Phthalates and Phenols, Leukocyte Telomere Length, and Breast Cancer Risk and Mortality in the Long Island Breast Cancer Study Project

Phthalates and phenols from the environment are synthetic chemicals that have been widely used as plasticizers, materials in medical devices and personal care products, food preservatives, and pesticides (1, 2). Phthalates and phenols are also endocrine-disrupting chemicals, have been inconsistently linked to breast cancer risk and mortality in epidemiologic studies (3–8). Many studies suggest that there are no adverse effects of low level or dietary exposure to phthalates or phenols (4, 6–9). That said, some phthalates and phenols were associated with increased breast cancer risk or mortality following breast cancer after stratifying study populations by factors such as menopausal status (3, 4), time after diagnosis (8), or body mass index (BMI; refs. 6, 8). In addition, some observed adverse health outcomes such as early puberty (10) are related to environmental exposure to phthalates or phenols (11, 12). Differing results for the total study population and for stratified populations suggest that potential effect modifiers exist between phthalate or phenol exposures and breast cancer risk or mortality.

A telomere is a segment of repeated DNA located at the end of a chromosome that protects it from damage. Leukocyte telomeres shorten with cell division so that length has been regarded as a biomarker of biological aging. Controlling for age, a shorter leukocyte telomere length (LTL) is associated with increased risk for a broad spectrum of adverse health outcomes (10, 13) including breast cancer (14–16). However, the direction of this effect is mixed because either shorter or longer LTL has been reported to be associated with an increased risk of breast cancer (14–16). We thereby hypothesized that LTL may play two possible roles in the association between environmental chemicals and breast cancer: as a mediator or as an effect modifier. Telomerase expression has been considered as a potential mechanism of the toxic effects of exposure to environmental chemicals (17, 18). Thus it has the potential to be a mediator of the association. In contrast, telomere shortening has been found to be a biomarker of increased body mass (19) and delayed menopause (20), which have been reported to be effect modifiers in the studies of phthalates, phenols, and breast cancer (3, 4, 6, 7). Therefore, in this study, we aimed to evaluate exposure to phthalates and phenols, LTL, and breast cancer by considering LTL both as a mediator and as an effect modifier. The results of this investigation can contribute to the knowledge of the role of biological aging played in the environmental origins of breast cancer.

This study utilized existing interview and laboratory data from the Long Island Breast Cancer Study Project (LIBCSP). As previously described in detail (21), the LIBCSP was initiated as a population-based case–control study in 1996 to 1997 and then continued with a follow-up of cases. Adult, English-speaking women newly diagnosed with breast cancer from the Long Island counties of Nassau and Suffolk in New York State were invited to join the study. Control women were selected among the English-speaking female residents of Nassau and Suffolk counties who did not have a personal history of breast cancer and whose ages were matched to cases within 5-year age distributions. As a result, 1,508 incident breast cases and 1,556 controls participated in the LIBCSP. The LIBCSP study protocol was approved by the Institutional Review Boards of all participating institutions, and written informed consent was obtained from participants prior to data collection.

The levels of phthalate and phenol analytes were measured in spot urine samples collected from breast cancer cases around the time of diagnosis and from controls at enrollment. In earlier work in the LIBCSP (6, 7), 11 phthalate metabolites [mono-ethyl phthalate (MEP), mono-n-butylphthalate phthalate (MnBP), mono-isobutyl phthalate (MiBP), mono (3-carboxypropyl) phthalate (MCPP), mono-Benzyl phthalate (MBzP), monoethylhexyl phthalate (MEHP), mono (2-ethyl-5-oxohexyl) phthalate (MEOHP), mono (2-ethyl-5-hydroxyhexyl) phthalate (MEHHP), mono (2-ethyl-5-carboxypentyl) phthalate (MECPP), monocarboxyoctyl phthalate (MCOP), and monocarboxy-isononyl phthalate (MCNP)] and seven phenols [2,5-dichlorophenol, benzophenone-3, bis-phenol A (BPA), methylparaben, propylparaben, butylparaben, and triclosan] were measured in urine samples using high-performance liquid chromatography electrospray ionization-isotope-dilution tandem mass spectrometry by the CDC laboratory. Laboratory methods have been described previously (6, 7). We followed the methods described in Teitelbaum and colleagues (22) to estimate the levels of di(2-ethylhexyl)phthalate (DEHP), lower molecular-weight phthalate (LMWP), and higher molecular-weight phthalate (HMWP) metabolites. ∑DEHP was estimated as the molar sum of MEHP, MEHHP, MEOHP, and MECPP. ∑LMWP was calculated as the molar sum of lower weight phthalates (MEP, MnBP, and MiBP) and ∑HMWP was calculated as the molar sum of higher weight phthalates (MBzP, MEHP, MEOHP, MEHHP, MECPP) separately. ∑Parabens was calculated as the molar sum of methylparaben, propylparaben, and butylparaben. The urinary concentrations of phthalate and phenol analytes were corrected for creatinine concentration (μg/g creatinine) before statistical analyses.

LTL was measured from the leukocyte DNA in blood samples collected from study participants at the same time as urine was collected. LTL was measured as the relative ratio of telomere (T) repeat copy number to a single-copy reference gene (S) copy number (T/S ratio) and calibrated with a five-point standard curve. The single-copy gene used as the reference gene encodes acidic ribosomal phosphoprotein P0 (36B4). Data were standardized against a pooled anonymous human leucocyte DNA sample. Details about LTL measurement in the LIBCSP have been described in Shen and colleagues (15).

Potential confounders of the associations between phthalates and phenols and breast cancer risk and mortality were selected from the published literature and previous LIBCSP publications (6, 7, 15). Covariates include age at diagnosis/enrollment (continuous), income (<$24,999, $25,000–$49,999, ≥$50,000), education (<high school, or high school graduate), BMI at diagnosis/enrollment (continuous), menopause status (premenopause, perimenopause, or postmenopause), menarche <12 years old (yes/no), parity and lactation (not parous, parous but no lactation, or parous and lactated), ever use hormone replacement therapy (yes/no), ever use oral contraceptive (yes/no), and smoking status (never, current, or past smoker). Notably, race and ethnicity were not considered because the majority (90%) of our study participants were non-Hispanic White women.

Participants of LIBCSP with available exposure, outcome, and covariates data were included in this analysis. Descriptive statistics were calculated for the variables examined and calculated separately for the original LIBCSP and the participants included in this study. Both phthalate and phenol analytes and LTL (T/S ratio) were natural log transformed and included as continuous variables in all statistical models.

We first examined the mediation hypothesis. The majority of phthalates and phenols were not significantly associated with LTL, thus there was no evidence of a mediation effect of LTL between phthalates and phenols and breast cancer risk nor mortality. All further analyses examined LTL as an effect modifier.

For the case–control analyses, the associations between individual phthalate or phenol analytes and breast cancer were estimated using multivariable logistic regression models including a cross-product interaction terms between phthalate/phenol analytes and continuous LTL. We calculated the ORs and 95% confidence intervals (CI) of breast cancer for one-unit (ln μg/g creatinine) increases of phthalate/phenol analytes at the 10th, 50th, and 90th percentile values of LTL given that LTL is a continuous variable and there is no standard for categorizing it.

For survival analyses, associations between phthalate or phenol analytes and LTL were analyzed separately for all-cause mortality and breast cancer-specific mortality. The proportional hazards assumption was examined using Schoenfeld residual for all study variables. Age at diagnosis and race violated the proportional hazards assumption in all-cause mortality models. Therefore, associations between phthalates/phenol analyte, LTL, and all-cause mortality were investigated using parametric survival models with Weibull distribution design. No violations were found in breast cancer–related mortality models. Associations between phthalates/phenol analyte, LTL, and breast cancer–specific mortality were investigated using Cox proportional hazards models. HRs and 95% CIs for the 10th, 50th, and 90th percentile values of LTL were estimated for each model.

We conducted sensitivity analyses to test the robustness of the modifying effect of LTL in two scenarios. In the first scenario, we replaced the interaction of LTL with an interaction of age to determine whether the identified modification effect of LTL is a surrogate of the modification effect of chronologic aging, given these two factors are closely related. In the second scenario, we excluded women had started breast cancer treatment (hormonal, chemo, or radiation) prior to biological sample donation since these therapies may change LTL.

All analyses were conducted using R version 3.6.3 (R Core Team 2014; ref. 23). Logistic regression was conducted using the package stats. ORs or HRs at a sequence of LTL values were calculated and illustrated using the interplot package. Parametric survival and Cox proportional hazards models were conducted using the package survival.

Characteristics of women included in this study and the entire LIBCSP population are reported in Table 1. A total of 1,268 (41%) women (687 breast cancer cases and 581 controls) had both urinary phthalate/phenol analytes and blood LTL measurements, and 1,158 (37%) women had complete data on all covariates. Cases were more likely than controls to donate adequate amounts of blood and urine samples (54% vs. 49% in controls, P < 0.001). Missing data were observed in several variables (Table 1). However, missing rates were similar in the study subset compared with the whole LIBCSP. Therefore, it was unlikely that the missing data in our study would bias the results.

ORs for a one-unit (ln μg/g creatinine) increase in analytes and breast cancer risk were estimated from the logistic regression models with an interaction between individual analyte and LTL. The ORs were estimated at LTL's 10th, 50th, and 90th percentiles (Table 2). Of the 22 phthalate and phenol analytes examined, 11 associations with breast cancer risk were modified by LTL (P < 0.05). For the majority of phthalate and phenol analytes, breast cancer risk was inversely associated with the metabolites when LTL were shorter and positively associated with phthalate or phenol analytes when LTL were longer.

Figure 1 presents the associations of phthalate/phenol analytes and breast cancer risk modified by LTL. In each panel, when the area of the 95% CI band is completely above or below the line of OR = 1, it indicates a statistically significant OR (P < 0.05). For the majority (17/22) of the phthalate/phenol analytes, concentrations were inversely associated with breast cancer risk at shorter LTL (T/S ratio <1) and became positively associated with breast cancer risk at longer LTL (T/S ratio >1). The trend of ORs and width of the 95% CIs were similar among the following phthalate analytes: MnBP, MBzP, MCPP, MECPP, MEHHP, MEHP, MEOHP, ∑DEHP, and ∑HMWP. Five analytes had different trends: MiBP, MEP, LMWP, 2,5-dichlorophenol, and BPA. MiBP, MEP, LMWP, and BPA were not associated with breast cancer risk at any LTL level. The OR for 2,5-dichlorophenol and breast cancer risk decreased with increasing LTL, significantly at relatively high LTL (T/S ratio >1).

The associations between urinary concentration of phthalates and phenols and breast cancer mortalities at the 10th, 50th, and 90th percentile levels of LTL are presented in the Supplementary Materials and Methods (Supplementary Table S1) for all-cause mortality, and Supplementary Table S2 for breast cancer–specific mortality. The interaction term of LTL was not statistically significant in any of the models. The phthalate and phenol analytes were not associated with mortalities at the three levels of LTL, including the two analytes (MEHP and propylparaben) found to be significant in our previous LIBCSP studies (6, 7). However, because only two of all analytes investigated were weakly associated with risk, we do not consider the association between phthalates/phenols and breast cancer mortalities to be established. Therefore, even though we did not observe a significant association after adjusting for an interaction term of LTL, we still consider the results to be consistent with our previous results.

In sensitivity analysis for the case–control study, we replaced the interaction terms of LTL with interaction terms of age at enrollment. The ORs for one unit (ln μg/g creatinine) increases in phthalate/phenol analytes and breast cancer risk by the age at enrollment are illustrated in Supplementary Fig. S1, and the estimated ORs and 95% CIs are presented in Supplementary Materials and Methods (Supplementary Table S3). For consistency with the layout of Fig. 1, the x-axis in Supplementary Fig. S1 starts at higher age because LTL is known to shorten with age. Compared with the modifying effect of LTL (Fig. 1; Supplementary Fig. S1), many of the interactions observed in the LTL models were not replicated in the sensitivity analysis (age models). For example, of 22 urinary analytes examined, only the association between MCNP and breast cancer risk was significantly modified by age (P_interaction: 0.03). This association between MCNP and breast cancer risk was not significantly modified by LTL (Table 2). In addition, the width of 95% CI bands in Supplementary Fig. S1 and in Fig. 1 suggested that no consensus was observed between the modifying effects of age and LTL.

Another sensitivity analysis also shows consistent results (Supplementary Table S4; Supplementary Fig. S2) from the main results. In the main results, nine phthalate metabolites were associated with breast cancer risk after considering multiple tests (Table 2), and six of the nine metabolite models were not materially different after excluding women who had breast cancer treatment (Supplementary Table S4). Three metabolites were no longer significantly associated with breast cancer risk after this exclusion. But the ORs were similar for modification by LTL comparing Fig. 1 and Supplementary Fig. S2. Likely, the attenuation of the results is due to the reduced sample size after exclusion.

We identified a significant modifying effect of LTL on the associations between multiple urinary and breast cancer risk in the LIBCSP. We found that 11 (9 after correcting for multiple tests) of the 22 investigated biomarkers' associations with breast cancer risk were significantly modified by LTL. This result is in the contrast to the results in the previous LIBCSP studies (6, 7) that did not consider biological aging as an effect modifier; only methylparaben and summed (∑) parabens were significantly associated with breast cancer risk (ORs were 1.09 for both methylparaben and ∑parabens). The direction of the associations found in our study suggests that women with longer LTL had increased risks to breast cancer associated with phthalates' and phenols' exposure. To our knowledge, this is the first study to find a modifying effect of a cellular aging biomarker on the association between exposure to environmental chemicals and breast cancer risk. No significant modifying effect of LTL was found for phthalate and phenol analytes and all-cause or breast cancer–specific mortality among the women with breast cancer.

In our results, women who were less biologically aged (longer LTL) were sensitive to the increased exposure of environmental chemical with respect to developing breast cancer. This finding may expand the definition of vulnerable populations in environmental health studies. Traditionally, chronologically older people, who are also biologically aged, are considered as more vulnerable to environmental exposures than the general population (24). In our results, women who were biologically older (shorter LTL) had an inverse association between phthalate and phenol analytes and breast cancer. Also in previous studies, an inverse association between phthalates and phenols and breast cancer has been repeatedly reported (4, 5, 25), including the two LIBCSP studies (6, 7). This inverse association itself is supported by experimental studies reporting protective effects of phthalates or phenols on breast cancer at a particular epigenetic status (26–30), but the underlying mechanism for the modifying effect of shorter LTL is not unknown. Our previous study reported an association between shorter LTL and increased breast cancer risk for women with low antioxidant intake (15), which suggested that factors related to oxidative stress might modify the association between LTL and breast cancer. A similar explanation might be made for phthalate and phenol exposures, which have been associated with oxidative stress (31). Hormonal mechanisms might also be considered, as aging is related to lower sex hormones (32). Endocrine activity of phthalate metabolites and phenols might affect sex hormone activity, perhaps the major risk factor for breast cancer (33). Further investigation should address biological aging as a modifying factor for endocrine disrupting chemicals such as phthalates and phenols as risk factors for breast cancer.

Our findings suggest that biological aging should be considered in future investigations of environmental exposures. Without investigating the interacting effect with biological aging, the carcinogenic effect of chemicals may be underestimated. Another important finding is the distinctly different results for the modification effects of LTL compared with age. The modifying effect of LTL seems independent from the effect of age, even though the shortening of LTL is highly correlated with a person's chronologic age. In addition, age's modifying effects were not substantial—only a limited number of phthalate or phenol analytes were significantly modified and the directions were not consistent across those analytes. Thus, biological aging may have stronger effects than chronologic aging on modifying the associations investigated in this study.

This study has several limitations. One is the possibility of reverse causality, often a consideration in cancer epidemiologic studies because the unrecognized disease may result in changes prior to the diagnosis of cancer. In our study, LTL of breast cancer cases was measured from blood samples collected shortly after the treatment began, therefore LTL may be affected by the development of breast cancer prior to that (34). The short half-lives of phthalate and phenol analytes pose a concern because a single spot urine sample was used to characterize exposure in the LIBCSP. However, multiple studies (35–37) have validated that spot urine sample represent chronic exposure to phthalates or phenols because the source of those chemicals is likely relatively constant (38, 39). Other limitations include the laboratory measurement of LTL. First, the length of the telomere was measured as a relative T/S ratio, which prevented us from comparing T/S ratios measured using different reference genes in other investigations. Moreover, the blood LTL measured in our study may not represent the telomere length at a specific organ such as breast (40). Future studies focusing on organ-specific telomere length are required to confirm the modifying role of telomere length plays in breast cancer.

LTL may be a notable effect modifier of the associations between urinary phthalate and phenol analytes and risk of breast cancer. Associations between our analytes and breast cancer risk are generally positive at longer LTL and negative at shorter LTL. Our study adds evidence to the hypothesis that the risk of breast cancer is influenced by both environmental chemicals and biological aging, and the effect may be stronger when women are less biologically aged.

X. Zhang and S.L. Teitelbaum report grants from NIH during the conduct of the study. A.I. Neugut reports grants and personal fees from Otsuka; personal fees from United Biosource Corp., GlaxoSmithKline, Eisai, Hospira, and EHE Intl. outside the submitted work. No disclosures were reported by the other authors.

X. Zhang: Conceptualization, formal analysis, methodology, writing–original draft, writing–review and editing. M.S. Wolff: Resources, methodology, writing–review and editing. J. Shen: Resources, data curation, writing–review and editing. H. Parada Jr: Resources, methodology, writing–review and editing. R.M. Santella: Resources, methodology, writing–review and editing. A.I. Neugut: Resources, supervision, methodology, writing–review and editing. J. Chen: Methodology, writing–review and editing. S.L. Teitelbaum: Resources, data curation, supervision, methodology, writing–review and editing.

We acknowledge the principal investigator of the Long Island Breast Cancer Study Project: Marilie Gammon, without whom this project would not be accomplished. We acknowledge the technical assistance of the Antonia Calafat, Manori Silva, Ella Samandar, James Preau, Xiaoyun Ye, and John Reidy (CDC, Atlanta, GA) in measuring the urinary concentrations of phthalate and phenol analytes.

This study was supported in part by funds from the NCI and the National Institute of Environmental Health Sciences (U01CA/ES66572, U01CA66572, U01ES019459, K01ES012645, UG3OD023320-02, and UH3OD023337). H. Parada Jr. was supported by the NCI (K01 CA234317), the SDSU/UCSD Comprehensive Cancer Center Partnership (U54 CA132384 and U54 CA132379), and by the Alzheimer's Disease Resource Center for Advancing Minority Aging Research at the University of California San Diego (P30AG059299). Xueying Zhang was supported by NIH ECHO programs (UG3OD023320 and UH3OD023337) and the Mount Sinai Transdisciplinary Center on Early Environmental Exposures (P30ES023515).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.