Serum perfluorooctane sulfonate and perfluorooctanoate and risk of postmenopausal breast cancer according to hormone receptor status: An analysis in the Prostate, Lung, Colorectal and Ovarian Cancer Screening Trial

Abstract

Per- and polyfluoroalkyl substances (PFAS) are highly persistent endocrine-disrupting chemicals that may contribute to breast cancer development; however, epidemiologic evidence is limited. We investigated associations between prediagnostic serum levels of perfluorooctane sulfonate (PFOS) and perfluorooctanoate (PFOA) and postmenopausal breast cancer risk, overall and by hormone receptor status, in a nested case-control study of 621 cases and 621 matched controls in the Prostate, Lung, Colorectal and Ovarian Cancer Screening Trial. PFOS and PFOA levels were determined based on serum metabolomic profiling performed using ultra-performance liquid chromatography-tandem mass spectrometry. We used multivariable conditional logistic regression to estimate odds ratios (ORs) and 95% confidence intervals (CIs) for the association between each PFAS and breast cancer risk, overall, by estrogen receptor (ER) or progesterone receptor (PR) status, and by joint ER/PR status. We found little evidence of association between PFOS or PFOA and breast cancer risk overall. However, in subtype-specific analyses, we observed statistically significant increased risks of ER+, PR+, and ER+/PR+ tumors for the third vs. lowest quartile of serum PFOS (ORs [95% CIs]=1.59 [1.01–2.50], 2.34 [1.29–4.23], and 2.19 [1.21–3.98], respectively) and elevated but non-statistically significant ORs for the fourth quartile. Conversely, for PFOA, modest positive associations with ER−, PR−, ER+/PR−, and ER−/PR− tumors were generally seen in the upper quartiles. Our findings contribute evidence supporting positive associations between serum PFOS and hormone receptor-positive tumors, and possibly between PFOA and receptor-negative tumors. Future prospective studies incorporating tumor hormone receptor status are needed to better understand the role of PFAS in breast cancer etiology.

1. INTRODUCTION

Per- and polyfluoroalkyl substances (PFAS) are a group of highly persistent synthetic chemicals manufactured since the 1940s and used in a wide range of industrial applications and consumer products.¹ Nearly the entire U.S. population has detectable levels of perfluorooctane sulfonate (PFOS) and perfluorooctanoate (PFOA)—two of the most common PFAS—in their blood.² In 2014, the International Agency for Research on Cancer (IARC) classified PFOA—the only PFAS evaluated by IARC to date—as “possibly carcinogenic to humans” (Group 2B), based in part on limited epidemiologic evidence of associations with kidney and testicular cancers.^3,4 As endocrine-disrupting chemicals with potential estrogenic properties, some PFAS, including PFOS and PFOA, have also been suggested to play a role in the development of breast cancer,⁵ although the epidemiologic evidence is relatively inconsistent and inconclusive.⁶ Recently, based on a growing body of literature since the IARC evaluation, a 2022 report by the U.S. National Academies of Sciences, Engineering, and Medicine stated that there is “limited or suggestive” evidence of an association between PFAS exposure and breast cancer risk.⁷ It should however be noted that most studies evaluating these associations to date relied on model-estimated exposure levels or PFAS measurements in blood collected after breast cancer diagnosis.⁶

In a recent nested case-control study with 194 case-control pairs of postmenopausal breast cancer in the French E3N cohort, Mancini et al. reported novel findings suggesting that the relationship between PFAS exposures and breast cancer may differ by tumor hormone receptor status.⁸ Specifically, they observed positive associations between prediagnostic serum PFOS concentrations and both estrogen receptor (ER)-positive and progesterone receptor (PR)-positive tumors, as well as increased risks of receptor-negative tumors associated with medium-to-low levels of PFOS and PFOA; however, the study did not evaluate associations by joint hormone receptor status, possibly due to sample size limitations.

We investigated associations between postmenopausal circulating levels of PFOS and PFOA and risk of developing breast cancer subtypes defined by tumor ER and PR status by leveraging data from our recent serum metabolomic profiling study of 621 postmenopausal breast cancer cases and 621 matched controls within the Prostate, Lung, Colorectal and Ovarian (PLCO) Cancer Screening Trial.

2. MATERIALS AND METHODS

2.1. Study design and population

The PLCO Cancer Screening Trial is a multicenter randomized trial that enrolled ~155,000 men and women 55–74 years of age across 10 study centers in the U.S. between 1993 and 2001; participants were assigned to either a screening or control arm and followed over time for cancer outcomes (<4% loss to follow-up), with blood samples collected from most (>90%) screening arm participants.⁹ Details of the design and population characteristics of our nested case-control study have been reported previously.^10,11 Briefly, this study included all incident primary invasive breast cancer cases diagnosed among women in the PLCO screening arm during follow-up through November 2013 who were postmenopausal at baseline and not using menopausal hormone therapy (MHT) at baseline. Eligible cases must have also provided prediagnostic serum samples one year after baseline (i.e., all postmenopausal at blood draw). To increase the number of hormone receptor-negative cases for subtype-specific analyses, all eligible cases with ER− and/or PR− breast tumors using MHT at baseline were also included. Using incidence density sampling, controls were selected from among women who were postmenopausal at baseline, alive and cancer-free (excluding non-melanoma skin cancer) at the time of case diagnosis, and were individually matched to cases by age at baseline (±2 years), date of blood draw (±3 months), and MHT use at baseline.

Incident cancer cases were ascertained through annual questionnaires, the National Death Index, physician reports, and next-of-kin reports, with further confirmation by review of hospital records at the screening centers. Information on demographics, lifestyle, and reproductive history of all cases and controls was obtained from the baseline PLCO questionnaire, and tumor hormone receptor status of cases was ascertained based on immunohistochemistry results from hospital records.

2.2. Serum PFOS and PFOA measurements

We used a non-targeted high-resolution mass spectrometry platform at Metabolon Inc. (Durham, NC) to measure concentrations of numerous molecular compounds in prediagnostic sera, as described previously.^10–12 Briefly, serum samples were extracted with methanol to precipitate proteins and analyzed using ultra-performance liquid chromatography-tandem mass spectrometry.¹² Of the 672 compounds identified from the analysis, PFOS and PFOA were the only two identifiable PFAS measured at detectable levels using this untargeted metabolomics approach. Based on data from another ongoing study (n=335 cancer-free PLCO participants), measurements of serum PFOS and PFOA from Metabolon were strongly correlated with standard, targeted measurements of these chemicals in the same subjects by the Centers for Disease Control and Prevention (CDC) laboratory using on-line solid phase extraction liquid chromatography-isotope dilution-tandem mass spectrometry¹³ (Spearman correlation coefficient=0.76 and 0.77 for PFOS and PFOA, respectively; unpublished data).

2.3. Statistical analysis

We fit conditional logistic regression models to estimate odds ratios (ORs) and 95% confidence intervals (CIs) relating serum PFOS or PFOA levels and breast cancer risk. ORs were estimated for breast cancer overall, by individual hormone receptor (ER and PR) status, and by joint ER/PR status. Due to small numbers, ER−/PR+ cases (n=6) and their corresponding matched controls were excluded from the joint ER/PR analyses. Levels of each PFAS, which were quantified as mass spectral peak intensities, were classified into quartiles based on the distribution among controls. In addition to conditioning on the matching factors, models were adjusted for age at blood draw, established breast cancer risk factors (age at menarche, age at first live birth and number of live births, age at menopause, duration of MHT use, first-degree family history of female breast cancer, personal history of benign breast disease, body mass index, smoking status, and vigorous physical activity), and variables whose removal resulted in a ≥10% change in the ORs (study center, race/ethnicity, and education). Models were additionally adjusted for natural log-transformed levels of PFOA (for PFOS models) or PFOS (for PFOA models); adjustment for PFOA or PFOS quartiles instead of continuous log-transformed levels yielded no meaningful changes in the results. Missing data on covariates (<2% for all variables except physical activity [8%]; Table S1) were replaced by the most frequent category among controls; for physical activity, a missing category was created. Tests for linear trend across quartiles were performed by modeling quartile-specific median PFOS or PFOA levels among controls as a continuous variable. Heterogeneity of associations across breast tumor subtypes was assessed using the Wald test for a cross-product term between hormone receptor status (ER, PR, or ER/PR status for each matched case-control pair) and serum PFOS or PFOA levels (continuous variable based on quartile-specific medians).

To explore the potential impact of duration of follow-up on our results, we conducted analyses stratified by number of years from blood draw to breast cancer diagnosis, based roughly on the median as a cutoff point (<6 and ≥6 years). Since menstruation is a route of PFAS elimination and PFAS levels are known to increase following menopause,^14,15 we also conducted analyses stratified by time since menopause (≤15 and >15 years), calculated by subtracting age at menopause (estimated using category medians) from age at blood draw. Additional sensitivity analyses were performed by excluding women who became postmenopausal ≤5 years before blood draw (10% of cases and 7% of controls).

All tests were two-sided, and statistical significance was evaluated at P<.05. All analyses were performed using SAS, version 9.4 (SAS Institute Inc., Cary, NC).

3. RESULTS

In this nested case-control study of 621 postmenopausal breast cancer cases and 621 controls, tumor ER, PR, and joint ER/PR status was available for 94%, 87%, and 86% of cases, respectively (participant characteristics presented in Table S1). Participants were on average 64 years of age at blood draw, and the median time from blood draw to case diagnosis was 5.6 years. Serum PFOS and PFOA levels were moderately correlated (Spearman correlation coefficient=0.60 among controls).

We observed little evidence of association between serum PFOS or PFOA levels and postmenopausal breast cancer risk overall (Table 1). However, when we evaluated PFOS associations by hormone receptor status, we observed statistically significant elevated risks of ER+, PR+, and ER+/PR+ tumors in relation to the third vs. lowest quartile (ORs [95% CIs]=1.59 [1.01–2.50], 2.34 [1.29–4.23], and 2.19 [1.21–3.98], respectively) and elevated but slightly weaker ORs for the fourth quartile, while findings were null for hormone receptor-negative tumors (Table 2). We observed an opposite pattern of results for PFOA; associations were null for ER+, PR+, and ER+/PR+ tumors, while modest positive associations with ER−, PR−, ER+/PR−, and ER−/PR− tumors were generally seen in the upper quartiles; risk estimates were not statistically significant, with the exception of an association with PR− tumors for the third quartile (OR=2.05, 95% CI=1.06–3.94). We observed generally stronger associations in these fully-adjusted models accounting for potential confounders compared with conditional models accounting for matching factors only (Tables S2 and S3).

TABLE 1.

Odds ratios (95% confidence intervals) for the associations between serum PFOS and PFOA levels and risk of postmenopausal breast cancer (N = 621 cases and 621 controls)

PFASa	ncontrols	ncases	Model 1b	Model 2c
PFOS
Quartile 1	156	145	1.00 (Ref)	1.00 (Ref)
Quartile 2	155	158	1.19 (0.83–1.70)	1.21 (0.84–1.74)
Quartile 3	155	167	1.35 (0.97–1.90)	1.39 (0.96–1.99)
Quartile 4	155	151	1.12 (0.79–1.59)	1.17 (0.77–1.79)
Ptrendd			.61	.58
PFOA
Quartile 1	156	147	1.00 (Ref)	1.00 (Ref)
Quartile 2	155	148	0.92 (0.65–1.31)	0.91 (0.64–1.30)
Quartile 3	155	162	1.09 (0.77–1.55)	1.07 (0.73–1.55)
Quartile 4	155	164	1.06 (0.75–1.50)	1.01 (0.66–1.55)
Ptrendd			.62	.83

Abbreviations: MHT, menopausal hormone therapy; PFAS, per- and polyfluoroalkyl substances; PFOA, perfluorooctanoate; PFOS, perfluorooctane sulfonate.

^aQuartiles are based on the distribution of PFOS or PFOA levels among controls.

^bConditional logistic regression models adjusted for age at blood draw (continuous; years), study center (Upper Midwest [Wisconsin and Minnesota], West/South [Colorado, Hawaii, Missouri, Utah, and Alabama], East [Washington DC, Michigan, and Pennsylvania]), race/ethnicity (non-Hispanic White, non-Hispanic Black, other), education (high school graduate or less, post-high school training or some college, college graduate or postgraduate), age at menarche (<12, 12–13, ≥14 years), age at first live birth and number of live births (nulliparous, <20 years and ≥1 birth, 20–29 years and 1–2 births, 20–29 years and ≥3 births, ≥30 years and ≥1 birth), age at menopause (<45, 45–49, 50–54, ≥55 years), duration of MHT use (never, ≤1, 2–5, 6–9, ≥10 years), first-degree family history of female breast cancer (no/unknown, yes), personal history of benign breast disease (no/unknown, yes), body mass index (<25, 25 to <30, ≥30 kg/m²), smoking status (never, former, current), and vigorous physical activity (<1, 1–3, ≥4 hours/week, missing). Models conditioned on case-control matched pairs, with controls individually matched to cases on age at baseline (±2 years), date of blood draw (±3 months), and MHT use at baseline.

^cAdjusted for all Model 1 covariates and additionally for natural log-transformed levels of PFOA (for the PFOS model) or PFOS (for the PFOA model).

^dTests for linear trend across quartiles performed by modeling quartile-specific median values of PFOS or PFOA among the controls as a continuous variable.

TABLE 2.

Odds ratios (95% confidence intervals) for the associations between serum PFOS and PFOA levels and risk of postmenopausal breast cancer according to hormone receptor status

PFASb	ER status			PR status			ER/PR statusa
	ER+ (n = 435)	ER− (n = 147)	P het c	PR+ (n = 299)	PR− (n = 241)	P het c	ER+/PR+ (n = 291)	ER+/PR− (n = 98)	ER−/PR− (n = 138)	P het c
PFOS
Quartile 1	1.00 (Ref)	1.00 (Ref)		1.00 (Ref)	1.00 (Ref)		1.00 (Ref)	1.00 (Ref)	1.00 (Ref)
Quartile 2	1.26 (0.81–1.95)	0.98 (0.39–2.47)		1.55 (0.90–2.67)	1.00 (0.52–1.92)		1.46 (0.84–2.54)	0.79 (0.19–3.38)	1.01 (0.38–2.63)
Quartile 3	1.59 (1.01–2.50)	1.13 (0.49–2.62)		2.34 (1.29–4.23)	0.91 (0.50–1.64)		2.19 (1.21–3.98)	0.32 (0.08–1.32)	1.12 (0.48–2.62)
Quartile 4	1.29 (0.77–2.15)	0.52 (0.18–1.55)		1.79 (0.92–3.48)	0.61 (0.29–1.31)		1.89 (0.97–3.69)	0.32 (0.06–1.86)	0.60 (0.19–1.83)
Ptrendd	.39	.20	.71	.14	.15	.66	.08	.12	.34	.25
PFOA
Quartile 1	1.00 (Ref)	1.00 (Ref)		1.00 (Ref)	1.00 (Ref)		1.00 (Ref)	1.00 (Ref)	1.00 (Ref)
Quartile 2	1.07 (0.68–1.66)	0.84 (0.36–1.95)		1.14 (0.66–1.96)	0.90 (0.47–1.70)		1.14 (0.66–1.97)	1.40 (0.32–6.11)	0.90 (0.38–2.10)
Quartile 3	1.01 (0.64–1.61)	2.08 (0.85–5.07)		1.02 (0.57–1.83)	2.05 (1.06–3.94)		0.99 (0.55–1.80)	1.88 (0.55–6.42)	2.23 (0.90–5.54)
Quartile 4	1.03 (0.61–1.75)	1.63 (0.63–4.20)		0.77 (0.39–1.52)	1.48 (0.75–2.93)		0.81 (0.40–1.62)	1.63 (0.45–5.87)	1.62 (0.62–4.23)
Ptrendd	.96	.19	.30	.31	.15	.21	.41	.50	.21	.44

Notes: Odds ratios and 95% confidence intervals were estimated from conditional logistic regression models adjusted for age at blood draw (continuous; years), study center (Upper Midwest [Wisconsin and Minnesota], West/South [Colorado, Hawaii, Missouri, Utah, and Alabama], East [Washington DC, Michigan, and Pennsylvania]), race/ethnicity (non-Hispanic White, non-Hispanic Black, other), education (high school graduate or less, post-high school training or some college, college graduate or postgraduate), age at menarche (<12, 12–13, ≥14 years), age at first live birth and number of live births (nulliparous, <20 years and ≥1 birth, 20–29 years and 1–2 births, 20–29 years and ≥3 births, ≥30 years and ≥1 birth), age at menopause (<45, 45–49, 50–54, ≥55 years), duration of MHT use (never, ≤1, 2–5, 6–9, ≥10 years), first-degree family history of female breast cancer (no/unknown, yes), personal history of benign breast disease (no/unknown, yes), body mass index (<25, 25 to <30, ≥30 kg/m²), smoking status (never, former, current), vigorous physical activity (<1, 1–3, ≥4 hours/week, missing), and natural log-transformed levels of PFOA (for PFOS models) or PFOS (for PFOA models). Models conditioned on case-control matched pairs, with controls individually matched to cases on age at baseline (±2 years), date of blood draw (±3 months), and MHT use at baseline.

Bold indicates statistical significance at P < .05.

Abbreviations: ER, estrogen receptor; MHT, menopausal hormone therapy; PFAS, per- and polyfluoroalkyl substances; PFOA, perfluorooctanoate; PFOS, perfluorooctane sulfonate; PR, progesterone receptor.

^aER−/PR+ cases were not analyzed due to small numbers (n = 6).

^bQuartiles are based on the distribution of PFOS or PFOA levels among controls.

^cP-values for heterogeneity of associations across tumor subtypes, estimated using a Wald test for a cross-product term between hormone receptor status (ER, PR, or ER/PR status for each matched case-control pair) and serum PFOS or PFOA levels (continuous variable based on quartile-specific median values).

^dTests for linear trend across quartiles performed by modeling quartile-specific median values of PFOS or PFOA among the controls as a continuous variable.

Associations of PFOS and PFOA with overall breast cancer risk remained null in analyses stratified by time to diagnosis or time since menopause (Table S4). However, we observed stronger positive associations between serum PFOS levels and risks of hormone receptor-positive tumors diagnosed ≥6 years after blood draw (Tables 3 and S5), including a statistically significant exposure-response trend for ER+/PR+ tumors (ORs=4.72 and 3.67 for the third and fourth quartiles, respectively; P_trend=.02). For PFOA, the positive association with ER− tumors was more prominent for cases diagnosed within 6 years, whereas the association with PR− tumors appeared stronger for those diagnosed ≥6 years after blood draw (Table S5); however, these results should be interpreted with caution due to limited sample sizes. In general, no notable differences by time since menopause were observed for associations with hormone receptor-positive tumors (Tables 3 and S6), while a stronger positive association between PFOA and PR− tumors was observed among those with >15 (vs. ≤15) years between menopause and blood draw (P_{heterogeneity}=.001; Table S6). Finally, analyses excluding women who became menopausal ≤5 years before blood draw yielded similar findings compared to our main analyses, though with stronger associations between PFOS and ER+/PR+ tumors in the upper quartiles and a statistically significant exposure-response trend (ORs [95% CIs]=2.27 [1.18–4.38] and 2.20 [1.05–4.61] for the third and fourth quartiles, respectively; P_trend=.03; Table S7).

TABLE 3.

Odds ratios (95% confidence intervals) for the associations between serum PFOS and PFOA levels and risk of ER+/PR+ breast cancer, stratified by time to diagnosis or time since menopause

PFASa	Time from blood draw to diagnosis			Time from menopause to blood draw
	<6 years (n = 152)	≥6 years (n = 139)	P het b	≤15 years (n = 146)	>15 years (n = 145)	P het b
PFOS
Quartile 1	1.00 (Ref)	1.00 (Ref)		1.00 (Ref)	1.00 (Ref)
Quartile 2	1.40 (0.61–3.19)	1.48 (0.59–3.73)		1.42 (0.34–5.99)	1.27 (0.42–3.81)
Quartile 3	2.02 (0.86–4.73)	4.72 (1.66–13.47)		2.45 (0.40–15.13)	3.28 (0.93–11.56)
Quartile 4	1.28 (0.47–3.42)	3.67 (1.20–11.25)		2.72 (0.45–16.56)	1.89 (0.53–6.80)
Ptrendc	.66	.02	.45	.25	.26	.36
PFOA
Quartile 1	1.00 (Ref)	1.00 (Ref)		1.00 (Ref)	1.00 (Ref)
Quartile 2	1.14 (0.49–2.66)	1.31 (0.53–3.24)		1.47 (0.31–7.02)	1.00 (0.34–2.92)
Quartile 3	1.39 (0.58–3.30)	1.02 (0.36–2.91)		2.41 (0.36–16.10)	0.92 (0.29–2.86)
Quartile 4	1.50 (0.52–4.38)	0.53 (0.17–1.70)		0.36 (0.03–3.75)	0.81 (0.20–3.28)
Ptrendc	.48	.15	.46	.22	.74	.81

Notes: Odds ratios and 95% confidence intervals were estimated from conditional logistic regression models adjusted for age at blood draw (continuous; years), study center (Upper Midwest [Wisconsin and Minnesota], West/South [Colorado, Hawaii, Missouri, Utah, and Alabama], East [Washington DC, Michigan, and Pennsylvania]), race/ethnicity (non-Hispanic White, non-Hispanic Black, other), education (high school graduate or less, post-high school training or some college, college graduate or postgraduate), age at menarche (<12, 12–13, ≥14 years), age at first live birth and number of live births (nulliparous, <20 years and ≥1 birth, 20–29 years and 1–2 births, 20–29 years and ≥3 births, ≥30 years and ≥1 birth), age at menopause (<45, 45–49, 50–54, ≥55 years), duration of MHT use (never, ≤1, 2–5, 6–9, ≥10 years), first-degree family history of female breast cancer (no/unknown, yes), personal history of benign breast disease (no/unknown, yes), body mass index (<25, 25 to <30, ≥30 kg/m²), smoking status (never, former, current), vigorous physical activity (<1, 1–3, ≥4 hours/week, missing), and natural log-transformed levels of PFOA (for PFOS models) or PFOS (for PFOA models). Models conditioned on case-control matched pairs, with controls individually matched to cases on age at baseline (±2 years), date of blood draw (±3 months), and MHT use at baseline.

Bold indicates statistical significance at P < .05.

Abbreviations: ER, estrogen receptor; MHT, menopausal hormone therapy; PFAS, per- and polyfluoroalkyl substances; PFOA, perfluorooctanoate; PFOS, perfluorooctane sulfonate; PR, progesterone receptor.

^aQuartiles are based on the distribution of PFOS or PFOA levels among controls.

^bP-values for heterogeneity of associations across strata, estimated using a Wald test for a cross-product term between time to diagnosis or time since menopause and serum PFOS or PFOA levels (continuous variable based on quartile-specific median values).

^cTests for linear trend across quartiles performed by modeling quartile-specific median values of PFOS or PFOA among the controls as a continuous variable.

4. DISCUSSION

To our knowledge, this is the largest prospective study to date to evaluate associations between prediagnostic serum PFOS and PFOA levels and breast cancer risk and the first to examine these associations by joint ER/PR status. Consistent with results reported by Mancini et al. from the E3N study,⁸ our findings from the PLCO Cancer Screening Trial suggest that higher levels of serum PFOS, but not PFOA, may be associated with elevated risks of hormone receptor-positive breast tumors among postmenopausal women, and provide novel suggestive evidence of a positive association between PFOS and double-positive (ER+/PR+) but not single-positive (ER+/PR−) tumors. Our results suggesting positive associations between serum PFOA and hormone receptor-negative tumors were also somewhat consistent with those from Mancini et al., although they observed elevated risks of ER− and PR− tumors for the second quartile of both PFOA and PFOS.⁸ Taken together, these findings suggest that while PFOS and PFOA may not be associated with breast cancer overall, they may differentially affect risks of different breast cancer subtypes, thus highlighting the importance of considering hormone receptor status when evaluating the role of PFAS in breast cancer etiology.

Notably, we observed stronger positive associations between PFOS and hormone receptor-positive tumors for cases with a longer duration between blood draw and diagnosis, indicating that reverse causation is unlikely to explain our findings. These results are also consistent with the potential long latency between exposure to endocrine-disrupting chemicals (e.g., PFAS) and breast cancer occurrence¹⁶ and further strengthened our findings, particularly for PFOS and receptor-positive tumors. Moreover, since menstruation is a route of excretion, PFAS levels increase over time following menopause,^14,15 which may complicate their associations with breast cancer. However, time since menopause appeared to have minimal impact on our main findings; in analyses excluding recently menopausal women, the observed associations were generally similar, although for ER+/PR+ tumors the association with PFOS levels in the top quartile and the exposure-response trend became statistically significant.

In vitro and in vivo experiments have demonstrated the ability of several PFAS, including PFOS and PFOA, to modulate hormone receptor (e.g., ER) activity and signaling,¹⁷ which add biologic plausibility for the observed associations, especially for PFOS and hormone receptor-positive tumors. Beyond hormone receptor-mediated effects, both PFOS and PFOA have been suggested to contribute to carcinogenesis through other mechanisms, such as oxidative stress and immunosuppression.¹⁷ It is unclear why PFOS and PFOA exhibited different patterns of associations with breast cancer subtypes. While both chemicals consist of an eight-carbon chain and have similarly long half-lives in human blood (~3–5 years),^18,19 differences in functional groups (sulfonic vs. carboxylic acid) and toxicokinetic properties may have contributed to different modes of action.²⁰ Experimental studies focusing on specific breast cancer subtypes are needed to further explore potential mechanisms underlying PFOS and PFOA carcinogenicity. Several previous epidemiologic studies have also evaluated associations between PFAS and breast cancer by hormone receptor status; however, they relied on post-diagnostic measurements of PFAS and reported inconsistent findings.^21–24 Furthermore, a recent case-cohort study in China measured plasma levels of six different PFAS and reported positive associations with breast cancer risk for PFOA, but not with PFOS or most other PFAS assessed; however, information on hormone receptor status was not available.²⁵ Additional prospective studies incorporating tumor hormone receptor status are needed to confirm findings from our study and Mancini et al.⁸ in order to clarify the relationship between PFOS, PFOA, and breast cancer risk, and to evaluate associations with other PFAS.

A major strength of our study was the larger sample size compared to previous prospective studies of circulating PFAS and breast cancer^6,8,25 and the availability of information on ER and PR status for most cases, which allowed us to evaluate associations with specific breast cancer subtypes defined by individual or joint hormone receptor status. Several study limitations should also be noted. First, serum PFOS and PFOA levels were not quantified using standard targeted methods, which precluded direct comparisons of absolute concentrations with other studies. However, as stated earlier, we observed strong correlations between measurements by Metabolon (untargeted metabolomics) and those by the CDC laboratory (targeted assay) among PLCO participants with PFOS and PFOA measured using both methods (unpublished data). Second, PFOS and PFOA were measured in serum samples collected at a single point in time; however, these measurements likely reflect long-term exposures given the long serum elimination half-lives of these chemicals.^18,19 Third, despite the relatively large sample size, our study had limited statistical power for analyses of rarer breast cancer subtypes, especially in stratified analyses. Finally, because all women were postmenopausal at blood draw in our study, our findings may not be generalizable to premenopausal breast cancer.

In conclusion, our study findings contribute to evidence supporting a positive association between serum PFOS levels and hormone receptor-positive breast tumors, and possibly between PFOA and hormone receptor-negative tumors. Future prospective studies should seek to incorporate tumor hormone receptor status where possible to better evaluate these highly persistent chemicals as potential breast carcinogens.

Novelty and Impact.

Perfluorooctane sulfonate (PFOS) and perfluorooctanoate (PFOA) are synthetic chemicals with widespread use in industrial applications and consumer products. They also are endocrine disruptors with suspected links to cancer. Here, the authors investigated potential associations between prediagnostic serum PFOS and PFOA levels and risk of breast cancer subtypes among postmenopausal women. No association was found between PFOS or PFOA and breast cancer risk overall. However, PFOS levels were positively associated with hormone receptor-positive breast tumors, and modest positive associations were observed between PFOA levels and receptor-negative tumors. The findings warrant further investigation in studies incorporating breast tumor hormone receptor status.

Abbreviations:

CDC

Centers for Disease Control and Prevention

CI

confidence interval

ER

estrogen receptor

IARC

International Agency for Research on Cancer

MHT

menopausal hormone therapy

OR

odds ratio

PFAS

per- and polyfluoroalkyl substances

PFOA

perfluorooctanoate

PFOS

perfluorooctane sulfonate

PLCO

Prostate, Lung, Colorectal and Ovarian

PR

progesterone receptor

DATA AVAILABILITY STATEMENT

Data from the PLCO Cancer Screening Trial are available from the National Cancer Institute Cancer Data Access System (https://cdas.cancer.gov/plco/) upon application and approval. Further information is available from the corresponding author upon request.