Evidence From Humans
Print this page
Occupational exposure to polychlorinated biphenyls and risk of breast cancer
Silver, S. R., Whelan, E. A., Deddens, J. A., Steenland, N. K., Hopf, N. B., Waters, M. A., Ruder, A. M., Prince, M. M., Yong, L. C., Hein, M. J., Ward, E. M. Environ Health Perspect. 2009. 117:2, 276-82.
Topic area
Environmental pollutant - PCBs Occupation
Study design
Retrospective cohort
Study Participants
Number of Cases
Overall: 281 Questionnaire subcohort: 201 Restricted subcohort: 145
Menopausal Status
The menopausal status of women included in this study is listed here.
No analysis based on menopausal status
Number in Cohort
Cohort: 5,752 Questionnaire subcohort: 3,952 Restricted subcohort: 3,141
Cohort participation rate
69% of cohort members or their proxies participate
Participant selection: Inclusion and exclusion criteria
Criteria used to select participants in the study.
The full cohort comprised 13,321 women employed at one of three capacitor manufacturing plants in NY, MA, and IN, but this analysis was restricted to women who had been employed for at least 1 year. These women were followed for breast cancer diagnosis through December 1998 in state cancer registries where the plants were located (MA 1982-1998; NY 1976-1998; IN 1987-1998), as well as in Florida and California cancer registries in an effort to capture individuals who moved. Additional cases were identified by self- or proxy-report from among those who responded to a questionnaire sent out in 1998-2000, and by mortality follow-up (MA & NY 1940-1998, IN 1957-1998) using records from SSA, IRS, NDI, US postal service and credit bureaus and NDI-plus. Analyses were conducted separately for the full cohort, the "questionnaire subcohort", and a further "restricted subcohort" of women in the questionnaire cohort who had complete data on smoking, parity, age at first live birth, family history of breast cancer, hormone use, and age started hormone use.
Comment about participation selection
Significant effort was made to collect questionnaire data, but about 20% of the cohort who were not known to be deceased at the end of follow-up did not respond to the questionnaire. Analyses conducted in the "questionnaire subcohort" and "restricted subcohort" include valuable information on individual risk factors, but the results may be biased if factors related to the exposure and outcome were different among those who returned the questionnaire/provided complete data and those who did not.
Exposure Investigated
Exposures investigated
PCB exposure for each worker was determined using a plant-specific job exposure matrix based on detailed work histories and modified to reflect changes in PCB usage during distinct "eras". The matrices drew from work history records, air samples, historic
How exposure was measured
Job history Other: Job exposure matrix (JEM)
Exposure assessment comment
Plant-specific job exposure matrices are a strength. A greater percentage of workers at the Massachusetts and New York plants fell into the highest quartile of estimated cumulative exposure, compared to workers at the Indiana plant. Estimated cumulative exposure and serum PCB levels among 221 workers who gave blood samples in 1977 were somewhat correlated (r = 0.37, p < 0.00001)
Breast cancer outcome investigated
Primary incident breast cancer
Ethnic groups with separate analysis
If this study provided a separate analysis by ethnic or racial group, the groups are listed here.
Predominantly white; separate analysis with nonwhite women
Confounders considered
Other breast cancer risk factors, such as family history, age at first birth, and hormone replacement therapy use, that were taken into account in the study.
Family history, ethnicity, smoking status, alcohol use, self vs proxy questionnaire completion, plant (MA, NY or IN, birth cohort, menopausal status, parity, age at first live birth, HRT status, and age at HRT use were considered in analyses of the restr
Genetic characterization included
If the study analyzed relationships between environmental factors and inherited genetic variations, this field will be marked Yes. No, if not.
Strength of associations reported
White women (131 cases)
Per 1,000 unit-years of cumulative PCB exposure: aHR 1.00 (95% CI 1.00 - 1.00)
Years exposed: aHR 1.02 (95% CI 0.99-1.05)
Categorical cumulative exposure (ref = 0 to < 0.16 per 1,000 unit-years):
0.16 to < 0.46 per 1,000 unit-years: aHR 1.36 (95% CI 0.83-2.23)
0.46 to < 1.6 per 1,000 unit-years: aHR 1.15 (95% CI 0.70-1.89)
≥ 1.6 per 1,000 unit-years: aHR 1.27 (95% CI 0.75-2.17)

Nonwhite women (14 cases)
Per 1,000 unit-years of cumulative PCB exposure: aHR 1.33 (95% CI 1.14 - 1.56)
Years exposed: aHR 1.13 (95% CI 1.03 - 1.23)
Categorical cumulative exposure (ref = < 0.47 per 1,000 unit-years, 4 cases):
0.47 to < 3.9 per 1,000 unit-years (3 cases): aHR 0.60 (95% CI 0.12-3.01)
3.9 to < 5.8 per 1,000 unit-years (4 cases): aHR 7.65 (95% CI 1.11-52.8)
≥ 5.8 per 1,000 unit-years (3 cases): aHR 22.3 (95% CI 2.38-209)
Results Comments
External comparisons to SEER data (1970-1998) with 10 year lag showed SIR 0.81 (95% CI 0.71–0.92) for the full cohort, SIR 0.80 (95% CI 0.70-0.90) for white women and SIR of 1.94 (95%CI 0.77-3.99; n = 7) for non-white women. However the general population may not be an appropriate comparison group (e.g. could lead to underestimation due to a healthy worker effect).
BACKGROUND: Despite the endocrine system activity exhibited by polychlorinated biphenyls (PCBs), recent studies have shown little association between PCB exposure and breast cancer mortality. OBJECTIVES: To further evaluate the relation between PCB exposure and breast cancer risk, we studied incidence, a more sensitive end point than mortality, in an occupational cohort. METHODS: We followed 5,752 women employed for at least 1 year in one of three capacitor manufacturing facilities, identifying cases from questionnaires, cancer registries, and death certificates through 1998. We collected lifestyle and reproductive information via questionnaire from participants or next of kin and used semiquantitative job-exposure matrices for inhalation and dermal exposures combined. We generated standardized incidence ratios (SIRs) and standardized rate ratios and used Cox proportional hazards regression models to evaluate potential confounders and effect modifiers. RESULTS: Overall, the breast cancer SIR was 0.81 (95% confidence interval, 0.72-0.92; n = 257), and regression modeling showed little effect of employment duration or cumulative exposure. However, for the 362 women of questionnaire-identified races other than white, we observed positive, statistically significant associations with employment duration and cumulative exposure; only smoking, birth cohort, and self- or proxy questionnaire completion had statistically significant explanatory power when added to models with exposure metrics. CONCLUSIONS: We found no overall elevation in breast cancer risk after occupational exposure to PCBs. However, the exposure-related risk elevations seen among nonwhite workers, although of limited interpretability given the small number of cases, warrant further investigation, because the usual reproductive risk factors accounted for little of the increased risk.
Author address
National Institute for Occupational Safety and Health, Division of Surveillance, Hazard Evaluations and Field Studies, Cincinnati, Ohio 45226, USA. SSilver@cdc.gov
Reviewers Comments
The analysis of nonwhite women is limited by the small number of cases. Furthermore, the majority of nonwhite workers were from the MA plant. Estimated mean exposures at this plant were more than 2x greater than at the NY plant an order of magnitude higher than at the Indiana plant. Unadjusted analyses of effects of cumulative exposure and years exposed were pretty consistent when stratified by cohort type (full, questionnaire subcohort, restricted subcohort). The authors note probable under ascertainment of incident breast cancer cases, especially nonfatal cases diagnosed during the years when follow-up was limited to mortality, a period which lasted until 1982 for MA, 1976 for NY, and 1987 for IN.
Privacy notice   |   Copyright statement