Stress and risk of breast cancer; findings from a large population-based incident case-control study

Abstract

The relationship between stress and breast cancer (BC) remains a topic of debate. We investigated the association between stress experienced within the past year and the risk of BC. Population-based incident case-control study of 600 newly diagnosed BC cases and 600 population controls (18–75 years) recruited in Isfahan, Iran between 2021 and 2023. Logistic regression model with short-term stress as main exposure was used to estimate odds ratios (ORs) and 95% confidence intervals (CI) after adjusting for confounders. Multiplicative interaction was tested between stress and menopausal status as well as other confounders. One-year stress level was significantly associated with BC risk. Compared with women reporting low stress levels, the odds of BC were higher among those with high stress levels in both the unadjusted model (OR = 3.10, 95%Cl: 2.45–3.92) and the fully adjusted model (OR = 3.38; 95%CI (2.56, 4.47)). Based on the statistical test for multiplicative interaction, the association appeared stronger among premenopausal women (adjusted OR = 4.88, 95% CI: 3.17–7.52) than among postmenopausal women (adjusted OR = 2.34, 95% CI: 1.53–3.58), with evidence of interaction by menopausal status (p for interaction = 0.018). Despite the inherent limitations of case-control studies, including potential recall and selection biases, the findings of the current study suggest that higher levels of stress experienced within the past year may be associated with an increased risk of breast cancer, particularly among premenopausal women. Further prospective studies are warranted to clarify the nature and direction of this association.

Introduction

Breast cancer (BC), the most common cancer in women, is considered the leading cause of cancer-related mortality worldwide^1,2. In 2022, with 1–5% increase in incidence rate, an estimated 2.3 million incident female BC cases were diagnosed worldwide, causing 670,000 deaths². While BC mortality rates are decreasing in countries with a very high Human Development Index (HDI), low-HDI countries are expected to experience a disproportionate increase in both incidence (38%) and mortality rates (38%) by 2050².The age-standardized incidence rate (ASIR) of BC in Iran is 32.1 per 100,000 women, with the crude incidence rate of 28.99 per 100,000³. Moreover, increasing trends in both the national burden⁴ and incidence rate^5,6 of BC have been reported over recent decades.

Evidence demonstrates several risk factors in BC development including being female, older age, and family history^7–9, with nearly one-quarter of BC cases linked to familial predisposition^9–11. Additionally, reproductive factors such as early menarche^9,12, late menopause, and late maternal age at first pregnancy, along with lifestyle-associated factors like high body mass index (BMI), have been identified as significant contributors⁹.

Individuals diagnosed with cancer often experience higher levels of psychological distress, including stress and anxiety, than the average level of general population^13,14. Epidemiological studies have investigated a possible link between stress and subsequent BC diagnosis but findings remain inconclusive. Psychological stress may contribute to cancer risk and progression by inducing inflammation, enhancing resistance to apoptosis, and impairing immunosurveillance¹⁴. Additionally, some studies suggest a link between the history of stressful life events and an increased risk of BC^15,16, others have found no significant association^17–19.

At the molecular level, there is substantial overlap between pro-tumorigenic pathways in breast cells and intracellular stress signaling^13,15,20. The hypothalamic-pituitary-adrenal axis and the sympathetic nervous system are key components of the neuroendocrine response to stress. Chronic stress triggers the release of corticotropin-releasing hormone from the hypothalamus, stimulating the pituitary gland to secrete adrenocorticotropic hormone, which in turn prompts the adrenal glands to release cortisol. Additionally, stress-related catecholamines e.g. epinephrine, norepinephrine, dopamine interacts with adrenergic and glucocorticoid receptors, promoting tumor initiation and metastasis, ultimately leading to poorer survival outcomes^13,21,22.

Most of earlier Iranian studies were hospital-based case-control studies^23,24, and therefore were more vulnerable to selection bias and may not fully represent the target population. In contrast, our study used a population-based incident case-control design, which provides a different and potentially more robust assessment of the past-year stress and breast cancer association. We also aimed to examine whether this association differs by menopausal status, given the possibility that hormonal and biological factors may modify the effect of stress on breast cancer risk. Understanding this association may contribute to strategies aimed at reducing BC incidence.

Materials and methods

This was a population-based incident case-control study that examined the potential association between stress experienced over the past year and the risk of developing BC. The target population included women aged 18 to 75 years living in 15 urban districts within Isfahan, Iran, between May 2021 to October 2023. A total of 600 women with newly diagnosed BC were included as the study cases, while 600 healthy women from the same study base served as controls.

Identification and enrollment of new breast cancer cases

Incident cases were identified from the Isfahan Breast Cancer Registry²⁵, with all diagnoses confirmed through histopathological examination. Date of BC diagnosis was used as the index date. A total of 653 eligible women with newly diagnosed BC were identified, and 600 (91.9%) agreed to participate in the study.

Control group selection

The control group consisted of women without any history of BC, selected across15 municipal districts in Isfahan, an area with approximately 780,000 women. To identify potential controls, the study utilized random digit dialing (RDD), a method designed to minimize selection bias²⁶. In households with more than one eligible woman, the Kish selection method was used to randomly choose respondents: this method is a well-established systematic approach for selecting participants when multiple candidates are present²⁷. Since the focus was exclusively on females, priority was given to older women in accordance with the Kish protocol. From each household, one woman was randomly selected using this method²⁷. From 2,742 RDD-generated phone numbers, based on the pre-determined RDD protocol, only contactable residential numbers of eligible women aged 18 to 75 years were included. As a result, 852 women met the eligibility criteria, and 600 controls agreed to participate in the study, yielding a response rate of 70.4%.

Data collection procedure

Data collection was done via structured interviews conducted over the phone by a team of five trained female interviewers. To reduce potential information bias, a comprehensive data collection protocol was provided. Participants were informed about the study’s objectives prior to the interviews. To ensure consistency and monitor the quality of data collection, random recordings of interviews were reviewed periodically. Given the phone-based approach, verbal informed consent was obtained from all participants before beginning the interviews. All research was performed in accordance with the Declaration of Helsinki. The study received approval from the Research Ethics Committee of Isfahan University of Medical Sciences (IR.MUI.REC.1399.010).

Measurement

Stress levels

Participants self-reported their stress level using a five-point Likert scale in response to the question: “Stress means a situation in which a person feels tense, restless, nervous or anxious or is unable to sleep at night because his/her mind is troubled all the time. During the past year, do you feel this kind of stress? with response options were 1 = Not at all, 2 = Slightly, 3 = Moderately, 4 = Much, and 5 = Very much. Responses were dichotomized into high stress (“much” and “very much”; 4–5) versus low stress (“not at all,” “little,” and “sometimes”; 1–3), consistent with standard recoding of 5-point Likert items for binary outcomes in stress epidemiology to identify elevated levels with predictive validity^6–8. This grouping was chosen a priori to distinguish participants with clearly elevated perceived stress from those reporting lower levels of stress. This item has content, criterion, and construct validity when applied to population-level assessments. Hence it is suggested that, this single-item and response scale can be an accurate substitute for longer psychological stress measurement scales in survey-based studies²⁸.

Confounders

Physical activity was assessed using the short form of International Physical Activity Questionnaire (IPAQ)²⁹, excluding walking and sedentary behaviors. Participants reported the frequency and duration of light and vigorous activities with the metabolic equivalent of task (MET) values of 3 and 8, respectively. The total MET minutes per week were categorized into three levels: 0–1000, 1001–2000, and > 2000. Family socioeconomic status (SES) was evaluated using a 10-step visual scale ladder reflecting the social hierarchy in Isfahan³⁰, with higher scores indicating better education, occupation, and income levels.

Energy intake was estimated through dietary assessment, where a trained interviewer guided participants to recall their food frequency and portion sizes, which were converted to daily intake (grams per day) and analyzed using the Nutritionist IV software.

Data on menopausal status (pre-menopausal, post-menopausal), marital status (married, single and widow), education years, first/second degree relative with BC (yes, no), menarche age, age at first delivery, BMI, benign breast diseases was also obtained.

Statistical analysis

We analyzed self-reported stress as a categorical variable: responses ‘very much’ and ‘much’ were classified as high stress level, and responses ‘not at all’, ‘little’ and ‘sometimes’ were considered low stress levels. Descriptive statistics were used to summarize the participant characteristics. Logistic regression models were used to estimate the crude and adjusted odds ratios (ORs) with 95% confidence intervals (CIs) for the association between stress level and BC. Based on existing literature, a number of variables were evaluated to verify if they fulfilled the criteria for confounding. Two multiple logistic regression models i.e., partially and fully adjusted models, were employed. The partially adjusted model was adjusted for menopause (yes, no), years of education, parity, first/second degree relative with BC (yes, no) (Table 2). The fully-adjusted model was further adjusted for age, BMI, marital status (married, single, widow), menarche age, age at first delivery, physical activity (MET level), and benign breast diseases (yes, no). Effect modification between stress and menopausal status as well as other study confounders was assessed entring interaction term in the logistic regression model. For each confounder, an adjusted logistic regression model was fitted including stress level, the confounder, the stress level × subgroup variable interaction term, and the covariates included in the fully adjusted model. P-values for interaction were obtained from the corresponding interaction term. Statistical significance was evaluated at a p-value threshold of < 0.05 (two-sided). All statistical analyses were performed using STATA version 12.

Results

Study participants’ characteristics are detailed in Table 1. The history of first/second degree relative with BC (39.8% vs. 28.8%, OR 1.63, 95% CI: 1.28–2.08, p < .001) and menopause (54.5% vs. 35.2%, OR 2.21, 95% CI: 1.75, 2.79, p < .001) was higher in cases than controls. Similarly, early menarche was more prevalent in cases (30.8%) than controls (25.0%) (OR 1.33, 95% CI: 1.03–1.72, p = .026). The proportion of those with more than 12 years of education was higher in cases than controls (OR 4.71, 95% CI: 3.40–6.53, p < .001). The odds of BC in those with BMI > 25 was significantly higher (OR 1.32, 95% CI: 1.03–1.68, p = .028) (Table 1).

Table 1.

Characteristics of breast cancer cases and controls, Isfahan, 2021–2023.

	Control		Case		Crude OR (95% CI)	p
	N	(%)	N	(%)
Age, year
<=45	303	50.7	208	34.7	Ref.
> 45	295	49.3	392	65.3	1.94 (1.53, 2.44)	< 0.001
Education years
< 6	203	33.8	92	15.3	Ref.
6 to 12	273	45.5	243	40.5	1.96 (1.45, 2.65)	< 0.001
> 12	124	20.7	265	44.2	4.71 (3.40, 6.53)	< 0.001
Marital status
Married	470	78.5	480	80.4	Ref.
Single	77	12.9	23	3.9	0.29 (0.18, 0.47)	< 0.001
Widow	52	8.7	94	15.7	1.77 (1.23, 2.54)	0.002
First/second degree relative with BC
No	427	71.2	361	60.2	Ref.
Yes	173	28.8	239	39.8	1.63 (1.28,2.08)	< 0.001
Menarche Age, year
≤ 12	150	25.0	184	30.8	1.33 (1.03,1.72)	0.026
> 12	449	75.0	413	69.2	Ref.
Age at First Delivery
≤ 18	112	23.6	166	30.6	1.43 (1.08,1.89)	0.013
> 18	362	76.4	376	69.4	Ref.
BMI, kg/m2
<=25	226	40.4	183	34.0	Ref.
> 25	333	59.6	355	66.0	1.32 (1.03, 1.68)	0.028
Physical activity MET level (Median)
< 4400	274	49.0	305	50.8	0.93 (0.74, 1.17)	0.536
> 4400	285	51.0	295	49.2	Ref.
Menopausal status
Pre-menopausal	389	64.8	273	45.5	Ref.
Post-menopausal	211	35.2	327	54.5	2.21 (1.75, 2.79)	< 0.001
Benign breast diseases
No	470	79.3	424	71.4	Ref.
Yes	123	20.7	170	28.6	1.53 (1.17, 2.00)	0.002
Family SES during adolescence (1–10), mean (SD)	4.5 (2.04)		5.31 (1.9)		1.22 (1.14, 1.29)	< 0.001

Multiple logistic regression models show the adjusted associations of stress with BC in Table 2. In the partially adjusted model, those with high stress levels, had significantly higher odds of BC than those with low stress levels, AOR 3.37 (95% CI: (2.50, 4.56), p < .001). In the fully adjusted model, the odds of BC in those with high stress levels were 3.38 (95% CI: 2.56–4.47, p < .001) times higher than those with low stress levels.

Table 2.

Crude and adjusted association of one-year stress level and risk of BC, Isfahan,2021–2023.

	Control	BC Case	Crude model	Partially adjusted Model	Fully adjusted Model
	N (%)	N (%)	OR (95% CI)	OR (95% CI)	OR (95% CI)
Stress experienced within the past year
Low	370 (61.7)	205 (34.2)	1	1	1
High	230 (38.3)	395 (65.8)	3.10 (2.45, 3.92)	3.37 (2.50, 4.56)	3.38 (2.56, 4.47)

Partially adjusted Model : Adjusted for menopausal status, years of education, parity, first/second degree relative with BC, entered by likelihood ratio forward selection variable. Fully adjusted Model: further adjusted by age, BMI, marital status, menarche age, age at first delivery, physical activity, and benign breast diseases.

Figure 1 presents the association between stress levels and risk of BC across the study main confounders. As noted, the observed association between stress level and BC is significantly stronger in pre-menopausal women than in post-menopausal (AOR 4.88, 95% CI: 3.17–7.52 vs. AOR 2.34, 95% CI: 1.53–3.58, p for interaction = 0.018). For other confounders, there was no evidence of interaction (Fig. 1).

Fig. 1.

Multiplicative interaction between high stress and selected confounders on breast cancer risk. Separate multivariable logistic regression models were fitted for each confounder, including stress level, the confounder, their interaction term (stress level × confounder), and all variables included in the fully adjusted model. The reported odds ratios (ORs) represent the association between high stress and breast cancer risk across strata of each confounder. P-values for interaction were obtained from the corresponding multiplicative interaction terms. BC: breast cancer; OR: odds ratio; CI: confidence interval.

Discussion

The findings of this population-based incident case-control study demonstrate that high stress experienced within the past year (vs. low level) significantly increase the odds of BC. This finding remained after adjustment for two sets of the confounders, as shown, in the partially and fully adjusted logistic models. In the subgroup analyses, we found that menopausal status may act as effect modifier for the stress-BC association.

The results of our study align with a meta-analysis comprising seven cohort studies involving 99,807 women, which found a 50% increased risk of BC development among women who experienced significant stressful life events³¹. Similarly, Peled et al. indicated that young women who experienced stressful life events had 62% increased odds of BC³². However, the literature is inconclusive in this regard³³, and a number of studies have not shown a significant association between stress and BC risk^34–37.

Stress may play an important role in cancer initiation and progression by releasing stress-related hormones including glucocorticoids and catecholamines^13,38. Moreover, modulating the tumor microenvironment, and fostering conditions conducive to cancer growth—such as increased pro-inflammatory cytokines, enhanced angiogenesis, and suppressed immune responses may be other pathways in stress-cancer relationship³⁸. Additionally, stress compromises immune surveillance by suppressing natural killer cells, cytotoxic T cells, and dendritic cell activity. Adrenergic signalling activates vascular endothelial growth factor (VEGF), leading to angiogenesis, which provides essential nutrients for tumor development^13,38,39. Chronic stress also increases cytokines such as IL-6, TNF-α, and IL-1β, which stimulate tumor growth³⁸. Moreover, stress activates inflammatory pathways, leading to oxidative stress, DNA damage, and mutations, thereby facilitating cancer initiation^13,38.

The stronger association observed among premenopausal women may suggest that menopausal status modifies the relationship between stress and breast cancer risk. A potential explanation is that stress-related neuroendocrine and inflammatory pathways may act differently in the hormonal environment before and after menopause⁴⁰. Chronic stress can activate the hypothalamic–pituitary–adrenal axis and sympathetic nervous system, leading to changes in cortisol, catecholamines, immune function, inflammation, angiogenesis, and tumour-related signalling pathways⁴¹. In addition, breast cancer risk is influenced by endogenous ovarian hormones, particularly oestrogen and progesterone⁴², and longer or higher exposure to these hormones has been linked to increased breast cancer risk^43,44. Therefore, it is biologically plausible that stress-related pathways may have different implications before and after menopause.

The primary strength of our study is its population-based design that enrolled only incident BC cases reducing the likelihood of major recall bias. The response rate amongst cases was excellent which reduces the possibility of major selection bias. Further, in addition to examining the association between stress and BC in adjusted analyses, we provide knowledge on effect modification of that association.

However, a number of limitations should be considered when interpreting the current study findings; First, although the reliability and validity indices of the single item stress question have been approved previously²⁸, measurement error cannot be excluded and could have impacted study findings. Second, recall bias is a well-recognized limitation in the case-control studies, can result in misclassification of exposure status. The extent to which this misclassification is differential or non-differential may determine both the estimated magnitude of the true OR and the direction of potential bias. However, recruiting incident cases can alleviate the magnitude of imposed recall bias. Third, the validity of the estimated ORs could also be distorted by potential selection bias. This is relevant given that RDD may access only those segments of the population that had active contactable residential phone-line²⁶ hence more chance of selecting participants who have higher income level and stable residence. This could introduce selection bias if these excluded individuals differ systematically concerning stress level. Nevertheless, assuming a substantial proportion of the target population has active landlines, multiple contact attempts to households enhanced the response rates and mitigated possible selection bias. Although the response rate among controls was lower than that among cases, it remained satisfactory⁴⁵. Furthermore, the response rate for incident BC cases reached approximately 92%, which likely minimizes the risk of a major selection bias within the case group. Forth, although reverse causation may threaten the inferred direction of the association between stress and breast cancer, the single-item stress measure asked cases to report their stress levels during the year preceding diagnosis (compared with “the past year” for controls). This approach helps establish temporal ordering and distinguishes our design from a simple “past year” recall. Finally, as higher educational attainment is associated with increased health-seeking behaviour, the observed association between stress and breast cancer may reflect detection bias rather than a true etiological relationship.

Conclusion

Considering inherent limitations of the case-control design, including potential recall and selection biases, our findings suggest that may be associated with an increased risk of breast cancer, particularly among premenopausal women. We also found menopausal status as possible effect modifier of the stress-breast cancer association. These findings underscore the possible role of psychosocial factors in BC development and may justify prospective, longitudinal studies to improve the validity and level of generated evidence. Implementing targeted interventions to address stress among women may represent a practical strategy to manage BC risk.

Author contributions

[M.Y]: statistical analysis, interpretation of data, support in manuscript preparation; [O. Kh.]: Data collection, drafting manuscript, participant recruitment; [S.H.J.]: methodological guidance, interpretation of biological mechanisms, critical manuscript review; [K.Z.]: scientific oversight, funding acquisition, critical manuscript review; [I.A.]: study conception and design, data quality control, data management and analysis, result interpretation, drafting and finalizing manuscript.

Funding

This work was supported by Isfahan University of Medical Sciences (grant number: 199061) and the Shams Charity Organization and Cancer Institute of Iran, Tehran University of Medical Sciences (grant number: 1400-2-244-53323). N. Houssami is funded by a National Breast Cancer Foundation (NBCF Australia) Chair in Breast Cancer Prevention (grant #EC-21-001).

Data availability

The dataset used in this study is available upon request. Dr. Ibrahim Abdollahpour, as the lead contact, maintained complete access to the data and was responsible for verifying the integrity and accuracy of the analytical procedure. He oversaw and performed all statistical analyses.

Declarations

Competing interests

The authors declare no competing interests.

Ethics approval and consent to participate

Verbal informed consent was obtained from all participants. The ethics committee of the Isfahan University of Medical Sciences approved this study (IR.MUI.REC.1399.010).