceptibility
Prince Edward Island Hospital
Ashley-Martin, 2012 81			Gene-environment: analysis stratified by CYP1A1 polymorphisms	Province-wide case-control study in Prince Edward Island, Canada with 207 cases and 621 controls frequency matched on age ±5 years, menopausal status, and family history of BC.	Fungicide exposure levels derived from 1991 Canadian Agricultural Census data and linked with participant’s postal codes using GIS.	Age, menopausal status, BC family history
Fungicide exposure (ref=< median)
Overall association	OR 0.72	(0.45, 1.12)
CYP1A1*2A variants	OR 0.79	(0.55, 1.12)
Northern States of Mexico
Gamboa-Loira, 2017100			Gene-environment: evaluated interaction between polymorphisms in AS3MT, FOLH1, MTHFD1, MTHFR, MTR, and MTRR genes (encode enzymes involved in one carbon metabolism and inorganic arsenic methylation pathways) and inorganic arsenic metabolism	Population-based case-control study in five northern Mexico states. Cases were identified in public tertiary care hospitals. Controls were selected randomly from residing area.	Urinary arsenic metabolites measured by HPLC in urine collected around interview. In all cases, urine samples were obtained before any cancer treatment was performed.	Adjusted for log-transformed TAs-AsB, AsB age, BMI, total breastfeeding, alcohol, creatinine, smoking, and age at first pregnancy	Genotypes examined: AS3MT c.860T>C, AS3MT c.529-56G>C, FOLH1 c.223T>C, MTHFD1 c.1958G>A, MTHFR c.665C>T, MTR c.2756A>G and MTRR c.66A>G A significant interaction (p =0.002) between MTR c.2756A>G polymorphism and %DMA on BC was found; BC risk related with %DMA was lower in AG+ GG carriers than in AA carriers. No other significant interactions were found.
%inorganic arsenic (iAs)
FOLH1 c.223T>C, CC	OR 2.33	(1.05, 5.18)
%monomethylarsonic/monomethylarsonic (MMAs)
All genotypes	OR 2.15	(1.67, 2.77)
%dimethylarsinic acid (DMA)
All genotypes	OR 0.56	(0.37, 0.86)
MTR c.2756A>G, AG+GG	OR 0.13	(0.04, −0.40)
%monomethylarsonic / inorganic arsenic (MMA/Ias)
All genotypes	OR 1.43	(1.21, 1.69)
%dimethylarsinic acid / monomethylarsonic (DMA/MMA)
All genotypes	OR 0.58	(0.48, 0.71)
Pineda-Belmontes, 2016 101			Gene-environment: evaluated interaction between selected polymorphisms of PPARγ Pro12Ala and PPARGC1B Ala203Pro and inorganic arsenic metabolism	Case-control study. Cases were identified in public tertiary care hospitals. Controls were selected randomly from residing area.	Urinary arsenic metabolites measured by HPLC in urine collected around interview. In all cases, urine samples were obtained before any cancer treatment was performed.	Controls age-matched with cases (±5 years). Adjusted for the variable AsT-AsB, AsB, and creatinine, age, age at 1st pregnancy, lactation throughout life, BMI, tobacco use, and alcohol consumption	The authors note that polymorphisms in PPAR gamma did not modify the association between arsenic methylation capacity and BC risk (no significant interactions observed)
PPARγ Prol2Ala, C/C (tertile 3 vs tertile 1)
inorganic arsenic	OR 0.58	(0.31-1.06)
monomethylarsonic	OR 3.28	(1.72-6.28)
dimethylarsinic acid	OR 0.65	(0.35-1.21)
primary methylation	OR 2.96	(1.57-5.59)
secondary methylation (vs tertile 2)	OR 0.33	(0.18-0.63)
total methylation, (vs tertile 2)	OR 1.51	(0.82-2.77)
PPARγ Pro12Ala, C/G+G/G (tertile 3 vs tertile 1)
inorganic arsenic	OR 0.19	(0.04-0.90)
monomethylarsonic	OR 5.58	(1.33-23.48)
dimethylarsinic acid	OR 2.25	(0.55-9.97)
primary methylation	OR 11.34	(2.28-56.42)
secondary methylation (vs tertile 2)	OR 0.34	(0.09-1.24)
total methylation, (vs tertile 2)	OR 3.6	(0.83-15.70)
PPARGC1B Ala203Pro, C/C
inorganic arsenic	OR 0.42	(0.21-0.82)
monomethylarsonic	OR 2.78	(1.44-5.37)
dimethylarsinic acid	OR 0.87	(0.46-1.66)
primary methylation	OR 3.60	(1.85-7.00)
secondary methylation (vs tertile 2)	OR 0.44	(0.23-0.84)
total methylation, (vs tertile 2)	OR 2.05	(1.07-3.93)
PPARGC1B Ala203Pro, C/G+G/G
inorganic arsenic	OR 0.77	(0.25-2.34)
monomethylarsonic	OR 8.64	(2.18-34.21)
dimethylarsinic acid	OR 0.61	(0.18-2.04)
primary methylation	OR 3.85	(1.06-13.95)
secondary methylation (vs tertile 2)	OR 0.12	(0.03-0.47)
total methylation, (vs tertile 2)	OR 0.98	(0.32-3.06)
Danish National Birth Cohort
Ghisari, 2017 84			Gene-environment: Analysis stratified by polymorphisms in CYP1A1, CYP1B1, COMT, CYP17A1, CYP19A1.	Nested case-control within the Danish National Birth Cohort (1996-2002). 178 cases and 233 controls age matched.	Serum concentrations of 16 PFASs measured through liquid chromatography tandem mass spectrometry. Blood serum samples collected during the 1st and 2nd pregnancy trimester.	Adjusted for age at blood drawing, BMI before pregnancy, gravidity, oral contraceptive use, menarche age, smoking, alcohol, maternal education and physical activity.	Significant interactions observed for COMT(Val158Met) *PFOSA (p-interaction =0.048) and for CYP19 *PFOA (p-interaction = 0.047), and borderline significant for PFOS * CYP19 (p-interaction =0.055). P-interaction not significant for other overall cross-product GXE terms, although individual RRs were significant. There was evidence of elevated but not statistically significant associations for PFOS and CYP1A1, CYP1B1, COMT, and CYP17 genotypes. There was no association of an elevated BC risk for PFHxS and all genotypes examined, but there was evidence of reduced BC risk for CYP1A1, CYP1B1, COMT, CYP17, and CYP19 genotypes.
PFOSA >median (ref= ≤median)
CYP1A1 Ile/Ile	RR 1.20	(0.94, 1.54)
CYP1B1 Leu/Leu	RR 1.70	(1.03, 2.82)
CYP1B1 Val/Val	RR 1.95	(1.10, 3.47)
CYP1B1 Leu/Val+Val/Val	RR 1.21	(0.89, 1.66)
COMT Val/Met	RR 1.20	(0.81, 1.77)
COMT Met/Met	RR 2.04	(1.27, 3.28)
COMT Val/Met+Met/Met	RR 1.46	(1.09, 1.93)
CYP17 A1A1	RR 2.02	(1.29, 3.16)
CYP17 A1A2	RR 1.11	(0.75, 1.64)
CYP19 CC	RR 2.08	(1.06, 4.09)
CYP19 CT	RR 1.25	(0.89, 1.74)
CYP19 CT+TT	RR 1.17	(0.89,1.52)
PFOS >median vs. ≤median
CYP19 CC	RR 6.42	(1.08, 38.3)
CYP19 CT	RR 1.16	(0.44, 3.10)
PFOA >median vs. ≤median
CYP19 CC	RR 7.24	(1.00, 52)
Greenlandic Inuit Women
Ghisari, 2014 83			Gene-environment: examined gene-environment interactions of polymorphisms in metabolism genes CYP1A1, CYP1B1, COMT, CYP17, and CYP19 and serum PFASs	Case-control study in Greenlandic Inuit women with 31 cases and 115 controls frequency matched by age and districts of residence from 2000-2003.	Serum measurement of PFASs (seven PFASs including perfluorooctane sulfonate (PFOS) and perfluorooctanoic acid (PFOA))	Adjusted by age and serum cotinine	Did not find significant interaction between the PFAS variables and gene polymorphisms, despite different BC risks for different strata of gene polymorphisms. Several genotype strata were not analyzed because of small sample size.
PFOS ≥median vs. <median
CYP1A1 Ile/Val + Val/Val	OR 12.1	(1.29, 115)
CYP1B1 Leu/Leu	OR 11.2	(1.8, 71.1)
COMT Val/Met/Met	OR 16.8	(1.68, 167)
CYP17 A1/A2 + A2/A2	OR 18.2	(1.67, 198.8)
CYP19_CT CC	OR 9.6	(1.48, 62.4)
CYP19_TTTA (TTTA)8-10	OR 29.3	(2.89, 298)
PFOA ≥median (ref= <median)
CYP1A1 Ile/Ile	OR 10.5	(0.12, 907)
CYP1A1 Ile/Val + Val/Val	OR 3.58	(0.81, 15.8)
CYP1B1 Leu/Leu	OR 2.1	(0.59, 7.6)
COMT Val/Val	OR 8.32	(0.29, 235)
COMT Val/Met + Met/Met	OR 3.73	(0.86, 16.3)
CYP17 A1/A1	OR 0.84	(0.093, 0.60)
CYP17 A1/A2 + A2/A2	OR 8.79	(1.22, 63.5)
CYP19_CT CC	OR 2.44	(0.57, 10.4)
CYP19_CT CT + TT	OR 4.99	(0.34, 74.7)
CYP19_TTTA (TTTA)8-10	OR 3.28	(0.74, 14.6)
Wielsoe, 201886			Gene-environment: evaluated interaction between SNPs in CYP1A1, CYP1A2, CYP1B1, CYP3A4 and COMT and exposure to persistent organic pollutants	Case-control study in Greenlandic Inuit women with 77 cases and 84 controls frequency matched by age and districts of residence from 2000-2003.	Serum measurement of PFAS (including PFOS and PFOA), PCBs, and OCPs	Only unadjusted analyses are reported, as none of the potential confounders	Did not find any other significant associations.
High exposure (>median) vs. low exposure ≤median
CYP17A1 -34T>C, ΣPFAA, TT	OR 8.55	(1.84, 39.1)
CYP19A1 *19C>T, ΣPFAA, CT + TT	OR 2.84	(1.09, 7.42)
CYP1A1 Ile462Val, ΣPFAA, Ile/Ile	OR 5.18	(1.15, 23.3)
CYP1B1 Leu432Val, ΣPFAA, Leu/Val + Val/Val	OR 6.19	(1.23, 31.3)
Mexico City 17 tertiary hospitals
Martinez-Nava, 2013 92			Gene-environment: examined interaction with polymorphisms in transcription factor genes PPARγ and PPARGC1B and phthalate metabolites	Case-control study within the study area population of 17 Mexican tertiary hospital units of Mexico City. 208 cases and 220 population controls matched by age and residency included.	Urinary phthalate metabolites from first-void urine analyzed through HPLC. Five phthalate metabolites assessed MEP, MBP, MiBP, MBzP, MCPP. Four DEHP metabolites assessed MEHP, MEHHP, MEOHP, MECPP.	adjusted by age, age at menarche, parity, menopausal status, and phthalate metabolites
Sum of DEHP metabolites, tertile 3 (ref=tertile 1)
PPARγ, C	OR 2.05	(1.36, 2.98)
PPARγ, G	OR 0.89	(0.26, 3.10)
PPARGC1B, G	OR 1.73	(1.17, 2.56)
PPARGC1B, C	OR 1.70	(0.50, 5.83)
DEP/MEP, tertile 3 (ref=tertile 1)
PPARγ, C	OR 1.87	(1.25, 2.78)
PPARγ, G	OR 4.31	(1.32, 14.11)
PPARGC1B, G	OR 1.82	(1.22, 2.72)
PPARGC1B, C	OR 4.13	(1.27, 13.48)
DEHP/MEHP, tertile 3 (ref=tertile 1)
PPARγ, C	OR 1.64	(1.12, 2.41)
PPARγ, G	OR 0.74	(0.23, 2.33)
PPARGC1B, G	OR 1.34	(0.91, 1.98)
PPARGC1B, C	OR 2.12	(0.65, 6.89)
DEHP/MEHHP, tertile 3 (ref=tertile 1)
PPARγ, C	OR 1.80	(1.23, 2.64)
PPARγ, G	OR 0.82	(0.25, 2.72)
PPARGC1B, G	OR 1.34	(0.91, 1.98)
PPARGC1B, C	OR 2.12	(0.65, 6.89)
DEHP/MECPP, tertile 3 (ref=tertile 1)
PPARγ, C	OR 2.35	(1.58, 3.50)
PPARγ, G	OR 1.52	(0.45, 5.12)
PPARGC1B, G	OR 1.98	(1.33, 2.94)
PPARGC1B, C	OR 2.68	(0.75, 9.54)

^a.OR = Odds Ratio, HR= Hazard Ratio, RR= relative risk, ROR= ratio of the odds ratio

^b.95% CI = 95% Confidence interval

Air pollution

Tables 2a and 2b summarize the 35 publications from 15 distinct epidemiologic studies that examined exposure to air pollution and BC risk, and that considered family history, early onset BC and/or genetic susceptibility. These studies examined air pollution exposure from PAHs ^6,10–23 or other aromatic DNA adducts^24,25,traffic-related air pollution and vehicular exhaust (as a proxy for PAH) ^26–30, indoor heating and cooking ^31,32 , annual ambient air concentrations ^33–35, ambient fine-particulate matter (PM_2.5, PM₁₀)^35–37, and nitrogen dioxide (NO₂) ^37,38, and consumption of grilled and/or smoked meat as a proxy for PAH and heterocyclic amine exposure^39–42. Fifteen publications (43%) used biospecimens to measure exposure to air pollution including plasma PAH-albumin adducts/PAH-DNA adducts ^{6,10–14,17–22}, urinary PAH metabolites ¹⁶, and aromatic DNA adducts (not specific to only PAH and include a wide range of bulky and hydrophobic adducts)^24,25, while ten publications (29%) used geographic assessment to measure air pollution exposure ^{27–29,33–38,43}.

Five publications considered BC family history in either the study design (4 publications) or analysis (1 publication). Four publications were from two cohorts enriched for underlying familial BC risk, the Breast Cancer Family Registry (BCFR) and the Sister Study. For example, using multigenerational pedigree data in the BCFR, a 4-fold increased BC risk was observed from PAH exposure in the top quantile (OR = 4.09, 95% CI = 1.38, 12.13) for women with a 10-year absolute risk of ≥3.4% compared to women in the lowest quantile of PAH exposure of average BC risk⁶. Publications from the Sister Study reported modest statistically significant associations (ORs: 1.11- 1.17) for overall indoor heating and cooking use, including having an indoor wood-burning stove/fireplace in the longest adult residence, or burning wood or natural gas/propane in the residence³¹, as well as an increased risk of ER+/PR+ BC subtype for women exposed to higher annual average concentrations of NO₂³⁷. A Canadian population-based case-control study not enriched for underlying familial risk, where 14% of controls and 19.8% of cases had a family history of BC¹⁵, reported that women with a first-degree family history and high level of occupational PAH exposure or a long duration of PAH exposure had greater than 2-fold increased BC risk (OR=2.27, 95% CI=1.34, 3.86 and OR = 2.79, 95% CI = 1.25, 6.24, respectively), compared to never exposed women; while women with no family history and high level of occupational PAH exposure had a 31% significant increased BC risk ¹⁵.

Over 60% of the publications that examined air pollution and BC risk considered early onset BC (22 publications in 14 epidemiologic studies) ^{6,12,13,15,16,23–28,30–36,38,39,42,43}. Less than 20% of these publications (four publications) were from epidemiologic studies that were enriched for early onset BC, either by oversampling for premenopausal women (National Enhanced Cancer Surveillance System; NECSS)³⁸, initially recruiting younger women aged 25–42 years (Nurses’ Health Study II; NHSII)^35,36, or limiting the study to pre-menopausal BC (population-based case control in western New York)²³, while the remainder of the publications stratified by menopausal status. Two of the four publications from enriched studies reported a statistically significant increased risk for premenopausal BC, with a 26% increased risk for residential exposure to NO₂ using satellite estimates in the NECSS³⁸, and an 82% and 91% increased risk for ever being occupationally exposed to PAH and benzene, respectively in western NY²³. The majority of studies examining premenopausal women had point estimates that were elevated but only the following studies were statistically significant for PAH ^{6,12,15,27,28} aromatic DNA –adducts²⁴, or dietary exposure to HCAs and PAHs (high intake of smoked meat)³⁹, with ORs ranging from 1.5–6.7 (see Table 2a).

Over 40% of air pollution publications (15 out of 35 publications), the majority from the Long Island Breast Cancer Study Project (LIBCSP), examined gene-environment interaction with PAH, with the majority assessing PAH exposure using a biomarker (9 publications) ^{10,11,14,17–22,27,29,32}. All but one publication (14 of 15 publications) reported a statistically significant association of increased BC risk (ORs ranging from 1.19–3.81) for high PAH exposure and variants in XPD^11,21, XRCC1¹⁹, IGHMBP2²⁰, FAS1377¹⁰, ERCC1¹¹, XPC¹⁸, GST^17,32, RARB²², ERCC2²⁹, and high consumption of grilled-smoked meat (as a proxy for PAH and HCA exposure) and variants in CYP1B1⁴⁰, SULT1A1³⁹, and NAT2⁴¹ (see Table 2a).

Persistent Endocrine Disrupting Chemicals

Tables 3a and 3b summarize the 32 publications from 27 distinct epidemiologic studies included in this review that examined persistent EDCs including organochlorine pesticides, DDT/DDT, PCBs, and/or total effective xenoestrogens burden (TEXB). Over 80% of publications (27 out of 32 EDC publications) used biomarkers (in serum, plasma or tissue) to measure exposure. Two studies used geographic assessment data to measure exposure ^44,45 and three studies used self-reported exposure data from questionnaires ^46–48.

19% of EDC publications (6 out of 32 publications) considered BC family history in either the study design (one publication)⁴⁶ or analysis (5 publications)^44,48–51. The one publication from an enriched cohort, the Sister Study, found no association between self-reported residential and farm exposure to pesticides in childhood and BC risk ⁴⁶. Of the five studies not enriched that examined modification by family history in analyses, three reported statistically significant overall associations with EDC biomarkers, including for TEXB-α (OR = 3.45, 95% CI = 1.45, 7.97) and TEXB-β (OR = 4.01, 95% CI = 1.88, 8.56)⁴⁹ , DDE (OR = 1.95, 95% CI = 1.10, 3.52)⁵¹, and PCB 105 (OR = 3.17, 95% CI = 1.51, 6.68) and PCB 118 (OR = 2.31, 95% CI = 1.11, 4.78)⁵⁰. The other two studies, both of which used self-reported exposure measures, found no overall associations that were statistically significant^44,48. None of the studies found a statistically significant multiplicative interaction by family history; only one publication provided results stratified by family history^44,48–51 (see Table 3a).

Over 80% of persistent EDC publications (28 out of 32 publications) considered early onset BC but only one study (Child Health and Development Studies; CHDS) was specifically enriched for young women, with all cases diagnosed younger than age 54 years. The CHDS used as a proxy for age at first DDT exposure, the age of the women in 1945 which was the year DDT was first introduced in the US. This study found that both timing of exposure and timing of outcome influenced DDT risk estimates. For example, DDT measured in serum (from blood samples collected during pregnancy) were associated with increased BC risk in women under age 50 (OR = 3.70, 95% CI = 1.22, 11.26), but was not associated with BC risk in women 50–54 years (OR = 0.92, 95% CI = 0.52, 1.63) for women whose approximate age at first exposure to DDT was less than 3 years; whereas later DDT exposure in childhood was associated with BC risk in women 50–54 years⁵². Sixteen case-control studies examined DDT/DDE measured in serum or tissue collected at study baseline and BC risk and considered early onset BC in the analysis by stratifying on menopausal status (where ≤34% of cases premenopausal at baseline ^{50,51,53–58}) or age (with between <23% - <50% ^53–58 of cases under 50 at baseline). Some studies reported elevated but non-statistically significant risk for premenopausal BC and serum DDE^50,51,53,55, but more studies that stratified by menopausal status in the analysis stage, reported no difference by menopausal status^59–65. It is important to note, as mentioned in previous reviews of DDT/DDE and BC risk⁶⁶, that some of these studies did not have extensive past exposure histories, so there may be potential exposure misclassification. Both the LIBCSP and the Sister study estimated this information based on questionnaire data and self-reported exposure to pesticides during different periods of life. In the LIBCSP, where 32% of cases were pre-menopausal at baseline, there was a statistically significant association between self-reported exposure to fogger trucks (a proxy for DDT exposure) and BC risk only in post-menopausal women (OR =1.25; 95% CI: 1.05, 1.50), and not pre-menopausal women (OR= 0.96; 95% CI: 0.73, 1.26)⁴⁷. In contrast, the Sister Study reported a modest increase in premenopausal BC risk (HR=1.3; 95% CI 0.9, 1.7) for women who were exposed to fogger trucks or planes and born 1941–1958, who would have been ages 0–18 during years of peak DDT use in the U.S., and a null finding for postmenopausal BC (HR = 1.0; 95% CI: 0.9, 1.2)⁴⁶.

In the CHDS, net PCB serum levels measured during the early postpartum window were associated with increased BC risk in women under 50 years (quantile 4 vs 1 OR = 2.81, 95% CI = 1.11, 7.09) ⁶⁷. When specific PCB congeners were assessed in the CHDS, PCB 203 was associated with increased BC risk (quantile 4 vs 1 OR = 6.34, 95% CI = 1.85, 21.73), while PCB 167 was associated with decreased BC risk (quantile 4 vs 1 OR = 0.24, 95% CI = 0.07, 0.79)⁶⁷. The authors also classified these by different biological activity according to previous recommendations⁶⁸, of note is that PCB 167 is dioxin like and anti-estrogenic while PCB 203 is Phenobarbital and a CYP1A and CYP2B inducer. Of twelve case-control studies that were not enriched for young women, two studies found a statistically significant positive associations overall and for premenopausal women for total PCB levels in blood samples collected at study baseline with BC risk ⁶⁹ and for PCB congeners 105 and 118 measured in breast tissue in a hospital-based case-control study in Canada⁵⁰ and the remaining 10 studies reported no statistically significant associations ^{53,56,58,59,61–64,70,71}. In the CECILE Study, a population-based case-control study, there was evidence of an inverse association between PCB 153 and overall BC risk (OR = 0.75, 95% CI = 0.57, 0.97); however, there was no association for premenopausal women⁵⁵. Additionally, a population-based study in Wisconsin examined the association of self-reported recent sport-caught fish consumption as a potential source of exposure to PCBs, DDT, PBDEs, and other halogenated hydrocarbons and BC risk, and reported a 70% increased risk for premenopausal BC who had any recent consumption of Great Lakes fish (RR 1.70; 95% CI: 1.16, 2.50)⁴⁸.

Over 12% of EDC publications (5 out of 32 publications) considered gene-environment interaction, specifically for PCBs^70,72–75. Four of these publications examined polymorphisms in the CYP1A1 gene^70,73–75, which is involved in the metabolism of steroids and several potentially genotoxic exposures including PCBs. Three publications reported significant gene-environment interactions, with elevated BC risks (ORs ranging from 2.93–2.96) for high PCB exposure and variants in CYP1A1 ^70,74,75. Two publications reported no evidence of gene-environment interaction for PCBs and variants in CYP1A1, CYP1B1, and COMT^72,73.

Pesticides/Fungicides/Insecticides

Tables 4a and 4b summarize the six publications from four epidemiologic studies that examined exposure to pesticides, fungicides, or organophosphate insecticide and BC risk. Exposures were measured through self-reported questionnaires^76–79 or geographic assessment^80,81.

Two publications considered family history either by examining the association in an enriched cohort (Sister Study)⁷⁶ or stratifying by first-degree family history (Agricultural Health Study; AHS)⁷⁸. In the AHS, women with a first-degree family history whose husbands used the insecticide parathion or the herbicide paraquat had a 4-fold increased risks for pre-menopausal BC, while no association was found for women with no family history whose husbands used the same chemicals⁷⁸. The Sister Study publication reported no association between having any occupational exposure to pesticides and BC risk ⁷⁶. Five publications considered early onset of BC in the analysis by stratifying based on menopausal status^76–79 or by age⁸⁰ at baseline. Three publications reported no association for pesticide exposure and premenopausal BC^76,77,80. In the AHS, organophosphate insecticide use among female spouses of pesticide applicators was associated with an overall elevated risk of breast cancer, but found that menopausal status did not modify the association between organophosphate use and BC risk⁷⁷. Similarly, a later publication in the AHS reported an increased risk for BC associated with organophosphate use, and this publication reported an even greater association for premenopausal BC, particularly for terbufos use⁷⁹. While over 15,000 women were pre-menopausal at enrollment, the number of exposed premenopausal BCs ranged from 4–105, depending on type of pesticide⁷⁷. One study assessed gene-environment interactions between fungicide exposure and a CYP1A1 polymorphism and found no statistically significant interactions⁸¹ (see Tables 4a and 4b).

Per-and polyfluoroalkyl substances (PFAS)

As summarized in Tables 4a and 4b, 5 publications^82–86 from 3 distinct epidemiologic studies examined exposure to per-and polyfluoroalkyl substances (PFAS) and BC. All 5 publications measured exposure to PFAS in blood specimens, with varying periods of measurement including: prior to the onset of BC ^82,84, at the time of BC diagnosis for cases but before treatment^83,86 or an average of 35 months after case diagnosis ⁸⁵.

Two publications considered early onset of BC by stratifying by menopausal status⁸⁵ or by age at baseline⁸², and both did not report statistically significant increased premenopausal BC risk associated with exposure. In CTS, where 5% of cases and 7% of controls were premenopausal at baseline, Hurley et al. found no significant association between exposure to six PFASs and BC risk for overall or premenopausal BC⁸⁵. In the Danish National Birth Cohort Study of premenopausal BC, higher perfluorohexane sulfonate (PFHxS) were associated with a significant decreased risk of BC in women aged 40 years or younger (quintile 5 vs. quintile 1 RR= 0.41, 95% CI=0.17, 0.96) , while higher levels of perfluorooctane sulfonamide (PFOSA) were associated with a significant increased risk of BC in women aged 40 years or younger (quantile 5 vs quantile 1 RR 2.45, 95% CI = 1.00, 6.00)⁸².

Three studies examined gene-environment interactions and serum concentrations of PFASs^83,84,86. These studies report an increased BC risk (ORs 1.7–18) for high serum levels of perfluorooctane sulfonate (PFOS) and perfluorooctanoic acid (PFOA) and variants in CYP1A1, COMT, or CYP17 A1 genes⁸³, and ΣPFAA exposure and variants in CYP17A1, CYP19A1, CYPA1, CPYB1 genes in Greenlandic Inuit women⁸⁶ and for high serum levels of PFOSA and variants in COMT ⁸⁴.

Organic Solvents

Tables 4a and 4b summarize 4 publications from 3 epidemiologic studies that examined exposure to organic solvents and BC risk^76,87–89. Self-reported occupational and workplace history questionnaires were used to measure cumulative lifetime exposure to organic solvents in all 4 publications^76,87–89.

Three publications considered BC family history in the study design^76,87 or analysis⁸⁹. The 2 publications from the enriched cohort, the Sister Study, reported a statistically significant 28%-39% increased risk of ER-positive BC for solvent exposure before 1980 or before their first birth⁸⁷, a 60% increased risk of BC occupational solvent exposure from 120 to ≤ 438 days⁷⁶, and 2.3-fold increased BC risk for high cumulative exposure to gasoline and petroleum products⁷⁶. The publication that examined this association in an unenriched cohort found that BC family history did not modify the association between organic solvents or benzene and BC risk⁸⁹.

Three publications considered early onset on BC in their analysis by stratifying based on menopausal status ^76,88,89 or by age ⁸⁹ at baseline. All 3 publications reported no statistically significant elevated association for solvent exposure and premenopausal BC, including exposure to solvents, gasoline or petroleum products, or benzene^76,88,89. However, all three of these publications report elevated, but not statistically significant associations for exposure to solvents and pre-menopausal BC, with ORs ranging from 1.3 to 1.57^76,88,89. For example, a large population based case-control study in Poland also assessed cumulative exposure to organic solvents, estimated as the as the product of the scores for exposure intensity, frequency, and duration, summed across all jobs, based on occupational data collected from questionnaires that were assessed by Polish industrial hygienists⁸⁹. This study reported elevated risk of premenopausal BC for women with high cumulative exposure (OR: 1.57, 95% CI: 0.99, 2.5), but not for post-menopausal BC⁸⁹.

Phthalates

Three publications^90–92 from 2 studies examined exposure to phthalates and BC risk while considering early onset BC in the analysis ^90,91 or gene-environment interaction ⁹² (Tables 4a and 4b). Exposure to phthalates were measured by linking drug ingredient data with the Danish National Prescription registry in Danish women⁹⁰ or via urine samples that were collected after BC diagnosis⁹¹.

The two publications that considered early onset BC stratified by menopausal status at baseline^90,91. In the Danish Nationwide Cohort Study, the association between cumulative DBP (≥10,000 mg vs no exposure) and BC risk was modestly stronger for premenopausal women (HR estimates not provided in published article)⁹⁰. Of premenopausal women in the Northern Mexico study, there was a statistically significant 4-fold increased risk of BC for women at the highest MEP exposure level and a negative association between MBzP and MCPP and BC risk⁹¹.

Gene-environment interaction between urinary phthalate metabolites and BC risk was examined in one publication in the Northern Mexico study ⁹². Statistically significant gene-environment interaction (ORs 1.64–2.35) was reported for polymorphisms in transcription factor genes PPARγ and PPARGC1B and phthalate metabolite MEP as well as DEHP metabolites MEHP, MEHHP, MECPP ⁹².

Personal Care Products

Five publications in five epidemiologic studies examined self-reported personal care and hair product use and BC ^93–97 (Tables 4a and 4b). One publication considering BC family history in the study design ⁹⁴, one publication considering early onset in the study design⁹⁵, and four publications considered early onset of BC in the analysis ^93,94,96,97. In the Sister Study, latent class analysis was used to identify mutually exclusive groups of women with differing patterns of personal care product use and a statically significant 15% increased risk of BC for White women who were frequent users for beauty care products was reported ⁹⁴. Further stratification by menopausal status revealed an increased risk of postmenopausal BC for white women who were frequent users for beauty care products and frequent users of skincare products, but no association for premenopausal women ⁹⁴. A population-based case-control study in Washington state of women aged 45 years or younger found a statistically significant 30% elevated BC risk for women who reported ever coloring their hair using any method (including use of rinses, semi-permanent or permanent dyes, bleaching then dyeing their hair, or frosting their hair)⁹⁵. This elevated risk was greater for women who reported any combination of two types (OR=1.8) or three types (OR = 3.1) of hair coloring⁹⁵. A publication from the Black Women’s Health Study, found no statistically significant association between hair relaxer use and BC in their overall analyses and after stratifying by age at baseline (<45 vs. ≥ 45 years), but report a 20% elevated BC risk for young women (<45 years) who used hair relaxers with unknown ingredients, but not for those with lye or no-lye⁹³. A population-based case-control study in Finland reported a statistically significant 23% increased BC risk for ever use of hair dyes for BC overall, and a 19% increased risk for pre-menopausal women although this was not statistically significant⁹⁶. In contrast, a prospective cohort of Chinese women reported no association for hair dye use and BC risk in overall analyses or in stratified analyses by menopausal status⁹⁷.

Polybrominated diphenyl ethers (PBDEs)

Only one study (CTS) that examined PBDE and BC risk was included in this review, which considered early onset BC in the analysis, using blood samples that were collected from cases an average of 35 months after BC diagnosis and reported no evidence of overall an association between serum concentrations of 19 PBDE congeners and BC risk and after stratifying by menopausal status⁸⁵.

Metals and Soldering Materials

As summarized in Tables 4a and 4b, 6 case-control studies^76,98–102 from 3 distinct studies examined exposure to metals and BC risk. Exposure to metals and soldering materials were measured through urinary biomarkers^100,101, self-reported occupational questionnaires⁷⁶, toenail clippings taken after BC diagnosis in cases⁹⁸ or by geographic assessment data^99,102.

Three studies considered BC family history in the design by using the Sister Study cohort data^76,98,102. There was no significant association between any occupational exposure to metals and soldering materials⁷⁶, airborne mercury¹⁰², or metallic air pollutants⁹⁸ and breast cancer risk. Two publications considered early onset BC in the analysis by stratifying on menopausal status^76,102 or by restricting the study sample to women under 50⁹⁸. Occupational exposure to soldering materials has been found to be associated with an 80% increased risk of premenopausal BC for women⁷⁶. There was no association for premenopausal BC and metallic air pollutants ¹⁰² or metal concentrations measured in toenail clippings ⁹⁸. The French E3N Cohort study reported an overall association between cumulative airborne cadmium exposure as measured by geographic assessment data and BC risk, but no statistically significant association for premenopausal BC⁹⁹. Two publications^100,101, both from a case-control study in Northern Mexico, examined gene-environment interactions for polymorphisms in AS3MT, FOLH1, MTHFD1, MTHFR, MTR, and MTRR genes (which encode enzymes involved in one carbon metabolism and inorganic arsenic methylation pathways)¹⁰⁰, or PPARγ¹⁰¹ and inorganic arsenic metabolism. One publication reported a decreased risk for BC to be association with inorganic arsenic exposure (measured as percentage dimethylarsinic acid; DMA) and MTR c.2756A>G polymorphism , and no significant interaction for the other genes assessed ¹⁰⁰, while another publication reported no significant gene-environment interaction for polymorphisms in PPARγ and inorganic arsenic exposure¹⁰¹.

Other Exposures including Occupational Exposures

Two publications from two studies (case-control study in Long Island, NY¹⁰³ and NECSS¹⁰⁴) assessed residential proximity to industrial plants. The case-control study in Long Island did not report significant increased risk for premenopausal BC, but found an increased risk for post-menopausal BC for women exposed to chemical facilities¹⁰³. The NECSS publication found significantly increased risk for premenopausal BC and residing in close proximity to a thermal power plant¹⁰⁴. The authors note that the fuel type used by thermal power plants includes coal, diesel, heavy fuel oil, light fuel oil, natural gas, nuclear, spent pulping liquor wood refuse, and waste heat and these plants release known or suspected carcinogens, including metals, radioactive elements, and polycyclic organic matter¹⁰⁴. It is important to note that both of these studies had small numbers of premenopausal women that were highly exposed (ranges from N=1–50), and are likely underpowered to fully examine these associations.

Three publications in two studies (case-control study in Canada¹⁰⁵ and in a cohort of female autoworkers^106,107) examined occupational exposures, and found significantly increased risk for premenopausal BC for women exposed to auto plastics manufacturing and canning¹⁰⁵ , and metalworking fluid¹⁰⁶. Additionally there was a statistically significant increase in overall BC risk for women in jobs with high exposures to carcinogens and endocrine disrupting chemicals, underscoring the value of collecting detailed occupational exposure histories in environmental epidemiologic studies¹⁰⁵.

Discussion

This systematic review specifically focused on studies that have examined ECE and BC risk in women at higher absolute BC risk. We found that although there have been several studies that considered associations of ECE by BC susceptibility, few studies of ECEs have specifically enriched for high-risk women. Enriched cohorts based on family history or early onset BC are a powerful way to test for gene-environment interactions, as population-based studies may be underpowered to detect interactions for those at highest absolute risk based on their family history¹⁰⁸. Of the 56 unique epidemiologic studies included in this systematic review, only 2 studies (<4%) were specifically enriched for underlying risk based on family history of BC and only 6 studies (11%) were enriched for early onset BC; these 8 studies resulted in less than a quarter of the publications included in this review (22%, 22 publications). Over 70% of publications that were from studies enriched for either family history (8 out of 10 publications) or early onset BC (8 out of 10 publications), or that considered underlying genetic susceptibility (20 out of 27 publications) reported statistically significant associations.

Type 1 Studies: Family History:

Of the 100 publications included, 18% (18 publications from 9 epidemiologic studies) considered BC family history in either the study design (by enriching the study sample for women at higher increased BC risk based on family history) or analysis (by examining effect modification by family history either through incorporating an interaction term in the analysis or stratifying by family history) (Table 1a, Table 1b). Of these, 10 publications came from two epidemiologic studies which specifically enriched for women at increased underlying familial or genetic BC risk, the BCFR¹⁰⁹ and the Sister Study¹¹⁰. Exposures examined in these publications include PAH-DNA adducts, ambient air pollution, vehicular traffic-related air pollution, organic solvents, pesticides, workplace chemicals, metals, and personal care product use ^{6,26,31,37,46,76,87,94,98,102}. Findings of note from these publications are statistically significantly elevated BC risks associated with: PAH-albumin adducts in women at increased familial risk in BCFR⁶, indoor wood burning and stove/fireplace use, particularly for women with ≥2 first-degree relatives³¹, occupational exposure to solvents before 1980 and prior to first birth⁸⁷, occupational exposure to soldering materials and industrial dyes/inks⁷⁶, airborne mercury exposure¹⁰², and frequent use of beauty or skincare products for White women⁹⁴ in the Sister Study. Additionally, the Sister Study also reported a modest but non-statistically significant increase for premenopausal BC risk for self-reported childhood exposure to fogger trucks or planes⁴⁶, but no association for childhood exposure to fogger trucks or planes for post-menopausal BC⁴⁶, residential exposure to PM₁₀, PM_2.5, and NO₂³⁷, childhood residential exposure to large roads (3+ lanes)²⁶, and airborne exposure to antimony, chromium, cobalt, manganese, and nickel¹⁰².

Eight publications considered family history in the study analysis including studies of pesticide use in the CTS⁴⁴, insecticide and herbicide use in AHS⁷⁸, total effective xenoestrogen burden the Multicase-Control on Cancer - Spain Study⁴⁹, DDE exposure in a case-control study in Colombia⁵¹, PCB exposure in a case-control study in Canada⁵⁰, sport-caught fish consumption in a population based case-control study in Wisconsin⁴⁸, organic solvents and benzene in the Polish Breast Cancer study ⁸⁹, and occupational exposure to PAHs in a population-based case-control study in Vancouver and Kingston in Canada¹⁵, where nearly 20% of cases (n=224) and 14% of controls (n=167) in this study had a positive BC family history¹⁵. With the exception of this study that focused on occupational PAH exposures¹⁵, studies were mostly underpowered to examine effect modification by family history, with three studies including less than 50 cases (n=10–49) with a family history ^49,78,89. For example, a large population-based case-control study in Poland reported no-association for occupational exposure to organic solvents in women with a family history, but found significant increased BC risk in women without a family history. However, the number of exposed cases with a family history was much lower (n=49) than those with no family history (n=373)⁸⁹. One publication from the CTS did not report findings from stratified models by family history, noting that small numbers hindered the analyses⁴⁴. Thus, most publications examining effect modification by BC family history have only very limited statistical power to detect effects within the subgroup of women with a family history and most publications only considered a binary definition of family history (ever/never) adding to the heterogeneity.

Type 2 Studies: Early onset:

Of the 100 included publications in this systematic review, 75% (75 publications from 47 epidemiologic studies) considered early onset BC in either the study design or analysis (Table 1a, Table 1b). Of these, 10 publications came from 6 epidemiologic studies that enriched for early onset BC either by initially recruiting younger women (NHSII)^35,36, by oversampling women with premenopausal BC (NECSS)³⁸, or by limiting case inclusion to women diagnosed with early on-set BC defined as either premenopausal BC (Danish National Birth Cohort⁸² and population-based case-control study of premenopausal BC in western New York²³) or using an age cut off: before age 45 years⁹⁵, before age 50 years of age ^67,111,112 or through age 54 years⁵² (CHDS and population-based case-control in Washington state). ECE examined in these publications include air pollution (PAH and NO₂), occupational exposure to PAH and benzene, hair product use, DDT, PCBs, and PFAS ^{23,35,36,38,52,67,82,95,111,112}. These publications reported statistically significant increased premenopausal BC risk associated with residential exposure to NO₂ using satellite estimates in NECSS ³⁸; early exposure to DDT^52,111,112 , and PCBs⁶⁷ in CHDS, occupational exposure to benzene²³, high levels of serum PFOSA in the Danish National Birth Cohort ⁸², and use of hair dyes⁹⁵. Two publications from NHSII reported no association for premenopausal BC and residential exposure to PM₁₀ and PM₂.₅, or hazardous air pollutants, although they found elevated but non-statistically significant risk of BC associated with living near major roadways^35,36.

Of the 75 publications that considered early onset BC, the majority did so in the study analysis (66 publications in 41 epidemiologic studies) through stratifying by menopausal status or age at BC diagnosis (Table 1b)^{6,12,13,15,16,26–28,30–34,44–47,49,53–58,69,70,76,77,80,85,88–91,93,94,98,99,102,113}. These studies included a range of individuals who were premenopausal or less than 50 years of age, with 5%-66% of the respective study samples including premenopausal or young women. Exposures included air pollution including diesel and gasoline engine exhausts, NO₂, PAH, aromatic adducts, pesticides, organophosphate pesticides, insecticides, DDE, DDT, PFAS, PBDE, PCBs, organochlorines, solvents, occupational exposures, phthalates, total effective xenoestrogen burden, metals, residential proximity to industrial facilities, occupational exposures, and hair dye use and chemical hair straighteners.

Seventeen of these publications reported statistically significant elevated premenopausal BC risk associated with detectable PAH adducts ^6,12, aromatic DNA adducts²⁴, occupational PAH exposure ¹⁵, residential traffic PAH exposure^27,28, smoked meat consumption³⁹, serum PCBs ⁶⁹, PCB congers 105 and 118 ⁵⁰, pesticide or insecticide exposure ^78,79, urinary phthalate metabolite DEP/MEP ⁹¹, occupational exposure to soldering material ⁷⁶, residential proximity to thermal power plants ¹⁰⁴, exposure to auto plastics manufacturing and canning ¹⁰⁵ , exposure to metalworking fluid ¹⁰⁶, and a weighted sum index of 10 airborne metals¹⁰².

Additionally, 33 publications reported elevated (OR>1.1) but not-statistically significant risk of premenopausal BC associations for childhood residential exposure²⁶, occupational exposure to engine exhaust³⁰, residential exposure to propylene oxide³³, residential traffic exposure (in non-smokers)²⁸, total suspended particles (TSP) at birth address⁴³, grilled/barbequed and smoked meat consumption⁴², plasma DDT/DDE^54,55 , tissue DDE⁵⁰, serum DDE⁵¹, childhood exposure to fogger trucks ⁴⁶, plasma/serum PCBs^55,70, airborne dioxin ⁴⁵, TEXB-α ⁴⁹, occupational pesticide exposure^76,80, insecticide, chlorpyrifos or terbufos use⁷⁷, PFNA¹¹³, occupational exposure to solvents^87–89, serum PBDE congeners BDE-100 and BDE-153¹¹³, dibutyl phthalate exposure⁹⁰, occupational exposure to metals⁷⁶, toenail concentrations of cadmium, cobalt, molybdenum, lead, tin, vanadium, airborne cobalt exposure¹⁰² , use of hair relaxer with unknown ingredients⁹³, ever use of hair dyes⁹⁶ and occupational exposure to industrial acids, dyes/inks, and glues/adhesives, stains⁷⁶. The number of exposed premenopausal BC cases in these publications ranged between 7 to over 1,500, although the majority of studies included a small number of exposed premenopausal BC cases (<50 women). Twenty-two publications reported no association with premenopausal BC for urinary PAH metabolites¹⁶, aromatic DNA adducts²⁵, indoor air pollution³¹, TSP at menarche address and first birth address⁴³, residential hazardous air pollutants^33,34, serum DDT^47,56–58, serum DDE⁷¹, serum PCBs^56,58,69,71, residential exposure to organochlorines⁴⁴, occupational exposure to benzenes⁸⁹, serum PBDE congener BDE-47¹¹³ , airborne cadmium exposure^99,102, airborne exposure to arsenic, chromium, lead, manganese, mercury, nickel and selenium¹⁰², hair dye use⁹⁷, and lye or no-lye hair relaxer use⁹³. Thus, although there are more studies of Type 2 than Type 1, Type 2 studies and analyses that were not specifically enriched for younger women ranged in statistical power with many having low power to detect signals from ECE.

Type 3: Genetic Susceptibility:

Of the 100 included publications, 27% (27 publications in 13 epidemiologic studies) considered underlying genetic susceptibility, with nearly half of those (13 out of 27 publications) from the Long Island Breast Cancer Study Project (LIBCSP)^{10,11,14,17–22,27,29,32,40,70,72–74,81,83,84,92}. All 13 publications in LIBCSP examined gene-environment interaction with an air pollution exposure (measured through PAH-DNA adduct biomarkers, geographic assessment of residential traffic exposure, self-reported burning of synthetic logs, and self-reported consumption of grilled/smoked meat)and genes involved in the following: DNA repair, apoptosis, tumor suppressor genes, toxin-metabolizing genes, phase I metabolizing enzyme genes, and steroid hormone genes, with over 80% of these publications reporting a statistically significant association of increased BC risk^{10,11,14,17–22,27,29,32,40}. Two additional publications not in LIBCSP assessed for dietary exposure to HCAs and PAHs and polymorphisms in SULT1A1 gene in a case-control study in in Sichuan Province, China³⁹, and for NAT2 polymorphisms in the German GENICA study⁴¹, and reported statistically significant association of increased BC risk.

Five publications in five epidemiologic studies examined gene-environment interaction with exposure to PCBs and metabolism genes (CYP1A1, CYP1B1, or COMT) in the Nurses’ Health Study ⁷³, Carolina Breast Cancer Study ⁷⁴, Diet Cancer and Health study ⁷², a case-control study in New Haven, CT hospitals ⁷⁰ and a case-control study in Western New York ⁷⁵, with three reporting significant associations for the genes they examined and two reported no association (as detailed in the results section of “Persistent Endocrine Disrupting Chemicals”). The different study findings could be due to several factors including: differences in PCB measurement including PCB measured in adipose tissue compared to serum, differences in which PCB congeners were included, or differences in the distribution of the genotypes between study populations as several studies noted they had limited numbers of women the rare homozygote . This illustrates that while there is substantially more consistency when focusing on high-risk individuals, epidemiological studies must really consider the different biological effects of different congeners as well as timing of exposures. Three publications in the Danish National Birth Cohort ⁸⁴ and Greenlandic Inuit ^83,86 examined gene-environment interaction with exposure to PFAS and metabolism genes (CYP1A1, CYP1B1, COMT, CYP17A1, and CYP19A1). One publication examined gene-environment interaction of exposure to fungicides and CYP1A1 in Prince Edward Island, Canada ⁸¹, and another publication examined phthalate metabolites and transcription factor genes (PPARγ, PPARGC1B) in Mexico City⁹². Two publications^100,101, both from a case-control study in Northern Mexico, examined gene-environment interactions for polymorphisms in AS3MT, FOLH1, MTHFD1, MTHFR, MTR, and MTRR ¹⁰⁰, or PPARγ¹⁰¹ and inorganic arsenic metabolism. Over 70% of publications that considered underlying genetic susceptibility (20 out of 27 publications) reported a statistically significant association of increased BC risk in subgroups of women with greater genetic susceptibility due to variants in carcinogen metabolism, DNA repair, response to oxidative stress, cellular apoptosis and tumor suppressor genes and exposure to PAHs, traffic-related air pollution, consumption of grilled/smoked meats, PCBs, phthalates, inorganic arsenic, and PFOSAs.

Measurement of Exposure

In addition to considering enrichment through design as well as stratification by absolute risk, there are additional methodological considerations that affect interpretation. Overall, we observed larger consistency for ECE findings when publications used biomarkers for exposure measurement rather than questionnaire-based measures of exposures. Of the 100 included publications, 53 used biomarkers to measure the ECE of interest, and 56% of these publications (30 out of 53 publications) supported statistically significant associations for the ECE. However, even when using biomarkers to assess exposure, it is important to consider the limitations associated with the particular measure. For example in the Northern Mexico study that examined urinary phthalate metabolites⁹¹, was based on a single urine sample of phthalates, which are rapidly metabolized and can result in a large amount of intraindividual variability. Additionally, as the authors noted, this study did not consider dietary patterns, which may be associated with BC risk, and these findings warrant further confirmation in studies that also consider dietary sources of phthalates in the analyses⁹¹.

For exposures that are more difficult to assess such as personal (not occupational) exposure to pesticides and personal care products, even in the studies with higher absolute risk, the results were less consistent. For example, publications included in this review of personal hair care products, some studies supported no association of elevated premenopausal BC risk for personal care and hair product use^94,97 while other studies reported elevated premenopausal BC risk for different hair product use^95,96. Moreover, findings from the Sister Study report significantly elevated overall BC risk for White women who frequently used beauty or skincare products; while there was no association observed for premenopausal White women, or Black women⁹⁴. This lack of consistency may be due to exposure misclassification rising from the difficulty of accurately reporting use of specific products, differences in product formulations, or measuring exposure at the wrong time window^93,94. Similarly, studies of personal exposure to pesticides consistently show no association between BC risk and measurements of late life organochlorines after the chemicals were banned^44,46,77. One study, the CHDS, which in addition to enriching for young women, measured levels of DDT from blood samples collected before the DDT-ban and during periods of heaviest DDT use and consistently reported an association with premenopausal BC and DDT^52,111,112. This highlights the importance of accurate exposure measurement, in addition to timing of the exposure and using enriched cohorts to study ECE and BC risk.

Conclusion

Overall, the association of ECE and BC risk have been primarily studied in average risk cohorts and outside the relevant WOS. Many epidemiological studies, particularly those using population-based ascertainment, do not include a substantial proportion of women with a BC family history, and are therefore not enriched for underlying genetic predisposition. This approach can impact the ability to identify ECE associations, if the risk gradient depends on underlying genetic susceptibility^109,114. Improving precision can be readily accomplished through sampling more individuals. However, increasing the sample size without ensuring that a sufficient number of individuals at higher risk are sampled limits the ability to investigate whether modifiable exposures can alter BC risk across the spectrum of risk. Future studies of environmental exposures and BC risk should consider more efficient study designs such as enriched cohorts. Particularly given the measurement challenges of environmental exposures earlier in life, it is essential for inference to have enough individuals at higher risk to detect signals between exposures and outcome given that measurement error may reduce the association closer to the null. This systematic review supports the link between various ECE and increased BC risk, and highlights the utility of epidemiologic studies conducted in high-risk populations.