Long-term Exposure to Fine Particulate Matter and Breast Cancer Incidence in the Danish Nurse Cohort Study

Traffic-related air pollution is a risk factor for lung cancer and classified as a carcinogen to humans. Experimental data provide some evidence of a link between a number of carcinogens in air pollution and breast cancer, including polycyclic aromatic hydrocarbons, which can cause mammary tumors in laboratory animals (1). Epidemiologic data are mixed, with case–control studies (2) detecting association and cohort studies (3–5) finding no association between air pollution and breast cancer incidence. The latest Canadian case–control study found a 26% to 32% increased risk of premenopausal, and 7% to 10% increased risk of postmenopausal breast cancer, per 10 ppb increase in nitrogen dioxide (NO₂; ref. 2). Three cohort studies did not detect association between breast cancer incidence and NO₂ (3), PM with diameter <2.5 μg/m³ (PM_2.5), or PM with diameter <10 μg/m³ (PM₁₀; refs. 3, 4), or nitrogen oxides NO_x (5). Only two studies had data on PM_2.5 and PM₁₀ (3, 4). Here, we examined the association between long-term exposure to PM_2.5, PM₁₀, and NO₂ at the residence and incidence of breast cancer in the nationwide, prospective Danish Nurse Cohort (DNC).

The DNC consists of 28,731 Danish female nurses (>44 years), recruited in 1993 (19,898) or 1999 (8,833) when they answered a questionnaire on working conditions, weight, height, lifestyle, and reproductive factors. Using the unique personal identifying number, we linked nurses to the Central Population Register to obtain residential address history (1971–2013) and vital status information and to the Danish Cancer Register for diagnosis of primary invasive breast cancer (International Classification of Diseases-10 code C50) between cohort baseline (1993 or 1999) and 2013, and any other cancer prior to cohort baseline. We used the Danish integrated air pollution modeling system (THOR and AirGIS) to estimate human exposure to outdoor air pollution at address level (6). We estimated annual mean concentration of PM_2.5, PM₁₀, and NO₂ at the residential addresses for nurses with complete information on residential address history 80% of the time between 1990 and 2013 and used the 3-year running mean of each pollutant concentrations as the main exposure proxy. We used the time-varying Cox proportional hazards regression to model association between breast cancer incidence and air pollution, with age as the underlying time scale, and stratified by year of birth, in three steps: age adjusted; adjusted additionally for smoking, alcohol consumption, physical activity, and body mass index (BMI); and additionally adjusted for age at first birth, parity, age at menarche, hormone therapy (HT), and oral contraceptive use (described in Table 1). Effect modification by smoking status, BMI, menopausal status at baseline, HT use, and urbanization level was examined by including interaction terms in the model. Effects are reported by HRs and 95% confidence intervals (CI), per interquartile range increase in 3-year mean of pollutant levels. The methods for establishing the cohort (7) and some prior accomplishments related to studying health effect of air pollution (6, 8) have been described previously.

Of the 28,731 nurses, we excluded 1,924 with cancer and 3 with inactive vital status before cohort baseline, and 3,927 with missing data on address or confounder, leaving 22,877 for final analyses. A total of 1,145 nurses developed breast cancer during mean follow-up of 16 years. Nurses with breast cancer were more likely nulliparous, HT users, and current smokers (Table 1). We found no association between breast cancer incidence and PM_2.5, PM₁₀, or NO₂ (Table 2), and no interaction by smoking status, BMI, or level of urbanization (results not shown). We detected stronger (P_interaction = 0.07) association between breast cancer and PM_2.5 in premenopausal (1.06; 0.94–1.18) than in postmenopausal (0.93; 0.85–1.05) women, and in past (1.19; 0.97–1.44) and current (1.05; 0.92–1.20) than in never HT users (0.95; 0.85–1.05; P = 0.06).

We found no association between long-term exposure to air pollution and breast cancer incidence in DNC in agreement with other cohort (2, 3, 5), and in contrast to a case–control study (2). We found moderately stronger associations in pre- than in postmenopausal women, in agreement with two studies (2, 3). We present novel findings of enhanced association between PM_2.5 and breast cancer in past and ever HT users, which needs replication.

Ours and most other recent studies (2–5) examined air pollution exposure during adulthood, relatively close to breast cancer onset, which may not be the relevant exposure window for the development of breast cancer. A study found significantly increased breast cancer risk with exposure to traffic emissions at first childbirth, but no risk associated with later exposure windows (9), in line with growing evidence of relevance of early smoking on breast cancer risk (10).

We benefited from having a nationwide cohort of female nurses, with large air pollution exposure contrasts, detailed information on breast cancer risk factors, objective validated information on breast cancer onset, and a state-of-the-art high-resolution air pollution model with data on PM. Weakness of our study is lack of data on air pollution levels during puberty and early adulthood.

Results of our study contribute to growing evidence of a lack of association between exposure to air pollution in adulthood and postmenopausal breast cancer, but association with early exposure windows cannot be excluded. Studies with data on exposure to air pollution early in life and premenopausal breast cancer are needed.

No potential conflicts of interest were disclosed.

Conception and design: Z.J. Andersen, K.K. Andersen, J. Brandt, E.V. Bräuner

Development of methodology: K.K. Andersen, J. Brandt, E.V. Bräuner

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): K.K. Andersen, J. Brandt, M. Ketzel, O. Hertel

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): L. Ravnskjær, K.K. Andersen, J. Brandt, M. Ketzel

Writing, review, and/or revision of the manuscript: Z.J. Andersen, L. Ravnskjær, S. Loft, J. Brandt, T. Becker, M. Ketzel, O. Hertel, E. Lynge, E.V. Bräuner

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): Z.J. Andersen, J. Brandt, T. Becker, M. Ketzel

Study supervision: Z.J. Andersen, E.V. Bräuner

We thank the board of the Danish Nurse Cohort for the permission to use the data.

Z.J. Andersen received a grant from Novo Nordisk Foundation (NNF6935) and J. Brandt received a grant for project NordForsk under the Nordic Programme on Health and Welfare - Project NordicWelfAir#75007.

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.