Thyroid disease and breast cancer risk in the Sister Study

Abstract

Background:

Thyroid disease (TD), and its treatments, are suspected risk factors for breast cancer (BC), attributed primarily to alterations in thyroid hormone-related metabolism. Prior literature is inconsistent, in part, due to small sample sizes. We, therefore, assessed the effects of hyper- and hypothyroidism on incident BC in a large cohort of US women.

Methods:

Data are from the Sister Study, a prospective cohort of 50,884 US women with no BC history at enrollment (2003–2009; ages 35–74 years) and follow-up through September 2021. We used time-varying Cox proportional hazards regression to estimate hazard ratios (HR) and 95% confidence intervals (CI) for the association between self-reported thyroid diseases and treatments and incident BC.

Results:

Of 45,783 eligible participants, 11,567 reported TD at baseline and over follow-up (hyperthyroidism: n=1,710; hypothyroidism: n=9,857) and 3,897 participants subsequently developed BC. Breast cancer HR was 1.15 (CI:0.95, 1.39) for those with hyperthyroidism and 0.93 (CI:0.85, 1.01) for those with hypothyroidism, compared to no thyroid disease, with similar patterns observed across racial and ethnic groups. Hyperthyroidism was associated with greater premenopausal BC hazard (HR=1.85, CI:1.11, 3.07) and differed from BC hazard for hypothyroidism, which appeared reduced (HR=0.75, CI:0.54, 1.05) but did not reach statistical significance (p_{heterogeneity}=0.06). BC hazards were higher, albeit not statistically significant, for those treated for hyperthyroidism but not hypothyroidism.

Conclusions:

Women diagnosed with hyperthyroidism or receiving related treatment may have elevated BC risk, particularly premenopausal BC.

Impact:

Although more research is needed, premenopausal women treated for hyperthyroidism may benefit from enhanced breast cancer screening.

Introduction

Diseases of the thyroid gland, which modulates growth, development, and cellular metabolism, are common endocrine disorders.(1) The two most common thyroid diseases are hyperthyroidism and hypothyroidism, both characterized by abnormal production of thyroid hormones.(2,3) According to a study using National Health and Nutrition Examination Survey (NHANES) data, the estimated prevalence of thyroid disease among US women was 7.6% between 2015 to 2018, with the highest prevalence among older non-Hispanic White women.(4)

Breast cancer is one of the most commonly diagnosed cancers and a leading cause of cancer death among US women.(5) Dysregulation of circulating estrogen and other reproductive hormones is thought to be a primary driver of breast cancer pathogenesis.(6) In vitro studies have reported that thyroid hormones can have estrogen-like effects, suggesting that thyroid hormones, such as triiodothyronine (T3), may impact cellular proliferation of breast tissue and subsequently increase breast cancer risk among individuals with hyperthyroidism.(7–9) Radioactive iodine (RAI) is a common therapy for hyperthyroidism and is administered systemically, but localizes in the thyroid. While traditional radiation exposure is a known risk factor for breast cancer, studies have been inconclusive on whether residual radiation exposure due to RAI may confer additional breast cancer risk.(10,11) Therefore, it is critical to better understand the relationship between thyroid disease and breast cancer risk, and how thyroid disease treatments may alter that risk.

Multiple observational studies and subsequent meta analyses have reported trends suggesting elevated breast cancer risk among women with hyperthyroidism and decreased or no breast cancer risk associated with hypothyroidism.(12–16) Prior research characterizing the association between RAI treatment for hyperthyroidism and breast cancer have also reported increased breast cancer risk particularly at higher doses.(10,11,14,15) While prior work has been relatively abundant, only a few studies have accounted for potential confounding due to known breast cancer risk factors (e.g., menopausal status, parity, health behaviors, etc.)(12,15,17,18), or have assessed the risk of hormone receptor-specific breast cancer associated with thyroid disease.(16,19) In the current study, we aimed to investigate the association of hyperthyroidism and hypothyroidism, as well as corresponding treatments, with incident breast cancer in a cohort of U.S. women.

Materials and Methods

Study Population

The Sister Study is a prospective cohort study of women (N=50,884), residing in the U.S., including Puerto Rico, that was developed to assess genetic and environmental risk factors of breast cancer and other chronic conditions. All participants were aged 35 to 74 years at enrollment (2003–2009) and had a sister diagnosed with breast cancer, but no personal breast cancer history.(20) Participants completed a baseline computer-assisted telephone interview and additional detailed questionnaires, which included information on sociodemographics, medical history, and health behaviors. Blood samples and anthropometric measurements were collected at an enrollment home visit. Health updates are collected annually, and detailed follow-up questionnaires are administered approximately every 2 to 3 years, with approximately 85% of participants responding to the most recent follow-up request. We use data collected through September 30, 2021 (data release 11.1). The Sister Study is overseen by the Institutional Review Board of the National Institutes of Health. All participants provided written informed consent at enrollment.

Participants were excluded due to study withdrawal (n=5); or if they additionally reported history of breast cancer at baseline or timing of breast cancer diagnosis was unknown (n=79); had a history of thyroid cancer (or timing of diagnosis was unknown; n=288), thyroid nodules (n=3,749), or other thyroid diseases excluding hyperthyroidism and hypothyroidism (n=196); had less than 0.1 years of follow-up time due to immediate breast cancer diagnosis or other censoring (n=269); or were missing data on key covariates (n=515; less than 0.5% for any given variable). This resulted in an analytic sample of 45,783 participants.

Breast Cancer Ascertainment

Participants reported information on incident invasive breast cancer and ductal carcinoma in situ (DCIS) diagnoses on follow-up questionnaires. Study staff were able to confirm breast cancer diagnosis and tumor characteristics, including estrogen receptor (ER) and tumor stage, for approximately 99% of the more than 82% of participants with breast cancer who provided medical records. Tumor receptor status remained unknown for less than 1.5% of participants with medical records validation and approximately 15% of participants with breast cancer overall. Due to the high concordance between self-reported breast cancer and medical records (positive predictive value of 98%), including breast cancer overall and self-reported ER status, we relied on self-reported data on breast cancer diagnoses and tumor characteristics when medical records were unavailable.(21)

Thyroid Disease Classification

Thyroid disease diagnosis was reported on the baseline questionnaire. Participants were asked “Has a doctor or other health professional ever told you that you had thyroid disease or thyroid problems?” In each subsequent detailed follow-up questionnaire, participants were similarly asked if they were diagnosed with a thyroid condition during the interim period since the preceding study visit. Participants responding affirmatively at any visit were subsequently asked to disclose the type of thyroid condition and age of diagnosis, as well as any corresponding treatments, including RAI treatment, thyroidectomy, antithyroid medication, and thyroid hormone replacement therapy. At enrollment and starting at follow-up questionnaire 4 (2017–2019), participants were asked specifically if they ever received RAI (regardless of whether they reported thyroid disease) or thyroid surgery and if so, the corresponding year. If participants reported both hyperthyroidism and hypothyroidism, their exposure classification was based on their first reported diagnosis (i.e., primary disease). Furthermore, if both hyper- and hypothyroidism were reported but the ages at diagnoses were unknown or indistinguishable, participants were classified as hyperthyroidism due to the potential for hyperthyroidism treatment-induced hypothyroidism.(22)

Covariate Assessment

Covariate selection was done a priori based on literature review. Covariates of interest were initially collected at baseline and included age at enrollment, self-identified race and ethnicity (non-Hispanic White, Black or African American, non-Black Hispanic, another race—including Asian, American Indian, multiracial and unknown), highest level of educational attainment (high school or less, some college, bachelor’s degree, master’s degree or higher), body mass index from examiner assessment (BMI; continuous), smoking status (never, past, current smoker), alcohol use (never, former, current nonregular, current regular), menopausal status (premenopausal, postmenopausal), hormonal contraceptive use (never, former, recent—past 5 years), parity (nulliparous, 1 child, 2 children, ≥3 children), and age at first pregnancy (nulligravid, ≤20 years old, 21–29 years old, ≥30 years old). Time-varying covariates were updated from detailed follow-up questionnaires including age at first pregnancy, parity, BMI (from self-reported height and weight), smoking status, alcohol use, menopausal status, and hormonal contraceptive use. Menopausal status was self-reported at baseline and during follow-up and was defined as the absence of menstrual periods for at least 12 months; age at menopause was subsequently defined as the age at last menstrual period before this interval.

Statistical Analyses

Characteristics of the study population were compared by thyroid disease status: hyperthyroidism, hypothyroidism, and no thyroid disease. We used time-varying Cox proportional hazards models with age as the timescale (i.e., age at baseline to age at breast cancer diagnosis, with censoring for death, end of follow-up, development of thyroid cancer or other thyroid disease, or loss to follow-up) to estimate hazard ratios (HR) and 95% confidence intervals (CI) for the association between hyperthyroidism or hypothyroidism and incident breast cancer. Thyroid disease status was treated as a time-varying exposure for which risk time was initially assigned based on thyroid disease status at baseline. Those reporting an incident thyroid disease diagnosis had their status re-assigned at the age at which they returned a follow-up questionnaire reporting thyroid disease information; this primary diagnosis persisted through the remainder of follow-up irrespective of any subsequent diagnoses of thyroid disease. Exposure status and other covariates were similarly updated, with each variable defined based on the participant’s self-reported status on their most recent follow-up questionnaire until end-of-follow-up. As such, the prospective nature of the data was retained. Participants with complete data at baseline, but who were missing one or more covariate values across follow-up had their values carried forward. All models were adjusted for covariates described above, except for menopausal status which was assessed in menopause-specific breast cancer models. We assessed for violations of the proportional hazards assumption using a Wald test of the interaction between age and the exposure of interest.

Stratified estimates were calculated for premenopausal and postmenopausal breast cancer, as well as ER-positive and ER-negative breast cancer. In menopause-stratified analyses, hazards for premenopausal breast cancer were estimated for the follow-up time between baseline and breast cancer diagnosis, end of follow-up, or onset of menopause, whichever occurred first. Conversely, hazards for postmenopausal breast cancer were assessed from onset of menopause or baseline, whichever occurred later, through breast cancer diagnosis or end of follow-up. For stratified models with competing disease outcomes (e.g., ER status), we censored participants without the outcome of interest at their competing diagnosis (e.g., when investigating ER-positive tumors, participants with ER- or unknown ER status were censored).

In models investigating the association between treatments for thyroid disease and breast cancer hazard, therapies for primary hyperthyroidism and hypothyroidism were modeled separately with referent groups comprised of individuals who did not have the respective primary thyroid disease.

In sensitivity analyses, we estimated the association between prevalent thyroid disease at baseline and breast cancer hazard to highlight potential differences between prevalent-only and time-varying (prevalent plus incident thyroid disease diagnoses) models. Models were stratified by race and ethnicity to identify potential race-specific patterns that could be attributed to historical differences in relevant environmental exposures and/or clinical practices.(23) We do not report breast cancer hazards for participants who self-identified as another race and ethnicity due to small sample sizes. Finally, since there are complicated relationships between obesity and thyroid disease, as well as obesity and breast cancer, we also stratified models by BMI category (underweight/healthy weight, overweight, obese). In stratified models, we used Wald tests for heterogeneity to test for group differences (p<0.05). All analyses were conducted using SAS 9.4 (Cary, NC; RRID:SCR_008567).

Data Availability

Requests for Sister Study data can be submitted by following the directions at http://sisterstudy.niehs.nih.gov/English/coll-data.htm.

Results

Among eligible participants, 863 reported prevalent hyperthyroidism, 5,869 reported prevalent hypothyroidism, and 4,835 developed thyroid disease over follow-up (primary disease: hyperthyroidism=847; hypothyroidism=3,988). Overall, participants contributed 535,846 person years for analyses with 3,897 participants developing invasive breast cancer (n=3,095) or DCIS (n=802) during follow-up.

Demographic and participant characteristics based on primary thyroid disease classification are presented in Table 1. Participants with hyperthyroidism (mean age: 56.0±8.9 years) or hypothyroidism (mean age: 57.5±8.9 years) before the end of follow-up were slightly older at baseline than those with no thyroid disease (mean age: 54.9±8.9 years). Furthermore, participants with hyperthyroidism were more likely to identify as Black or African American (12%) or non-Black Hispanic (7%) compared to those with hypothyroidism (4%; 4%, respectively) and no thyroid disease (10%; 5%, respectively). Individuals with thyroid disease were more likely to be classified as obese and be postmenopausal at enrollment compared to those with no thyroid disease.

Table 1.

Baseline characteristics of eligible Sister Study participants by primary thyroid disease diagnosis at end of follow-up (N=46,298)

	Primary Thyroid Disease Diagnosis at End of Follow-up
	None (n = 34,603)		Hyperthyroidism (n = 1,733)		Hypothyroidism (n = 9,962)
	N	%	N	%	N	%
Age at enrollment, mean (SD)	54.9	8.9	56.0	8.9	57.5	8.9
Race and Ethnicity
Non-Hispanic white	28,458	82	1,342	77	8,990	90
Black/African American	3,610	10	212	12	348	4
Non-Black Hispanic	1,633	5	129	7	360	4
Other	889	3	50	3	263	3
Missing	13		0		1
Educational attainment
High school/GED or less	5,304	15	348	20	1,433	14
Some college or associate’s degree	11,517	33	612	35	3,406	34
Bachelor’s degree	9,448	27	425	25	2,679	27
Master’s degree or higher	8,324	24	348	20	2,443	25
Missing	10		0		1
Body mass index at Baseline
Underweight/Healthy weight (<25.0)	14,576	42	667	39	3,631	37
Overweight (25.0–29.9)	10,884	32	553	32	3,188	32
Obese (30.0+)	9,048	26	502	29	3,116	31
Missing	95		11		27
Menopause status
Pre-menopausal	12,585	36	532	31	2,637	27
Post-menopausal	22,010	64	1,201	69	7,325	74
Missing	8		0		0
Ever used hormonal birth control
Never	4,905	14	260	15	1,536	16
Prior	25,863	75	1,314	76	7,633	77
Recent (Past 5 years)	3,656	11	152	9	737	7
Missing	179		7		56
Parity
Nulliparous	6,233	18	328	19	1,887	19
1	5,132	15	273	16	1,311	13
2	12,790	37	625	36	3,572	36
3 or more	10,427	30	505	29	3,184	32
Missing	21		2		8
Age at first pregnancy
Nulligravid	4,269	12	224	13	16	0.2
<20 years	8,999	26	530	31	1,318	15
20–<30 years	16,776	49	798	46	2,532	28
30+ years	4,467	13	177	10	5,079	57
Missing	65		4		16
Smoking status
Never smoked	19,651	57	941	54	5,522	56
Past smoker	11,970	35	595	34	3,812	38
Current smoker	2,972	9	196	11	624	6
Missing	10		1		4
Alcohol use
Never	1,241	4	96	6	398	4
Former	4,988	14	340	20	1,593	16
Current non-regular	2,788	8	166	10	826	8
Current regular	25,576	74	1,130	65	7,144	72
Missing	10		1		1

We estimated the association between thyroid disease and incident breast cancer overall, by ER tumor status, and by menopausal status (Table 2). After adjusting for age (as the timescale), educational attainment, race and ethnicity, smoking, BMI, hormonal contraceptive use, parity, and age at first pregnancy, the HR for breast cancer was 1.15 (95% CI: 0.95, 1.39) for those with hyperthyroidism and 0.93 (95% CI: 0.85, 1.01) for those with hypothyroidism compared to those with no thyroid disease. When we evaluated the association between prevalent thyroid disease and incident breast cancer, the hazard for hyperthyroidism remained constant and the estimate for hypothyroidism was slightly attenuated compared to time-varying models that included both prevalent and time-varying thyroid disease diagnoses (Supplemental Table 1). Estimates of breast cancer hazard for both hyperthyroidism and hypothyroidism were less precise, which can be attributed to the approximately 4,800 participants who developed thyroid disease after baseline and were more accurately classified in time-varying analyses.

Table 2.

Association of time-varying thyroid disease and incident breast cancer, overall, by ER status, and by menopausal status

		Overall (N=45,783)		ER-positive (N=45,783)		ER-negative (N=45,783)		Premenopausal breast cancer (N=15,631)			Postmenopausal breast cancer (N=43,084)
	PY	Cases	HR (95% CI)a	Cases	HR (95% CI) a	Cases	HR (95% CI) a	PY	Cases	HR (95% CI) a	PY	Cases	HR (95% CI) a
Hyperthyroidism	18,834	113	1.15 (0.95, 1.39)	82	1.21 (0.97, 1.50)	12	1.02 (0.71, 1.47)	1,377	16	1.85 (1.11, 3.07)	11,643	97	1.09 (0.89, 1.34)
Hypothyroidism	117,578	665	0.93 (0.85, 1.01)	495	0.96 (0.87, 1.06)	77	0.85 (0.72, 1.01)	7,792	38	0.75 (0.54, 1.05)	82,875	628	0.94 (0.86, 1.03)
No Thyroid Disease	399,434	3,119	1.00 (Ref)	2,246	1.00 (Ref)	398	1.00 (Ref)	81,423	523	1.00 (Ref)	352,714	2614	1.00 (Ref)

CI: Confidence interval; ER: Estrogen receptor; HR: Hazard ratio; PY: Person years

^aAdjusted for race and ethnicity, educational attainment, self-reported BMI-continuous, smoking status, alcohol use, birth control use, parity, age at first pregnancy

Wald tests for heterogeneity: ER status p_{heterogeneity} = 0.4; Pre-/Post- menopausal breast cancer p_{heterogeneity} = 0.06

We also observed estimates similar to overall estimates when assessing hazards for ER positive breast cancer: HR=1.21 (95%CI: 0.97, 1.50) for hyperthyroidism and HR=0.96 (95%CI: 0.87, 1.06) for hypothyroidism, compared to no thyroid disease. Estimates for ER-negative breast cancer were largely null for both hyperthyroidism (HR=1.02, 95%CI: 0.71, 1.47) and hypothyroidism (HR=0.85, 95%CI: 0.72, 1.01). There was no statistically significant difference in hazard estimates for ER-positive versus ER-negative breast cancer (p_{heterogeneity} = 0.4). The hazard for premenopausal breast cancer was 1.85 times as high (95%CI: 1.11, 3.07) for participants with hyperthyroidism compared to those with no thyroid disease. Conversely, hypothyroidism was associated with a lower hazard for premenopausal breast cancer (HR=0.75, 95%CI: 0.54, 1.05), albeit not statistically significant. Estimates for postmenopausal breast cancer hazard differed from premenopausal breast cancer hazard but did not reach statistical significance (p_{heterogeneity} = 0.06).

Models stratified by race and ethnicity are presented in Supplemental Table 2. Estimates suggest elevated hazards for breast cancer due to hyperthyroidism across all racial and ethnic groups but did not reach statistical significance. However, associations of hyper- or hypothyroid disease with breast cancer did not differ significantly by race and ethnicity. (p_{heterogeneity} = 0.7), likely due to the limited number of participants identified as Black or non-Black Hispanic with a thyroid disease diagnosis. In models stratified by baseline BMI category, we observed patterns similar to overall BC hazards among participants identified as overweight or obese at baseline, with increased BC hazards observed for hyperthyroidism and decreased BC hazards observed for hypothyroidism, albeit not statistically significant (Supplemental Table 3). Hazards for BC were largely null for both hyperthyroidism (HR=0.92; 95%CI: 0.65, 1.31) and hypothyroidism (HR=0.90; 95%CI: 0.78, 1.05) among participants who were underweight or healthy weight at baseline. There were no statistically significant differences in estimated effects between participants based on BMI category (p for heterogeneity = 0.6).

We next assessed breast cancer hazards associated with common treatments for hyperthyroidism—RAI treatment, thyroidectomy, and antithyroid medication—and hypothyroidism—thyroid hormone replacement therapy. Approximately 20% of all participants reported receiving RAI treatment (n=463), thyroidectomy (n=107), antithyroid medication (n=119), and/or thyroid hormone replacement therapy (n=8,726). Overall, 649 participants who received any of these common thyroid disease treatments were subsequently diagnosed with breast cancer (approximately 17% of all participants who developed breast cancer). Hazards for breast cancer were elevated, albeit not statistically significantly, for participants who reported ever receiving common treatments for hyperthyroidism, including RAI treatment alone (HR=1.21; 95%CI: 0.89, 1.64), having a thyroidectomy alone (HR=1.47; 95%CI: 0.69, 3.14), treatment with antithyroid medication alone (HR=1.62; 95%CI: 0.85, 3.09), or multiple therapies (HR=1.59; 95%CI: 0.60, 4.21) compared to participants who did not have primary hyperthyroidism after adjusting for covariates of interest and other treatments for hyperthyroidism (Table 3). Hyperthyroidism with no reported treatment was not associated with breast cancer hazard (HR=1.03; 95%CI: 0.80, 1.32) compared to participants without primary hyperthyroidism. We did not observe an association between untreated primary hypothyroidism or hypothyroidism treated with thyroid hormone replacement therapy and breast cancer hazard compared to participants without primary hypothyroidism (Table 3).We observed no violations of the proportional hazards assumption for any models (all p >0.05).

Table 3.

The association of time-varying thyroid disease treatments with incident breast cancer, by primary thyroid disease type (N=45,221).

	PY	Cases	HR (95% CI)a
Primary Hyperthyroidism
No hyperthyroidism	97,404	3,771	1.00 (Reference)
Hyperthyroidism – No reported treatment	2,142	64	1.03 (0.80, 1.32)
RAI only	904	42	1.21 (0.89, 1.64)
Thyroidectomy only	188	7	1.47 (0.69, 3.14)
Antithyroid medication only	213	9	1.62 (0.85, 3.09)
Multiple therapies	91	4	1.59 (0.60, 4.21)
Primary Hypothyroidism
No hypothyroidism	78,511	3,154	1.00 (Reference)
Hypothyroidism – No reported treatment	2,954	94	0.95 (0.77, 1.16)
Thyroid hormone replacement therapy	19,477	649	0.94 (0.86, 1.03)

CI: Confidence interval; HR: Hazard ratio; PY: Person years; RAI: Radioactive iodine

^aAdjusted for race and ethnicity, educational attainment, self-reported body mass index, smoking status, alcohol use, birth control use, parity, age at first pregnancy

Discussion

Results from the present study, conducted in a large prospective cohort of US women, support the association between diagnosis of hyperthyroidism and incident breast cancer overall, and for ER-positive disease, although estimates did not reach statistical significance. However, we did observe greater hazard for premenopausal versus postmenopausal breast cancer. Conversely, diagnosis of hypothyroidism was not associated with incident breast cancer in our overall cohort. These patterns were consistent across groups defined by self-reported race and ethnicity. Our study also examined breast cancer hazards associated with treatments for hyperthyroidism, observing that participants who received RAI alone, thyroidectomy alone, antithyroid medication alone, or multiple therapies, but not untreated hyperthyroidism, had elevated breast cancer hazards compared to participants without primary hyperthyroidism, albeit not reaching statistical significance.. Conversely, thyroid hormone replacement therapy, a common treatment for hypothyroidism, was not associated with breast cancer hazard. These results suggest that thyroid disease, and in particular hyperthyroidism and related treatments, may represent an important risk factor for incident breast cancer.

Circulating thyroid hormones (i.e., T₃ and T₄) are key components in endocrine signaling. In addition to regulating metabolic effects of the thyroid, liver, heart, and other major organs, thyroid hormones have also been shown to affect cellular growth and differentiation in breast tissue.(24) For example, studies have suggested that thyroid hormones may interact with estrogen receptors in breast cancer tumor cells resulting in anti-apoptotic effects.(25) Furthermore, previous studies have reported increased breast cancer risk among women with high levels of T₃ and T₄.(18,26,27) These biologic mechanisms support the overall risk of breast cancer associated with hyperthyroidism and the lack of additional risk or decreased risk associated with hypothyroidism observed in our study as well as previous research from the UK Biobank and the Danish National Health Service.(15,19) A recent meta-analysis on thyroid dysfunction and cancer incidence reported similar findings, namely a positive pooled risk ratio (RR) for breast cancer and hyperthyroidism (RR: 1.20; 95%CI: 1.04, 1.38) but a null or slightly inverse RR for breast cancer and hypothyroidism (RR=0.73; 95%CI: 0.43, 1.24).(28) Overall, these findings are fairly consistent with our results even though few studies account for potential confounding from traditional breast cancer risk factors or lifestyle factors known to be associated with both thyroid disease and breast cancer risk, such as obesity. (12,15,17,18,29)

While we observed estimates that suggested greater postmenopausal breast cancer hazard among participants with hyperthyroidism (albeit not statistically significant), as reported in other studies(12,19), greater hazard of premenopausal breast cancer was also associated with hyperthyroidism in our cohort, which has not been commonly reported. A recent meta-analysis of premenopausal breast cancer reported a RR of 1.04 (95%CI: 0.75, 1.45), which included results from three cohort studies from Korea, Sweden, and the United Kingdom and one case control study of Italian women.(13,15,30,31) Conversely, hazards for premenopausal breast cancer appeared lower among participants with hypothyroidism, albeit not statistically significant, which aligns with results observed in other studies, including the Women’s Health Initiative, a large US cohort of postmenopausal women.(13,16) The Sister Study represents an enriched cohort such that all participants have at least one first degree relative with breast cancer, which may contribute to the observed differences in association between hyperthyroidism and premenopausal breast cancer. Furthermore, observed differences may also be in part due to the fact that previous studies have often adjusted for or stratified by baseline menopause status, ignored menopause status in their analyses, restricted study samples to predominantly postmenopausal women, and/or not evaluated time-varying premenopausal breast cancer risk as we do in this study, in which 82% of participants who were premenopausal at baseline experienced menopause over follow-up with an average time to menopause of 5.2 years. The three studies we observed that reported the differential effects of both hyperthyroidism and hypothyroidism on pre- and post- menopausal breast cancer were conducted in non-US cohorts or case control studies and reported mixed results for the effects of hyperthyroidism on premenopausal breast cancer risk.(13,17,18)

We did not observe statistically significant differences by race and ethnicity in breast cancer risk associated with diagnosis of thyroid disease, potentially due to limited case numbers. Larger and more diverse studies are needed to investigate the potential social impacts attributed to race and ethnicity on the association between thyroid disease and breast cancer hazard.

Previous studies that have focused on breast cancer risk associated with RAI treatment have reported both slightly increased(11,14,15,32) and decreased(10) risk, which have largely been attributed to differences in RAI dosage. We observed potentially elevated breast cancer hazards, albeit not statistically significant, among participants receiving common treatments for hyperthyroidism, including RAI treatment, thyroidectomy, or antithyroid medication, alone or in combination, compared to participants who did not receive these therapeutics. RAI treatment and thyroidectomy may reflect access to advanced medical care and/or thyroid disease severity since definitive treatments are predominantly prescribed in cases of recurring hyperthyroid disease and severe disease manifestations.(3) However, these estimates should be considered with caution as they are likely under powered due to the small sample size and may not directly reflect clinical severity of thyroid disease nor dose of RAI treatment, which were not collected in the Sister Study.

There are some limitations of this study to be considered. The data in this study are self-reported, with potential for misclassification of thyroid disease, timing, disease severity, and treatments. Further, thyroid disease is a unique condition that may change over time; for example, patients may develop hyperthyroidism, receive treatment, and subsequently be diagnosed with hypothyroidism due to treatment-induced suppression of thyroid function.(22) We classified participants who reported both hyperthyroidism and hypothyroidism by the end of follow-up (approximately 10% of participants with any thyroid disease) according to the thyroid disease that was first reported. Ascertainment of thyroid disease treatment was not uniform across follow-up questionnaires and did not capture dosage or frequency. Medications were classified by medication type: antithyroid medication and thyroid hormone replacement. Due to the potential for misclassification by therapeutic and thyroid disease diagnosis (e.g., participants who 1) report RAI but only report hypothyroidism; or 2) are classified as hyperthyroidism but later develop hypothyroidism and report therapies for both), we did not restrict therapeutic models to participants with specific thyroid disease diagnoses. While RAI therapy and thyroidectomy are both used in the treatment of thyroid cancer, participants with thyroid cancer were censored at diagnosis over follow-up, and unreported thyroid cancer is likely rare due to the straightforward nature of its diagnosis. Further, prior research has reported up to a 20-year latency period between RAI exposure and breast cancer risk.(33) While we were unable to incorporate a latency period into our models due to limitations of the data, the average time between RAI exposure and end of follow-up was approximately 18 years and only 5 participants developed breast cancer within 5 years of exposure. Finally, Sister Study participants all have a family history of breast cancer, which may limit the generalizability of our findings, though our results remain internally valid.(34)

Despite these limitations, this work utilizes a large prospective cohort of US women to assess the relationship between thyroid disease and breast cancer risk. Our large, geographically and racially and ethnically diverse prospective cohort includes a moderate proportion of premenopausal women, who are underrepresented in previous studies of thyroid disease and breast cancer risk. Although misclassification due to undiagnosed thyroid disease is possible, the effects are likely minimal due to the high utilization of preventative healthcare in the cohort; approximately 90% of participants reported having a physical exam in the past 24 months at baseline and 94% reported having general health coverage at the time of the first follow-up visit. Further, previous studies have often utilized prevalent disease models which only consider thyroid disease status at baseline, even though risk of thyroid disease increases with age. We consider time-varying status of thyroid disease and other covariates to reduce misclassification and better capture the dynamic relationships between hyperthyroidism, hypothyroidism, and incident breast cancer while controlling for likely confounders. While estimates were not largely changed in prevalent analyses, estimates were more precise in time-varying models.

This study continues to support a positive association between hyperthyroidism and breast cancer risk. We reported elevated risk of breast cancer overall, and ER-positive breast cancer, albeit not statistically significant, with evidence that the association is most apparent for premenopausal disease. We observed no overall association between hypothyroidism and breast cancer risk. Future research could be improved by ensuring medical record confirmed classification of thyroid disease and its treatments among larger and more diverse cohorts, including targeted consideration of premenopausal women.