Environment and Breast Cancer: Science Review
About
Exposure to multiple sources of polycyclic aromatic hydrocarbons and breast cancer incidence
White, A. J., Bradshaw, P. T., Herring, A. H., Teitelbaum, S. L., Beyea, J., Stellman, S. D., Steck, S. E., Mordukhovich, I., Eng, S. M., Engel, L. S., Conway, K., Hatch, M., Neugut, A. I., Santella, R. M., Gammon, M. D. Environ Int. 2016. 89-90, 185-92.
Topic area
Environmental pollutant - PAHs Air pollution, vehicle exhaust
Environmental pollutant - PAHs Air pollution, vehicle exhaust
Study design
Population-based case-control
Population-based case-control
Funding agency
NIH
NIH
Study Participants
Number of Cases
1,508
1,508
Menopausal Status
The menopausal status of women included in this study is listed here.
No analysis based on menopausal status
Number of Controls
Controls 1,556
Controls 1,556
Participant selection: Inclusion and exclusion criteria
Criteria used to select participants in the study.
Female residents of Nassau and Suffolk Counties (Long Island), NY, participating in the Long Island Breast Cancer Study Project, age 20 or older, English-speaking, newly diagnosed with in situ or invasive breast cancer in 1996-1997. Cases identified by regional hospital pathology laboratories. Controls had no breast cancer history and were matched by 5-year age group, identified by random-digit-dialing or Medicare records (for women 65 and older).
Exposure Investigated
Exposures investigated
Multiple sources of PAH exposure modeled together in a hierarchical regression. PAH sources considered were: Active smoking (ever, current, before first pregnancy), residential environmental tobacco smoke (ETS) (ever, from spouse), lifetime grilled/smoked
Multiple sources of PAH exposure modeled together in a hierarchical regression. PAH sources considered were: Active smoking (ever, current, before first pregnancy), residential environmental tobacco smoke (ETS) (ever, from spouse), lifetime grilled/smoked
How exposure was measured
Questionnaire, in-person GIS/geographic
Questionnaire, in-person GIS/geographic
Exposure assessment comment
Authors used 1995 estimates to determine PAH exposure from traffic which may not capture the relevant etiologic period; however exposure estimates from 1995 were correlated with estimates for the period 1960-1990 (not used in this analysis because of degree of missing data). Estimates for PAH exposure "profiles" such as PAHs from "indoor sources" were determined by linear combination of the relevant coefficients from the hierarchical regression model. This assumes no interaction between PAH sources. Among 1,508 cases and 1,556 controls in the LIBCSP, 87% of participants provided at least one address that could be successfully geocoded to the street level. Missing data for the high vehicular traffic variable (which required geocoded address) were imputed by logistic regression with predictors age, income, education and specifying non-informative normal prior distributions for the coefficients (mean = 0, variance = 1000).
Authors used 1995 estimates to determine PAH exposure from traffic which may not capture the relevant etiologic period; however exposure estimates from 1995 were correlated with estimates for the period 1960-1990 (not used in this analysis because of degree of missing data). Estimates for PAH exposure "profiles" such as PAHs from "indoor sources" were determined by linear combination of the relevant coefficients from the hierarchical regression model. This assumes no interaction between PAH sources. Among 1,508 cases and 1,556 controls in the LIBCSP, 87% of participants provided at least one address that could be successfully geocoded to the street level. Missing data for the high vehicular traffic variable (which required geocoded address) were imputed by logistic regression with predictors age, income, education and specifying non-informative normal prior distributions for the coefficients (mean = 0, variance = 1000).
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.
Age at menarche, parity, lifetime alcohol intake, education, income, age
Genetic characterization included
If the study analyzed relationships between environmental factors and inherited genetic variations, this field will be marked “Yes.” “No”, if not.
No
Strength of associations reported
These strengths of association are reported from one hierarchical logistic regression model specified in a Bayesian framework. ORs are thus reported with a 95% credible interval, rather than a 95% confidence interval. The interpretation for each of these credible intervals (CrI) is that there is a 95% chance that the true parameter lies within the reported interval.
Current active smoking (yes vs no): aOR 1.09 (95% CrI 0.91-1.30)
ETS from spouse (ever vs never): aOR 1.20 (95% CrI 1.03-1.40)
High lifetime intake of total grilled/bbq and smoked meats (55+ servings/year vs <55 servings/year): aOR 1.16 (95%CrI 0.97-1.38)
High lifetime intake of grilled/bbq and beef, pork and lamb (14+ servings/year vs <14 servings/year): aOR 1.14 (95% CrI 0.96-1.34)
High lifetime intake of smoked beef, pork, and lamb (22+ servings/year vs <22 servings/year): aOR 1.09 (95% CrI 0.95-1.26)
1995 estimated PAH exposure from vehicular traffic (≥ 95th vs < 95th percentile: aOR 1.25 (95% CrI 0.85-1.76)
Synthetic log burning (ever vs never): aOR 1.29 (95% CrI 1.06-1.57)
Exposure to PAH from "indoor sources"* vs those with "low PAH exposure"**: aOR 1.45 (95% CrI 1.02-2.04)
*Indoor sources = active smokers, residential ETS exposure from spouse, high overall intake of grilled/smoked foods and indoor stove/fireplace users.)
**Low PAH exposure = non-active smokers, no residential ETS, low intake of grilled/smoked foods, low exposure to PAHs from vehicular traffic, and no indoor stove/fireplace use
These strengths of association are reported from one hierarchical logistic regression model specified in a Bayesian framework. ORs are thus reported with a 95% credible interval, rather than a 95% confidence interval. The interpretation for each of these credible intervals (CrI) is that there is a 95% chance that the true parameter lies within the reported interval.
Current active smoking (yes vs no): aOR 1.09 (95% CrI 0.91-1.30)
ETS from spouse (ever vs never): aOR 1.20 (95% CrI 1.03-1.40)
High lifetime intake of total grilled/bbq and smoked meats (55+ servings/year vs <55 servings/year): aOR 1.16 (95%CrI 0.97-1.38)
High lifetime intake of grilled/bbq and beef, pork and lamb (14+ servings/year vs <14 servings/year): aOR 1.14 (95% CrI 0.96-1.34)
High lifetime intake of smoked beef, pork, and lamb (22+ servings/year vs <22 servings/year): aOR 1.09 (95% CrI 0.95-1.26)
1995 estimated PAH exposure from vehicular traffic (≥ 95th vs < 95th percentile: aOR 1.25 (95% CrI 0.85-1.76)
Synthetic log burning (ever vs never): aOR 1.29 (95% CrI 1.06-1.57)
Exposure to PAH from "indoor sources"* vs those with "low PAH exposure"**: aOR 1.45 (95% CrI 1.02-2.04)
*Indoor sources = active smokers, residential ETS exposure from spouse, high overall intake of grilled/smoked foods and indoor stove/fireplace users.)
**Low PAH exposure = non-active smokers, no residential ETS, low intake of grilled/smoked foods, low exposure to PAHs from vehicular traffic, and no indoor stove/fireplace use
Results Comments
The effect estimates from the single hierarchical model including all the measured sources (shown above) were generally somewhat reduced (especially for dietary measures and synthetic log burning) compared to models that considered each exposure individually. No significant associations found for ever active smoking, smoking before first pregnancy, every residential ETS, ever stove/fireplace use, or ever wood burning in the hierarchical model. ORs were significantly elevated for ingested PAHs, indoor stove/fireplace use, and "indoor source" PAHs.
The effect estimates from the single hierarchical model including all the measured sources (shown above) were generally somewhat reduced (especially for dietary measures and synthetic log burning) compared to models that considered each exposure individually. No significant associations found for ever active smoking, smoking before first pregnancy, every residential ETS, ever stove/fireplace use, or ever wood burning in the hierarchical model. ORs were significantly elevated for ingested PAHs, indoor stove/fireplace use, and "indoor source" PAHs.
Author address
Department of Epidemiology, University of North Carolina, Chapel Hill, NC, USA. Electronic address: whitea@unc.edu. Department of Nutrition, University of North Carolina, Chapel Hill, NC, USA. Department of Biostatistics, University of North Carolina, Cha
Department of Epidemiology, University of North Carolina, Chapel Hill, NC, USA. Electronic address: whitea@unc.edu. Department of Nutrition, University of North Carolina, Chapel Hill, NC, USA. Department of Biostatistics, University of North Carolina, Cha
Reviewers Comments
Authors note that results were similar (though with a higher effect estimate for high intake of grilled/smoked meats) when analyses were limited to postmenopausal women with ER+ tumors, but did not show data. Very few women in this study population (1.3% of cases and 1.7% of controls) were characterized as having a "low PAH exposure" profile.
Authors note that results were similar (though with a higher effect estimate for high intake of grilled/smoked meats) when analyses were limited to postmenopausal women with ER+ tumors, but did not show data. Very few women in this study population (1.3% of cases and 1.7% of controls) were characterized as having a "low PAH exposure" profile.