Stressful Life Events, Social Support, and Incident Breast Cancer by Estrogen Receptor Status

Abstract

Chronic stress affects immune function and hormonal signaling and has been hypothesized to be associated with breast cancer, although results from the few prior studies are mixed and have not examined potential differences by estrogen receptor (ER) status. Using the Women’s Health Initiative Study, we included 76,951 postmenopausal women followed for events for a median of 16.7 years to investigate the association between baseline self-reported stressful life events and incident breast cancer by ER status and whether the association was modified by social support. We generated Cox proportional hazards models adjusting for demographic, clinical, lifestyle/behavioral, and social factors to estimate hazard ratios (HR) and 95% confidence intervals (95%CI). The mean age was 63 (SD 7.3), and majority of participants were White race (83.5%) and married or in a marriage-like relationship (63.0%). In analyses stratified by ER status, there was no relationship between stressful life events and ER-positive breast cancer. In contrast, compared to women in the lowest quartile, those in higher quartiles had an increased risk of ER-negative breast cancer, where those in quartile 4 had the highest risk (Quartile 4 vs Quartile 1; HR=1.30; 95%CI 1.01–1.68; P_trend=0.050). Moreover, associations were stronger for the highest versus lowest quartile of stressful life events among widowed women (HR=2.39; 95%CI 1.29–4.44; P_interaction<0.001). Association between stressful life events and ER-negative breast cancer was not modified by social support. In this cohort of postmenopausal women, higher experiences of pre-diagnostic stressful life events were associated with increased risk of ER-negative breast cancer.

Introduction

Epidemiologic studies have linked psychosocial stress to disease risk, progression, and mortality (1–3). Psychosocial stress encompasses the dynamic interaction between individuals and their social environment, where adverse external events result in feelings of threat and subsequent emotional tension or discomfort (1,2,4). Chronic exposure to stress has been documented to influence inflammation, hormonal changes, suppressed immune function, and endothelial dysfunction (4–6). Additionally, studies have linked chronic psychosocial stress to excess adiposity, emotional distress, and engagement in unhealthy behaviors (e.g., tobacco use, alcohol consumption) (1,2). Prior studies have documented an association between chronic experiences of psychosocial stress and increased risk of chronic disease and poor mental health (2,7,8).

Although many mechanisms suggest that chronic exposure to stress may be associated with breast cancer, results from prior studies are inconsistent with many reporting a null association for breast cancer risk (5,9–12). However, breast cancer is a heterogeneous disease, where studies have established risk factors can vary by estrogen receptor (ER) status, such as breastfeeding and time since last pregnancy (13–15). Additionally, recent descriptive studies have shown that less established risk factors such as experiences of social adversity are associated with increased incidence of ER-negative breast cancer (16,17). However, study of the relationship between psychosocial stress and cancer remains limited.

Studies have documented that greater social support provide resources during periods of stress (i.e., reducing negative feelings about oneself), thereby buffering the harmful effects psychosocial stress may have on disease risk (2,18,19). However, investigation of whether social support modifies the association between psychosocial stress and breast cancer remains limited.

The present study investigated the relationship between stressful life events, which measures the accumulation and perceived severity of a stressful event, and incident breast cancer by ER status among a large cohort of postmenopausal women. We further evaluated whether this association is modified by social support. Our findings aim to provide evidence on whether chronic psychosocial stress and incident breast cancer vary by ER status. Based on findings on social adversity and breast cancer, we hypothesized that higher experiences of stressful life events would be associated with increased risk of ER-negative breast cancer. We further hypothesized that the stressful life events-ER-negative breast cancer relationship will be buffered with high social support.

Methods

Study Cohort

The present study used data from the Women’s Health Initiative (WHI). WHI is a large prospective cohort study composed of clinical trials (including hormone therapy [2 trials], low-fat diet patterns, and vitamin D and calcium supplementation) and an observational study. Details of the WHI study has been previously described (20). Briefly, between 1993 and 1998 postmenopausal women aged 50-79 years were recruited from forty clinical sites across the U.S. into one or more randomized clinical trials (WHI-CT: n=68,132) or the observational study (WHI-OS: m=93,676). WHI-OS included women that were either ineligible or did not want to participate in the WHI-CT. The WHI ended in 2003-2004, and participants were invited to continue follow-up in the WHI Extension Study 1 (2005-2010) and Extension Study 2 (2010-2015). The study was conducted according to the Declaration of Helsinki. All participants provided written informed consent and Institutional Review Board approval were obtained at each study institution.

The present study included women participating in the WHI-OS only. Women were excluded if they were previously diagnosed with cancer (n=12,827), diagnosed with non-invasive breast cancer (n=1,169), no information or unknown breast cancer estrogen receptor and/or progesterone receptor status (n=377), missing information on baseline stressful life events (n=1,701), and missing follow-up data (n=382). To account for potential undiagnosed breast cancer at baseline, women with a breast cancer diagnosis within the first year of follow-up were excluded (n=269). Our final analytical cohort included 76,951 individuals (Supplemental Figure 1).

Measures

Participants’ experiences of stressful life events and social support were collected via questionnaire (Supplemental Table 1) at baseline. The life events questionnaire was adapted from the Beta-Blocker Heart Attack Trial (21). At baseline, participants indicated whether any of eleven life changes had occurred over the past year. Sample items included, “spouse died”, “physically abused”, and “major conflict with children or grandchildren.” Participants were then asked to appraise each of the eleven life events that occurred based on the degree of upset that it caused on a scale of 1 (did not upset me) to 3 (upset me very much), where the total score ranged from 0 to 33, with higher scores indicate greater experiences of stressful life events. Scores were categorized into approximate quartiles (Quartile 1: ≤1; Quartile 2: 2; Quartile 3: 3-4; and Quartile 4: >4 or more) based on the response distribution.

Social support was evaluated using the 9-items from the Medical Outcomes Study Social Support Survey among participants that completed the questionnaires (n=75,196) (22). Participants were asked to rate their perceived social support on a scale of 1 (none of the time) to 5 (all of the time), whether specific types of support (emotional support, affection, tangible support, and positive interaction) were available. The total score ranged from 9 to 45, with higher scores indicating greater social support. Social Support scores were categorized into approximate quartiles (Quartile 1= ≤32, Quartile 2= >32-≤37, Quartile 3= >37-≤43, and Quartile 4= >43.) based on the distribution of responses.

Covariates

Social function, which measures the ability to participate in desired social activities without limitation was measured using the RAND 36-Item Health Survey Quality of Life subscales on social functioning (23). Scores range from 0 to 100 with a higher score indicating a more favorable health state. Scores were categorized into approximate tertiles (Tertile 1: <87.5, Tertile 2: ≥87.5 and <100, and Tertile 3: 100) based on distributions of participants’ scores. Social strain was assessed using the 4 items derived from previously validated measure of negative aspects of social relationships (24). Participants rated their perceived social strain on a scale from 1 (none) to 5 (all), yielding a social strain score ranging 4-20, with higher scores indicating greater social strain. Scores were categorized into approximate tertiles (Tertile 1: ≤5, Tertile 2: 6-7, and Tertile 3: ≥8) based on the study population distribution. Finally, depressive symptoms were measured using the 8-item Burnham short version of the Center for Epidemiologic Studies-Depression (CES-D) (25). Scores ranged from 0 to 1 with a higher score indicating a greater depressive symptomatology.

We created a causal diagram guided by conceptual models of psychosocial stress and cancer to identify potential confounders of the association between stressful life events and incident breast cancer (4,26). We grouped covariates into four groups of factors collected at baseline: demographic, health and clinical, lifestyle/behavioral, and social factors. Demographic factors included age at study entry (continuous), race/ethnicity (Alaskan Native/American Indian, Asian/Pacific Islander, Black, Hispanic/Latina, White, other/unknown), educational attainment (less than high school, high school diploma/GED, vocational/training school or Associate/some college, college graduate, graduate/professional education), annual family income (<$35,000, $35,000 to $49,999, $50,000 to $74,999, ≥ $75,000 or don’t know), and marital/relationship status (never married, married or marriage-like relationship, divorced/separated or widowed). Health and clinical factors included: body mass index [weight(kg)/height(m)²] (continuous), age at menarche (<12, 12-13, or ≥14), age at first birth (no term pregnancy, <20, 20-29, or ≥30), parity (never pregnant/no term pregnancy, 1, 2, 3, 4, or ≥5), benign breast disease (yes or no), age at menopause (continuous), hysterectomy (yes or no), bilateral oophorectomy (yes or no), oral contraceptive use duration (continuous), hormone therapy usage (unopposed estrogen and/or estrogen plus progesterone) (never used, past user, or current user), family history of breast cancer (yes or no), and depressive symptoms (continuous). Lifestyle/behavioral factors included number of months breastfed (never, 1-6, 7-12, ≥13), pack years of smoking (continuous), alcohol serving per week (continuous), and recreational physical activity (metabolic equivalent-hours per week) (continuous). Social factors included: social strain (in tertile) and social function (in tertile). Moreover, an estimated invasive female breast cancer risk score was calculated based on the Gail model derived from the following covariates: age, race/ethnicity, age at menarche, age at first live child, number of first-degree relatives with breast cancer, and the number of previous breast biopsy examinations. Details on the Gail model has been previously described (27,28).

Breast Cancer

Participants were contacted annually in the WHI-OS arm to obtain information on clinical outcomes using in-person, mailed, or telephone questionnaires (20,29). Information from medical records were forwarded to the WHI coordinating center for central adjudication and coding of breast cancer characteristics. Breast cancer was confirmed centrally by a trained physician adjudicator following review of pathology reports and medical records. ER and PR status was determined by local laboratory confirmation.

Statistical Analysis

Cox proportional regression was used to estimate hazard ratios (HRs) and 95% confidence intervals (95%CI) for the association between stressful life events and incident breast cancer overall and by ER status. Time to breast cancer was computed from one year after enrollment date to first breast cancer diagnosis. Participants without breast cancer or absent ER status being evaluated were censored at date of last follow-up visit or death. Covariates were selected if their inclusion in the models resulted in a 10% change in estimated main effect of stressful life events. In Model 1, we adjusted for Gail model risk score, body mass index, parity, age at menopause, pack years of smoking, number of months of breastfeeding, alcohol servings per week, total energy expenditure from recreational physical activity, family income, marital/relationship status, and hormone therapy. Based on prior findings indicating mental health and social factors are associated with stressful life events and may influence breast cancer risk, in Model 2 we adjusted for Model 1+ depressive symptoms, social strain, and social function (16,17,30–34). The variables depressive symptoms, social strain, and social function may be possible mediators, which could be a response to experiences of stressful life events, or actual confounders, thereby utilizing this sequential modeling approach will allow for evaluating the absence and presence of these covariates. Using Model 2, we further examined interaction by income, marital/relationship status, social strain, and social support. Tests for interaction were performed using the likelihood ratio tests comparing model with the cross-product terms to one without. Tests for trend was performed across quartiles of stressful life events by entering the categories as a continuous variable. We examined the assumption of proportional hazards by including an interaction between stressful life events and the log of time in the Cox proportional hazards models. There was no evidence that the assumption of proportional hazards was violated. All statistical tests were two-tailed, and analyses were performed using SAS version 9.4.

Data Availability Statement

The datasets generated during this study may be made available upon appropriate request from the Biologic Specimen and Data Repository Information Coordinating Center (BioLINCC) at the National Heart, Lung, and Blood Institute (Bethesda, MD). Additionally, the analytical methods for this study are available from the corresponding author upon appropriate request.

Results

Women in the highest quartile of stressful life events were younger, more likely to be Black racial group, less likely to be married or in a marriage-like relationship, and lower annual family income than the lowest quartile (Table 1). They were also less likely to engage in recreational physical activity, have a higher body mass index, and greater pack years of smoking. Moreover, those in the highest quartile were more likely to experience greater social strain, and depressive symptoms, as well as lower social support.

Table 1.

Selected Baseline Participant Characteristics by Stressful Life Events Category, in the Women’s Health Initiative, (n=76,951)

Stressful Life Events a
	Quartile 1	Quartile 2	Quartile 3	Quartile 4
Age, y, mean ± SD	63.98 ± 7.28	63.81 ± 7.29	62.73 ± 7.31	62.15 ± 7.31
Race/Ethnicity, %
Alaskan Native/American Indian	0.3	0.4	0.4	0.7
Asian/Pacific Islander	3.7	3.0	2.7	2.4
Black	5.7	6.9	8.8	12.6
Hispanic/Latina	3.0	2.9	3.9	6.1
White	86.1	85.6	82.5	76.4
Other/unknown	1.2	1.2	1.7	1.8
Highest Educational Attainment, %
Less than high school	4.3	4.2	4.9	7.0
High School Diploma/GED	16.2	16.3	16.7	16.7
Vocational/Training School or Associate/Some college	33.5	35.9	38.4	40.8
College graduate	12.9	12.1	10.7	9.6
Graduate/Professional education	33.1	31.5	29.3	25.9
Marital/Relationship status, %
Never married	5.5	4.9	3.8	3.3
Married or Marriage-like relationship	66.3	65.4	63.1	53.5
Divorced or separated	12.6	13.8	16.4	22.7
Widowed	15.6	15.9	16.7	20.5
Annual family income, %
<$35,000	32.9	35.5	38.1	47.7
$35,000 to $49,999	19.8	20.3	19.8	18.1
$50,000 to $74,999	21.2	20.1	19.9	16.6
≥ $75,000	23.0	21.1	19.2	14.3
Don’t Know	3.1	3.0	3.0	3.3
Family history of female relative with breast cancer, %	40.1	39.1	39.3	37.2
Age at menarche, %
<12	20.6	21.5	22.6	24.4
12-13	56.4	55.9	55.0	52.5
≥14	23.0	22.6	22.4	23.1
Age at menopause, y, mean ± SD	48.65 ± 6.07	48.42 ± 6.25	48.03 ± 6.43	47.57 ± 6.67
Body mass index, mean ± SD b	26.62 ± 5.51	26.94 ± 5.64	27.46 ± 5.85	28.37 ± 6.43
Total energy expenditure from recreational physical activity (MET-hours/week), mean ± SD	14.77 ± 14.67	13.91 ± 14.17	13.22 ± 13.91	12.51 ± 14.47
Parity
Never been pregnant/ no term pregnancy	14.0	13.2	11.3	9.7
1	8.9	8.8	8.9	9.4
2	27.1	26.4	26.4	25.2
3	24.4	24.2	24.4	23.9
4	13.9	14.4	14.7	15.6
≥5	11.7	13.0	14.3	16.2
Age at first birth, y, %
Never had term pregnancy	3.3	3.4	3.1	3.4
<20	11.0	12.5	15.8	19.5
20-29	75.5	74.7	72.5	68.6
≥30	10.2	9.4	8.6	8.5
Unopposed Hormone therapy Usage, % c
Never used	39.7	38.6	37.6	40.8
Past user	13.3	13.8	13.7	14.3
Current user	47.0	47.6	48.7	44.9
Number of months breastfed, %
Never breastfed	50.5	49.5	48.2	47.4
1-6 months	24.9	25.6	26.5	26.8
7-12 Months	10.9	11.2	11.2	11.5
≥13	13.7	13.7	14.1	14.3
Oral contraceptive use duration, y, mean ± SD	2.10 ± 4.22	2.08 ± 4.19	2.25 ± 4.27	2.25 ± 4.30
Hysterectomy
Yes	36.5	39.4	41.5	44.3
No	63.5	60.6	58.5	55.7
Bilateral Oophorectomy
Yes	17.7	19.5	19.3	19.9
No	82.3	80.5	80.7	80.1
Benign Breast Disease, %
Yes	22.3	22.7	21.9	22.2
No	77.7	77.3	78.1	77.8
Pack years of smoking, mean ± SD	9.20 ± 17.82	9.67 ± 18.21	10.02 ± 18.34	10.52 ± 19.10
Alcohol Servings Per week, mean ± SD	2.61 ± 5.10	2.59 ± 5.09	2.46 ± 5.50	2.21 ± 5.07
Social Strain, % d
Tertile 1	55.1	47.0	36.9	26.8
Tertile 2	25.3	27.7	28.2	25.6
Tertile 3	19.6	25.3	34.9	47.6
Social Support, % e
Quartile 1	22.3	24.8	30.6	40.3
Quartile 2	22.8	23.5	23.9	23.4
Quartile 3	29.7	29.9	27.8	23.3
Quartile 4	25.2	21.8	17.7	13.0
Social Function, % f,
Tertile 1	13.9	18.9	26.1	38.6
Tertile 2	9.6	11.8	12.9	14.7
Tertile 3	76.5	69.3	61.0	46.7
Depressive symptoms, score, mean ± SD g	0.01 ± 0.07	0.03 ± 0.10	0.04 ± 0.13	0.11 ± 0.21
Gail model score for breast cancer risk, mean ± SD h	1.83 ± 0.96	1.82 ± 1.00	1.74 ± 1.01	1.65 ± 1.03

Abbreviations: SD, standard deviation; GED, general equivalency diploma; MET, metabolic equivalent of task

^aStressful life events: Quartile 1= ≤1, Quartile 2= 2, Quartile 3= 3-4, Quartile 4= >4.

^bBody mass index was calculated as weight (kg)/height (m)²

^cHormone therapy includes unopposed estrogen and/or estrogen plus progesterone)

^dSocial Strain score: Tertile 1= ≤5, Tertile 2= 6-7, Tertile 3= ≥8.

^eSocial Support score: Quartile 1= ≤32, Quartile 2= >32-≤37, Quartile 3= >37-≤43, Quartile 4= >43 (n=75,196)

^fSocial Functioning score: Tertile 1= <87.5, Tertile 2= ≥87.5 - <100, Tertile 3= 100.

^gHigher score indicating a greater likelihood of depression

^hGail model risk of breast cancer score (age, ethnicity, age at menarche, age of the mother at the birth of her first live child, number of first-degree relatives with breast cancer, and the number of previous breast biopsy examinations)

In multivariable analysis, no association was observed between experiences of stressful life events and incident breast cancer for both models (Table 2). When stratified by ER status, compared to women in the lowest quartile, those in higher quartiles had an increased risk of ER-negative breast cancer in the fully adjusted model, where those in quartile 4 had the highest risk (Model 2: Quartile 4 vs Quartile 1; HR=1.30; 95%CI 1.01–1.68; P_trend=0.050) (Table 3). No association was observed for ER-positive breast cancer.

Table 2.

Adjusted-Hazard Ratios for the Associations between Stressful Life Events and Incident Breast Cancer in the Women’s Health Initiative, (n=76,951)

	Incident breast cancer events	Model 1 HR (95%CI)	Model 2 HR (95%CI)
Stressful life events a
Quartile 1	1,561	Reference	Reference
Quartile 2	1,518	1.07 (0.99, 1.16)	1.07 (0.99, 1.15)
Quartile 3	778	1.00 (0.92, 1.10)	1.01 (0.92, 1.11)
Quartile 4	827	1.00 (0.91, 1.09)	1.01 (0.92, 1.12)
P trend		0.772	0.944

^aStressful life events: Quartile 1= ≤1, Quartile 2= 2, Quartile 3= 3-4, Quartile 4= >4.

Model 1: Adjusted for Gail model score for breast cancer risk (age, race/ethnicity, age at menarche, age of the mother at the birth of her first live child, number of first-degree relatives with breast cancer, and the number of previous breast biopsy examinations), parity, body mass index, age at menopause, pack years of smoking, number of months breastfed, alcohol servings per week, total energy expenditure from recreational physical activity, annual family income, marital/relationship status, hormone therapy.

Model 2: Model 1+, depressive symptoms, social strain, social function

Abbreviations: 95%CI, 95% confidence interval; HR, hazard ratio; ER-, estrogen receptor negative, H=ER+, estrogen receptor positive; MET, metabolic equivalent of task

Note: Unopposed Hormone therapy usage includes unopposed estrogen and/or estrogen plus progesterone

Table 3.

Adjusted Hazard Ratios for the Associations between Stressful Life Events and Incident Breast Cancer by Estrogen Receptor Status in the Women’s Health Initiative, (n=76,951)

	Incident ER− Breast cancer events	Model 1 HR (95%CI)	Model 2 HR (95%CI)	Incident ER+ breast cancer events	Model 1 HR (95%CI)	Model 2 HR (95%CI)
Stressful life events a
Quartile 1	196	Reference	Reference	2,099	Reference	Reference
Quartile 2	213	1.18 (0.95, 1.45)	1.21 (0.98, 1.50)	1,965	1.06 (0.97, 1.15)	1.05 (0.96, 1.14)
Quartile 3	110	1.10 (0.85, 1.42)	1.19 (0.92, 1.54)	1,074	0.99 (0.90, 1.09)	0.98 (0.89, 1.09)
Quartile 4	128	1.17 (0.92, 1.50)	1.30 (1.01, 1.68)	1,152	0.97 (0.88, 1.07)	0.97 (0.87, 1.08)
P trend		0.255	0.050		0.434	0.476

^aStressful life events: Quartile 1= ≤1, Quartile 2= 2, Quartile 3= 3-4, Quartile 4= >4.

Model 1: Adjusted for Gail model score for breast cancer risk (age, race/ethnicity, age at menarche, age of the mother at the birth of her first live child, number of first-degree relatives with breast cancer, and the number of previous breast biopsy examinations), parity, body mass index, age at menopause, pack years of smoking, number of months breastfed, alcohol servings per week, total energy expenditure from recreational physical activity, annual family income, marital/relationship status, hormone therapy.

Model 2: Model 1+, depressive symptoms, social strain, social function

Abbreviations: 95%CI, 95% confidence interval; HR, hazard ratio; ER−, estrogen receptor negative, H=ER+, estrogen receptor positive; MET, metabolic equivalent of task

Note: Unopposed Hormone therapy usage includes unopposed estrogen and/or estrogen plus progesterone

We further assessed effect modification of stressful life events and ER-positive and ER-negative breast cancer by participant’s characteristics (Supplemental Table 2). In women with an income ranging $35,000 to $49,999, higher quartiles of stressful life events were associated with greater risk of ER-negative breast cancer when compared to the lowest quartile, albeit statistical significance was only observed for quartile 2 (Quartile 2 vs Quartile 1: HR=1.68; 95%CI 1.02–2.77) (P _interaction <0.001). When stratifying by marital status, widowed women in the highest quartile of stressful life events had an increased risk of ER-negative breast cancer than those in the lowest quartile (Quartile 4 vs Quartile 1: HR=2.39; 95%CI 1.29–4.44) (P _interaction <0.001). Finally, there was no statistically significant interaction for social strain on the relationship between stressful life events and breast cancer (P _interaction=0.395). Moreover, social support did not modify the relationship between stressful life events and ER-positive and ER-negative breast cancer (Table 4).

Table 4.

Adjusted Hazard Ratios for the Association of Stressful Life Events and Incident Breast Cancer by Social Support and Estrogen Receptor Status in the Women’s Health Initiative, (n= 75,196)

		Stressful Life Events b
	Incident breast cancer events	Quartile 1	Quartile 2	Quartile 3	Quartile 4	P interaction
		HR (95%CI) a	HR (95%CI)	HR (95%CI)	HR (95%CI)
ER−
Social Support c						0.473
Quartile 1	161	Reference	1.38 (0.85, 2.26)	1.56 (0.92, 2.65)	1.24 (0.73, 2.11)
Quartile 2	147	Reference	1.06 (0.68, 1.64)	0.82 (0.46, 1.47)	1.66 (1.02, 2.72)
Quartile 3	185	Reference	1.47 (0.98, 2.18)	1.67 (1.05, 2.66)	1.82 (1.12, 2.96)
Quartile 4	141	Reference	1.10 (0.72, 1.67)	0.77 (0.41, 1.43)	0.73 (0.37, 1.44)
ER+
Social Support						0.381
Quartile 1	977	Reference	0.90 (0.75, 1.08)	0.85 (0.68, 1.05)	1.03 (0.85, 1.25)
Quartile 2	972	Reference	1.04 (0.88, 1.23)	1.01 (0.82, 1.24)	0.88 (0.71, 1.09)
Quartile 3	1,161	Reference	1.15 (0.99, 1.33)	1.03 (0.85, 1.24)	0.93 (0.76, 1.15)
Quartile 4	858	Reference	1.05 (0.89, 1.24)	1.06 (0.85, 1.32)	0.95 (0.73, 1.22)

^aAdjusted for Gail model score for breast cancer risk (age, race/ethnicity, age at menarche, age of the mother at the birth of her first live child, number of first-degree relatives with breast cancer, and the number of previous breast biopsy examinations), parity, body mass index, age at menopause, pack years of smoking, number of months breastfed, alcohol servings per week, total energy expenditure from recreational physical activity ,annual family income, marital/relationship status, hormone therapy, depressive symptoms, social strain, social function.

^bStressful life events: Quartile 1= ≤1, Quartile 2= 2, Quartile 3= 3-4, Quartile 4= >4.

^cSocial Support: Quartile 1= ≤32, Quartile 2= >32-≤37, and Quartile 3= >37-≤43, Quartile 4= >43.

Abbreviations: 95%CI, 95% confidence interval; HR, hazard ratio; MET, metabolic equivalent of task

Discussion

In this large cohort study of over 75,000 postmenopausal women, we identified differential associations in risk between stressful life events and incident breast cancer by ER status. Specifically, we demonstrated the contribution of higher stressful life events to increased risk of ER-negative breast cancer, while no association was observed for ER-positive breast cancer. We further observed no significant interaction for social support on the relationship between stressful life events and breast cancer ER status.

Our findings of a positive association between stressful life events and ER-negative breast cancer are consistent with evidence that suggested a relationship between social adversity and risk of ER-negative breast cancer (16,17,35). For instance, a descriptive epidemiologic study examined whether being born in a state with legal Jim Crow laws (legal discrimination practiced in twenty-one U.S. states and the District of Columbia) was associated with breast cancer ER status (17). The study observed higher odds of ER-negative tumors among Black woman born in a state when Jim Crow was legal (born before 1965) compared to Black women born in a non-Jim Crow state. Another study examined whether being born during the Great Chinese Famine (1959-1962) was associated with risk of breast cancer (16). The study results suggests that exposure to the Famine may increase likelihood of ER-negative breast cancer compared to women born after (after January 1, 1962) (16). These study findings are consistent with the present study, where higher experiences of stressful life events were associated with elevated risk of ER-negative breast cancer.

We demonstrate a relationship between stressful life events with ER-negative breast cancer and no association with ER-positive breast cancer. One potential explanation for our findings is attributed to alterations in mitochondrial function. Emerging evidence suggests that prolonged psychosocial stress can disrupt mitochondrial function resulting in stress-induced molecular changes and subsequent disease risk (36–40). Mitochondria, a subcellular organelle with its own genome, senses and responds to chronic psychosocial stress resulting in mitochondrial dysfunction and the production of reactive oxygen species with pro-tumorigenic effects.(36,40) Evidence suggests that mitochondrial biology is differential by breast cancer subtype and that dysfunction is higher in basal-like tumors in contrast to luminal A or ER-positive tumors (36,41). Nonetheless, further studies are needed to understand the biological mechanism between experiences of chronic psychosocial stress and increased risk of breast cancer by subtype. Interestingly, our results suggest an inverse association for ER-positive breast cancer among women experiencing higher stressful life events, though not statistically significant. Although additional research in this area is needed, previous studies have suggested that higher levels of chronic stress disrupt the hypothalamic–pituitary–adrenal axis, resulting in lower estradiol levels or chronic impairment of estrogen synthesis, potentially reducing ER-positive breast cancer among women at higher risk of estrogen-sensitive cancers (e.g., postmenopausal women) (42–44). Overall, our study adds to the limited body of literature describing the associations between lifetime stress and breast cancer risk by demonstrating an association between high psychosocial stress and increased risk of ER-negative breast cancer.

In our secondary analyses stratifying by participant characteristics, we observed the relationship between stress and ER-negative breast cancer varied by marital/relationship status. For instance, women who were widowed and in the highest quartile of stressful life events had an increased risk of ER-negative breast cancer. Previous studies have documented widowed individuals often report worse self-rated health, compromised biological function, and greater stress, all potential contributors to ER-negative breast cancer (45,46). We further observed that higher levels of stressful life events were more strongly associated with ER-negative breast cancer in women reporting lower household incomes, specifically those with annual incomes ranging $35,000 to $49,999. Prior studies have suggested that individuals of lower socioeconomic status are disproportionately exposed to social adversity and environmental factors that may contribute to increased risk of ER-negative breast cancer (16,30,47–49). Further, while the interaction was not statistically significant, we observed an elevated hazard for ER-negative breast cancer among women in the highest quartile of stressful life events reporting lower social support. Previous studies have suggested strong social support can buffer the deleterious effect psychosocial stress has on health, where lower social support has no influence (2,19,50).

The present study findings have important clinical and public health implications, given that ER-negative breast cancer is a more aggressive and harder to treat subtype compared with ER-positive breast cancer (51–53). Therefore, interventions are needed to prevent the risk of ER-negative breast cancer among women experiencing high psychosocial stress. Interventions can include clinicians screening for psychosocial stress in their assessment with the aim of providing referrals to resources that can help mitigate the adverse health effects.

The strengths of this study include being a large prospective cohort with extended follow-up time allowing for evaluation of breast cancer risk by ER status. The use of previously validated stressful life events which measures both the accumulation and perceptions of each major life events that occurred over the past year at study baseline. Additionally, stressful life events were measured prior to breast cancer diagnosis. Finally, breast cancer cases and corresponding ER status were confirmed using medical records and pathology reports that were centrally adjudicated, providing a more objective examination of the associations.

Several limitations must be noted. First, stressful life events were only measured at baseline, whereas repeat measures would have provided a more comprehensive understanding of the longitudinal role psychosocial stress has on incident breast cancer. Second, the stressful life events scale was used as a composite, where there is a chance that not all individual stressors assessed are equal. For instance, some stressors may be more impactful and detrimental to incident breast cancer than others. Third, given the prospective study design, any misclassification of self-reported stressful life events will lead to nondifferential misclassification, which tends to bias the association towards the null. Fourth, despite the Gail model’s attempt to develop race-specific risk calculations to obviate the need for race correction, the model does not account for socio-contextual factors known to affect risk that varies by race and ethnicity (e.g., socioeconomic position, social adversity) (27,28). Fifth, confounding variables were measured at baseline, resulting in potential misclassification due to lack of examination changes of these factors during follow-up. Sixth, the study included majority White women (83.5%), limiting the generalizability of our findings to other racial and ethnic groups and our ability to compare across groups. Finally, WHI solely includes older postmenopausal women; therefore, findings might not be generalizable to younger premenopausal women.

Conclusion

In this prospective analysis of postmenopausal women, experiencing higher stressful life events were associated with incident ER-negative breast cancer. Our findings provide evidence that high psychosocial stress can influence risk of ER-negative breast cancer, thereby potentially providing insight on why some groups are more likely to be diagnosed with this subtype than others. These results highlight the critical need to address the health implications of experiencing high psychosocial stress and to further explore social drivers of ER-negative breast cancer.

Prevention Relevance.

Epidemiologic studies on the association between psychosocial stress and breast cancer risk remain inconsistent, while investigation of whether the association differs by estrogen receptor (ER) status are limited. In this prospective cohort of postmenopausal women, high experiences of stressful life events were positively associated with ER-negative disease but not ER-positive.

Abbreviations

ER

Estrogen Receptor

WHI

Women’s Health Initiative

Data Availability

The full data set is stored at the Biologic Specimen and Data Repository Information Coordinating Center (BioLINCC) at the National Heart, Lung, and Blood Institute (Bethesda, MD). The data is available for verification of these results with approval from the WHI Publications and Presentations committee and BioLINCC.