Abstract
The role of metals in breast cancer is of interest because of their carcinogenic and endocrine-disrupting capabilities. Evidence from epidemiologic studies remains elusive, and prior studies have not investigated metal mixtures. In a case cohort nested within the Sister Study (enrolled in 2003–2009; followed through September 2017), we measured concentrations of 15 metals in toenails collected at enrollment in a race/ethnicity-stratified sample of 1,495 cases and a subcohort of 1,605 women. We estimated hazard ratios and 95% confidence intervals for each metal using Cox regression and robust variance. We used quantile g-computation to estimate the joint association between multiple metals and breast cancer risk. The average duration of follow-up was 7.5 years. There was little evidence supporting an association between individual metals and breast cancer. An exception was molybdenum, which was associated with reduced incidence of overall breast cancer risk (third tertile vs. first tertile: hazard ratio = 0.82, 95% confidence interval: 0.67, 1.00). An inverse association for antimony was observed among non-Hispanic Black women. Predefined groups of metals (all metals, nonessential metals, essential metals, and metalloestrogens) were not strongly associated with breast cancer. This study offers little support for metals, individually or as mixtures, as risk factors for breast cancer. Mechanisms for inverse associations with some metals warrant further study.
Keywords: breast cancer, metals, mixtures
Abbreviations
- CI
confidence interval
- ER
estrogen receptor
- HR
hazard ratio
- IQR
interquartile range
While numerous lifestyle and reproductive risk factors have been identified for breast cancer (1), the role of environmental exposures remains unestablished. The International Agency for Research on Cancer (Lyon, France) has classified numerous metals as either established (arsenic, cadmium, chromium, and nickel) or probable (lead) human carcinogens (2, 3), and certain metals (aluminum, cadmium, copper, cobalt, nickel, lead, tin, and chromium) have been shown to mimic the activity of estrogen and activate estrogen receptors (ERs) (4–6).
Most previous studies of biomarker-measured metals and breast cancer have considered lead, cadmium, or arsenic (7–17), with the strongest evidence being seen for cadmium (18). Prospective studies have been limited (7, 9, 13, 17, 19–22), and little is known about the impact of other metals. Previous studies have largely not considered associations by race/ethnicity. Historical and current social and political factors have resulted in documented differences in the sources and burden of exposure to environmental chemicals, including metals, across racial/ethnic groups. Racial/ethnic minorities have been shown to live closer to industrial sources of pollution (23), to live in areas with higher arsenic and lead soil contamination (24), and to have higher biomarker concentrations of cadmium, lead, and mercury (25). Because of racism and residential segregation, race/ethnicity also correlates with a number of other factors, such as nonmetallic environmental exposures or social stressors at the community and/or individual level, which may enhance susceptibility to metal exposure (26–28). For example, it has been documented that stressors can amplify vulnerability to contaminants by increasing or decreasing absorption, detoxification, or recovery from environmental exposures, as well as influence the development of disease (28). Therefore, examining the relationship between metals and breast cancer in a diverse population and stratifying analyses by race/ethnicity are important considerations.
Although sources or concentrations of certain metals tend to be correlated, previous studies of metal biomarkers have not addressed the impact of complex mixtures of metals on breast cancer. Although individual metals could confer a modest increase in risk, a stronger joint effect of exposure to multiple metals is plausible. Determining the association between multiple metals and breast cancer is critical to understanding the potential impact of interventions that may affect exposure to multiple metals. A few classifications of metals that may be relevant to consider as groups include essential metals, which have a known physiological function in humans (chromium, cobalt, copper, iron, manganese, molybdenum, nickel, selenium, and zinc); nonessential metals, which have no known physiological function in humans (aluminum, antimony, arsenic, cadmium, lead, and tin); and metalloestrogens, which can mimic the activity of estrogens and activate the ERs (aluminum, antimony, cadmium, chromium, cobalt, copper, lead, nickel, and tin) (6, 29).
Our objective in this study was to examine whether metal concentrations measured in toenails were associated with subsequent hazard of breast cancer, overall and by race/ethnicity, ER status, and menopausal status. We further evaluated whether increasing the concentrations of multiple metals in predefined groups was related to breast cancer.
METHODS
Study design and population
The Sister Study is an ongoing prospective cohort study of 50,884 women living in the United States and Puerto Rico who were aged 35–74 years when they enrolled (2003–2009) (30). Participants were eligible if they had a sister who had been diagnosed with breast cancer but had no prior breast cancer themselves at enrollment.
Computer-assisted telephone interviews and written questionnaires were completed at enrollment. Participants provided toenail clippings from each toe, which were collected during enrollment. Participants have completed annual health updates and follow-up questionnaires for assessment of changes in health and risk factor information. Approximately 90% of participants responded to the most recent detailed follow-up questionnaire (September 2014–August 2016) preceding the end of follow-up for this analysis.
Medical records were obtained for 82% of the women reporting a breast cancer diagnosis. There was high agreement (99%) between self-reports and medical records, including information on ER status, so self-reports were used when medical records were not obtained (31).
All participants provided written informed consent. The Sister Study was approved by the institutional review board of the National Institute of Environmental Health Sciences (Research Triangle Park, North Carolina). We utilized data from Sister Study Data Release 7.2, with follow-up through September 15, 2017.
We selected a race/ethnicity-stratified case cohort in which to measure toenail metal concentrations (see Web Figure 1, available at https://doi.org/10.1093/aje/kwab204). This consisted of 1,499 cases with invasive or ductal carcinoma in situ breast cancer (245 non-Hispanic Black and 1,254 non-Hispanic White) and a random subcohort of 1,607 women (542 non-Hispanic Black and 1,065 non-Hispanic White). Because 107 women from the subcohort (42 non-Hispanic Black and 65 non-Hispanic White) developed breast cancer and are counted in both the case and subcohort numbers, the case-cohort sample size was 2,999 individuals. Three individuals were excluded because of incomplete processing of the toenails. After excluding women missing information on parity/breastfeeding (n = 2) or smoking (n = 1), there were 1,495 cases (243 non-Hispanic Black and 1,252 non-Hispanic White) and 1,605 subcohort women (541 non-Hispanic Black and 1,064 non-Hispanic White); 107 women from the subcohort who developed breast cancer were counted in both groups, leading to a final analytical sample size of 2,993.
Exposure assessment
The processing and measurement of metal concentrations in toenails was conducted at RTI International (Research Triangle Park, North Carolina). Concentrations of 15 metals (aluminum, arsenic, cadmium, cobalt, chromium, copper, iron, manganese, molybdenum, nickel, lead, antimony, selenium, tin, and zinc) were measured in discrete analytical batches using a Thermo Fisher Scientific (Waltham, Massachusetts) iCAP Q quadrupole inductively coupled plasma–mass spectrometry system. Additional details on specimen cleaning, processing, and quality control are provided in the Web Appendix. The numbers of samples below the limit of quantification and the limit of detection are provided in Web Table 1.
We considered metal concentrations modeled for an interquartile range (IQR) increase and with 3 categories based on tertile cutpoints using the inverse-sampling-probability–weighted exposure distribution in the overall subcohort. We used the cutpoints calculated in the overall subcohort for all analyses for comparability across subgroups.
We identified outliers or influential observations by calculating DFBETA statistics (32) and examining, for each metal, a plot of the DFBETAs by toenail concentration. There appeared to be 1–8 outliers per metal. Removing these observations did not change our overall or race/ethnicity-stratified results (data not shown), and we retained them in analyses.
Statistical analyses
We estimated hazard ratios (HRs) and 95% confidence intervals (CIs) for the association between each metal and breast cancer using Cox proportional hazards regression with age as the time scale. Women were followed from age at enrollment until age at breast cancer diagnosis, study withdrawal, last follow-up, or death. We analyzed the data using the case-cohort method of Prentice and Self (33) with a robust variance estimator. Because sampling was stratified by race/ethnicity, we used sampling weights derived from the inverse probability of sampling given in Web Figure 1. We conducted a joint Wald test to assess the strength of the association across the tertile-based categories of each metal and breast cancer. We also checked for a linear exposure response by using a linear term for the tertile-category–specific median value for each exposure. Conclusions were similar to those of the joint Wald test, so only the Wald test results are shown. We tested the proportional hazards assumption in the overall case-cohort sample with a joint Wald test for interaction terms between the categories of each metal and age-time. We conducted a sensitivity analysis excluding the first year of follow-up when examining associations for overall breast cancer.
We also examined associations stratified by race/ethnicity, ER status (ER-positive vs. ER-negative), menopausal status, and invasiveness (ductal carcinoma in situ vs. invasive breast cancer). Heterogeneity by race/ethnicity, ER status, and invasiveness, separately, was examined using Wald tests of interaction terms between the metal and the characteristic (34). There were 216 cases with missing information on ER status who were excluded from ER status analyses. ER-negative cases in the subcohort were censored at the age of diagnosis in the ER-positive model and vice versa. For ER status, we also conducted a supplemental case-only analysis for etiological heterogeneity (35). For analyses of premenopausal breast cancer, we censored follow-up at menopause. Follow-up for postmenopausal breast cancer started at age at enrollment or age at menopause, whichever occurred later.
We identified confounders that we considered to be possibly associated with both metal concentrations and breast cancer risk based on a directed acyclic graph (Web Figure 2). These included education (high school diploma/equivalent or less, some college, or college degree or higher), race/ethnicity (non-Hispanic White or non-Hispanic Black women), body mass index (weight (kg)/height (m)2; <25.0, 25.0–29.9, or ≥30.0), smoking status (never, past, or current smoker), and parity/breastfeeding (nulliparous, parous and did not breastfeed, or parous and breastfed).
We used quantile g-computation to examine associations between a simultaneous 1-quantile increase in multiple metal concentrations, which we refer to as the overall mixture effect, and breast cancer (36). The method was implemented by categorizing all metal concentrations into tertiles, 
, coded 0, 1, 2, and fitting an adjusted Cox proportional hazards model
 |
where 
represents the (unspecified) hazard at the referent level of all covariates at time 
, 
represents log HRs for confounders Z, and 
is the change in the log HR of breast cancer for a 1-tertile change in all 
exposures in the group. The a priori–defined metal groups that we examined were: 1) all metals; 2) nonessential metals (aluminum, antimony, arsenic, cadmium, lead, and tin) (29); 3) essential metals (chromium, cobalt, copper, iron, manganese, molybdenum, nickel, selenium, and zinc) (29); and 4) metalloestrogens (aluminum, antimony, cadmium, chromium, copper, cobalt, nickel, lead, and tin) (37).
Individual metal analyses were conducted in SAS 9.4 (SAS Institute, Inc., Cary, North Carolina), and quantile g-computation was implemented in R 3.6.0 (R Foundation for Statistical Computing, Vienna, Austria).
RESULTS
Follow-up averaged 7.5 years in duration (range, 0.1–13.5). Most women had a college degree or higher, were parous, had breastfed, and were never smokers. The average age at enrollment was 55.9 years. The distributions of education and body mass index were similar between the cases and the random subcohort, but cases were slightly more likely to be past smokers compared with subcohort members within each racial/ethnic group (Table 1). Non-Hispanic Black case women were more likely than subcohort members to be parous, but many had not breastfed. Non-Hispanic Black case women and subcohort members were more likely to have never smoked, more likely to have a body mass index greater than or equal to 30.0, and slightly younger compared with non-Hispanic White women.
Table 1.
Characteristics of Participants Included in a Case-Cohort Substudy of Toenail Metal Concentrations, by Race/Ethnicity, Sister Study, 2003–2017a
| Race/Ethnicity | ||||||||
|---|---|---|---|---|---|---|---|---|
| Non-Hispanic Black | Non-Hispanic White | |||||||
| Characteristic |
Random Subcohort
(n = 541) |
Breast Cancer Cases
(n = 243) |
Random Subcohort
(n = 1,064) |
Breast Cancer Cases
(n = 1,252) |
||||
| No. | % | No. | % | No. | % | No. | % | |
| Education | ||||||||
| High school or less | 67 | 12 | 22 | 9 | 140 | 13 | 179 | 14 |
| Some college | 183 | 34 | 87 | 36 | 350 | 33 | 404 | 32 |
| College degree or higher | 291 | 54 | 134 | 55 | 574 | 54 | 669 | 53 |
| Body mass indexb | ||||||||
| <25.0 | 91 | 17 | 40 | 16 | 443 | 42 | 482 | 39 |
| 25.0–29.9 | 166 | 31 | 72 | 30 | 346 | 33 | 413 | 33 |
| ≥30.0 | 284 | 53 | 131 | 54 | 275 | 26 | 357 | 29 |
| Parity/breastfeeding | ||||||||
| Nulliparous | 103 | 19 | 28 | 12 | 202 | 19 | 240 | 19 |
| Parous, did not breastfeed | 210 | 39 | 98 | 40 | 245 | 23 | 302 | 24 |
| Parous, breastfed | 228 | 42 | 117 | 48 | 617 | 58 | 710 | 57 |
| Smoking status | ||||||||
| Never smoker | 350 | 65 | 149 | 61 | 590 | 55 | 631 | 50 |
| Past smoker | 139 | 26 | 78 | 32 | 378 | 36 | 526 | 42 |
| Current smoker | 52 | 10 | 16 | 7 | 96 | 9 | 95 | 8 |
| Tumor invasivenessa | ||||||||
| Ductal carcinoma in situ | 69 | 28 | 272 | 22 | ||||
| Invasive | 174 | 72 | 980 | 78 | ||||
| Tumor stagea | ||||||||
| 0 | 57 | 30 | 243 | 22 | ||||
| I | 77 | 41 | 584 | 52 | ||||
| II | 46 | 24 | 224 | 20 | ||||
| III | 10 | 5 | 61 | 5 | ||||
| IV | 0 | 0 | 7 | 0.6 | ||||
| Missing data | 53 | 133 | ||||||
| Tumor gradea | ||||||||
| Predominately well-differentiated | 28 | 17 | 258 | 25 | ||||
| Moderately differentiated/balanced | 61 | 38 | 474 | 47 | ||||
| Poorly differentiated | 72 | 45 | 280 | 28 | ||||
| Missing data | 82 | 240 | ||||||
| Age at toenail collection, yearsc | 53.3 (8.2) | 55.2 (8.6) | 56.1 (8.7) | 57.2 (8.8) | ||||
| Duration of follow-up, yearsc,d | 8.5 (2.3) | 5.0 (2.8) | 9.6 (2.4) | 5.5 (3.1) | ||||
a A total of 42 non-Hispanic Black cases and 65 non-Hispanic White cases were selected into the subcohort and are included in both columns; information on invasiveness, stage, and grade for these individuals is listed only in the case columns.
b Weight (kg)/height (m)2.
c Values are expressed as mean (standard deviation).
d Time between sample collection and diagnosis for cases; time between sample collection and end of follow-up for the subcohort.
The mean toenail concentrations of each metal are shown in Web Table 2. Compared with non-Hispanic White women, non-Hispanic Black women on average had higher concentrations of nonessential metals and lower concentrations of essential metals, with a few exceptions. The weakest Spearman correlation across metals was between tin and zinc (r = −0.002), and the strongest was between iron and manganese (r = 0.73) (Web Figure 3).
For 14 metals, there were no detected violations of the proportional hazards assumption. A violation was observed for selenium (P = 0.02), but it did not persist after stratification by menopausal status; stratifying results by menopausal status may be more appropriate for selenium.
When considered individually, we found little evidence that the 15 metals were strongly associated with overall breast cancer risk (Table 2). Overall patterns were similar in the sensitivity analysis excluding the first year of follow-up (Web Table 3). Additionally, we did not observe an association between groups of metal mixtures and breast cancer (Table 3). This was true for a simultaneous increase in all 15 metals (HR = 0.96, 95% CI: 0.77, 1.19), nonessential metals (HR = 1.01, 95% CI: 0.84, 1.21), essential metals (HR = 0.95, 95% CI: 0.76, 1.19), and metalloestrogens (HR = 0.98, 95% CI: 0.80, 1.20).
Table 2.
Associations Between Individual Metals and Breast Cancer Risk in the Overall Study Population, Sister Study, 2003–2017
| Metal, Tertile, and Concentration (ng/g) |
No. of
Cases |
No. in
Subcohort |
HR a | 95% CI | Bivariate Wald P Value for Categorized Metal b |
|---|---|---|---|---|---|
| Aluminum | 0.5 | ||||
| T1 (<7,819.9) | 475 | 548 | 1.00 | Referent | |
| T2 (7,819.9–15,303.0) | 525 | 547 | 1.12 | 0.93, 1.36 | |
| T3 (>15,303.0) | 495 | 510 | 1.07 | 0.89, 1.30 | |
| Per IQR increase (12,315.0) | 1,495 | 1,605 | 0.99 | 0.96, 1.03 | |
| Antimony | 0.5 | ||||
| T1 (<9.8) | 504 | 487 | 1.00 | Referent | |
| T2 (9.8–19.0) | 519 | 547 | 1.04 | 0.86, 1.25 | |
| T3 (>19.0) | 472 | 571 | 0.93 | 0.77, 1.13 | |
| Per IQR increase (15.2) | 1,495 | 1,605 | 0.99 | 0.97, 1.02 | |
| Arsenic | 0.1 | ||||
| T1 (<48.0) | 516 | 591 | 1.00 | Referent | |
| T2 (48.0–78.7) | 529 | 532 | 1.16 | 0.96, 1.40 | |
| T3 (>78.7) | 450 | 482 | 0.96 | 0.79, 1.16 | |
| Per IQR increase (46.3) | 1,495 | 1,605 | 1.02 | 0.98, 1.05 | |
| Cadmium | 0.5 | ||||
| T1 (<7.3) | 510 | 563 | 1.00 | Referent | |
| T2 (7.3–15.9) | 475 | 542 | 0.90 | 0.74, 1.09 | |
| T3 (>15.9) | 510 | 500 | 1.00 | 0.83, 1.21 | |
| Per IQR increase (14.8) | 1,495 | 1,605 | 0.99 | 0.97, 1.02 | |
| Chromium | 0.5 | ||||
| T1 (<363.6) | 524 | 568 | 1.00 | Referent | |
| T2 (363.6–788.5) | 492 | 548 | 0.93 | 0.77, 1.13 | |
| T3 (>788.5) | 479 | 489 | 0.89 | 0.73, 1.08 | |
| Per IQR increase (701.0) | 1,495 | 1,605 | 1.00 | 0.98, 1.03 | |
| Cobalt | 1.0 | ||||
| T1 (<5.2) | 524 | 583 | 1.00 | Referent | |
| T2 (5.2–10.5) | 488 | 537 | 1.00 | 0.82, 1.21 | |
| T3 (>10.5) | 483 | 485 | 1.01 | 0.83, 1.22 | |
| Per IQR increase (8.9) | 1,495 | 1,605 | 1.00 | 0.98, 1.01 | |
| Copper | 0.9 | ||||
| T1 (<2,929.7) | 552 | 656 | 1.00 | Referent | |
| T2 (2,929.7–3,686.5) | 463 | 506 | 0.96 | 0.79, 1.17 | |
| T3 (>3,686.5) | 480 | 443 | 1.01 | 0.83, 1.23 | |
| Per IQR increase (1,281.0) | 1,495 | 1,605 | 1.00 | 0.98, 1.01 | |
| Iron | 0.8 | ||||
| T1 (<9,060.6) | 506 | 563 | 1.00 | Referent | |
| T2 (9,060.6–15,912.8) | 505 | 546 | 1.06 | 0.88, 1.29 | |
| T3 (>15,912.8) | 484 | 496 | 1.03 | 0.85, 1.25 | |
| Per IQR increase (11,093.0) | 1,495 | 1,605 | 1.00 | 0.97, 1.03 | |
| Lead | 0.9 | ||||
| T1 (<85.5) | 493 | 533 | 1.00 | Referent | |
| T2 (85.5–198.7) | 490 | 528 | 1.03 | 0.85, 1.25 | |
| T3 (>198.7) | 512 | 544 | 1.04 | 0.86, 1.26 | |
| Per IQR increase (190.9) | 1,495 | 1,605 | 1.00 | 0.99, 1.01 | |
| Manganese | 0.8 | ||||
| T1 (<143.4) | 519 | 590 | 1.00 | Referent | |
| T2 (143.4–313.0) | 504 | 542 | 1.07 | 0.88, 1.29 | |
| T3 (>313.0) | 472 | 473 | 1.02 | 0.84, 1.24 | |
| Per IQR increase (282.1) | 1,495 | 1,605 | 1.00 | 0.96, 1.04 | |
| Molybdenum | 0.03 | ||||
| T1 (<4.4) | 481 | 449 | 1.00 | Referent | |
| T2 (4.4–8.4) | 550 | 546 | 1.04 | 0.86, 1.27 | |
| T3 (>8.4) | 464 | 610 | 0.82 | 0.67, 1.00 | |
| Per IQR increase (6.1) | 1,495 | 1,605 | 1.00 | 0.98, 1.03 | |
| Nickel | 0.3 | ||||
| T1 (<206.5) | 498 | 500 | 1.00 | Referent | |
| T2 (206.5–624.6) | 458 | 518 | 0.90 | 0.75, 1.10 | |
| T3 (>624.6) | 539 | 587 | 1.04 | 0.86, 1.26 | |
| Per IQR increase (805.7) | 1,495 | 1,605 | 1.00 | 1.00, 1.01 | |
| Selenium | 0.7 | ||||
| T1 (<1,194.4) | 480 | 520 | 1.00 | Referent | |
| T2 (1,194.4–1,449.4) | 480 | 517 | 1.02 | 0.84, 1.24 | |
| T3 (>1,449.4) | 535 | 568 | 1.09 | 0.90, 1.32 | |
| Per IQR increase (400.0) | 1,495 | 1,605 | 1.01 | 0.96, 1.07 | |
| Tin | 0.3 | ||||
| T1 (<58.5) | 525 | 470 | 1.00 | Referent | |
| T2 (58.5–119.1) | 487 | 540 | 0.91 | 0.75, 1.10 | |
| T3 (>119.1) | 483 | 595 | 0.85 | 0.70, 1.04 | |
| Per IQR increase (97.7) | 1,495 | 1,605 | 1.02 | 0.99, 1.05 | |
| Zinc | 0.4 | ||||
| T1 (<103,062.9) | 525 | 630 | 1.00 | Referent | |
| T2 (103,062.9–118,777.8) | 455 | 496 | 0.90 | 0.74, 1.10 | |
| T3 (>118,777.8) | 515 | 479 | 1.02 | 0.84, 1.24 | |
| Per IQR increase (25,538.0) | 1,495 | 1,605 | 1.02 | 0.96, 1.09 |
Abbreviations: CI, confidence interval; HR, hazard ratio; IQR, interquartile range; T, tertile.
a Adjusted for age, education, race/ethnicity, body mass index, smoking status, and parity/breastfeeding.
b Magnitudes of the P values were used to compare the strength of the evidence for the 2 tertile estimates between the metals.
Table 3.
Quantile g-Computation Estimates of Breast Cancer Risk for a 1-Quantile Increase in Toenail Concentrations of Specific Metal Groups, Sister Study, 2003–2017
| Metal Group | HR a | 95% CI |
|---|---|---|
| All metals | 0.96 | 0.77, 1.19 |
| Nonessential metalsb | 1.01 | 0.84, 1.21 |
| Essential metalsc | 0.95 | 0.76, 1.19 |
| Metalloestrogensd | 0.98 | 0.80, 1.20 |
Abbreviations: CI, confidence interval; HR, hazard ratio.
a Adjusted for age, education, race/ethnicity, parity/breastfeeding, smoking status, and body mass index.
b Includes aluminum, antimony, arsenic, cadmium, lead, and tin. Results were adjusted for age, education, race/ethnicity, parity/breastfeeding, smoking status, body mass index, and the essential metals.
c Includes chromium, cobalt, copper, iron, manganese, molybdenum, nickel, selenium, and zinc. Results were adjusted for age, education, race/ethnicity, parity/breastfeeding, smoking status, body mass index, and the nonessential metals.
d Includes aluminum, antimony, cadmium, chromium, cobalt, copper, lead, nickel, and tin. Results were adjusted for age, education, race/ethnicity, parity/breastfeeding, smoking status, body mass index, and the nonmetalloestrogens.
There were a few exceptions to the generally null findings. The strongest evidence was for molybdenum, which was inversely associated with any breast cancer when classified categorically (third tertile vs. first: HR = 0.82, 95% CI: 0.67, 1.00) (Table 2). This association was particularly evident for ER-negative breast cancer both categorically (third tertile vs. first: HR = 0.57, 95% CI: 0.38, 0.88) and per IQR increase (HR = 0.86, 95% CI: 0.73, 1.01) (compared with ER-positive breast cancer, P values for heterogeneity were 0.1 and 0.05, respectively) (Table 4). The contrast between the associations with ER-positive and ER-negative breast cancer was strongly supported (P = 0.01) by the supplemental case-only analysis (Web Table 4). There was also evidence of an inverse association for an IQR increase in copper concentration for ER-negative breast cancer (HR = 0.88, 95% CI: 0.78, 0.99), which differed from the estimate for ER-positive breast cancer (HR = 1.00, 95% CI: 0.99, 1.01) (P for heterogeneity = 0.04).
Table 4.
Associations Between Toenail Concentrations of Individual Metals and Breast Cancer Risk, by Estrogen Receptor Status, Sister Study, 2003–2017
| ER Status a | |||||||||
|---|---|---|---|---|---|---|---|---|---|
| ER-Positive | ER-Negative | ||||||||
| Metal and Tertile |
No. of
Cases |
No. in
Subcohort |
HR c | 95% CI |
No. of
Cases |
No. in
Subcohort |
HR c | 95% CI |
P for
Heterogeneityb |
| Aluminum | 0.8 | ||||||||
| T1 | 343 | 548 | 1.00 | Referent | 64 | 548 | 1.00 | Referent | |
| T2 | 391 | 547 | 1.13 | 0.91, 1.39 | 59 | 547 | 1.03 | 0.69, 1.53 | |
| T3 | 359 | 510 | 1.05 | 0.86, 1.30 | 63 | 510 | 1.09 | 0.74, 1.60 | |
| Per IQR increase | 1,093 | 1,605 | 0.98 | 0.94, 1.02 | 186 | 1,605 | 0.99 | 0.94, 1.05 | 0.7 |
| Antimony | 0.7 | ||||||||
| T1 | 373 | 487 | 1.00 | Referent | 64 | 487 | 1.00 | Referent | |
| T2 | 381 | 547 | 1.03 | 0.84, 1.27 | 66 | 547 | 1.05 | 0.71, 1.54 | |
| T3 | 339 | 571 | 0.92 | 0.74, 1.13 | 56 | 571 | 0.79 | 0.52, 1.18 | |
| Per IQR increase | 1,093 | 1,605 | 0.99 | 0.97, 1.01 | 186 | 1,605 | 0.93 | 0.87, 0.99 | 0.09 |
| Arsenic | 0.1 | ||||||||
| T1 | 361 | 591 | 1.00 | Referent | 77 | 591 | 1.00 | Referent | |
| T2 | 393 | 532 | 1.21 | 0.98, 1.49 | 60 | 532 | 0.88 | 0.60, 1.29 | |
| T3 | 339 | 482 | 1.01 | 0.81, 1.24 | 49 | 582 | 0.67 | 0.45, 1.02 | |
| Per IQR increase | 1,093 | 1,605 | 0.97 | 0.94, 1.01 | 186 | 1,605 | 1.01 | 0.95, 1.07 | 0.3 |
| Cadmium | 0.8 | ||||||||
| T1 | 375 | 563 | 1.00 | Referent | 64 | 563 | 1.00 | Referent | |
| T2 | 344 | 542 | 0.91 | 0.74, 1.13 | 63 | 542 | 0.92 | 0.62, 1.36 | |
| T3 | 374 | 500 | 1.00 | 0.81, 1.23 | 59 | 500 | 0.90 | 0.60, 1.34 | |
| Per IQR increase | 1,093 | 1,605 | 0.99 | 0.96, 1.02 | 186 | 1,605 | 0.98 | 0.93, 1.04 | 0.8 |
| Chromium | 0.5 | ||||||||
| T1 | 367 | 568 | 1.00 | Referent | 71 | 568 | 1.00 | Referent | |
| T2 | 358 | 548 | 0.94 | 0.76, 1.16 | 63 | 548 | 1.00 | 0.68, 1.47 | |
| T3 | 368 | 489 | 0.94 | 0.76, 1.16 | 52 | 489 | 0.79 | 0.53, 1.17 | |
| Per IQR increase | 1,093 | 1,605 | 1.01 | 0.99, 1.03 | 186 | 1,605 | 0.99 | 0.91, 1.07 | 0.6 |
| Cobalt | 0.6 | ||||||||
| T1 | 366 | 583 | 1.00 | Referent | 67 | 583 | 1.00 | Referent | |
| T2 | 355 | 537 | 1.01 | 0.82, 1.24 | 64 | 537 | 1.07 | 0.72, 1.58 | |
| T3 | 372 | 485 | 1.07 | 0.87, 1.32 | 55 | 485 | 0.92 | 0.61, 1.37 | |
| Per IQR increase | 1,093 | 1,605 | 1.00 | 0.98, 1.01 | 186 | 1,605 | 0.98 | 0.92, 1.05 | 0.6 |
| Copper | 0.3 | ||||||||
| T1 | 379 | 656 | 1.00 | Referent | 79 | 656 | 1.00 | Referent | |
| T2 | 339 | 506 | 1.01 | 0.81, 1.25 | 53 | 506 | 0.76 | 0.50, 1.14 | |
| T3 | 375 | 443 | 1.10 | 0.88, 1.36 | 54 | 443 | 0.80 | 0.53, 1.21 | |
| Per IQR increase | 1,093 | 1,605 | 1.00 | 0.99, 1.01 | 186 | 1,605 | 0.88 | 0.78, 0.99 | 0.04 |
| Iron | 0.5 | ||||||||
| T1 | 361 | 563 | 1.00 | Referent | 62 | 563 | 1.00 | Referent | |
| T2 | 365 | 546 | 1.05 | 0.86, 1.30 | 68 | 546 | 1.27 | 0.86, 1.87 | |
| T3 | 367 | 496 | 1.07 | 0.87, 1.31 | 56 | 596 | 1.02 | 0.69, 1.52 | |
| Per IQR increase | 1,093 | 1,605 | 1.00 | 0.97, 1.04 | 186 | 1,605 | 0.96 | 0.89, 1.04 | 0.3 |
| Lead | 0.3 | ||||||||
| T1 | 356 | 533 | 1.00 | Referent | 63 | 533 | 1.00 | Referent | |
| T2 | 355 | 528 | 1.04 | 0.85, 1.29 | 67 | 528 | 1.08 | 0.73, 1.58 | |
| T3 | 382 | 544 | 1.08 | 0.88, 1.33 | 56 | 544 | 0.84 | 0.56, 1.26 | |
| Per IQR increase | 1,093 | 1,605 | 1.00 | 0.99, 1.01 | 186 | 1,605 | 1.00 | 0.98, 1.02 | 1.0 |
| Manganese | 1.0 | ||||||||
| T1 | 380 | 590 | 1.00 | Referent | 66 | 590 | 1.00 | Referent | |
| T2 | 365 | 542 | 1.04 | 0.84, 1.28 | 64 | 542 | 1.07 | 0.72, 1.58 | |
| T3 | 348 | 473 | 0.99 | 0.81, 1.23 | 56 | 473 | 0.99 | 0.67, 1.48 | |
| Per IQR increase | 1,093 | 1,605 | 1.00 | 0.95, 1.04 | 186 | 1,605 | 0.94 | 0.86, 1.04 | 0.3 |
| Molybdenum | 0.1 | ||||||||
| T1 | 343 | 449 | 1.00 | Referent | 69 | 449 | 1.00 | Referent | |
| T2 | 411 | 546 | 1.12 | 0.91, 1.38 | 67 | 546 | 0.91 | 0.62, 1.34 | |
| T3 | 339 | 610 | 0.89 | 0.72, 1.10 | 50 | 610 | 0.57 | 0.38, 0.88 | |
| Per IQR increase | 1,093 | 1,605 | 1.01 | 0.98, 1.03 | 186 | 1,605 | 0.86 | 0.73, 1.01 | 0.05 |
| Nickel | 0.5 | ||||||||
| T1 | 379 | 500 | 1.00 | Referent | 56 | 500 | 1.00 | Referent | |
| T2 | 326 | 518 | 0.86 | 0.70, 1.06 | 62 | 518 | 1.08 | 0.72, 1.63 | |
| T3 | 388 | 587 | 1.00 | 0.81, 1.23 | 68 | 587 | 1.20 | 0.80, 1.80 | |
| Per IQR increase | 1,093 | 1,605 | 1.01 | 1.00, 1.01 | 186 | 1,605 | 0.99 | 0.98, 1.01 | 0.1 |
| Selenium | 0.4 | ||||||||
| T1 | 348 | 520 | 1.00 | Referent | 67 | 520 | 1.00 | Referent | |
| T2 | 352 | 517 | 1.05 | 0.85, 1.30 | 50 | 517 | 0.79 | 0.53, 1.20 | |
| T3 | 393 | 568 | 1.13 | 0.92, 1.39 | 69 | 568 | 1.04 | 0.71, 1.54 | |
| Per IQR increase | 1,093 | 1,605 | 1.02 | 0.97, 1.07 | 186 | 1,605 | 0.96 | 0.86, 1.07 | 0.3 |
| Tin | 0.5 | ||||||||
| T1 | 383 | 470 | 1.00 | Referent | 68 | 470 | 1.00 | Referent | |
| T2 | 367 | 540 | 0.96 | 0.78, 1.18 | 59 | 540 | 0.79 | 0.53, 1.16 | |
| T3 | 343 | 595 | 0.87 | 0.71, 1.08 | 59 | 595 | 0.72 | 0.49, 1.08 | |
| Per IQR increase | 1,093 | 1,605 | 1.02 | 0.99, 1.06 | 186 | 1,605 | 1.00 | 0.94, 1.06 | 0.5 |
| Zinc | 0.2 | ||||||||
| T1 | 362 | 630 | 1.00 | Referent | 76 | 630 | 1.00 | Referent | |
| T2 | 343 | 496 | 0.97 | 0.79, 1.21 | 48 | 496 | 0.70 | 0.46, 1.05 | |
| T3 | 388 | 479 | 1.11 | 0.90, 1.36 | 62 | 479 | 0.86 | 0.58, 1.27 | |
| Per IQR increase | 1,093 | 1,605 | 1.02 | 0.95, 1.10 | 186 | 1,605 | 0.94 | 0.83, 1.08 | 0.3 |
Abbreviations: CI, confidence interval; ER, estrogen receptor; HR, hazard ratio; IQR, interquartile range; T, tertile.
a Separate models stratified by ER status.
b Heterogeneity P value calculated by the Wald statistic in a model with an interaction term.
c Adjusted for age, education, race/ethnicity, body mass index, smoking status, and parity/breastfeeding.
The strongest evidence for heterogeneity between ductal carcinoma in situ and invasive breast cancer was for arsenic, where there was an inverse association for ductal carcinoma in situ (third tertile vs. first: HR = 0.63, 95% CI: 0.46, 0.88) but no evident association for invasive breast cancer (third tertile vs. first: HR = 1.07, 95% CI: 0.87, 1.32) (P for heterogeneity = 0.01) (Web Table 5).
In this diverse case-cohort sample, associations for many metals and breast cancer did not vary by race/ethnicity (Table 5). However, across categories of exposure, antimony was inversely associated with breast cancer among non-Hispanic Black women (third tertile vs. first: HR = 0.56, 95% CI: 0.37, 0.84) but demonstrated a negligible association among non-Hispanic White women (HR = 0.96, 95% CI: 0.78, 1.18) (P for heterogeneity = 0.05). Additionally, the point estimates across categories of zinc in non-Hispanic Black women suggested a monotonic, dose-related association (second tertile vs. first: HR = 1.12 (95% CI: 0.76, 1.65); third tertile vs. first: HR = 1.38 (95% CI: 0.94, 2.03)), which was reflected per IQR increase (HR = 1.16, 95% CI: 1.01, 1.34), although the estimates were not notably different from those for non-Hispanic White women (P values for heterogeneity were 0.3 and 0.1, respectively).
Table 5.
Associations Between Toenail Concentrations of Individual Metals and Breast Cancer Risk, by Race/Ethnicity, Sister Study, 2003–2017
| Race/Ethnicity a | |||||||||
|---|---|---|---|---|---|---|---|---|---|
| Non-Hispanic White | Non-Hispanic Black | ||||||||
| Metal and Tertile | No. of Cases |
No. in
Subcohort |
HR c | 95% CI | No. of Cases |
No. in
Subcohort |
HR c | 95% CI |
P for
Heterogeneityb |
| Aluminum | 0.8 | ||||||||
| T1 | 384 | 351 | 1.00 | Referent | 91 | 197 | 1.00 | Referent | |
| T2 | 437 | 353 | 1.13 | 0.92, 1.40 | 88 | 194 | 1.02 | 0.71, 1.47 | |
| T3 | 431 | 360 | 1.09 | 0.89, 1.38 | 64 | 150 | 0.96 | 0.65, 1.42 | |
| Per IQR increase | 1,252 | 1,064 | 1.00 | 0.96, 1.04 | 243 | 541 | 0.97 | 0.92, 1.03 | 0.6 |
| Antimony | 0.05 | ||||||||
| T1 | 433 | 366 | 1.00 | Referent | 71 | 121 | 1.00 | Referent | |
| T2 | 424 | 352 | 1.05 | 0.86, 1.29 | 95 | 195 | 0.86 | 0.57, 1.29 | |
| T3 | 395 | 346 | 0.96 | 0.78, 1.18 | 77 | 225 | 0.56 | 0.37, 0.84 | |
| Per IQR increase | 1,252 | 1,064 | 1.00 | 0.97, 1.02 | 243 | 541 | 0.98 | 0.95, 1.02 | 0.6 |
| Arsenic | 0.2 | ||||||||
| T1 | 392 | 342 | 1.00 | Referent | 124 | 249 | 1.00 | Referent | |
| T2 | 449 | 354 | 1.19 | 0.97, 1.47 | 80 | 178 | 0.91 | 0.64, 1.29 | |
| T3 | 411 | 368 | 0.98 | 0.80, 1.21 | 39 | 114 | 0.66 | 0.43, 1.02 | |
| Per IQR increase | 1,252 | 1,064 | 1.02 | 0.99, 1.04 | 243 | 541 | 0.92 | 0.75, 1.12 | 0.3 |
| Cadmium | 0.8 | ||||||||
| T1 | 420 | 348 | 1.00 | Referent | 90 | 215 | 1.00 | ||
| T2 | 391 | 353 | 0.89 | 0.73, 1.10 | 84 | 189 | 1.02 | 0.71, 1.46 | |
| T3 | 441 | 363 | 0.99 | 0.80, 1.21 | 69 | 137 | 1.04 | 0.70, 1.54 | |
| Per IQR increase | 1,252 | 1,064 | 0.99 | 0.96, 1.02 | 243 | 541 | 0.99 | 0.98, 1.00 | 0.7 |
| Chromium | 0.8 | ||||||||
| T1 | 422 | 346 | 1.00 | Referent | 102 | 222 | 1.00 | Referent | |
| T2 | 409 | 353 | 0.94 | 0.76, 1.16 | 83 | 195 | 0.85 | 0.59, 1.22 | |
| T3 | 421 | 365 | 0.89 | 0.72, 1.09 | 58 | 124 | 0.92 | 0.61, 1.37 | |
| Per IQR increase | 1,252 | 1,064 | 1.01 | 0.97, 1.06 | 243 | 541 | 0.99 | 0.97, 1.02 | 0.5 |
| Cobalt | 0.2 | ||||||||
| T1 | 402 | 342 | 1.00 | Referent | 122 | 241 | 1.00 | Referent | |
| T2 | 418 | 355 | 1.03 | 0.84, 1.27 | 70 | 182 | 0.71 | 0.49, 1.01 | |
| T3 | 432 | 367 | 1.02 | 0.84, 1.26 | 51 | 118 | 0.78 | 0.52, 1.18 | |
| Per IQR increase | 1,252 | 1,064 | 0.99 | 0.98, 1.01 | 243 | 541 | 1.14 | 1.06, 1.22 | <0.001 |
| Copper | 0.5 | ||||||||
| T1 | 401 | 324 | 1.00 | Referent | 151 | 332 | 1.00 | Referent | |
| T2 | 409 | 362 | 0.97 | 0.78, 1.20 | 54 | 144 | 0.82 | 0.57, 1.19 | |
| T3 | 442 | 378 | 1.01 | 0.82, 1.25 | 38 | 65 | 1.12 | 0.70, 1.80 | |
| Per IQR increase | 1,252 | 1,064 | 1.00 | 0.98, 1.01 | 243 | 541 | 1.03 | 0.89, 1.20 | 0.7 |
| Iron | 0.5 | ||||||||
| T1 | 402 | 349 | 1.00 | Referent | 104 | 214 | 1.00 | Referent | |
| T2 | 423 | 352 | 1.09 | 0.89, 1.34 | 82 | 194 | 0.83 | 0.59, 1.19 | |
| T3 | 427 | 363 | 1.05 | 0.86, 1.29 | 57 | 133 | 0.84 | 0.56, 1.26 | |
| Per IQR increase | 1,252 | 1,064 | 1.00 | 0.97, 1.03 | 243 | 541 | 0.94 | 0.81, 1.09 | 0.5 |
| Lead | 0.2 | ||||||||
| T1 | 404 | 356 | 1.00 | Referent | 89 | 177 | 1.00 | Referent | |
| T2 | 410 | 356 | 1.05 | 0.85, 1.29 | 80 | 172 | 0.91 | 0.62, 1.33 | |
| T3 | 438 | 352 | 1.07 | 0.87, 1.31 | 74 | 192 | 0.71 | 0.49, 1.04 | |
| Per IQR increase | 1,252 | 1,064 | 1.00 | 0.98, 1.01 | 243 | 541 | 1.00 | 1.00, 1.00 | 0.5 |
| Manganese | 0.3 | ||||||||
| T1 | 399 | 340 | 1.00 | Referent | 120 | 250 | 1.00 | Referent | |
| T2 | 420 | 355 | 1.09 | 0.88, 1.34 | 84 | 187 | 0.86 | 0.61, 1.21 | |
| T3 | 433 | 369 | 1.05 | 0.85, 1.29 | 39 | 104 | 0.74 | 0.48, 1.15 | |
| Per IQR increase | 1,252 | 1,064 | 1.00 | 0.96, 1.04 | 243 | 541 | 0.87 | 0.70, 1.08 | 0.2 |
| Molybdenum | 0.9 | ||||||||
| T1 | 447 | 376 | 1.00 | Referent | 34 | 73 | 1.00 | Referent | |
| T2 | 457 | 352 | 1.05 | 0.86, 1.28 | 93 | 194 | 1.05 | 0.64, 1.73 | |
| T3 | 348 | 336 | 0.81 | 0.66, 1.00 | 116 | 274 | 0.86 | 0.54, 1.38 | |
| Per IQR increase | 1,252 | 1,064 | 1.00 | 0.96, 1.05 | 243 | 541 | 1.00 | 0.98, 1.03 | 0.9 |
| Nickel | 0.7 | ||||||||
| T1 | 445 | 364 | 1.00 | Referent | 53 | 136 | 1.00 | Referent | |
| T2 | 384 | 359 | 0.89 | 0.73, 1.09 | 74 | 159 | 1.07 | 0.70, 1.64 | |
| T3 | 423 | 341 | 1.03 | 0.83, 1.26 | 116 | 246 | 1.26 | 0.85, 1.86 | |
| Per IQR increase | 1,252 | 1,064 | 1.00 | 1.00, 1.01 | 243 | 541 | 1.01 | 1.00, 1.02 | 0.09 |
| Selenium | 0.7 | ||||||||
| T1 | 404 | 359 | 1.00 | Referent | 76 | 161 | 1.00 | Referent | |
| T2 | 410 | 359 | 1.03 | 0.84, 1.26 | 70 | 158 | 0.91 | 0.61, 1.36 | |
| T3 | 438 | 346 | 1.10 | 0.90, 1.36 | 97 | 222 | 0.94 | 0.64, 1.38 | |
| Per IQR increase | 1,252 | 1,064 | 1.01 | 0.96, 1.06 | 243 | 541 | 1.01 | 0.85, 1.20 | 1.0 |
| Tin | 0.5 | ||||||||
| T1 | 479 | 370 | 1.00 | Referent | 46 | 100 | 1.00 | Referent | |
| T2 | 409 | 354 | 0.91 | 0.75, 1.12 | 78 | 186 | 0.96 | 0.60, 1.51 | |
| T3 | 364 | 340 | 0.84 | 0.68, 1.03 | 119 | 255 | 1.07 | 0.69, 1.64 | |
| Per IQR increase | 1,252 | 1,064 | 1.03 | 1.00, 1.06 | 243 | 541 | 1.00 | 0.98, 1.03 | 0.3 |
| Zinc | 0.3 | ||||||||
| T1 | 403 | 332 | 1.00 | Referent | 122 | 298 | 1.00 | Referent | |
| T2 | 398 | 363 | 0.88 | 0.72, 1.09 | 57 | 133 | 1.12 | 0.76, 1.65 | |
| T3 | 451 | 369 | 0.99 | 0.81, 1.22 | 64 | 110 | 1.38 | 0.94, 2.03 | |
| Per IQR increase | 1,252 | 1,064 | 1.01 | 0.94, 1.09 | 243 | 541 | 1.16 | 1.01, 1.34 | 0.1 |
Abbreviations: CI, confidence interval; HR, hazard ratio; IQR, interquartile range T, tertile.
a Separate models stratified by race/ethnicity.
b Joint test of heterogeneity from a model with an interaction term.
c Adjusted for age, education, race/ethnicity, body mass index, smoking status, and parity/breastfeeding.
Patterns for copper and selenium were inverse for premenopausal breast cancer and elevated for postmenopausal breast cancer when considered categorically (Web Table 6; P values for heterogeneity were 0.01 and 0.02, respectively).
DISCUSSION
Using a prospective case-cohort design and a panel of 15 metals that were measured in toenail clippings from women across the United States, we observed a reduced incidence of breast cancer in women with higher concentrations of molybdenum. For the remaining metals, we observed little evidence of an association with breast cancer risk overall. The joint associations between overall breast cancer and multiple metals in predefined groups (all metals, nonessential metals, essential metals, and metalloestrogens) were negligible. Certain metals showed differences across strata by race/ethnicity, ER status, or menopausal status.
Most prior evidence that metals may be associated with breast cancer comes from studies of cadmium. However, this evidence of increased risk came solely from retrospective case-control studies where cadmium was measured after diagnosis in the cases, so temporality of the association could not be established (8, 10–12, 14). In contrast, but similarly to our findings, researchers in other prospective studies reported no evidence of an association (7, 9, 13).
To our knowledge, only 1 prior epidemiologic study has prospectively examined concentrations of other metals measured in toenails in relation to breast cancer risk (19). None of the metals examined (arsenic, copper, chromium, iron, and zinc) were related to breast cancer risk, similarly to the findings of our study. A strength of our study was that we measured a large panel of metals, which allowed us to go beyond the more commonly studied metals and explore associations among metals for which less is known.
The finding that molybdenum was inversely associated with breast cancer overall and most notably for ER-negative breast cancer was unexpected. Molybdenum serves as a cofactor for multiple enzymes that break down toxic sulfites and aldehydes and convert xanthine to uric acid, an antioxidant in the blood (38, 39). The largest source of molybdenum in the general population is diet, through legumes, cereal grains, leafy vegetables, and milk (40). In a case-control study of bean fiber, beans, and grains, Sangaramoorthy et al. (41) found reduced odds of breast cancer in women with high intake, with the protection being particularly notable for ER-negative breast cancer. Though we considered molybdenum and not diet, the findings of that study align well with our results. To our knowledge, only 1 prior epidemiologic study of molybdenum and breast cancer has been conducted; in a case-control study of disease-discordant sister pairs within the Sister Study and the Two Sister Study, O’Brien et al. (42) observed an elevated but imprecise association for molybdenum and young-onset (diagnosis age <50 years) breast cancer. The present study was independent and included all breast cancer diagnoses rather than only retrospectively ascertained young-onset cases. Additionally, the toenail clippings in the prior study were collected from the cases years after diagnosis and treatment, in contrast to the present prospective study, where toenails were collected at baseline, prior to the diagnosis of cases.
There was also some evidence that the association for an IQR increase in copper concentration depended on ER status. Copper may increase cell proliferation and angiogenesis (43). Copper depletion agents, which have been shown to induce apoptosis and suppress angiogenesis (44, 45), are being investigated as a potential treatment for triple-negative breast cancer (46). Therefore, our finding of an inverse association between an IQR increase in copper concentration and ER-negative breast cancer was somewhat unexpected. However, our study examined breast cancer incidence rather than progression, and prior epidemiologic studies of breast cancer incidence have not examined whether the associations for copper and breast cancer risk vary by ER status.
An important strength of our study was the race/ethnicity-stratified design, which oversampled non-Hispanic Black women. This gave us better statistical power to examine associations by race/ethnicity, which we hypothesized could be an important consideration, since sources and levels of exposure and cofactors may differ across racial/ethnic groups (23–25). We noted that concentrations of most of the nonessential metals were higher, and those of most of the essential metals were lower, among non-Hispanic Black women than among non-Hispanic White women. We found the strongest evidence of heterogeneity by race/ethnicity for antimony; there was an inverse association among non-Hispanic Black women but a negligible association among non-Hispanic White women. Exposure to antimony in the general population comes from contamination of air and water by power plants and industrial facilities (47). Some studies have suggested that non-Hispanic Black women may be more likely to live near industrial sources of pollution (23), which may explain the finding that concentrations of antimony were slightly higher in non-Hispanic Black women in our study. However, the reason we observed an inverse association between antimony and breast cancer specifically among non-Hispanic Black women is unclear.
Patterns across categories of zinc suggested an unexpected positive association with breast cancer among non-Hispanic Black women. Zinc is an essential element that is a cofactor in certain detoxifying enzymes and has been considered for chemoprevention (48, 49). Data from the Third National Health and Nutrition Examination Survey (1988–1994) have shown that zinc intake is lower in non-Hispanic Black women than in non-Hispanic White women (50), consistent with the toenail concentrations observed in our study. Prior studies have not examined the relationship between zinc and breast cancer stratified race/ethnicity, so additional work in large, diverse populations would be valuable.
An innovative aspect of our study was that we applied a mixtures approach, quantile g-computation, to evaluate the effect of simultaneously increasing all metals in each group by a tertile. In addition to looking at all 15 metals combined, we specified a priori–defined groups with potential biological plausibility. The method accounts for co-metal confounding and resists undue influence of outliers through its use of quantization. We were unable to account for either synergistic effects or nonlinearities in this application of quantile g-computation for the case-cohort study design.
We used data from the Sister Study, which contains a well-educated, mostly postmenopausal cohort of women who all have a first-degree family history of breast cancer. While these characteristics may influence the representativeness of our population, this study design results in rapid case accrual, improving our statistical power without impairing internal validity. Extensive covariate information was collected at baseline and we determined our confounder adjustment set based on a directed acyclic graph, but the possibility of unmeasured confounding remains. We conducted many analyses and comparisons in this work, so we cannot exclude the possibility that some of our findings were false-positive. We did not have data on metal speciation, which may have attenuated results for certain metals, such as arsenic and chromium, where different species have different biological activity (51–53). To our knowledge, there are no consistently agreed upon toxic or harmful concentrations for each of these metals, particularly when measured in toenails, that we could use to categorize the concentrations. Toenails are a stable and practical matrix for measuring metal concentrations, because they are easy to collect and store and are less susceptible to fluctuations than other kinds of biospecimens such as urine or blood. Concentrations based on toenail clippings are thought to reflect 4–6 months of exposure in the 6–12 months before collection (based on a study of lead, manganese, cadmium, nickel, and arsenic) (54) but may be adequate proxies for exposure occurring over even longer periods of time, as some studies have reported correlations in toenail metal concentrations over 4–10 (based on a study of most of the same metals included here) or 1–7 (based on a study of lead, arsenic, cadmium, and manganese) years (55, 56). It is possible that certain metals may act during the initiation or promotion stages (or both) of carcinogenesis (57, 58), but the mechanisms and exact time frame that would be etiologically relevant for individual metals, including molybdenum, are not known. However, given that this previous literature found that metal concentrations measured in toenails may be correlated over many years, toenail metal concentrations may be more adequate proxies for etiologically relevant periods of exposure for breast cancer than other biomarkers, such as concentrations in urine or blood, which represent a shorter window of time.
Using a prospective case-cohort design with a large panel of metals measured in toenails, we found that molybdenum was inversely associated with breast cancer, particularly the ER-negative subtype. There was little evidence that other individual metals or groups of metals were associated with breast cancer overall, although there was some evidence that a few metals may be associated with risk in subgroups of women. To our knowledge, this was the first study to evaluate this research question considering associations by race/ethnicity, and it will be important for additional research to consider diverse populations.
Supplementary Material
ACKNOWLEDGMENTS
Author affiliations: Epidemiology Branch, National Institute of Environmental Health Sciences, Research Triangle Park, North Carolina, United States (Nicole M. Niehoff, Katie M. O’Brien, Alexander P. Keil, Alexandra J. White); Department of Epidemiology, Gillings School of Global Public Health, University of North Carolina, Chapel Hill, North Carolina, United States (Alexander P. Keil); RTI International, Research Triangle Park, North Carolina, United States (Keith E. Levine, Chamindu Liyanapatirana, Laura G. Haines); Program Operations Branch, Division of the National Toxicology Program, National Institute of Environmental Health Sciences, Research Triangle Park, North Carolina, United States (Suramya Waidyanatha); and Biostatistics and Computational Biology Branch, National Institute of Environmental Health Sciences, Research Triangle Park, North Carolina, United States (Clarice R. Weinberg).
This work was funded by the Intramural Research Program of the National Institutes of Health (NIH), Office of Women’s Health Research, National Institute of Environmental Health Sciences (NIEHS) (projects Z01-ES044005 and Z1AES103332-02). The work was also supported by the Intramural Research Program of the NIH, NIEHS (project Z1AES103316-05) and was performed for the National Toxicology Program at the NIEHS under contract HHSN273201400022C (RTI International, Research Triangle Park, North Carolina).
We appreciate the helpful comments of Drs. Jack Taylor and Jacob Kresovich at the National Institute of Environmental Health Sciences.
Requests for deidentified Sister Study data, including the data used in this manuscript, can be made through the Sister Study website (https://sisterstudy.niehs.nih.gov/English/data-requests.htm).
Conflict of interest: none declared.
REFERENCES
- 1. American Cancer Society . Breast Cancer Facts and Figures 2019–2020. Atlanta, GA: American Cancer Society; 2019. https://www.cancer.org/research/cancer-facts-statistics/breast-cancer-facts-figures.html. Accessed July 2, 2020. [Google Scholar]
- 2. International Agency for Research on Cancer . Arsenic, Metals, Fibres, and Dusts. (IARC Monographs on the Evaluation of Carcinogenic Risks to Humans, vol. 100C). Lyon, France: International Agency for Research on Cancer; 2012. [PMC free article] [PubMed] [Google Scholar]
- 3. International Agency for Research on Cancer . Inorganic and Organic Lead Compounds. (IARC Monographs on the Evaluation of Carcinogenic Risks to Humans, vol. 87). Lyon, France: International Agency for Research on Cancer; 2006. [PMC free article] [PubMed] [Google Scholar]
- 4. Byrne C, Divekar SD, Storchan GB, et al. Metals and breast cancer. J Mammary Gland Biol Neoplasia. 2013;18(1):63–73. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5. Darbre PD. Metalloestrogens: an emerging class of inorganic xenoestrogens with potential to add to the oestrogenic burden of the human breast. J Appl Toxicol. 2006;26(3):191–197. [DOI] [PubMed] [Google Scholar]
- 6. Martin MB, Reiter R, Pham T, et al. Estrogen-like activity of metals in MCF-7 breast cancer cells. Endocrinology. 2003;144(6):2425–2436. [DOI] [PubMed] [Google Scholar]
- 7. Adams SV, Shafer MM, Bonner MR, et al. Urinary cadmium and risk of invasive breast cancer in the Women’s Health Initiative. Am J Epidemiol. 2016;183(9):815–823. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8. Wei XL, He JR, Cen YL, et al. Modified effect of urinary cadmium on breast cancer risk by selenium. Clin Chim Acta. 2015;438:80–85. [DOI] [PubMed] [Google Scholar]
- 9. Eriksen KT, McElroy JA, Harrington JM, et al. Urinary cadmium and breast cancer: a prospective Danish cohort study. J Nat Cancer Inst. 2017;109(2):djw204. [DOI] [PubMed] [Google Scholar]
- 10. Strumylaite L, Kregzdyte R, Bogusevicius A, et al. Association between cadmium and breast cancer risk according to estrogen receptor and human epidermal growth factor receptor 2: epidemiological evidence. Breast Cancer Res Treat. 2014;145(1):225–232. [DOI] [PubMed] [Google Scholar]
- 11. Nagata C, Nagao Y, Nakamura K, et al. Cadmium exposure and the risk of breast cancer in Japanese women. Breast Cancer Res Treat. 2013;138(1):235–239. [DOI] [PubMed] [Google Scholar]
- 12. McElroy JA, Shafer MM, Trentham-Dietz A, et al. Cadmium exposure and breast cancer risk. J Natl Cancer Inst. 2006;98(12):869–873. [DOI] [PubMed] [Google Scholar]
- 13. Gaudet MM, Deubler EL, Kelly RS, et al. Blood levels of cadmium and lead in relation to breast cancer risk in three prospective cohorts. Int J Cancer. 2019;144(5):1010–1016. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14. Gallagher CM, Chen JJ, Kovach JS. Environmental cadmium and breast cancer risk. Aging. 2010;2(11):804–814. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15. McElroy JA, Shafer MM, Gangnon RE, et al. Urinary lead exposure and breast cancer risk in a population-based case-control study. Cancer Epidemiol Biomarkers Prev. 2008;17(9):2311–2317. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16. López-Carrillo L, Hernández-Ramírez RU, Gandolfi AJ, et al. Arsenic methylation capacity is associated with breast cancer in northern Mexico. Toxicol Appl Pharmacol. 2014;280(1):53–59. [DOI] [PubMed] [Google Scholar]
- 17. Marciniak W, Derkacz R, Muszyńska M, et al. Blood arsenic levels and the risk of familial breast cancer in Poland. Int J Cancer. 2020;146(10):2721–2727. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18. Larsson SC, Orsini N, Wolk A. Urinary cadmium concentration and risk of breast cancer: a systematic review and dose-response meta-analysis. Am J Epidemiol. 2015;182(5):375–380. [DOI] [PubMed] [Google Scholar]
- 19. Garland M, Morris JS, Colditz GA, et al. Toenail trace element levels and breast cancer: a prospective study. Am J Epidemiol. 1996;144(7):653–660. [DOI] [PubMed] [Google Scholar]
- 20. Hunter DJ, Morris JS, Stampfer MJ, et al. A prospective study of selenium status and breast cancer risk. JAMA. 1990;264(9):1128–1131. [PubMed] [Google Scholar]
- 21. van den Brandt PA, Goldbohm RA, van’t Veer P, et al. Toenail selenium levels and the risk of breast cancer. Am J Epidemiol. 1994;140(1):20–26. [DOI] [PubMed] [Google Scholar]
- 22. van Noord PA, Maas MJ, van der Tweel I, et al. Selenium and the risk of postmenopausal breast cancer in the DOM cohort. Breast Cancer Res Treat. 1993;25(1):11–19. [DOI] [PubMed] [Google Scholar]
- 23. Mohai P, Lantz PM, Morenoff J, et al. Racial and socioeconomic disparities in residential proximity to polluting industrial facilities: evidence from the Americans’ Changing Lives Study. Am J Public Health. 2009;99(suppl 3):S649–S656. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24. Davis HT, Aelion CM, Liu J, et al. Potential sources and racial disparities in the residential distribution of soil arsenic and lead among pregnant women. Sci Total Environ. 2016;551-552:622–630. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25. Shim YK, Lewin MD, Ruiz P, et al. Prevalence and associated demographic characteristics of exposure to multiple metals and their species in human populations: the United States NHANES, 2007–2012. J Toxicol Environ Health A. 2017;80(9):502–512. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26. Gee GC, Payne-Sturges DC. Environmental health disparities: a framework integrating psychosocial and environmental concepts. Environ Health Perspect. 2004;112(17):1645–1653. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27. Schulz AJ, Kannan S, Dvonch JT, et al. Social and physical environments and disparities in risk for cardiovascular disease: the healthy environments partnership conceptual model. Environ Health Perspect. 2005;113(12):1817–1825. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28. Morello-Frosch R, Lopez R. The riskscape and the color line: examining the role of segregation in environmental health disparities. Environ Res. 2006;102(2):181–196. [DOI] [PubMed] [Google Scholar]
- 29. Tchounwou PB, Yedjou CG, Patlolla AK, et al. Heavy metal toxicity and the environment. Exp Suppl. 2012;101:133–164. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30. Sandler DP, Hodgson ME, Deming-Halverson SL, et al. The Sister Study cohort: baseline methods and participant characteristics. Environ Health Perspect. 2017;125(12):127003. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31. D’Aloisio AA, Nichols HB, Hodgson ME, et al. Validity of self-reported breast cancer characteristics in a nationwide cohort of women with a family history of breast cancer. BMC Cancer. 2017;17(1):692. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32. Wicklin R. Influential observations in a linear regression model: the DFBETAS statistics [Web log post]. https://blogs.sas.com/content/iml/2019/06/17/influence-regression-dfbeta.html. Published July 17, 2019. Accessed April 28, 2020.
- 33. Prentice RL, Self SG. Aspects of the use of relative risk models in the design and analysis of cohort studies and prevention trials. Stat Med. 1988;7(1-2):275–287. [DOI] [PubMed] [Google Scholar]
- 34. Xue X, Kim MY, Gaudet MM, et al. A comparison of the polytomous logistic regression and joint Cox proportional hazards models for evaluating multiple disease subtypes in prospective cohort studies. Cancer Epidemiol Biomarkers Prev. 2013;22(2):275–285. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35. Begg CB, Zhang ZF. Statistical analysis of molecular epidemiology studies employing case-series. Cancer Epidemiol Biomarkers Prev. 1994;3(2):173–175. [PubMed] [Google Scholar]
- 36. Keil AP, Buckley JP, O’Brien KM, et al. A quantile-based g-computation approach to addressing the effects of exposure mixtures. Environ Health Perspect. 2020;128(4):47004. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37. Wallace DR. Nanotoxicology and metalloestrogens: possible involvement in breast cancer. Toxics. 2015;3(4):390–413. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38. Mendel RR, Bittner F. Cell biology of molybdenum. Biochim Biophys Acta. 2006;1763(7):621–635. [DOI] [PubMed] [Google Scholar]
- 39. Ames BN, Cathcart R, Schwiers E, et al. Uric acid provides an antioxidant defense in humans against oxidant- and radical-caused aging and cancer: a hypothesis. Proc Natl Acad Sci U S A. 1981;78(11):6858–6862. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 40. Agency for Toxic Substances and Disease Registry, US Department of Health and Human Services . Toxicological Profile for Molybdenum. Atlanta, GA: Agency for Toxic Substances and Disease Registry; 2020. https://www.atsdr.cdc.gov/ToxProfiles/tp212-p.pdf. Accessed July 16, 2020. [PubMed] [Google Scholar]
- 41. Sangaramoorthy M, Koo J, John EM. Intake of bean fiber, beans, and grains and reduced risk of hormone receptor-negative breast cancer: the San Francisco Bay Area Breast Cancer Study. Cancer Med. 2018;7(5):2131–2144. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 42. O’Brien KM, White AJ, Jackson BP, et al. Toenail-based metal concentrations and young-onset breast cancer. Am J Epidemiol. 2019;188(4):34–43. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 43. Denoyer D, Masaldan S, La Fontaine S, et al. Targeting copper in cancer therapy: ‘copper that cancer’. Metallomics. 2015;7(11):1459–1476. [DOI] [PubMed] [Google Scholar]
- 44. Karginova O, Weekley CM, Raoul A, et al. Inhibition of copper transport induces apoptosis in triple-negative breast cancer cells and suppresses tumor angiogenesis. Mol Cancer Ther. 2019;18(5):873–885. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 45. Cui L, Gouw AM, LaGory EL, et al. Mitochondrial copper depletion suppresses triple-negative breast cancer in mice. Nat Biotechnol. 2021;39(3):357–367. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 46. Chan N, Willis A, Kornhauser N, et al. Influencing the tumor microenvironment: a phase II study of copper depletion using tetrathiomolybdate in patients with breast cancer at high risk for recurrence and in preclinical models of lung metastases. Clin Cancer Res. 2017;23(3):666–676. [DOI] [PubMed] [Google Scholar]
- 47. Agency for Toxic Substances and Disease Registry, US Department of Health and Human Services . Toxicological Profile for Antimony. Atlanta, GA: Agency for Toxic Substances and Disease Registry; 2019. https://www.atsdr.cdc.gov/toxprofiles/tp23.pdf. Accessed July 16, 2020. [Google Scholar]
- 48. Dhawan DK, Chadha VD. Zinc: a promising agent in dietary chemoprevention of cancer. Indian J Med Res. 2010;132(6):676–682. [PMC free article] [PubMed] [Google Scholar]
- 49. Jouybari L, Kiani F, Akbari A, et al. A meta-analysis of zinc levels in breast cancer. J Trace Elem Med Biol. 2019;56:90–99. [DOI] [PubMed] [Google Scholar]
- 50. Briefel RR, Bialostosky K, Kennedy-Stephenson J, et al. Zinc intake of the U.S. population: findings from the Third National Health and Nutrition Examination Survey, 1988–1994. J Nutr. 2000;130(5):1367S–1373S. [DOI] [PubMed] [Google Scholar]
- 51. Templeton DM. Speciation in metal toxicity and metal-based therapeutics. Toxics. 2015;3(2):170–186. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 52. Kitchin KT, Ahmad S. Oxidative stress as a possible mode of action for arsenic carcinogenesis. Toxicol Lett. 2003;137(1–2):3–13. [DOI] [PubMed] [Google Scholar]
- 53. Wang Y, Su H, Gu Y, et al. Carcinogenicity of chromium and chemoprevention: a brief update. Onco Targets Ther. 2017;10:4065–4079. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 54. Grashow R, Zhang J, Fang SC, et al. Toenail metal concentration as a biomarker of occupational welding fume exposure. J Occup Environ Hyg. 2014;11(6):397–405. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 55. O’Brien KM, White AJ, Sandler DP, et al. Do post-breast cancer diagnosis toenail trace element concentrations reflect prediagnostic concentrations? Epidemiology. 2019;30(1):112–119. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 56. Wu AC, Allen JG, Coull B, et al. Correlation over time of toenail metals among participants in the VA Normative Aging Study from 1992 to 2014. J Expo Sci Environ Epidemiol. 2019;29(5):663–673. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 57. Serra M, Columbano A, Ammarah U, et al. Understanding metal dynamics between cancer cells and macrophages: competition or synergism? Front Oncol. 2020;10:646. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 58. Mulware SJ. The mammary gland carcinogens: the role of metal compounds and organic solvents. Int J Breast Cancer. 2013;2013:640851. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.