Residential Proximity to Carcinogenic Industrial Air Emissions and Breast Cancer Incidence in a United States-Wide Prospective Cohort

Abstract

Purpose: To evaluate whether residential proximity to air emissions of correlated industrial carcinogens, both singly and in combination, are associated with breast cancer incidence. Methods: Using the US Environmental Protection Agency's Toxics Release Inventory, we estimated the 10-year annual average air releases of 26 known or probable carcinogens within 3, 5, and 10 km of Sister Study participants' baseline residences (n = 50,343, 2003–2009). We used Cox proportional hazards regression to estimate adjusted hazard ratios (HR) and 95% confidence intervals (CI) for the association between ambient concentrations of each individual compound and incident breast cancer. To assess mixtures, we applied an exposure continuum mapping (ECM) framework to identify latent mixture profiles via a self-organizing map and assessed whether these profiles were related to incident breast cancer by estimating a joint exposure-response function with generalized additive models. Results: During follow-up (mean = 11.6 years), 4,282 breast cancer cases were identified. The exposure prevalence for compounds emitted within 3-km of participants' residences varied from <1%–15%. For individual compounds, HRs for the association between quantiles of emission levels and breast cancer were largely null, except for cadmium, vinyl chloride and asbestos (e.g., asbestos 3-km HR>median vs. no exposure = 2.61, 95% CI: 1.37–5.09). Our application of ECM identified 49 profiles that explained 79% of the variance in emission patterns observed within 3-km of participant residences. Profiles revealed that relatively high levels of exposure to several compounds were rare (<1%) and most participants resided in locations defined by low emissions patterns. Estimation of a joint exposure-response surface indicated that breast cancer was not significantly related to changing patterns of the composition of emission mixtures (P = 0.31). Conclusions: Preliminary results suggest that breast cancer incidence may be related to emissions of certain individual industrial carcinogens, particularly asbestos. Examination of complex exposure scenarios revealed that high simultaneous exposure to many compounds was rare, and that our identified latent patterns of these emissions were not associated with breast cancer.