University of Southern California Dissertations and Theses

Air pollution and breast cancer survival in California teachers: using address histories and individual-level data

(USC Thesis Other)

Air pollution and breast cancer survival in California teachers: using address histories and individual-level data

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content

AIR POLLUTION AND BREAST CANCER SURVIVAL IN CALIFORNIA
TEACHERS: USING ADDRESS HISTORIES AND INDIVIDUAL-LEVEL
DATA

By

Minhao Wang

A Thesis Presented to the
FACULTY OF THE USC KECK SCHOOL OF MEDICINE
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
MASTER OF SCIENCE
(APPLIED BIOSTATISTICS AND EPIDEMIOLOGY)

May 2021

Copyright 2021 Minhao Wang

ii
Table of Contents
List of Tables ................................................................................................................................. iii
List of Figures ................................................................................................................................. v
Abstract .......................................................................................................................................... vi
Introduction ..................................................................................................................................... 1
Methods........................................................................................................................................... 3
Study population ......................................................................................................................... 3
Air pollution exposure assignment ............................................................................................. 5
Covariates ................................................................................................................................... 7
Outcome ...................................................................................................................................... 8
Statistical analysis ....................................................................................................................... 8
Results ........................................................................................................................................... 11
Discussion ..................................................................................................................................... 21
References ..................................................................................................................................... 25
Appendix ....................................................................................................................................... 28

iii
List of Tables
Table 1 Demographic, tumor, treatment, and health behavior characteristics of CTS
participants newly diagnosed with first primary breast cancer from 1995-2015, by stage of
diagnosis. ................................................................................................................................... 11
Table 2 Air pollution exposure summaries by stage at diagnosis .......................................... 13
Table 3 Adjusted* HRs (95%CI) for all-cause and breast cancer-specific mortality
associated with 1 SD§ increase in air pollution average of 12 months after diagnosis , stratified
by stage and baseline hazard by age group at diagnosis ............................................................... 14
Table 4 Adjusted* HRs (95%CI) for all-cause mortality associated with 1 SD§ increase in
each air pollutant summary, stratified by stage and baseline hazard by age group at diagnosis .. 15
Table 5 Adjusted* HRs (95%CI) for breast cancer-specific mortality associated with 1 SD§
increase in each air pollutant summary, stratified by stage at diagnosis ...................................... 16
Table 6 Secondary analysis only using address at diagnosis: adjusted* HRs (95%CI) for all-
cause and breast cancer-specific mortality associated with 1 SD§ increase in air pollution
average of 12 months after diagnosis, stratified by stage and baseline hazard by age group at
diagnosis ................................................................................................................................... 18
Table 7 Secondary analysis only adjusting for CCR covariates: adjusted* HRs (95%CI) for
all-cause and breast cancer-specific mortality associated with 1 SD§ increase in air pollution
average of 12 months after diagnosis, stratified by stage and baseline hazard by age group at
diagnosis ................................................................................................................................... 19
Table S 1 Correlation coefficient of each air pollutant average change over follow-up with its
relative air pollution exposure summaries. ................................................................................... 31
Table S 2 Diagnosis address-based air pollution exposure summaries by stage at diagnosis .. 32
Table S 3 Secondary analysis only using address at diagnosis: adjusted* HRs (95%CI) for all-
cause mortality associated with 1 SD§ increase in each air pollutant summary stratified by stage
and baseline hazard by age group at diagnosis. ............................................................................ 33
Table S 4 Secondary analysis only using address at diagnosis: adjusted* HRs (95%CI) for
breast cancer-specific mortality associated with 1 SD§ increase in each air pollutant summary
stratified by stage at diagnosis. ..................................................................................................... 35
Table S 5 Secondary analysis only adjusting for CCR covariates: adjusted* HRs (95%CI) for
all-cause mortality associated with 1 SD§ increase in each air pollutant summary, stratified by
stage and baseline hazard by age group at diagnosis. ................................................................... 37

iv
Table S 6 Secondary analysis only adjusting for CCR covariates: adjusted* HRs (95%CI) for
breast cancer-specific mortality associated with 1 SD§ increase in each air pollutant summary,
stratified by stage at diagnosis. ..................................................................................................... 39

v
List of Figures
Figure 1 Flow chart of study participants ....................................................................................... 4
Figure S 1 Missingness of PM2.5 during follow-up ..................................................................... 28
Figure S 2 Missingness and time point of censoring for participants met the 75% completed air
pollution data criteria for (a) NO2, (b) O3, (c) PM10, and (d) PM2.5. ............................................. 29

vi
Abstract
Background: Breast cancer has been a leading cause of cancer death in women for
decades. Air pollution, a complex mixture of compounds, can affect carcinogenesis in the human
breast. Some studies have investigated the relationship between air pollution exposures and
breast cancer incidence, but little information is available on the relationship between air
pollution exposures after diagnosis and breast cancer survival. Research on this relationship
could inform air pollution interventions to better protect people diagnosed with breast cancer in
the future.
Methods: We identified 6,539 participants of the California Teachers Study (CTS) with
first primary breast cancer newly diagnosed between 1995 to 2015 with high-quality individual-
level data from the CTS and data from linkage to the California Cancer Registry (CCR).
Residential addresses were geocoded and ambient air pollution exposures after diagnosis,
including nitrogen dioxide (NO2, ppb), ozone (O3, ppb), particulate matter with diameter <10 𝜇 m
(PM10, 𝜇 g/m
3
), and 2.5 𝜇 m (PM2.5, 𝜇 g/m
3
), were assigned using inverse distance weighted
averages of central site monitor data. Follow-up time was calculated from the date of diagnosis
to either the date of death, the date of study end (December 31, 2015), the date of moving out of
California, or the date of starting missing air pollution assignment, whichever occurred first. In
our primary analysis, Cox proportional hazards models and Fine and Gray competing risk
models were used to estimate hazard ratios (HRs) relating air pollutant exposure averages to all-
cause mortality and breast cancer-specific mortality respectively, by tumor stage at diagnosis
(localized, regional, and distant). In secondary analysis, we evaluated the sensitivity of our
results to using only data from the CCR, which has less individual-level data (e.g., no smoking
history and only address at diagnosis).

vii
Results: After adjusting for covariates, the hazard for breast-cancer specific mortality for
participants diagnosed at the regional stage was 1.14 (95% CI: 1.00, 1.29) times higher per SD
increase in average NO2 over the 12 months after diagnosis. We didn’t detect other significant
associations between mortality with average O3, PM10, or PM2.5 over 12 months after diagnosis.
Statistically significant associations between follow-up period average air pollutants and all-
cause mortality were attenuated when the follow-up period was truncated to a maximum of 5
years. Using limited individual-level data and address only at diagnosis produced results similar
to the primary analysis.
Conclusions: Our analyses do not provide clear evidence for associations of air pollution
exposures after diagnosis with breast cancer survival. However, our findings motivate future
analyses using time-varying annual average exposures since results suggest a bias in analyses
using follow-up period average exposures for pollutants with downwards trends during the study
period, which could potentially be remedied using time-varying exposures.

1
Introduction
Breast cancer has been the most frequently diagnosed cancer and the second leading
cause of cancer death in the United States women for decades, following lung cancer. In
particular, the state of California sees a large number of breast cancer cases and deaths. It is
estimated that 10.9% of breast cancer new cases and 10.8% of breast cancer deaths in the United
States will occur in the state of California in 2021.
1

While many previous epidemiological studies have focused on established risk factors for
incident breast cancer (e.g., aging, alcohol use, smoking, and family history of breast cancer
etc.),
2-4
ambient air pollution has emerged as a new modifiable risk factor of concern for breast
cancer incidence.
5,6
Air pollution is a mixture of many compounds, including polycyclic aromatic
hydrocarbons (PAHs), metals, and benzene, which might influence breast carcinogenesis.
Among them, PAHs is the most well-studied compound, which can bind to DNA and form
adducts in breast tissue. Evidence have shown that PAHs and metal are active estrogenic species,
interfering with estrogen receptor signaling,
7,8
and can increase the oxidative stress.
9
It has been
reported that inhaled toxicants can reach human breast tissue in 30 minutes by a study which
measured nicotine and cotinine in breast fluid after smoking.
10
Outdoor air pollution has been
classified as a Group 1 carcinogen to humans by the International Agency for Research on
Cancer (IARC) in 2013.
11
Based on these adverse effects of air pollution, it is reasonable to
speculate on its potential association with breast cancer incidence.
A review, focusing on the role of air pollution on breast cancer, strongly suggested that
nitrogen dioxide (NO2, ppb) and nitrogen oxides (NOx, ppb) increased the risk of breast cancer
incidence, while particulate matter (PM) had less evidence.
12
PAHs and particulate matter with

2
diameter <2.5 𝜇 m (PM2.5, 𝜇 g/m
3
) were suggested to increase the risk of female breast cancer
incidence in urban areas.
13
Additionally, motor vehicle density was found positively associated
with breast cancer incidence rates.
14

With increasing evidence on the association between air pollution exposure and breast
cancer incidence, we may further question whether the biological mechanisms under the
development of cancer might also cause cancer progression and mortality. Existing evidence
supports our hypothesis. NO2, ozone (O3, ppb), particulate matter with diameter <10 𝜇 m (PM10,
𝜇 g/m
3
), and PM2.5 have been observed associated with all-cause mortality and lung cancer-
specific mortality.
15
A study addressing the role of PM2.5 on hepatocellular carcinoma (HCC)
survival also suggests the adverse effect of PM2.5 on cancer survival.
16
Other previous studies
consistently implied PM2.5 is associated with shorter survival in breast cancer
17,18
and other
cancers.
19

In this study, we hypothesized that exposure to higher levels of air pollution after
diagnosis of breast cancer is associated with reduced all-cause survival and breast cancer-specific
survival. We assessed the association in a population-based cohort study of California female
teachers. The state of California makes up about 12% population in the United States and has
wide range of air pollution exposures, which accounts for large cases of breast cancer diagnosed
in California and being a suitable region for our study. The study was conducted by relating air
pollution exposure histories to all-cause mortality and breast cancer-specific mortality, by stage
and pollutant.

3
Methods
Study population
The California Teachers Study (CTS) was designed to study the high incidence of breast
cancer in female teachers and was initially supported by the State of California through revenue
from cigarette sales tax. The CTS is a closed cohort study. The 133,477 participants recruited at
baseline in 1995-1996 include female teachers, administrators, school nurses, and other active
members of the California State Teachers’ Retirement System (CalSTRS) living throughout
California. At baseline, their median age was 53. They represent various socioeconomic
backgrounds, with a majority being white (87%). A full description of CTS cohort can be found
elsewhere.
20
Participants have provided health and behavior information through 6 CTS
questionnaires, administered during the following years: 1995-1996, 1997-1998, 2000-2002,
2005-2008, 2012-2015, and 2017-2019.
21

The CTS regularly links with the California Cancer Registry (CCR) to collect cancer-
related data on CTS participants. The CCR collects information about almost all cancers
diagnosed in California, as a statewide population-based cancer registry. CCR contain high-
quality individual-level data on patient demographics, date of diagnosis, tumor characteristics at
diagnosis, treatment occurring <6 months after diagnosis, and routine follow-up on vital status.
Patient vital status is updated through information sharing with reporting hospitals and linkage
with a variety of administrative records.
22

CTS participants eligible for inclusion in this study were the 8,192 women newly
diagnosed with first primary breast cancer during the study period of Nov 21
st
1995 to Dec 31
st

2015. After excluding participants with data error (N=4), death on the date of breast cancer

4
diagnosis (N=7), diagnosis of breast cancer on Dec 31
st
, 2015 (N=2), non-carcinoma cases
(histology ICD-O-3 >8800) (N=14), and in situ (N=1581) or unknown breast cancer stage
(N=45), 6,539 participants remained for analysis. Additional exclusions based on air pollution
exposure assignment availability are listed below in Figure 1.

Figure 1 Flow chart of study participants

Censored (N=2)
Move out of CA (N=366)
Died of breast cancer (N=569)
Died of other causes (N=896)
End of study (Dec 31
st
, 2015)
(N=4,706)
The original CTS population enrolled at baseline
(N=133,477)
Newly diagnosed with primary breast cancer during the study (Nov 21
st
,
1995 to Dec 31
st
, 2015)
(N=8,192)
Included in analysis
(N=6,539)
Non-carcinoma cases (histology ICD-O-
3 > 8800)(N=14)
In situ stage (N=1,581)
Unknown breast cancer stage (N=45)
Data error (N=4)
Participant died on the date of breast cancer
diagnosis (N=7)
Participant diagnosed as breast cancer on
12/31/2015 (N=2)
Participants moved out of California before
breast cancer diagnosis (N=134)
Customary exclusion (N=7)
Participants without breast cancer
(N=111,717)
Participants live outside of California at
baseline (N=8,332)
Prevalent breast cancer at baseline
(N=4,270)
Participants had other cancer preceding
breast cancer (N=825)

5
Air pollution exposure assignment
Residential address histories were available on CTS participants from: completed CTS
questionnaires; participant’s self-report through call, mail, or e-mail; subproject recruitment
information; USPS address records; or linkage with a credit agency.
5,20
We ranked these
multiple sources’ reliability into 5 levels, with CTS questionnaires as the highest and linkages
with credit agency as the lowest, and kept the most reliable address or addresses for each person-
month. Within remaining addresses, address records with the same geocoded location as the
other address(es) were deleted in the same person-month. If multiple addresses were kept for a
given participant-month, the mean exposure across locations was assigned. If a participant’s
address was missing during the follow-up or the last known address ended before Dec 31
st
, 2015,
the formerly known address was carried forward until the next known address was available or to
the end of follow up, and monthly exposure values from the extended location were used to fill
in exposures during the missing time period. After these processing steps, all study participants
had address locations for their whole follow-up period and no participants were excluded due to
missing address histories.
Air pollution data in California was obtained from the U.S. Environmental Protection
Agency’s (EPA) Air Quality System (AQS).
23
Hourly measurements of NO2, O3, PM10, and
PM2.5 were summarized as 24-hour averages for NO2, PM10, and PM2.5, average 8-hour daily
maximum for O3. Monthly average exposures were assigned to each participant’s residence
location from the month of diagnosis through to the end of follow-up. Exposures were spatially
interpolated to a participant’s residence location using inverse distance weighting, with the decay
parameter equal to the inverse of the square of the distance of the residence from each
monitoring stations, from the up to 4 closest air quality monitoring stations with a maximum

6
radius of 50km from the residence locations, requiring that the nearest monitor be within 25km.
Analyses including PM2.5 consider only the subset of participants diagnosed after January 1
st

2000 since PM2.5 routine monitoring did not start until 1998.
Air pollution assignments were not available at all participant locations at all study period
months. For participants who had missing monthly data during their follow-up (PM2.5 for
illustration here), their missingness is shown in Figure S1, each line representing 1 participant.
There were long periods of missingness for some participants, while some participants had
sporadic missing months or unavailable address at the end of study. After carrying forward the
previously known address and calculating percent of completeness on pollutant exposure data,
only participants with air pollution assignments available for ≥75% of the months of their
follow-up period were included in the final analysis. This resulted in the exclusion of 247, 69,
105, and 33 participants for NO2, O3, PM10, and PM2.5 respectively. For the remaining
participants (up to 25% of follow-up months missing), their follow-up period was censored at the
first month of any missing exposure, resulting in new exposure availability-based censoring of
164, 71, 120, and 53 participants for NO2, O3, PM10, and PM2.5 respectively. Note that after
imposing the exposure availability-based censoring, we had to drop additional participants since
censoring occurred in the month of diagnosis or month immediately following diagnosis (40 for
NO2, 42 for O3, 35 for PM10, 36 for PM2.5). Figure S2 displays remaining participants’
missingness, if available, and when they were censored in our analysis by pollutant.
We calculated four summaries of participants’ exposure histories: (1) average of the 12
months before diagnosis (some participants had less than 12 months available for the calculation
due to missingness: 98 for NO2, 91 for O3, 97 for PM10, and 386 for PM2.5), (2) average of the 12
months after diagnosis (a group of participants were assigned this summary calculated from less

7
than 12 months because of short follow-up or early censoring: 389 for NO2, 385 for O3, 387 for
PM10, and 366 for PM2.5), (3) average of the whole follow-up period after diagnosis, and (4)
average of the follow-up period, truncating to a maximum of 5 years. For sensitivity analyses, we
also calculated the change over the participant’s follow-up period, defined as the difference
between the final 12 months of the participant’s follow-up period and the first 12 months after
diagnosis. This summary quantifies the magnitude of air pollution decline (or, less frequently,
increase) throughout a participant’s follow-up period.
Covariates
From the CCR, individual-level data are available on: age at diagnosis (categorized as: 20
to 49, 50 to 69, over 70 years old), race/ethnicity (non-Hispanic White, non-Hispanic African
American, Hispanic, non-Hispanic Asian/Pacific Islander and other/mixed), marital status
(married/living together, divorced/separated, widowed, never married, and unknown), year of
diagnosis (categorized as:1995 to 2000, 2001 to 2005, 2006 to 2010, and 2011 to 2015), and
initial treatment (surgery, radiation, chemotherapy, and/or hormone therapy versus none), breast
cancer subtype (ER/PR positive, both negative, and other/unknown). Based on address at
diagnosis, the CCR has geocode participants to provide information on area-level: rural-urban
commuting area (RUCA, dichotomized as: metropolitan core, non-metropolitan core) based on
census tract and socioeconomic status (SES, categorized into quintiles) based on block group.
The RUCA code, at the census tract-level, is based on the size and direction of the main
commuting traffic, ranging from 1 (urban) to 10 (rural), calculated from population density,
urbanization and daily commuting measurements.
24
SES, at the census block group level, used
valid area-level measures from Census 2000 data and American Community Survey 5-years
estimates from 2007 to 2011. Additional individual-level health/behavior data are available from

8
the CTS on: body mass index (BMI, ≤18.4 as underweight, 18.5-24.9 as normal, 25.0-29.9 as
overweight, ≥30.0 kg/m
2
as obese, unknown) exercise (<1, 1-2.9, 3-4.9, ≥5 hours/week,
unknown), alcohol use, smoke exposure (no exposure, passive exposure, former self-exposure,
and current self-exposure), family history of breast cancer. BMI, exercise, and alcohol use data
were obtained from the most recent questionnaire before diagnosis. Smoke exposure and family
history of breast cancer were available from the baseline questionnaire in 1995 to 1996.
Outcome
The International Classification of Disease, 9
th
Revision (ICD-9) and 10
th
Revision (ICD-
10) codes were assigned to deaths for 1995-1998 and deaths after 1998 respectively. We
considered both all-cause mortality and breast cancer-specific mortality (ICD-9 codes 233.0 and
174.0-174.9; ICD-10 codes D05.0-D05.9 and C50.0-C50.9). Survival time was calculated from
the date of diagnosis to the date of death or censoring due to: study end (December 31, 2015),
participant moving out of California, or missing air pollution assignment. In sensitivity analyses
focusing on survival in the period immediately following diagnosis, we censored all participants
to have a maximum of five years of follow-up.
Statistical analysis
To study air pollution exposure associations with all-cause mortality, Cox proportional
hazards models were fit separately for each stage at diagnosis (localized, regional, distant).
Preliminary analyses suggested that aforementioned covariates potentially confounded the risk of
mortality. Thus, each model including one air pollution exposure summary at a time, adjusted for
the aforementioned covariates and stratified by baseline hazard by age group at diagnosis. To
study air pollution exposures with breast-cancer specific mortality, we fit Fine and Gray
competing risk models.
25
The Fine and Gray model accounts for competing risks by modifying

9
the risk set at time t, which continually includes individuals who had already had competing
events (any deaths except causing by breast cancer in this study) prior to time t in the risk set,
along with those free of events. Again, we fit separate models for each stage at diagnosis,
adjusting for covariates. Since baseline hazard stratification is not allowed for Fine and Gray
competing risk models in Stata, these models were additionally adjusted by age group at
diagnosis.
Our primary models of interest were those using air pollution exposures in the 12 months
after diagnosis, since they targeted our exposure period of interest (after diagnosis) and were not
subject to potential bias from the declines in air pollution observed over the study period
(especially PM2.5, PM10, and NO2 but not O3). We conducted sensitivity analyses considering
exposures in the 12 months prior to diagnosis. Additional sensitivity analyses were aimed to
understand the potential bias from long-term trend of declining air pollution by including:
average exposure over the whole follow-up period (or truncated to up to 5 years of follow-up), or
including both the 12 months after diagnosis as well as the change over the follow-up period
(difference between the final 12 months of the participant’s follow-up period and the first 12
months after diagnosis).
Secondary analyses addressed questions about the sensitivity of our results to using the
rich individual-level data available in the CTS versus the more limited individual-level data that
would have been available if we only had access to information in the CCR. First, we evaluated
the sensitivity of our results to having access to residential address history rather than only
address at diagnosis. Specially, we re-calculated a participant’s exposure summaries using
monthly exposures based only on the address at diagnosis (assuming no moves) and then we re-
ran our primary models. Second, we evaluated the sensitivity to adjustment for individual-level

10
potential confounding variables from the CTS by re-fitting our primary models by including only
CCR covariates and excluding the CTS covariates (BMI, exercise, alcohol use, smoke exposure,
and family history of breast cancer). To quantify the sensitivity of our results to these
modifications, we reported the percent change in exponentiate of standard deviation (SD)
rescaled 𝛽 estimates.
All air pollutant hazard ratios (HRs) are scaled to a 1 SD change and can be interpreted
as: the hazard ratio for mortality associated with per SD increase in continuous air pollution
exposure. Air pollution exposure assignment and graphics were conducted by SAS version 9.4
(SAS Institute, Cary, NC). Other data analyses were performed using Stata IC 16.0 (StataCorp
LLC, College Station, TX).

11
Results
Table 1 shows characteristics of our study population of CTS participants diagnosed with
first primary breast cancer in California from 1995 to 2015, by stage. These women were on
average 65.0 years old at diagnosis, primarily non-Hispanic White (88.7%), over half
married/living with partner (62.4%), mostly living in an urban area (87.6%). The majority
(79.9%) did not have family history of breast cancer. Most were diagnosed at localized stage
(70.2%), while 27.3% were diagnosed at regional stage and 2.5% were diagnosed at distant site.
During the follow-up period, there were 1,465 deaths (22.4% of 6,539 participants). Of these
1,465 deaths, 569 were caused by breast cancer. We observed higher 5-year survival rate for
participants with localized or regional stages at diagnosis than those diagnosed at distant stage.
The 5-year survival rates for participants were 92.2% for those diagnosed at localized stage,
85.3% for those diagnosed at regional stage, and 39.3% for those diagnosed at distant stage.
Table 1 Demographic, tumor, treatment, and health behavior characteristics of CTS participants newly diagnosed with first
primary breast cancer from 1995-2015, by stage of diagnosis.
Characteristic (mean±SD or %)
Localized stage
(N=4,591)
Regional
(N=1,783)
Distant site(s)
(N=165)
Total (N=6,539)
Follow-up period (years) 8.8±5.5 8.6±5.5 3.5±3.5 8.6±5.6
Age at diagnosis (CCR§) (years) 65.7±11.4 63.2±11.8 65.2±13.4 65.0±11.6
Age group at diagnosis (CCR)
20-49
50-69
70+

348 (7.6)
2,566 (55.9)
1,677 (36.5)

209 (11.7)
1,073 (60.2)
501 (28.1)

19 (11.5)
83 (50.3)
63 (38.2)

576 (8.8)
3,722 (56.9)
2,241 (34.3)
Race/Ethnicity (CCR§)
non-Hispanic White (ref)
non-Hispanic African American
Hispanic
non-Hispanic Asian/Pac. IslK
Other/Mixed

4,078 (88.8)
98 (2.1)
125 (2.7)
169 (3.7)
121 (2.6)

1,576 (88.4)
53 (3.0)
65 (3.6)
53 (3.0)
36 (2.0)

149 (90.3)
6 (3.6)
6 (3.6)
2 (1.2)
2 (1.2)

5,803 (88.7)
157 (2.4)
196 (3.0)
224 (3.4)
159 (2.4)
Marital status (CCR&CTS¶)
Married/Living together
Divorced/Separated
Widowed
Never married
Missing/Unknown

2,881 (62.8)
517 (11.3)
658 (14.3)
492 (10.7)
43 (0.9)

1,107 (62.1)
219 (12.3)
244 (13.7)
196 (11.0)
17 (1.0)

91 (55.1)
26 (15.8)
20 (12.1)
26 (15.8)
2 (1.2)

4,079 (62.4)
762 (11.6)
922 (14.1)
714 (10.9)
62 (1.0)
ER/PR status (CCR§)
ER+ or PR+
ER- & PR-
Other/Missing

3,673(80.0)
501(10.9)
417(9.1)

1,381(77.4)
258(14.5)
144(8.1)

120(72.7)
20(12.1)
25(15.2)

5,174 (79.1)
799 (11.9)
586 (9.0)

12
Treatment: Surgery (CCR§)
Yes
No
Unknown

4,494 (97.9)
97 (2.1)
0

1,688 (94.7)
94 (5.3)
1 (0.1)

65 (39.4)
100 (60.6)
0

6,247 (95.5)
291 (4.5)
1 (0.0)
Treatment: Radiation (CCR§)
Yes
No
Unknown

2,324 (50.6)
2,265 (49.3)
2 (0.1)

448 (25.1)
1,335 (74.9)
0

39 (23.6)
126 (76.4)
0

2,811 (43.0)
3,726 (57.0)
2 (0.0)
Treatment: Chemotherapy (CCR§)
Yes
No
Unknown

895(19.5)
3,640 (79.3)
56 (1.2)

1,159(65.0)
585(32.8)
39 (2.2)

84 (50.9)
77 (46.7)
4 (2.4)

2,138 (32.7)
4,302 (65.8)
99 (1.5)
Treatment: Hormone Therapy (CCR§)
Yes
No
Unknown

1,841 (40.1)
2,588 (56.4)
162 (3.5)

433 (24.3)
1,265 (70.9)
85 (4.8)

73 (44.2)
89 (53.9)
3 (1.8)

2,347 (35.9)
3,942 (60.3)
250 (3.8)
Year of diagnosis (CCR§)
1995-2000
2001-2005
2006-2010
2011-2015

1,312 (28.6)
1,152 (25.1)
1,074 (23.4)
1,053 (22.9)

533 (29.9)
517 (29.0)
405 (22.7)
328 (18.4)

27 (16.4)
46 (27.9)
49 (29.7)
43 (26.1)

1,872 (28.6)
1,715 (26.2)
1,528 (23.4)
1,424 (21.8)
SES (CCR§)
Lowest SES
Lower-middle SES
Middle SES
Higher-middle SES
Highest SES
Missing/Unknown

163 (3.6)
457 (10.0)
847 (18.4)
1,301 (28.3)
1,816 (39.6)
7 (0.2)

64 (3.6)
204 (11.4)
311 (17.4)
528 (29.6)
674 (37.8)
2 (0.1)

10 (6.1)
15 (9.1)
30 (18.2)
55 (33.3)
55 (33.3)
0

237 (3.6)
676 (10.3)
1,188 (18.2)
1,884 (28.8)
2,545 (38.9)
9 (0.1)
Urban/Rural (CCR§)
Not Urban
Urban

577 (12.6)
4,014 (87.4)

212 (11.9)
1,571 (88.1)

23 (13.9)
142 (86.1)

812 (12.4)
5,727 (87.6)
BMI before diagnosis (CTS#)
Underweight
Normal
Overweight
Obese
Unknown

90 (2.0)
2,346 (51.1)
1,325 (28.9)
704 (15.3)
126 (2.7)

29 (1.6)
875 (49.1)
494 (27.7)
318 (17.84)
67 (3.76)

0
56 (33.9)
51 (30.9)
45 (27.3)
13 (7.9)

119 (1.8)
3,277 (50.1)
1,870 (28.6)
1,067 (16.3)
206 (3.2)
Exercise before diagnosis (hrs/wk) (CTS#)
<1
1-2.9
3-4.9
>=5
Missing/Unknown

914 (19.9)
1,224 (26.7)
905 (19.7)
1,519 (33.1)
29 (0.6)

371 (20.8)
521 (29.2)
337 (18.9)
540 (30.3)
14 (0.8)

27 (16.4)
61 (37.0)
22(13.3)
53 (32.1)
2 (1.2)

1,312 (20.1)
1,806 (27,6)
1,264 (19.3)
2,112 (32.30)
45 (0.7)
Alcohol before diagnosis (CTS#)
No
Yes
Missing/unknown

1,373 (29.9)
3,020 (65.8)
198 (4.3)

560 (31.4)
1,144 (64.2)
79 (4.4)

52 (31.5)
103 (62.4)
10 (6.1)

1,985 (30.4)
4,267 (65.2)
287 (4.4)
Smoke exposure (CTS# Q1)
No exposure
Passive exposure
Former (self-exposure)
Current (self-exposure)
Missing

728 (15.9)
2,012 (43.8)
1,525 (33.2)
261 (5.7)
65 (1.4)

302 (16.9)
751 (42.1)
588 (33.0)
123 (6.9)
19 (1.1)

29 (17.6)
78 (47.3)
47 (28.5)
8 (4.8)
3 (1.8)

1,059 (16.2)
2,841 (43.4)
2,160 (33.0)
392 (6.0)
87 (1.3)
Family history of breast cancer (CTS# Q1)
No
Yes
Adopted/Missing

3,645 (79.4)
778 (16.8)
168 (3.7)

1,454 (81.5)
267 (15.0)
62 (3.5)

124 (75.2)
32 (19.4)
9 (5.4)

5,223 (79.9)
1,077 (16.5)
239 (3.6)
All-cause death 865 (18.8) 491 (27.5) 109 (66.1) 1,465 (22.4)
Breast cancer-specific death 181 (3.8) 294 (16.5) 94 (57.0) 569 (8.7)
5-year survival rate 92.2 85.3 39.3 89.0
§ Characteristic information is available from CCR.

13
# Characteristic information is available from CTS.
¶ Marital status is available from CCR. Anyone with unknown marital status used information from CTS.

Average air pollution exposures in the 12 months after diagnosis were on average: 16.5
ppb for NO2, 39.9 ppb for O3, 26.8 𝜇 g/m
3
for PM10, and 12.1 𝜇 g/m
3
for PM2.5 (Table 2). There is
a clear pattern of decline on 4 different time frame averages for NO2, PM10, and PM2.5, either by
stage or in total. This results in mean exposure concentrations for these pollutants being slightly
lower when calculated for longer durations.
Table 2 Air pollution exposure summaries by stage at diagnosis
Air pollution exposure (mean±SD or %)

Localized stage
(N=4,591)
Regional
(N=1,783)
Distant site(s)
(N=165)
Total
(N=6,539)
NO2 (ppb)
% available
12 months before the diagnosis average
12 months after diagnosis average
5 years truncated follow-up average
Whole follow-up average

95.4
16.9±7.8
16.3±7.4
15.4±6.8
14.0±5.7

96.2
17.6±7.9
16.9±7.7
16.0±7.1
14.5±6.2

93.9
15.7±7.6
15.6±8.0
15.1±7.7
14.8±7.7

95.6
17.0±7.8
16.5±7.5
15.5±6.9
14.1±5.9
O3 (ppb)
% available
12 months before the diagnosis average
12 months after diagnosis average
5 years truncated follow-up average
Whole follow-up average

98.2
39.9±7.9
39.8±7.9
40.1±7.6
40.6±7.2

98.9
40.0§±8.1
40.0§±8.1
40.4±7.8
40.7±7.4

96.4
39.4§±8.2
39.4§±8.7
39.4§±8.4
39.4§±8.4

98.3
39.9§±8.0
39.9§±8.0
40.2±7.6
40.6±7.3
PM10 ( 𝜇 g/m
3
)
% available
12 months before the diagnosis average
12 months after diagnosis average
5 years truncated follow-up average
Whole follow-up average

97.6
27.0±9.1
26.6±9.0
26.0±8.4
24.8±7.7

98.4
27.8±9.5
27.4±9.2
26.8±8.8
25.6±8.2

98.2
26.2±9.1
25.7±9.0
25.3±8.5
25.0±8.5

97.8
27.2±9.2
26.8±9.0
26.2±8.5
25.0±7.8
PM2.5 ( 𝜇 g/m
3
)*
% available
12 months before the diagnosis average
12 months after diagnosis average
5 years truncated follow-up average
Whole follow-up average

98.5
12.2±4.4
11.9±4.3
11.2±3.6
10.7±3.0

99.3
12.7±4.8
12.4±4.5
11.7±3.8
11.1±3.4

96.6
11.9±4.2
11.7±4.8
11.4±4.5
11.2±4.4

98.6
12.3±4.5
12.1±4.4
11.4±3.7
10.8±3.2
* PM2.5 data were reported only for participants who were diagnosed breast cancer in 2000 or later. The total sample size is 5070,
with 3551, 1373, 146 for three stages respectively.
§Summaries at the same stage of the same air pollutant appear to have identical mean due to rounding.

We observed little evidence for strong associations between average ambient air pollution
exposures in 12 months after diagnosis and all-cause mortality or breast cancer-specific mortality
(Table 3). Statistically significant associations can be found in all the following tables with p-
value in red. The only statistically significantly association was between NO2 and breast cancer-
specific mortality for participants diagnosed at regional stage (p=0.045). After adjusting for

14
covariates, the hazard for breast-cancer specific mortality for participants diagnosed at the
regional stage was 1.14 (95% CI: 1.00, 1.29) times higher per SD increase in average NO2 over
the 12 months after diagnosis.
Table 3 Adjusted* HRs (95%CI) for all-cause and breast cancer-specific mortality associated with 1 SD§ increase in air
pollution average of 12 months after diagnosis , stratified by stage and baseline hazard by age group at diagnosis
Air pollutant Stage at diagnosis Sample size
All-cause mortality Breast cancer-specific mortality
HR (95%CI) p-value HR (95%CI) p-value
NO2
Localized 4,381 0.99 (0.92, 1.06) 0.730 1.03 (0.86, 1.23) 0.749
Regional 1,716 1.07 (0.97, 1.17) 0.200 1.14 (1.00, 1.29) 0.045
Distant 155 1.01 (0.75, 1.36) 0.956 0.88 (0.58, 1.34) 0.545
O3
Localized 4,506 1.00 (0.93, 1.07) 0.948 0.98 (0.85, 1.14) 0.829
Regional 1,763 0.92 (0.83, 1.02) 0.099 0.89 (0.78, 1.01) 0.073
Distant 159 0.92 (0.71, 1.21) 0.564 0.83 (0.63, 1.09) 0.173
PM10
Localized 4,483 0.96 (0.89, 1.03) 0.289 0.98 (0.84, 1.15) 0.811
Regional 1,754 0.98 (0.89, 1.08) 0.712 1.03 (0.91, 1.17) 0.590
Distant 162 1.23 (0.94, 1.62) 0.135 1.11 (0.80, 1.54) 0.531
PM2.5#
Localized 3,497 0.94 (0.86, 1.04) 0.215 1.00 (0.81, 1.25) 0.965
Regional 1,363 1.02 (0.91, 1.15) 0.721 1.07 (0.91, 1.26) 0.414
Distant 141 1.10 (0.80, 1.50) 0.559 1.14 (0.75, 1.73) 0.530
*Adjusted for race/ethnicity, marital status, ER/PR status, initial treatment, year of diagnosis, SES, rural-urban commuting area,
BMI, exercise, alcohol, smoke exposure, and family history of breast cancer. Breast cancer-specific mortality models additionally
adjusted for age group at baseline instead of stratified by it. Statistically significant associations, at alpha level of 0.05, are shown
in red.
§ SD values for NO2, O3, PM10, and PM2.5 are 7.5 ppb, 8.0 ppb, 9.0 𝜇 g/m
3
,

and 4.4 𝜇 g/m
3
respectively.
# PM2.5 data were reported only for participants who were diagnosed breast cancer in 2000 or later.
In sensitivity analyses (subject to potential bias from declines in air pollution over the
study period), we observed statistically significant associations between follow-up period
average air pollutants and all-cause mortality (Table 4). These associations were attenuated when
the follow-up period was truncated to a maximum of 5 years. In models that included both the 12
months after diagnosis and the average change over follow-up, we observed statistically
significant positive HR for NO2, PM10 and PM2.5, with larger HR for the average change in
exposure, but also HR for the 12 month after diagnosis which were larger and have smaller p-
values than in Table 3. The correlation of average change in exposure (both for whole follow-up
period and the truncated one) and respective different exposure summaries was shown in Table
S1. Average change over whole follow-up period was highly correlated with average of 12
months after diagnosis for NO2 and PM2.5 (Pearson’s R=-0.69 for NO2, -0.75 for PM2.5).

15
Correlations were attenuated in the truncated follow-up period. Similar patterns of strong,
significant associations were found for breast cancer-specific mortality (Table 5).
Table 4 Adjusted* HRs (95%CI) for all-cause mortality associated with 1 SD§ increase in each air pollutant summary, stratified
by stage and baseline hazard by age group at diagnosis
Pollutant
Localized Regional Distant
HR (95%CI) p-value HR (95%CI) p-value HR (95%CI) p-value
NO2
Whole follow-up period
Whole follow-up
1.34
(1.24, 1.44)
<0.001
1.50
(1.37, 1.65)
<0.001
1.19
(0.95, 1.50)
0.138
12 months after diagnosis
0.99
(0.92, 1.06)
0.730
1.07
(0.97, 1.17)
0.200
1.01
(0.75, 1.36)
0.956
12 months after diagnosis¶

Average change over follow-up

2.20
(2.00, 2.42)
3.51
(3.17, 3.90)
<0.001

<0.001

2.25
(2.00, 2.53)
4.16
(3.60, 4.81)
<0.001

<0.001

1.05
(0.79, 1.41)
4.50
(2.41, 8.38)
0.721

<0.001

12 months before diagnosis
0.99
(0.92, 1.07)
0.801
1.05
(0.95, 1.16)
0.347
0.88
(0.64, 1.21)
0.417
Follow-up period truncated to a maximum of 5 years
Follow-up truncated to 5yrs
1.14
(0.99, 1.30)
0.065
1.32
(1.14, 1.53)
<0.001
1.14
(0.85, 1.54)
0.386
12 months after diagnosis
1.01
(0.88, 1.16)
0.886
1.16
(1.01, 1.35)
0.041
1.00
(0.72, 1.38)
0.987
12 months after diagnosis¶

Average change over follow-up

1.46
(1.25, 1.70)
2.08
(1.81, 2.39)
<0.001

<0.001

1.75
(1.49, 2.06)
2.45
(2.09, 2.88)
<0.001

<0.001

1.00
(0.72, 1.38)
3.40
(2.04, 5.65)
0.994

<0.001

12 months before diagnosis
1.03
(0.90, 1.18)
0.682
1.13
(0.98, 1.31)
0.105
0.84
(0.60, 1.19)
0.330
O3
Whole follow-up period
Whole follow-up
0.90
(0.84, 0.97)
0.007
0.85
(0.76, 0.93)
0.001
0.88
(0.69, 1.12)
0.289
12 months after diagnosis
1.00
(0.93, 1.07)
0.948
0.92
(0.83, 1.02)
0.099
0.92
(0.71, 1.21)
0.564
12 months after diagnosis¶

Average change over follow-up

0.79
(0.73, 0.87)
0.69
(0.64, 0.74)
<0.001

<0.001

0.78
(0.69, 0.87)
0.76
(0.69, 0.830
<0.001

<0.001

0.85
(0.65, 1.12)
0.50
(0.33, 0.77)
0.249

0.002

12 months before diagnosis
0.99
(0.92, 1.06)
0.711
0.93
(0.85, 1.03)
0.174
0.97
(0.74, 1.27)
0.814
Follow-up period truncated to a maximum of 5 years
Follow-up truncated to 5yrs
0.94
(0.83, 1.06)
0.310
0.93
(0.81, 1.08)
0.349
0.83
(0.61, 1.11)
0.203
12 months after diagnosis
0.98
(0.87, 1.11)
0.771
0.99
(0.86, 1.15)
0.912
0.89
(0.65, 1.21)
0.447
12 months after diagnosis¶

Average change over follow-up

0.93
(0.82, 1.06)
0.86
(0.76, 0.97)
0.256

0.011

0.91
(0.78, 1.06)
0.78
(0.68, 0.89)
0.209

<0.001

0.81
(0.59, 1.12)
0.53
(0.34, 0.84)
0.197

0.007

12 months before diagnosis
1.00
(0.88, 1.12)
0.943
1.00
(0.87, 1.16)
0.966
1.00
(0.73, 1.37)
0.993
PM10
Whole follow-up period
Whole follow-up
1.15
(1.07, 1.22)
<0.001
1.23
(1.12, 1.34)
<0.001
1.42
(1.12, 1.81)
0.004
12 months after diagnosis 0.96 0.289 0.98 0.712 1.23 0.135

16
(0.89, 1.03) (0.89, 1.08) (0.94, 1.62)
12 months after diagnosis¶

Average change over follow-up

1.20
(1.10, 1.30)
1.47
(1.38, 1.57)
<0.001

<0.001

1.31
(1.18, 1.45)
1.82
(1.67, 1.98)
<0.001

<0.001

1.50
(1.12, 2.01)
1.90
(1.33, 2.71)
0.007

<0.001

12 months before diagnosis
0.95
(0.89, 1.02)
0.189
0.99
(0.90, 1.08)
0.759
1.07
(0.79, 1.46)
0.644
Follow-up period truncated to a maximum of 5 years
Follow-up truncated to 5yrs
0.94
(0.83, 1.07)
0.351
1.18
(1.03, 1.34)
0.014
1.37
(1.02, 1.83)
0.036
12 months after diagnosis
0.93
(0.82, 1.05)
0.232
1.11
(0.98, 1.27)
0.106
1.20
(0.88, 1.63)
0.254
12 months after diagnosis¶

Average change over follow-up

0.97
(0.84, 1.11)
1.09
(0.97, 1.21)
0.619

0.143

1.21
(1.05, 1.39)
1.24
(1.11, 1.40)
0.007

<0.001

1.44
(1.03, 2.01)
1.71
(1.22, 2.40)
0.031

0.002

12 months before diagnosis
0.94
(0.83, 1.07)
0.349
1.13
(0.99, 1.28)
0.073
1.02
(0.72, 1.43)
0.930
PM2.5#
Whole follow-up period
Whole follow-up
1.19
(1.08, 1.31)
<0.001
1.38
(1.24, 1.54)
<0.001
1.23
(0.96, 1.57)
0.101
12 months after diagnosis
0.94
(0.86, 1.04)
0.215
1.02
(0.91, 1.15)
0.721
1.10
(0.80, 1.50)
0.559
12 months after diagnosis¶

Average change over follow-up

1.50
(1.32, 1.70)
1.80
(1.60, 2.03)
<0.001

<0.001

2.00
(1.72, 2.33)
2.73
(2.32, 3.22)
<0.001

<0.001

1.37
(0.96, 1.95)
1.86
(1.14, 3.02)
0.079

0.012

12 months before diagnosis
0.95
(0.87, 1.04)
0.293
1.01
(0.90, 1.13)
0.853
0.81
(0.55, 1.21)
0.306
Follow-up period truncated to a maximum of 5 years
Follow-up truncated to 5yrs
1.04
(0.89, 1.21)
0.638
1.23
(1.06, 1.44)
0.008
1.19
(0.88, 1.62)
0.257
12 months after diagnosis
0.92
(0.79, 1.07)
0.258
1.11
(0.95, 1.31)
0.182
1.09
(0.77, 1.56)
0.615
12 months after diagnosis¶

Average change over follow-up

1.16
(0.97, 1.40)
1.37
(1.20, 1.56)
0.105

<0.001

1.44
(1.20, 1.75)
1.52
(1.29, 1.80)
<0.001

<0.001

1.37
(0.94, 2.01)
1.83
(1.16, 2.89)
0.105

0.009

12 months before diagnosis
0.93
(0.81, 1.08)
0.353
1.12
(0.96, 1.31)
0.155
0.87
(0.56, 1.35)
0.535
*Adjusted for race/ethnicity, marital status, ER/PR status, initial treatment, year of diagnosis, SES, rural-urban commuting area,
BMI, exercise, alcohol, smoke exposure, and family history of breast cancer. Statistically significant associations, at alpha level
of 0.05, are shown in red.
§ SD values for each air pollution summary by pollutant can be found in Table 2.
# PM2.5 data were reported only for participants who were diagnosed breast cancer in 2000 or later.
¶ Average of 12 months after diagnosis was additionally adjusted for exposure change over the follow-up period.

Table 5 Adjusted* HRs (95%CI) for breast cancer-specific mortality associated with 1 SD§ increase in each air pollutant
summary, stratified by stage at diagnosis
Pollutant
Localized Regional Distant
HR (95%CI) p-value HR (95%CI) p-value HR (95%CI) p-value
NO2
Whole follow-up period
Whole follow-up
1.43
(1.19, 1.71)
<0.001
1.61
(1.43, 1.82)
<0.001
1.01
(0.74, 1.37)
0.974
12 months after diagnosis
1.03
(0.86, 1.23)
0.749
1.14
(1.00, 1.29)
0.045
0.88
(0.58, 1.34)
0.545
12 months after diagnosis¶

Average change over follow-up
1.84
(1.48, 2.30)
3.00
<0.001

<0.001
2.11
(1.80, 2.48)
3.72
<0.001

<0.001
0.87
(0.58, 1.31)
2.30
0.500

0.005

17
(2.48, 3.64) (3.05, 4.53) (1.28, 4.14)
12 months before diagnosis
1.06
(0.89, 1.26)
0.517
1.12
(0.98, 1.27)
0.088
0.79
(0.49, 1.26)
0.319
Follow-up period truncated to a maximum of 5 years
Follow-up truncated to 5yrs
1.28
(0.96, 1.69)
0.091
1.35
(1.14, 1.61)
0.001
1.00
(0.66, 1.53)
0.991
12 months after diagnosis
1.13
(0.86, 1.50)
0.382
1.20
(1.02, 1.42)
0.032
0.90
(0.55, 1.47)
0.667
12 months after diagnosis¶

Average change over follow-up

1.64
(1.20, 2.24)
2.12
(1.72, 2.61)
0.002

<0.001

1.75
(1.43, 2.15)
2.37
(1.95, 2.89)
<0.001

<0.001

0.85
(0.52, 1.40)
2.69
(1.46, 4.96)
0.530

0.002

12 months before diagnosis
1.13
(0.87, 1.47)
0.351
1.17
(0.99, 1.39)
0.064
0.77
(0.46, 1.29)
0.325
O3
Whole follow-up period
Whole follow-up
0.88
(0.75, 1.02)
0.095
0.82
(0.71, 0.94)
0.004
0.80
(0.62, 1.03)
0.082
12 months after diagnosis
0.98
(0.85, 1.14)
0.829
0.89
(0.78, 1.01)
0.073
0.83
(0.63, 1.09)
0.173
12 months after diagnosis¶

Average change over follow-up

0.79
(0.65, 0.95)
0.69
(0.59, 0.81)
0.015

<0.001

0.77
(0.66, 0.90)
0.77
(0.68, 0.88)
0.001

<0.001

0.77
(0.57, 1.03)
0.50
(0.33, 0.76)
0.075

0.001

12 months before diagnosis
1.01
(0.87, 1.16)
0.928
0.86
(0.76, 0.99)
0.032
0.86
(0.65, 1.14)
0.301
Follow-up period truncated to a maximum of 5 years
Follow-up truncated to 5yrs
0.93
(0.75, 1.14)
0.466
0.93
(0.78, 1.12)
0.449
0.75
(0.54, 1.04)
0.083
12 months after diagnosis
0.98
(0.79, 1.20)
0.815
0.98
(0.83, 1.17)
0.837
0.79
(0.57, 1.10)
0.166
12 months after diagnosis¶

Average change over follow-up

0.91
(0.73, 1.15)
0.84
(0.71, 1.01)
0.440

0.064

0.92
(0.76, 1.11)
0.83
(0.72, 0.95)
0.381

0.008

0.70
(0.48, 1.03)
0.42
(0.25, 0.73)
0.069

0.002

12 months before diagnosis
1.03
(0.85, 1.25)
0.742
0.98
(0.82, 1.17)
0.813
0.89
(0.61, 1.28)
0.515
PM10
Whole follow-up period
Whole follow-up
1.14
(0.98, 1.33)
0.096
1.27
(1.14, 1.43)
<0.001
1.22
(0.93, 1.60)
0.147
12 months after diagnosis
0.98
(0.84, 1.15)
0.811
1.03
(0.91, 1.17)
0.590
1.11
(0.80, 1.54)
0.531
12 months after diagnosis¶

Average change over follow-up

1.21
(1.02, 1.43)
1.54
(1.35, 1.76)
0.030

<0.001

1.34
(1.18, 1.53)
1.71
(1.55, 1.89)
<0.001

<0.001

1.23
(0.86, 1.76)
1.60
(1.14, 2.26)
0.247

0.007

12 months before diagnosis
0.98
(0.84, 1.15)
0.810
1.01
(0.89, 1.15)
0.827
1.04
(0.74, 1.46)
0.813
Follow-up period truncated to a maximum of 5 years
Follow-up truncated to 5yrs
0.98
(0.78, 1.24)
0.893
1.22
(1.05, 1.43)
0.011
1.24
(0.88, 1.75)
0.227
12 months after diagnosis
0.96
(0.76, 1.20)
0.698
1.17
(1.01, 1.36)
0.040
1.10
(0.74, 1.64)
0.644
12 months after diagnosis¶

Average change over follow-up

1.05
(0.82, 1.34)
1.22
(1.06, 1.40)
0.712

0.006

1.27
(1.08, 1.49)
1.24
(1.11, 1.38)
0.004

<0.001

1.26
(0.82, 1.92)
1.64
(1.16, 2.32)
0.296

0.005

12 months before diagnosis
0.96
(0.77, 1.19)
0.706
1.16
(0.99, 1.36)
0.066
0.97
(0.66, 1.43)
0.882
PM2.5#

18
Whole follow-up period
Whole follow-up
1.25
(0.99, 1.58)
0.056
1.41
(1.20, 1.67)
<0.001
1.25
(0.94, 1.64)
0.120
12 months after diagnosis
1.00
(0.81, 1.25)
0.965
1.07
(0.91, 1.26)
0.414
1.14
(0.75, 1.73)
0.530
12 months after diagnosis¶

Average change over follow-up

1.71
(1.28, 2.29)
2.19
(1.69, 2.85)
<0.001

<0.001

1.85
(1.48, 2.31)
2.33
(1.87, 2.89)
<0.001

<0.001

1.40
(0.94, 2.08)
1.93
(1.20, 3.10)
0.100

0.006

12 months before diagnosis
0.99
(0.81, 1.21)
0.918
1.03
(0.88, 1.21)
0.686
0.88
(0.53, 1.49)
0.643
Follow-up period truncated to a maximum of 5 years
Follow-up truncated to 5yrs
1.06
(0.79, 1.43)
0.677
1.25
(1.01, 1.56)
0.044
1.22
(0.82, 1.83)
0.330
12 months after diagnosis
0.94
(0.70, 1.25)
0.670
1.13
(0.92, 1.40)
0.241
1.12
(0.67, 1.87)
0.666
12 months after diagnosis¶

Average change over follow-up

1.27
(0.91, 1.78)
1.54
(1.27, 1.88)
0.166

<0.001

1.49
(1.13, 1.97)
1.55
(1.25, 1.90)
0.004

<0.001

1.41
(0.86, 2.32)
2.05
(1.21, 3.49)
0.174

0.008

12 months before diagnosis
0.99
(0.76, 1.29)
0.966
1.11
(0.91, 1.36)
0.290
0.93
(0.50, 1.75)
0.832
*Adjusted for race/ethnicity, marital status, ER/PR status, initial treatment, year of diagnosis, SES, rural-urban commuting area,
BMI, exercise, alcohol, smoke exposure, family history of breast cancer, and age group at baseline. Statistically significant
associations, at alpha level of 0.05, are shown in red.
§ SD values for each air pollution summary by pollutant can be found in Table 2.
# PM2.5 data were reported only for participants who were diagnosed breast cancer in 2000 or later.
¶ Average of 12 months after diagnosis was additionally adjusted for exposure change over the follow-up period.

In the secondary analysis, results from exposure summaries based only on the address at
diagnosis (Table 6) were not dramatically different from results using exposure summaries
calculated based on residential address histories (Table 3) . For example, the adjusted hazard for
breast cancer-specific mortality for participants diagnosed at the regional stage was 1.14 (95%
CI: 1.00, 1.29) times higher per SD increase in average NO2 over the 12 months after diagnosis
(p=0.045). Similarly, results for models including other air pollutant summaries also had
negligible differences (compare Tables S3-S4 to Tables 4-5). The percent change in HRs
(exponentiated β) were all less than 2%, for both all-cause mortality and breast cancer-specific
mortality. The negligible differences might result from the limited number of movers in final
analysis. Over the entire follow-up, there were only around 19% movers for NO2, O3, PM10
analysis (16.8% for PM2.5).
Table 6 Secondary analysis only using address at diagnosis: adjusted* HRs (95%CI) for all-cause and breast cancer-specific
mortality associated with 1 SD§ increase in air pollution average of 12 months after diagnosis, stratified by stage and baseline
hazard by age group at diagnosis

19
Air pollutant Stage at diagnosis Sample size
All-cause mortality Breast cancer-specific mortality
HR (95%CI) p-value HR (95%CI) p-value
NO2
Localized 4,390 0.99 (0.92, 1.06) 0.696 1.03 (0.86, 1.23) 0.741
Regional 1,723 1.07 (0.97, 1.17) 0.199 1.14 (1.00, 1.29) 0.045
Distant 155 1.01 (0.75, 1.36) 0.667 0.88 (0.58, 1.34) 0.545
O3
Localized 4,508 1.00 (0.93, 1.07) 0.935 0.98 (0.85, 1.14) 0.836
Regional 1,763 0.92 (0.83, 1.02) 0.099 0.89 (0.78, 1.01) 0.073
Distant 159 0.92 (0.71, 1.21) 0.564 0.83 (0.63, 1.09) 0.173
PM10
Localized 4,494 0.96 (0.89, 1.04) 0.310 0.98 (0.84, 1.15) 0.800
Regional 1,756 0.98 (0.89, 1.08) 0.686 1.04 (0.92, 1.17) 0.563
Distant 162 1.23 (0.94, 1.62) 0.135 1.11 (0.80, 1.54) 0.531
PM2.5#
Localized 3,499 0.94 (0.86, 1.04) 0.224 1.01 (0.81, 1.25) 0.944
Regional 1,363 1.02 (0.91, 1.15) 0.694 1.08 (0.91, 1.27) 0.383
Distant 141 1.10 (0.80, 1.50) 0.559 1.14 (0.75, 1.73) 0.530
*Adjusted for race/ethnicity, marital status, ER/PR status, initial treatment, year of diagnosis, SES, rural-urban commuting area,
BMI, exercise, alcohol, smoke exposure, and family history of breast cancer. Breast cancer-specific mortality models additionally
adjusted for age group at baseline instead of stratified by it. Statistically significant associations, at alpha level of 0.05, are shown
in red.
§ SD values for NO2, O3, PM10, and PM2.5 are 7.5 ppb, 8.0 ppb, 9.0 𝜇 g/m
3
,

and 4.4 𝜇 g/m
3
respectively (shown in Table S2).
# PM2.5 data were reported only for participants who were diagnosed breast cancer in 2000 or later.
When only adjusting for covariates from CCR, the hazard for breast cancer-specific
mortality for participants diagnosed at regional stages was similar to the primary analysis: 1.13
(95% CI: 1.00, 1.28) times higher per SD increase in the average NO2 over the 12 months after
diagnosis (Table 7). Compared with HR estimates from our primary analysis (Table 3), only the
HR for all-cause mortality for participants at distant stage with PM10 average after 12 months
after diagnosis changed over 10% (from 1.23 to 1.11). We observed little difference in the HR
estimates from models adjusting only CCR covariates (Table S5 and S6) and versus those from
models including extra rich individual-level potential confounders (Table 4 and 5). 94.2% HRs
for all-cause mortality and 92.5% for breast Cancer-specific mortality had magnitude of
difference less than 10%.
Table 7 Secondary analysis only adjusting for CCR covariates: adjusted* HRs (95%CI) for all-cause and breast cancer-specific
mortality associated with 1 SD§ increase in air pollution average of 12 months after diagnosis, stratified by stage and baseline
hazard by age group at diagnosis
Air pollutant Stage at diagnosis Sample size
All-cause mortality Breast cancer-specific mortality
HR (95%CI) p-value HR (95%CI) p-value
NO2
Localized 4,381 0.99 (0.92, 1.07) 0.819 1.04 (0.87, 1.24) 0.668
Regional 1,716 1.06 (0.96, 1.17) 0.228 1.13 (1.00, 1.28) 0.048
Distant 155 1.04 (0.79, 1.37) 0.788 0.91 (0.64, 1.31) 0.616
O3
Localized 4,506 1.00 (0.93, 1.08) 0.936 0.98 (0.84, 1.13) 0.742
Regional 1,763 0.91 (0.82, 1.00) 0.057 0.89 (0.78, 1.01) 0.063
Distant 159 0.94 (0.75, 1.17) 0.570 0.91 (0.71, 1.15) 0.433
PM10
Localized 4,483 0.97 (0.90, 1.04) 0.412 0.98 (0.85, 1.14) 0.829
Regional 1,754 0.98 (0.89, 1.07) 0.612 1.03 (0.91, 1.16) 0.695

20
Distant 162 1.11 (0.87, 1.42)¶ 0.397 1.04 (0.77, 1.40) 0.795
PM2.5#
Localized 3,497 0.95 (0.86, 1.04) 0.277 1.01 (0.82, 1.24) 0.915
Regional 1,363 1.02 (0.91, 1.14) 0.728 1.05 (0.90, 1.23) 0.521
Distant 141 1.12 (0.85, 1.48) 0.426 1.20 (0.82, 1.74) 0.343
*Adjusted for race/ethnicity, marital status, ER/PR status, initial treatment, year of diagnosis, SES, and rural-urban commuting
area. Breast cancer-specific mortality models additionally adjusted for age group at baseline instead of stratified by it.
Statistically significant associations, at alpha level of 0.05, are shown in red.
§ SD values for NO2, O3, PM10, and PM2.5 are 7.5 ppb, 8.0 ppb, 9.0 𝜇 g/m
3
,

and 4.4 𝜇 g/m
3
respectively.
# PM2.5 data were reported only for participants who were diagnosed breast cancer in 2000 or later.
¶%change of this estimate was 10.7% (>10.0%).

21
Discussion
In this study, we found that participants who were diagnosed with first primary breast
cancer at a regional stage and who lived in areas with higher NO2 exposure for 12 months after
diagnosis had a higher risk of breast cancer-specific mortality, after adjusting for potential
confounders. We also found statistically significant associations of follow-up period average air
pollutants with all-cause mortality and breast cancer-specific mortality. However, these
associations were attenuated when the follow-up period was truncated to a maximum of 5 years,
which demonstrates a potential bias from representing exposures using follow-up period
averages during a time when air pollution levels were declining, tending to result in lower
averages for participants with longer survival. Additional adjustment for individual-level
covariates from CTS or utilizing residential address histories provided us with similar results as
if these individual-level data were not available. Based on these results, a registry-based study
with lower cost and higher efficiency could be valid if existing data are limited.
Our study had several strengths. The CTS, a large population-based cohort was originally
designed to study breast cancer incidence and includes a large number of breast cancer cases in
California female teachers between 1995 to 2015. California also has high quality air pollution
monitoring networks and wide range of air pollutants exposures, which allows for assigning
exposure data to participants by person-month. As participants diagnosed with breast cancer may
relocate during their potentially long survival time after breast cancer diagnosis, we utilized
residential address histories available in the CTS, to minimize exposure misclassification caused
by participants relocating. The CTS also contains individual-level data on important risk factors

22
for breast cancer and mortality (e.g., BMI, alcohol consumption, exercise, smoking, and family
history of breast cancer).
An important limitation of our study is that we only model time-constant averages of air
pollutant with all-cause mortality or breast cancer-specific mortality. Individuals diagnosed with
breast cancer tend to have longer survival than those diagnosed with other cancers, like lung or
liver by comparing the 5-year survival rates from previous studies.
15-18
Since ambient air
pollution levels have been declining over the last 10 to 20 years in California, participants were
assigned with lower whole follow-up period average if they survived longer (and lived in areas
with decreasing air pollution). This has the potential to induce spurious associations that air
pollution is more positively associated with mortality, which is consistent with the observed
attenuation of large, statistically significant associations when the follow-up was truncated to a
maximum of 5 years. For this reason, we conducted our primary analysis using the air pollutant
average over the first 12 months after diagnosis, which is less influenced by the decline trend of
air pollution. However, this approach is still not informative enough to demonstrate the exposure
histories for participants over their follow-up period, especially for those who had survived for
years.
Previous studies have suggested an association between air pollution exposure and breast
cancer survival.
17-19
High exposure to PM10
17
and PM2.5
17,18
was indicated to be associated with
increased breast cancer-specific mortality. Hu et al
17
found that the adjusted hazard of breast
cancer-specific mortality overall is 1.13 times (95%CI: 1.02, 1.25) higher per 10 𝜇 g/m
3
increase
in PM10 and 1.86 times (95%CI: 1.12, 3.10) higher per 5 𝜇 g/m
3
increase in PM2.5. The important
limitation of their study is they used survival period averages of county-level air pollution data,
which were not representative for each participant as our study, assigning air pollution data by

23
person-level. Implementing averages of long-period air pollution data also suffered from the
potential bias that we mentioned before. Compared with their results, we also detected consistent
significant associations between averages of whole follow-up period PM10 (in regional stage but
not in localized or distant stages) and PM2.5 (in regional stage but not in localized or distant
stages). Tagliabue G, Borgini A, Tittarelli A, et al
18
categorized median of ground level PM2.5
concentrations within 10*10 km area around participant’s residence over 3 years starting from
the year before the one when participants were diagnosed to year after the year of diagnosis into
quartiles. They discovered that the three upper quartiles of PM2.5 all had increased risk of breast
cancer-specific mortality compared with the lowest quartile (adjusted HRs ranging from 1.72
(95%CI: 1.08 to 2.75) to 1.82 (95%CI: 1.15 to 2.89)) after baseline hazard stratifications on
covariates. Median of PM2.5 concentrations over 3 years around the diagnosis was still not
informative enough to indicate the exposure histories for participants over their follow-up period.
Focusing on our results of primary analysis, the average PM2.5 over 12 months after diagnosis
was not statistically significant associated with breast cancer-specific mortality. The possible
reasons for the differences are that we adjusted for additional demographic and lifestyle
covariates while they only stratified by limited characteristics at diagnosis, or a benefit of using
satellite-derived data for PM2.5 assignments. Moreover, these two studies were unable to adjust
for lifestyle individual-level covariates and test the sensitivity of excluding them. These studies
have called attention to study the association between ambient air pollution and breast cancer
survival though there were some limitations.
In summary, the significant associations between ambient air pollution and breast cancer
survival from our study, as well as those consistent findings with previous studies need to be
interpreted with caution owing to limitations. However, the relationship between air pollution

24
and breast cancer survival is still worthwhile to investigate in the future as the literature is
currently limited and there may be adverse effects of air pollutants, to which women with breast
cancer are universally exposed, but for which there may be possible interventions. In future
analyses, we will develop models with time-varying 12-month average exposures rather than
time-constant follow-up period summaries of exposure histories since diagnosis. Evidence
supporting an association between air pollution exposures after diagnosis and breast cancer
survival would provide important information for agencies regulating air pollution and could
motivate future studies of filter-based interventions aiming to reduce personal air pollution
exposures for women diagnosed with breast cancer.

25
References
1 Siegel RL, Miller KD, Fuchs HE, Jemal A. Cancer Statistics, 2021. CA Cancer J Clin. 2021
Jan;71(1):7-33. doi: 10.3322/caac.21654. Epub 2021 Jan 12. PMID: 33433946.
2 Winters S, Martin C, Murphy D, Shokar NK. Breast Cancer Epidemiology, Prevention, and
Screening. Prog Mol Biol Transl Sci. 2017;151:1-32. doi: 10.1016/bs.pmbts.2017.07.002.
Epub 2017 Oct 10. PMID: 29096890.
3 Sun YS, Zhao Z, Yang ZN, Xu F, Lu HJ, Zhu ZY, Shi W, Jiang J, Yao PP, Zhu HP. Risk
Factors and Preventions of Breast Cancer. Int J Biol Sci. 2017 Nov 1;13(11):1387-1397. doi:
10.7150/ijbs.21635. PMID: 29209143; PMCID: PMC5715522.
4 Barnard ME, Boeke CE, Tamimi RM. Established breast cancer risk factors and risk of
intrinsic tumor subtypes. Biochim Biophys Acta. 2015 Aug;1856(1):73-85. doi:
10.1016/j.bbcan.2015.06.002. Epub 2015 Jun 10. PMID: 26071880.
5 Garcia E, Hurley S, Nelson DO, Hertz A, Reynolds P. Hazardous air pollutants and breast
cancer risk in California teachers: a cohort study. Environ Health. 2015 Jan 30;14:14. doi:
10.1186/1476-069X-14-14. PMID: 25636809; PMCID: PMC4417287.
6 Liu R, Nelson DO, Hurley S, Hertz A, Reynolds P. Residential exposure to estrogen
disrupting hazardous air pollutants and breast cancer risk: the California Teachers Study.
Epidemiology. 2015 May;26(3):365-73. doi: 10.1097/EDE.0000000000000277. PMID:
25760782; PMCID: PMC5101045.
7 Byrne C, Divekar SD, Storchan GB, Parodi DA, Martin MB. Metals and breast cancer. J
Mammary Gland Biol Neoplasia. 2013 Mar;18(1):63-73. doi: 10.1007/s10911-013-9273-9.
Epub 2013 Jan 22. PMID: 23338949; PMCID: PMC4017651.
8 Sievers CK, Shanle EK, Bradfield CA, Xu W. Differential action of monohydroxylated
polycyclic aromatic hydrocarbons with estrogen receptors α and β. Toxicol Sci. 2013
Apr;132(2):359-67. doi: 10.1093/toxsci/kfs287. Epub 2012 Sep 18. PMID: 22989670;
PMCID: PMC3595519.
9 Wang T, Feng W, Kuang D, Deng Q, Zhang W, Wang S, He M, Zhang X, Wu T, Guo H.
The effects of heavy metals and their interactions with polycyclic aromatic hydrocarbons on
the oxidative stress among coke-oven workers. Environ Res. 2015 Jul;140:405-13. doi:
10.1016/j.envres.2015.04.013. Epub 2015 May 15. PMID: 25956561.
10 Hill P, Wynder EL. Nicotine and cotinine in breast fluid. Cancer Lett. 1979 Apr;6(4-5):251-
4. doi: 10.1016/s0304-3835(79)80042-7. PMID: 436120.

26
11 Loomis D, Grosse Y, Lauby-Secretan B, El Ghissassi F, Bouvard V, Benbrahim-Tallaa L,
Guha N, Baan R, Mattock H, Straif K; International Agency for Research on Cancer
Monograph Working Group IARC. The carcinogenicity of outdoor air pollution. Lancet
Oncol. 2013 Dec;14(13):1262-3. doi: 10.1016/s1470-2045(13)70487-x. PMID: 25035875.
12 White AJ, Bradshaw PT, Hamra GB. Air pollution and Breast Cancer: A Review. Curr
Epidemiol Rep. 2018 Jun;5(2):92-100. doi: 10.1007/s40471-018-0143-2. Epub 2018 Mar 27.
PMID: 30271702; PMCID: PMC6159940.
13 Parikh PV, Wei Y. PAHs and PM2.5 emissions and female breast cancer incidence in metro
Atlanta and rural Georgia. Int J Environ Health Res. 2016 Aug;26(4):458-66. doi:
10.1080/09603123.2016.1161178. Epub 2016 Mar 17. PMID: 26983363.
14 Chen F, Bina WF. Correlation of white female breast cancer incidence trends with nitrogen
dioxide emission levels and motor vehicle density patterns. Breast Cancer Res Treat. 2012
Feb;132(1):327-33. doi: 10.1007/s10549-011-1861-z. Epub 2011 Nov 11. PMID: 22076479.
15 Eckel SP, Cockburn M, Shu YH, Deng H, Lurmann FW, Liu L, Gilliland FD. Air pollution
affects lung cancer survival. Thorax. 2016 Oct;71(10):891-8. doi: 10.1136/thoraxjnl-2015-
207927. Epub 2016 Aug 4. PMID: 27491839; PMCID: PMC5400105.
16 Deng H, Eckel SP, Liu L, Lurmann FW, Cockburn MG, Gilliland FD. Particulate matter air
pollution and liver cancer survival. Int J Cancer. 2017 Aug 15;141(4):744-749. doi:
10.1002/ijc.30779. Epub 2017 Jun 7. PMID: 28589567; PMCID: PMC5505313.
17 Hu H, Dailey AB, Kan H, Xu X. The effect of atmospheric particulate matter on survival of
breast cancer among US females. Breast Cancer Res Treat. 2013 May;139(1):217-26. doi:
10.1007/s10549-013-2527-9. Epub 2013 Apr 17. PMID: 23592372.
18 Tagliabue G, Borgini A, Tittarelli A, van Donkelaar A, Martin RV, Bertoldi M, Fabiano S,
Maghini A, Codazzi T, Scaburri A, Favia I, Cau A, Barigelletti G, Tessandori R, Contiero P.
Atmospheric fine particulate matter and breast cancer mortality: a population-based cohort
study. BMJ Open. 2016 Nov 14;6(11):e012580. doi: 10.1136/bmjopen-2016-012580. PMID:
28076275; PMCID: PMC5129133.
19 Wong CM, Tsang H, Lai HK, Thomas GN, Lam KB, Chan KP, Zheng Q, Ayres JG, Lee SY,
Lam TH, Thach TQ. Cancer Mortality Risks from Long-term Exposure to Ambient Fine
Particle. Cancer Epidemiol Biomarkers Prev. 2016 May;25(5):839-45. doi: 10.1158/1055-
9965.EPI-15-0626. PMID: 27197138; PMCID: PMC5505442.
20 Bernstein L, Allen M, Anton-Culver H, Deapen D, Horn-Ross PL, Peel D, Pinder R,
Reynolds P, Sullivan-Halley J, West D, Wright W, Ziogas A, Ross RK. High breast cancer
incidence rates among California teachers: results from the California Teachers Study
(United States). Cancer Causes Control. 2002 Sep;13(7):625-35. doi:
10.1023/a:1019552126105. PMID: 12296510.
21 California Teachers Study website: https://www.calteachersstudy.org/study-population

27
22 California Cancer registry website: https://www.ccrcal.org/learn-about-ccr/about-cancer-
registries/
23 database [US Environmental Protection Agency. Air Quality System Data Mart [internet
database]. http://www.epa.gov/ttn/airs/aqsdatamart (accessed 1 Mar 2020).
24 RUCA codes: www.ers.usda.gov/data-products/rural-urban-commuting-area-codes.aspx
25 Lau B, Cole SR, Gange SJ. Competing risk regression models for epidemiologic data. Am J
Epidemiol. 2009 Jul 15;170(2):244-56. doi: 10.1093/aje/kwp107. Epub 2009 Jun 3. PMID:
19494242; PMCID: PMC2732996.

28
Appendix
Figure S 1 Missingness of PM2.5 during follow-up

Participants

29
Figure S 2 Missingness and time point of censoring for participants met the 75% completed air pollution data criteria for (a) NO2,
(b) O3, (c) PM10, and (d) PM2.5.
(a)

(b)

Participants
Participants

30
(c)

(d)

Participants
Participants

31
Table S 1 Correlation coefficient of each air pollutant average change over follow-up with its relative air pollution exposure
summaries.
Air pollutant
summary
average
NO2 average change O3 average change PM10 average change PM2.5 average change
Pearson’s
R
Spearman’s
R
Pearson’s
R
Spearman’s
R
Pearson’s
R
Spearman’s
R
Pearson’s
R
Spearman’s
R
Whole follow-up period*
Whole
follow-up
-0.42 -0.34 -0.19 -0.23 -0.21 -0.23 -0.43 -0.39
12 months
after
diagnosis
-0.69 -0.59 -0.46 -0.47 -0.53 -0.50 -0.75 -0.65
12 months
before
diagnosis
-0.68 -0.59 -0.39 -0.40 -0.40 -0.38 -0.65 -0.53
Follow-up period truncated to a maximum of 5 years§
Follow-up
truncated to
5 years
-0.41 -0.39 -0.09 -0.11 -0.11 -0.11 -0.49 -0.43
12 months
after
diagnosis
-0.55 -0.52 -0.31 -0.30 -0.38 -0.33 -0.67 -0.61
12 months
before
diagnosis
-0.53 -0.50 -0.25 -0.25 -0.20 -0.18 -0.54 -0.46
*The average change over whole follow-up period is difference between the final 12 months of the participant’s follow-up period
and the first 12 months after diagnosis.
§The average change over follow-up period truncated to a maximum of 5 years id the difference between the final 12 months of
the participant’s truncated follow-up period and the first 12 months after diagnosis.

32
Table S 2 Diagnosis address-based air pollution exposure summaries by stage at diagnosis
Air pollution exposure (mean±SD or %)

Localized stage
(N=4,591)
Regional
(N=1,783)
Distant site(s)
(N=165)
Total
(N=6,539)
NO2 (ppb)
% available
12 months before the diagnosis average
12 months after diagnosis average
5 years truncated follow-up average
Whole follow-up average

95.7
16.9±7.8
16.3±7.4
15.4±6.8
14.0±5.8

96.7
17.6±7.9
16.9±7.7
16.0±7.1
14.6±6.2

93.9
15.7±7.6
15.6±8.0
15.1±7.8
14.9±7.7

95.9
17.0±7.8
16.5±7.5
15.6±7.0
14.2±5.9
O3 (ppb)
% available
12 months before the diagnosis average
12 months after diagnosis average
5 years truncated follow-up average
Whole follow-up average

98.2
39.9±7.9
39.8±7.9
40.1±7.6
40.5±7.2

98.9
40.0±8.1
40.0±8.1
40.4±7.8
40.7±7.5

96.4
39.4±8.2
39.4±8.7
39.4±8.5
39.4±8.5

98.3
39.9±8.0
39.9±8.0
40.2±7.7
40.6±7.4
PM10 ( 𝜇 g/m
3
)
% available
12 months before the diagnosis average
12 months after diagnosis average
5 years truncated follow-up average
Whole follow-up average

97.9
27.0±9.1
26.6±9.0
26.0±8.4
24.8±7.7

98.5
27.8±9.5
27.4±9.2
26.8±8.8
25.6±8.2

98.2
26.2±9.1
25.7±9.0
25.4±8.7
25.1±8.6

98.1
27.2±9.2
26.8±9.0
26.2±8.5
25.0±7.9
PM2.5 ( 𝜇 g/m
3
)*
% available
12 months before the diagnosis average
12 months after diagnosis average
5 years truncated follow-up average
Whole follow-up average

98.5
12.2±4.4
11.9±4.3
11.3±3.6
10.7±3.0

99.3
12.7±4.7
12.4±4.5
11.7±3.8
11.1±3.4

96.6
11.9±4.2
11.7±4.8
11.4±4.6
11.3±4.5

98.7
12.3±4.5
12.0±4.4
11.4±3.7
10.8±3.2
* PM2.5 data were reported only for participants who were diagnosed breast cancer in 2000 or later. The total sample size is 5070,
with 3551, 1373, 146 for three stages respectively.
§Summaries at the same stage of the same air pollutant appear to have identical mean due to rounding.

33
Table S 3 Secondary analysis only using address at diagnosis: adjusted* HRs (95%CI) for all-cause mortality associated with 1
SD§ increase in each air pollutant summary stratified by stage and baseline hazard by age group at diagnosis.
Pollutant
Localized Regional Distant
HR (95%CI) p-value HR (95%CI) p-value HR (95%CI) p-value
NO2
Whole follow-up period
Whole follow-up
1.33
(1.24, 1.43)
<0.001
1.50
(1.37, 1.65)
<0.001
1.18
(0.93, 1.49)
0.168
12 months after diagnosis
0.99
(0.92, 1.06)
0.696
1.07
(0.97, 1.17)
0.199
1.01
(0.75, 1.36)
0.956
12 months after diagnosis¶

Avg change over follow-up

2.19
(1.99, 2.41)
3.52
(3.17, 3.90)
<0.001

<0.001

2.25
(2.00, 2.54)
4.17
(3.61, 4.82)
<0.001

<0.001

1.05
(0.79, 1.41)
4.50
(2.41, 8.38)
0.725

<0.001

12 months before diagnosis
0.99
(0.92, 1.06)
0.770
1.05
(0.95, 1.16)
0.346
0.88
(0.64, 1.21)
0.417
Follow-up period truncated to a maximum of 5 years
Follow-up truncated to 5yrs
1.15
(1.01, 1.32)
0.038
1.32
(1.14, 1.52)
<0.001
1.13
(0.84, 1.53)
0.406
12 months after diagnosis
1.01
(0.88, 1.16)
0.900
1.16
(1.01, 1.35)
0.041
1.00
(0.72, 1.38)
0.987
12 months after diagnosis¶

Avg change over follow-up

1.46
(1.25, 1.70)
2.09
(1.82, 2.40)
<0.001

<0.001

1.75
(1.49, 2.06)
2.45
(2.09, 2.88)
<0.001

<0.001

1.00
(0.72, 1.38)
3.40
(2.04, 5.65)
0.994

<0.001

12 months before diagnosis
1.03
(0.90, 1.18)
0.692
1.13
(0.98, 1.31)
0.104
0.84
(0.60, 1.19)
0.330
O3
Whole follow-up period
Whole follow-up
0.90
(0.84, 0.97)
0.007
0.84
(0.76, 0.93)
0.001
0.87
(0.68, 1.11)
0.259
12 months after diagnosis
1.00
(0.93, 1.07)
0.935
0.92
(0.83, 1.02)
0.099
0.92
(0.71, 1.21)
0.564
12 months after diagnosis¶

Avg change over follow-up

0.79
(0.73, 0.87)
0.69
(0.64, 0.74)
<0.001

<0.001

0.78
(0.69, 0.87)
0.76
(0.69, 0.83)
<0.001

<0.001

0.85
(0.65, 1.12)
0.50
(0.33, 0.77)
0.249

0.002

12 months before diagnosis
0.99
(0.92, 1.06)
0.707
0.93
(0.85, 1.03)
0.175
0.97
(0.74, 1.27)
0.814
Follow-up period truncated to a maximum of 5 years
Follow-up truncated to 5yrs
0.94
(0.83, 1.06)
0.316
0.94
(0.81, 1.08)
0.387
0.82
(0.61, 1.11)
0.199
12 months after diagnosis
0.98
(0.88, 1.10)
0.775
0.99
(0.87, 1.14)
0.912
0.90
(0.67, 1.19)
0.447
12 months after diagnosis¶

Avg change over follow-up

0.93
(0.83, 1.05)
0.86
(0.76, 0.97)
0.256

0.011

0.91
(0.79, 1.05)
0.78
(0.68, 0.89)
0.209

<0.001

0.82
(0.61, 1.11)
0.53
(0.34, 0.84)
0.197

0.007

12 months before diagnosis
1.00
(0.88, 1.12)
0.947
1.00
(0.87, 1.16)
0.966
1.00
(0.73, 1.37)
0.993
PM10
Whole follow-up period
Whole follow-up
1.16
(1.08, 1.24)
<0.001
1.22
(1.11, 1.33)
<0.001
1.40
(1.10, 1.79)
0.007
12 months after diagnosis
0.96
(0.89, 1.04)
0.310
0.98
(0.89, 1.08)
0.686
1.23
(0.94, 1.62)
0.135
12 months after diagnosis¶

Avg change over follow-up

1.20
(1.11, 1.30)
1.49
(1.39, 1.59)
<0.001

<0.001

1.31
(1.18, 1.45)
1.82
(1.68, 1.98)
<0.001

<0.001

1.50
(1.12, 2.01)
1.90
(1.33, 2.71)
0.007

<0.001

34
12 months before diagnosis
0.95
(0.88, 1.02)
0.179
0.98
(0.89, 1.08)
0.669
1.07
(0.79, 1.46)
0.644
Follow-up period truncated to a maximum of 5 years
Follow-up truncated to 5yrs
0.96
(0.84, 1.09)
0.516
1.18
(1.03, 1.34)
0.014
1.35
(1.01, 1.81)
0.042
12 months after diagnosis
0.93
(0.82, 1.05)
0.238
1.12
(0.98, 1.27)
0.101
1.20
(0.88, 1.63)
0.254
12 months after diagnosis¶

Avg change over follow-up

0.97
(0.84, 1.11)
1.09
(0.97, 1.21)
0.615

0.144

1.21
(1.05, 1.39)
1.25
(1.11, 1.40)
0.007

<0.001

1.44
(1.03, 2.01)
1.71
(1.22, 2.40)
0.031

0.002

12 months before diagnosis
0.94
(0.83, 1.07)
0.359
1.11
(0.98, 1.27)
0.106
1.02
(0.72, 1.43)
0.930
PM2.5#
Whole follow-up period
Whole follow-up
1.20
(1.09, 1.32)
<0.001
1.36
(1.21, 1.51)
<0.001
1.21
(0.95, 1.55)
0.129
12 months after diagnosis
0.94
(0.86, 1.04)
0.224
1.02
(0.91, 1.15)
0.694
1.10
(0.80, 1.50)
0.559
12 months after diagnosis¶

Avg change over follow-up

1.50
(1.32, 1.70)
1.80
(1.60, 2.03)
<0.001

<0.001

2.00
(1.72, 2.33)
2.73
(2.31, 3.23)
<0.001

<0.001

1.37
(0.96, 1.95)
1.86
(1.14, 3.02)
0.079

0.012

12 months before diagnosis
0.95
(0.87, 1.04)
0.306
1.02
(0.91, 1.14)
0.764
0.81
(0.55, 1.21)
0.306
Follow-up period truncated to a maximum of 5 years
Follow-up truncated to 5yrs
1.04
(0.90, 1.21)
0.608
1.24
(1.06, 1.45)
0.007
1.19
(0.87, 1.61)
0.274
12 months after diagnosis
0.92
(0.79, 1.07)
0.263
1.12
(0.96, 1.32)
0.161
1.09
(0.77, 1.56)
0.615
12 months after diagnosis¶

Avg change over follow-up

1.16
(0.97, 1.40)
1.37
(1.20, 1.56)
0.105

<0.001

1.46
(1.20, 1.76)
1.53
(1.29, 1.81)
<0.001

<0.001

1.37
(0.94, 2.01)
1.83
(1.16, 2.89)
0.105

0.009

12 months before diagnosis
0.93
(0.81, 1.08)
0.360
1.13
(0.97, 1.32)
0.124
0.87
(0.56, 1.35)
0.535
*Adjusted for race/ethnicity, marital status, ER/PR status, initial treatment, year of diagnosis, SES, rural-urban commuting area,
BMI, exercise, alcohol, smoke exposure, and family history of breast cancer. Statistically significant associations, at alpha level
of 0.05, are shown in red.
§ SD values for each air pollution summary by pollutant can be found in Table S2.
# PM2.5 data were reported only for participants who were diagnosed breast cancer in 2000 or later.
¶ Average of 12 months after diagnosis was additionally adjusted for exposure change over the follow-up period.

35
Table S 4 Secondary analysis only using address at diagnosis: adjusted* HRs (95%CI) for breast cancer-specific mortality
associated with 1 SD§ increase in each air pollutant summary stratified by stage at diagnosis.
Pollutant
Localized Regional Distant
HR (95%CI) p-value HR (95%CI) p-value HR (95%CI) p-value
NO2
Whole follow-up period
Whole follow-up
1.42
(1.18, 1.70)
<0.001
1.62
(1.44, 1.83)
<0.001
0.99
(0.73, 1.36)
0.970
12 months after diagnosis
1.03
(0.86, 1.23)
0.741
1.14
(1.00, 1.29)
0.045
0.88
(0.58, 1.34)
0.545
12 months after diagnosis¶

Avg change over follow-up

1.84
(1.48, 2.29)
3.00
(2.48, 3.64)
<0.001

<0.001

2.11
(1.80, 2.48)
3.72
(3.05, 4.53)
<0.001

<0.001

0.87
(0.58, 1.31)
2.30
(1.28, 4.14)
0.500

0.005

12 months before diagnosis
1.06
(0.89, 1.27)
0.508
1.12
(0.98, 1.27)
0.088
0.79
(0.49, 1.26)
0.319
Follow-up period truncated to a maximum of 5 years
Follow-up truncated to 5yrs
1.27
(0.96, 1.69)
0.095
1.35
(1.13, 1.60)
0.001
1.00
(0.65, 1.53)
0.988
12 months after diagnosis
1.13
(0.86, 1.50)
0.381
1.20
(1.02, 1.42)
0.032
0.90
(0.55, 1.47)
0.667
12 months after diagnosis¶

Avg change over follow-up

1.64
(1.20, 2.23)
2.13
(1.73, 2.62)
0.002

<0.001

1.75
(1.43, 2.15)
2.37
(1.95, 2.89)
<0.001

<0.001

0.85
(0.52, 1.40)
2.69
(1.46, 4.96)
0.530

0.002

12 months before diagnosis
1.13
(0.87, 1.47)
0.348
1.17
(0.99, 1.39)
0.064
0.77
(0.46, 1.29)
0.325
O3
Whole follow-up period
Whole follow-up
0.89
(0.76, 1.04)
0.147
0.82
(0.71, 0.93)
0.003
0.80
(0.62, 1.03)
0.078
12 months after diagnosis
0.98
(0.85, 1.14)
0.836
0.89
(0.78, 1.01)
0.073
0.83
(0.63, 1.09)
0.173
12 months after diagnosis¶

Avg change over follow-up

0.79
(0.65, 0.95)
0.69
(0.59, 0.81)
0.015

<0.001

0.77
(0.66, 0.90)
0.77
(0.68, 0.88)
0.001

<0.001

0.77
(0.57, 1.03)
0.50
(0.33, 0.76)
0.075

0.001

12 months before diagnosis
1.01
(0.88, 1.16)
0.921
0.86
(0.76, 0.99)
0.032
0.86
(0.65, 1.14)
0.301
Follow-up period truncated to a maximum of 5 years
Follow-up truncated to 5yrs
0.94
(0.77, 1.16)
0.574
0.94
(0.79, 1.12)
0.499
0.75
(0.54, 1.04)
0.085
12 months after diagnosis
0.98
(0.81, 1.19)
0.824
0.98
(0.84, 1.16)
0.837
0.80
(0.59, 1.10)
0.166
12 months after diagnosis¶

Avg change over follow-up

0.92
(0.74, 1.14)
0.84
(0.71, 1.01)
0.440

0.064

0.92
(0.77, 1.10)
0.83
(0.72, 0.95)
0.381

0.008

0.72
(0.51, 1.03)
0.42
(0.25, 0.73)
0.069

0.002

12 months before diagnosis
1.03
(0.85, 1.25)
0.732
0.98
(0.82, 1.17)
0.813
0.89
(0.61, 1.28)
0.515
PM10
Whole follow-up period
Whole follow-up
1.13
(0.97, 1.32)
0.104
1.28
(1.14, 1.44)
<0.001
1.21
(0.92, 1.59)
0.179
12 months after diagnosis
0.98
(0.84, 1.15)
0.800
1.04
(0.92, 1.17)
0.563
1.11
(0.80, 1.54)
0.531
12 months after diagnosis¶

Avg change over follow-up

1.20
(1.01, 1.43)
1.54
(1.35, 1.76)
0.038

<0.001

1.35
(1.18, 1.54)
1.72
(1.55, 1.90)
<0.001

<0.001

1.23
(0.86, 1.76)
1.60
(1.14, 2.26)
0.247

0.007

36
12 months before diagnosis
0.98
(0.84, 1.15)
0.807
1.01
(0.89, 1.14)
0.931
1.04
(0.74, 1.46)
0.813
Follow-up period truncated to a maximum of 5 years
Follow-up truncated to 5yrs
0.99
(0.79, 1.25)
0.953
1.23
(1.05, 1.44)
0.009
1.23
(0.87, 1.73)
0.236
12 months after diagnosis
0.96
(0.76, 1.20)
0.699
1.17
(1.01, 1.37)
0.038
1.10
(0.74, 1.64)
0.644
12 months after diagnosis¶

Avg change over follow-up

1.04
(0.81, 1.34)
1.21
(1.06, 1.40)
0.738

0.006

1.27
(1.08, 1.50)
1.24
(1.11, 1.39)
0.004

<0.001

1.26
(0.82, 1.92)
1.64
(1.16, 2.32)
0.296

0.005

12 months before diagnosis
0.96
(0.77, 1.20)
0.709
1.14
(0.98, 1.34)
0.095
0.97
(0.66, 1.43)
0.882
PM2.5#
Whole follow-up period
Whole follow-up
1.26
(1.00, 1.58)
0.051
1.39
(1.18, 1.65)
<0.001
1.24
(0.93, 1.64)
0.138
12 months after diagnosis
1.01
(0.81, 1.25)
0.944
1.08
(0.91, 1.27)
0.383
1.14
(0.75, 1.73)
0.530
12 months after diagnosis¶

Avg change over follow-up

1.71
(1.28, 2.29)
2.19
(1.69, 2.85)
<0.001

<0.001

1.87
(1.50, 2.33)
2.35
(1.88, 2.93)
<0.001

<0.001

1.40
(0.94, 2.08)
1.93
(1.20, 3.10)
0.100

0.006

12 months before diagnosis
0.99
(0.81, 1.21)
0.947
1.04
(0.89, 1.22)
0.618
0.88
(0.53, 1.49)
0.643
Follow-up period truncated to a maximum of 5 years
Follow-up truncated to 5yrs
1.07
(0.80, 1.44)
0.637
1.26
(1.01, 1.56)
0.040
1.22
(0.82, 1.82)
0.334
12 months after diagnosis
0.94
(0.71, 1.25)
0.679
1.14
(0.93, 1.41)
0.209
1.12
(0.67, 1.87)
0.666
12 months after diagnosis¶

Avg change over follow-up

1.27
(0.91, 1.78)
1.54
(1.27, 1.88)
0.166

<0.001

1.51
(1.15, 1.99)
1.55
(1.26, 1.92)
0.003

<0.001

1.41
(0.86, 2.32)
2.05
(1.21, 3.49)
0.174

0.008

12 months before diagnosis
1.00
(0.77, 1.30)
0.980
1.13
(0.93, 1.38)
0.228
0.93
(0.50, 1.75)
0.832
*Adjusted for race/ethnicity, marital status, ER/PR status, initial treatment, year of diagnosis, SES, rural-urban commuting area,
BMI, exercise, alcohol, smoke exposure, family history of breast cancer, and age group at diagnosis. Statistically significant
associations, at alpha level of 0.05, are shown in red.
§ SD values for each air pollution summary by pollutant can be found in Table S2.
# PM2.5 data were reported only for participants who were diagnosed breast cancer in 2000 or later.
¶ Average of 12 months after diagnosis was additionally adjusted for exposure change over the follow-up period.

37
Table S 5 Secondary analysis only adjusting for CCR covariates: adjusted* HRs (95%CI) for all-cause mortality associated with
1 SD§ increase in each air pollutant summary, stratified by stage and baseline hazard by age group at diagnosis.
Pollutant
Localized Regional Distant
HR (95%CI) p-value HR (95%CI) p-value HR (95%CI) p-value
NO2
Whole follow-up period
Whole follow-up
1.34
(1.25, 1.43)
<0.001
1.49
(1.36, 1.63)
<0.001
1.25
(1.01, 1.55)
0.043
12 months after diagnosis
0.99
(0.92, 1.07)
0.819
1.06
(0.96, 1.17)
0.228
1.04
(0.79, 1.37)
0.788
12 months after diagnosis¶

Avg change over follow-up

2.20
(2.01, 2.42)
3.49
(3.15, 3.86)
<0.001

<0.001

2.09
(1.87, 2.33)
3.89
(3.40, 4.46)
<0.001

<0.001

1.12
(0.85, 1.48)
4.00
(2.34, 6.82) ‡
0.411

<0.001

12 months before diagnosis
1.00
(0.93, 1.07)
0.903
1.04
(0.95, 1.15)
0.379
0.94
(0.70, 1.25)
0.655
Follow-up period truncated to a maximum of 5 years
Follow-up truncated to 5yrs
1.14
(1.00, 1.30)
0.052
1.31
(1.13, 1.51)
<0.001
1.17
(0.89, 1.54)
0.268
12 months after diagnosis
1.02
(0.89, 1.16)
0.797
1.16
(1.00, 1.33)
0.047
1.02
(0.75, 1.38)
0.906
12 months after diagnosis¶

Avg change over follow-up

1.45
(1.25, 1.68)
2.03
(1.78, 2.33)
<0.001

<0.001

1.68
(1.44, 1.96)
2.32
(1.99, 2.70)
<0.001

<0.001

1.05
(0.77, 1.42)
3.06
(2.04, 4.59) ‡
0.769

<0.001

12 months before diagnosis
1.04
(0.91, 1.18)
0.598
1.12
(0.97, 1.29)
0.113
0.89
(0.65, 1.22)
0.476
O3
Whole follow-up period
Whole follow-up
0.90
(0.84, 0.97)
0.008
0.84
(0.76, 0.93)
0.001
0.91
(0.74, 1.12)
0.356
12 months after diagnosis
1.00
(0.93, 1.08)
0.936
0.91
(0.82, 1.00)
0.057
0.94
(0.75, 1.17)
0.570
12 months after diagnosis¶

Avg change over follow-up

0.79
(0.73, 0.86)
0.68
(0.64, 0.74)
<0.001

<0.001

0.77
(0.69, 0.86)
0.76
(0.70, 0.84)
<0.001

<0.001

0.87
(0.68, 1.10)
0.68
(0.49, 0.94) ‡
0.243

0.020

12 months before diagnosis
0.99
(0.93, 1.07)
0.861
0.93
(0.85, 1.03)
0.162
1.00
(0.80, 1.25)
0.997
Follow-up period truncated to a maximum of 5 years
Follow-up truncated to 5yrs
0.95
(0.85, 1.08)
0.445
0.92
(0.80, 1.06)
0.258
0.86
(0.67, 1.11)
0.244
12 months after diagnosis
1.00
(0.88, 1.12)
0.955
0.97
(0.84, 1.12)
0.667
0.91
(0.70, 1.17)
0.445
12 months after diagnosis¶

Avg change over follow-up

0.94
(0.83, 1.07)
0.87
(0.77, 0.97)
0.372

0.015

0.90
(0.77, 1.04)
0.80
(0.70, 0.91)
0.154

0.001

0.83
(0.63, 1.08)
0.66
(0.45, 0.96) ‡
0.170

0.028

12 months before diagnosis
1.01
(0.90, 1.14)
0.858
0.99
(0.86, 1.14)
0.890
1.01
(0.78, 1.30)
0.935
PM10
Whole follow-up period
Whole follow-up
1.16
(1.08, 1.25)
<0.001
1.23
(1.12, 1.34)
<0.001
1.33
(1.06, 1.67)
0.014
12 months after diagnosis
0.97
(0.90, 1.04)
0.412
0.98
(0.89, 1.07)
0.612
1.11
(0.87, 1.42) ‡
0.397
12 months after diagnosis¶

Avg change over follow-up

1.20
(1.11, 1.30)
1.44
(1.35, 1.54)
<0.001

<0.001

1.29
(1.16, 1.42)
1.79
(1.65, 1.94)
<0.001

<0.001

1.38
(1.06, 1.80)
1.85
(1.34, 2.54)
0.017

<0.001

38
12 months before diagnosis
0.96
(0.90, 1.03)
0.302
0.98
(0.90, 1.08)
0.744
1.06
(0.81, 1.38)
0.674
Follow-up period truncated to a maximum of 5 years
Follow-up truncated to 5yrs
0.96
(0.85, 1.09)
0.570
1.17
(1.03, 1.33)
0.018
1.24
(0.95, 1.61) ‡
0.116
12 months after diagnosis
0.95
(0.84, 1.08)
0.429
1.11
(0.97, 1.26)
0.122
1.09
(0.83, 1.43)
0.547
12 months after diagnosis¶

Avg change over follow-up

0.99
(0.86, 1.13)
1.08
(0.97, 1.21)
0.864

0.159

1.19
(1.04, 1.37)
1.23
(1.10, 1.38)
0.012

<0.001

1.34
(0.99, 1.81)
1.61
(1.20, 2.16)
0.057

0.001

12 months before diagnosis
0.97
(0.86, 1.10)
0.611
1.12
(0.98, 1.27)
0.091
1.03
(0.76, 1.39)
0.851
PM2.5#
Whole follow-up period
Whole follow-up
1.21
(1.10, 1.33)
<0.001
1.37
(1.23, 1.53)
0.005
1.24
(1.01, 1.53)
0.130
12 months after diagnosis
0.95
(0.86, 1.04)
0.277
1.02
(0.91, 1.14)
0.159
1.12
(0.85, 1.48)
0.440
12 months after diagnosis¶

Avg change over follow-up

1.51
(1.33, 1.72)
1.77
(1.58, 1.97)
<0.001

<0.001

1.96
(1.69, 2.27)
2.66
(2.27, 3.11)
<0.001

<0.001

1.37
(1.03, 1.83)
1.80
(1.18, 2.74)
0.058

0.011

12 months before diagnosis
0.97
(0.89, 1.06)
0.496
1.01
(0.91, 1.13)
0.134
0.88
(0.63, 1.23)
0.556
Follow-up period truncated to a maximum of 5 years
Follow-up truncated to 5yrs
1.07
(0.92, 1.24)
0.387
1.24
(1.07, 1.45)
0.005
1.22
(0.94, 1.57)
0.130
12 months after diagnosis
0.94
(0.81, 1.09)
0.410
1.12
(0.96, 1.31)
0.159
1.12
(0.84, 1.04)
0.440
12 months after diagnosis¶

Avg change over follow-up

1.19
(1.00, 1.43)
1.36
(1.20, 1.54)
0.055

<0.001

1.44
(1.20, 1.74)
1.51
(1.28, 1.77)
<0.001

<0.001

1.34
(0.99, 1.80)
1.61
(1.12, 2.33) ‡
0.058

0.011

12 months before diagnosis
0.96
(0.83, 1.11)
0.608
1.12
(0.96, 1.31)
0.134
0.89
(0.62, 1.29)
0.556
*Adjusted for race/ethnicity, marital status, ER/PR status, initial treatment, year of diagnosis, SES, rural-urban commuting area,
BMI, exercise, alcohol, smoke exposure, and family history of breast cancer. Statistically significant associations, at alpha level
of 0.05, are shown in red.
§ SD values for each air pollution summary by pollutant can be found in Table 2.
# PM2.5 data were reported only for participants who were diagnosed breast cancer in 2000 or later.
¶ Average of 12 months after diagnosis was additionally adjusted for exposure change over the follow-up period.
‡%change of this estimate was greater than 10.0%.

39
Table S 6 Secondary analysis only adjusting for CCR covariates: adjusted* HRs (95%CI) for breast cancer-specific mortality
associated with 1 SD§ increase in each air pollutant summary, stratified by stage at diagnosis.
Pollutant
Localized Regional Distant
HR (95%CI) p-value HR (95%CI) p-value HR (95%CI) p-value
NO2
Whole follow-up period
Whole follow-up
1.44
(1.21, 1.71)
<0.001
1.59
(1.42, 1.79)
<0.001
1.03
(0.79, 1.35)
0.831
12 months after diagnosis
1.04
(0.87, 1.24)
0.668
1.13
(1.00, 1.28)
0.048
0.91
(0.64, 1.31)
0.616
12 months after diagnosis¶

Avg change over follow-up

1.81
(1.46, 2.26)
2.91
(2.41, 3.51)
<0.001

<0.001

2.00
(1.71, 2.33)
3.53
(2.93, 4.27)
<0.001

<0.001

0.94
(0.66, 1.34)
2.01
(1.25, 3.22) ‡
0.748

0.004

12 months before diagnosis
1.07
(0.90, 1.27)
0.441
1.11
(0.98, 1.26)
0.093
0.87
(0.58, 1.29)
0.483
Follow-up period truncated to a maximum of 5 years
Follow-up truncated to 5yrs
1.25
(0.96, 1.64)
0.102
1.33
(1.13, 1.56)
0.001
1.03
(0.71, 1.49)
0.871
12 months after diagnosis
1.12
(0.85, 1.46)
0.416
1.18
(1.01, 1.38)
0.038
0.93
(0.62, 1.40)
0.723
12 months after diagnosis¶

Avg change over follow-up

1.61
(1.19, 2.18)
2.09
(1.70, 2.55)
0.002

<0.001

1.69
(1.40, 2.03)
2.26
(1.89, 2.69)
<0.001

<0.001

0.96
(0.64, 1.45) ‡
2.28
(1.49, 3.49) ‡
0.858

<0.001

12 months before diagnosis
1.12
(0.87, 1.43)
0.383
1.16
(0.99, 1.36)
0.071
0.85
(0.55, 1.31)
0.466
O3
Whole follow-up period
Whole follow-up
0.87
(0.75, 1.02)
0.088
0.82
(0.72, 0.94)
0.004
0.90
(0.71, 1.13) ‡
0.369
12 months after diagnosis
0.98
(0.84, 1.13)
0.742
0.89
(0.78, 1.01)
0.063
0.91
(0.71, 1.15)
0.433
12 months after diagnosis¶

Avg change over follow-up

0.79
(0.65, 0.95)
0.70
(0.60, 0.82)
0.014

<0.001

0.77
(0.66, 0.90)
0.79
(0.70, 0.88)
0.001

<0.001

0.86
(0.67, 1.12) ‡
0.75
(0.54, 1.03) ‡
0.266

0.079

12 months before diagnosis
1.00
(0.87, 1.15)
0.993
0.87
(0.77, 0.99)
0.037
0.95
(0.75, 1.21)
0.683
Follow-up period truncated to a maximum of 5 years
Follow-up truncated to 5yrs
0.91
(0.74, 1.12)
0.370
0.93
(0.78, 1.10)
0.396
0.81
(0.61, 1.07)
0.140
12 months after diagnosis
0.96
(0.78, 1.18)
0.679
0.97
(0.82, 1.14)
0.697
0.83
(0.63, 1.10)
0.193
12 months after diagnosis¶

Avg change over follow-up

0.90
(0.71, 1.12)
0.84
(0.71, 1.01)
0.343

0.057

0.92
(0.76, 1.10)
0.85
(0.75, 0.97)
0.344

0.014

0.76
(0.55, 1.04)
0.65
(0.45, 0.95) ‡
0.082

0.027

12 months before diagnosis
1.02
(0.84, 1.23)
0.850
0.97
(0.82, 1.15)
0.743
0.93
(0.71, 1.22)
0.592
PM10
Whole follow-up period
Whole follow-up
1.14
(0.98, 1.33)
0.080
1.27
(1.13, 1.42)
<0.001
1.17
(0.90, 1.51)
0.246
12 months after diagnosis
0.98
(0.85, 1.14)
0.829
1.03
(0.91, 1.16)
0.695
1.04
(0.77, 1.40)
0.795
12 months after diagnosis¶

Avg change over follow-up

1.21
(1.03, 1.43)
1.55
(1.36, 1.76)
0.023

<0.001

1.33
(1.16, 1.51)
1.70
(1.54, 1.87)
<0.001

<0.001

1.17
(0.85, 1.61)
1.60
(1.08, 1.94)
0.341

0.013

40
12 months before diagnosis
0.98
(0.84, 1.15)
0.833
1.01
(0.89, 1.14)
0.904
1.06
(0.78, 1.43)
0.729
Follow-up period truncated to a maximum of 5 years
Follow-up truncated to 5yrs
0.98
(0.78, 1.22)
0.831
1.21
(1.04, 1.41)
0.014
1.14
(0.83, 1.56)
0.424
12 months after diagnosis
0.95
(0.76, 1.18)
0.641
1.16
(0.99, 1.35)
0.064
1.01
(0.71, 1.42)
0.966
12 months after diagnosis¶

Avg change over follow-up

1.04
(0.82, 1.32)
1.20
(1.05, 1.38)
0.768

0.010

1.25
(1.06, 1.46)
1.23
(1.11, 1.37)
0.007

<0.001

1.19
(0.81, 1.75)
1.53
(1.14, 2.05)
0.375

0.005

12 months before diagnosis
0.95
(0.76, 1.17)
0.606
1.14
(0.98, 1.33)
0.093
1.00
(0.70, 1.43)
0.985
PM2.5#
Whole follow-up period
Whole follow-up
1.26
(1.01, 1.56)
0.041
1.37
(1.17, 1.61)
<0.001
1.31
(1.05, 1.65)
0.019
12 months after diagnosis
1.01
(0.82, 1.24)
0.915
1.05
(0.90, 1.23)
0.521
1.20
(0.82, 1.74)
0.343
12 months after diagnosis¶

Avg change over follow-up

1.66
(1.28, 2.13)
2.07
(1.65, 2.58)
<0.001

<0.001

1.76
(1.42, 2.17)
2.19
(1.78, 2.69)
<0.001

<0.001

1.49
(1.10, 2.03)
1.97
(1.30, 2.98)
0.011

0.001

12 months before diagnosis
1.00
(0.82, 1.22)
0.996
1.02
(0.88, 1.19)
0.798
1.00
(0.69, 1.46) ‡
0.984
Follow-up period truncated to a maximum of 5 years
Follow-up truncated to 5yrs
1.04
(0.78, 1.37)
0.802
1.24
(1.00, 1.53)
0.050
1.31
(0.97, 1.77)
0.078
12 months after diagnosis
0.92
(0.70, 1.21)
0.565
1.11
(0.91, 1.36)
0.307
1.22
(0.81, 1.83)
0.335
12 months after diagnosis¶

Avg change over follow-up

1.21
(0.87, 1.68)
1.45
(1.21, 1.75)
0.260

<0.001

1.45
(1.11, 1.90)
1.52
(1.25, 1.85)
0.006

<0.001

1.46
(1.03, 2.08)
1.76
(1.22, 2.53)‡
0.034

0.002

12 months before diagnosis
0.98
(0.76, 1.27)
0.894
1.10
(0.90, 1.33)
0.349
1.02
(0.65, 1.59)
0.940
*Adjusted for race/ethnicity, marital status, ER/PR status, initial treatment, year of diagnosis, SES, rural-urban commuting area,
BMI, exercise, alcohol, smoke exposure, family history of breast cancer, and age group at diagnosis. Statistically significant
associations, at alpha level of 0.05, are shown in red.
§ SD values for each air pollution summary by pollutant can be found in Table 2.
# PM2.5 data were reported only for participants who were diagnosed breast cancer in 2000 or later.
¶ Average of 12 months after diagnosis was additionally adjusted for exposure change over the follow-up period.
‡%change of this estimate was greater than 10%.

Abstract (if available)

Abstract Background: Breast cancer has been a leading cause of cancer death in women for decades. Air pollution, a complex mixture of compounds, can affect carcinogenesis in the human breast. Some studies have investigated the relationship between air pollution exposures and breast cancer incidence, but little information is available on the relationship between air pollution exposures after diagnosis and breast cancer survival. Research on this relationship could inform air pollution interventions to better protect people diagnosed with breast cancer in the future. ❧ Methods: We identified 6,539 participants of the California Teachers Study (CTS) with first primary breast cancer newly diagnosed between 1995 to 2015 with high-quality individual-level data from the CTS and data from linkage to the California Cancer Registry (CCR). Residential addresses were geocoded and ambient air pollution exposures after diagnosis, including nitrogen dioxide (NO₂, ppb), ozone (O₃, ppb), particulate matter with diameter <10 µm (PM⏨, µg/m³), and 2.5 µm (PM₂.₅, µg/m³), were assigned using inverse distance weighted averages of central site monitor data. Follow-up time was calculated from the date of diagnosis to either the date of death, the date of study end (December 31, 2015), the date of moving out of California, or the date of starting missing air pollution assignment, whichever occurred first. In our primary analysis, Cox proportional hazards models and Fine and Gray competing risk models were used to estimate hazard ratios (HRs) relating air pollutant exposure averages to all-cause mortality and breast cancer-specific mortality respectively, by tumor stage at diagnosis (localized, regional, and distant). In secondary analysis, we evaluated the sensitivity of our results to using only data from the CCR, which has less individual-level data (e.g., no smoking history and only address at diagnosis). ❧ Results: After adjusting for covariates, the hazard for breast-cancer specific mortality for participants diagnosed at the regional stage was 1.14 (95% CI: 1.00, 1.29) times higher per SD increase in average NO₂ over the 12 months after diagnosis. We didn’t detect other significant associations between mortality with average O₃, PM⏨, or PM₂.₅ over 12 months after diagnosis. Statistically significant associations between follow-up period average air pollutants and all-cause mortality were attenuated when the follow-up period was truncated to a maximum of 5 years. Using limited individual-level data and address only at diagnosis produced results similar to the primary analysis. ❧ Conclusions: Our analyses do not provide clear evidence for associations of air pollution exposures after diagnosis with breast cancer survival. However, our findings motivate future analyses using time-varying annual average exposures since results suggest a bias in analyses using follow-up period average exposures for pollutants with downwards trends during the study period, which could potentially be remedied using time-varying exposures.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Disparities in colorectal cancer survival among Latinos in California

PDF

An analysis of disease-free survival and overall survival in inflammatory breast cancer

PDF

Predictive factors of breast cancer survival: a population-based study

PDF

Examining exposure to extreme heat and air pollution and its effects on all-cause, cardiovascular, and respiratory mortality in California: effect modification by the social deprivation index

PDF

Assessment of the mortality burden associated with ambient air pollution in rural and urban areas of India

PDF

Association between informed decision-making and mental health-related quality of life in long term prostate cancer survivors

PDF

Associations between isoflavone soy protein (ISP) supplementation and breast cancer in postmenopausal women in the Women’s Isoflavone Soy Health (WISH) clinical trial

PDF

Racial/ethnic differences in colorectal cancer patient experiences, health care utilization and their association with mortality: findings from the SEER-CAHPS data

PDF

Screening and association testing of coding variation in steroid hormone coactivator and corepressor genes in relationship with breast cancer risk in multiple populations

PDF

Instability of heart rate and rating of perceived exertion during high-intensity interval training in breast cancer patients undergoing anthracycline chemotherapy

PDF

Incidence and survival rates of the three major histologies of renal cell carcinoma

PDF

Disparities in exposure to traffic-related pollution sources by self-identified and ancestral Hispanic descent in participants of the USC Children’s Health Study

PDF

Diet quality and pancreatic cancer incidence in the multiethnic cohort

PDF

Pathogenic variants in cancer predisposition genes and risk of non-breast multiple primary cancers in breast cancer patients

PDF

Analysis of factors associated with breast cancer using machine learning techniques

PDF

Sensor-based mobile health approaches for personal air pollution and pediatric asthma studies

PDF

A novel risk-based treatment strategy evaluated in pediatric head and neck non-rhabdomyosarcoma soft tissue sarcomas (NRSTS) patients: a survival analysis from the Children's Oncology Group study...

PDF

Association of maternal and environmental factors with infant feeding behaviors in a birth cohort study

PDF

Air pollution, smoking, and multigenerational DNA methylation Signatures: a study of two southern California cohorts

PDF

Arm lymphedema in a multi-ethnic cohort of female breast cancer survivors

Asset Metadata

Creator Wang, Minhao (author)

Core Title Air pollution and breast cancer survival in California teachers: using address histories and individual-level data

School Keck School of Medicine

Degree Master of Science

Degree Program Applied Biostatistics and Epidemiology

Publication Date 04/26/2021

Defense Date 03/17/2021

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag Air pollution,breast cancer,cancer survival,OAI-PMH Harvest

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Eckel, Sandrah Proctor (committee chair), Garcia, Erika (committee member), Liu, Lihua (committee member)

Creator Email minhaowa@usc.edu,monica96w@gmail.com

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c89-454198

Unique identifier UC11668918

Identifier etd-WangMinhao-9546.pdf (filename),usctheses-c89-454198 (legacy record id)

Legacy Identifier etd-WangMinhao-9546.pdf

Dmrecord 454198

Document Type Thesis

Rights Wang, Minhao

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA