University of Southern California Dissertations and Theses

The role of pesticide exposure in breast cancer

(USC Thesis Other)

The role of pesticide exposure in breast cancer

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content

THE ROLE OF PESTICIDE EXPOSURE

IN BREAST CANCER

by

Carrie Tayour

A Dissertation Presented to the
FACULTY of THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(EPIDEMIOLOGY)

May 2015

Copyright 2015 Carrie Tayour

Dedication

This dissertation is dedicated to my parents Charles and Rose Marie Nagy
for their unconditional love and support, to my sister Lesley Nagy who inspires me
every step of the way, and to my husband Latif Tayour for his understanding and
encouragement.
ii

Acknowledgements

I would like to thank my mentor and dissertation committee chair Myles
Cockburn, PhD for his advice, expertise, direction, motivation, and time. I
sincerely thank my committee members Anna Wu, PhD, John Wilson, PhD, Bryan
Langholz, PhD, and Meredith Franklin, PhD for their invaluable insights ,
knowledge and guidance.
This work was possible with the help of wonderful, efficient and detail -
oriented staff in the participant recruitment and data collection by Carmen Harr
and Letech Caldera- Huerta, as well as administrative support from Marlene
Caldera. I would like to thank Kaveh Shahabi for his work improving the
computing power and functional ease of the software used in the exposure
assessment. I would also like to specially thank Dr. Beate Ritz and the UCLA staff
for providing me with data fro m the Parkinson’s disease study and all their time
and effort collecting the data.

iii

Table of Contents

Dedication ii
Acknowledgements iii
List of Tables vi
List of Figures xi
Abbreviations xv
Abstract xvi
Chapter 1
1. Exposure to Pesticides and Breast Cancer Risk 1
1.1 Introduction 1
1.2 Pesticides as Endocrine Disruptors 2
1.3 Previous Epidemiologic Studies of Breast Cancer Risk from Pesticide
Exposure 4
1.4 Future Direction in Studies Assessing Pesticide Exposure in Breast
Cancer Risk 17
1.5 New and Improved Methods for Pesticide Exposure Assessment 26
1.6 Previous Studies of Pesticide Exposure and Risk of Other Diseases 36
Chapter 1 References 41

Chapter 2
2. Registry-Based Case-Control Study of Breast Cancer Risk from Ambient
Exposure to Pesticides 52
2.1 Abstract 52
2.2 Introduction 53
2.3 Materials and Methods 56
2.4 Results 63
2.5 Discussion 66
Chapter 2 References 85

Chapter 3
3. Case-Control Study of Breast Cancer Risk from Ambient Exposure to
Pesticides 93
3.1 Abstract 93
3.2 Introduction 94
3.3 Materials and Methods 95
3.4 Results 105
3.5 Discussion 108
Chapter 3 References 140
iv

Chapter 4
4. Impact of Geocoding Certainty on the Pesticide Exposure Assessment
in a Case-Control Study of Breast Cancer Risk 147
4.1 Abstract 147
4.2 Introduction 148
4.3 Materials and Methods 158
4.4 Results 163
4.5 Discussion 167
Chapter 4 References 184

Chapter 5
5. Summary and Discussion of Findings 188
5.1 Future Directions 191
5.2 Public Health Impact 198
Chapter 5 References 200

Bibliography 203

Appendix
Pesticide Exposure and Breast Cancer Pilot Study Questionnaire 223

v

List of Tables

Table 1.1 Studies of Breast Cancer Risk and Pesticide Exposure Based
on Self-Reported Exposure to Pesticides.
19

Table 1.2 Studies of Breast Cancer Risk and Pesticide Exposure in
Agricultural Workers.
21

Table 1.3 Ecologic or Regression Studies of Breast Cancer Risk and
Pesticide Exposure.
23

Table 1.4 Studies of Breast Cancer Risk and Pesticide Exposure Based
on GIS-Based Proxy Measures of Exposure.
24

Table 1.5 Previous Studies Assessing Disease Risk and Exposure to
Pesticides Based on GIS -Methods That Utilize California
Pesticide Use Reporting and Land-Use Data.
40

Table 2.1 Comparison of Characteristics of Breast Cancer Cases and
Other Types of Cancer Controls in Fresno, Tulare, and Kern
Counties, 1988–2012.
75

Table 2.2 Prevalence of Exposure and Measures of Association for
Women Diagnosed with Breast Cancer Compared to Several
Types of Cancer Registry Control Groups, Including All
Women Diagnosed at Sites Other Than the Breast, All
Diagnoses Except Lung Cancers, and All Diagnoses Except
Ovarian, Uterine and Other Reproductive C ancers, Fresno,
Tulare, and Kern Counties, 1988–2012.
76

Table 2.3 Measures of Association between Breast Cancer and
Exposure to Selected Pesticides Based on Residential
Address at the Time of Diagnosis Using Linked PUR and
Land-Use Data for 1974 –2012, in Fresno, Tulare, and Kern
Counties, Assuming 10 -Year or 15 -Year Latency Periods or
No Latency Period.
77

vi

List of Tables, Continued

Table 2.4 Measures of Association for Estrogen Receptor Breast
Cancer Subtypes and Exposure to Selected Pesticides Based
on Residential Address at the Time of Diagnosis Using
Linked PUR and Land -Use Data for 1974 –2012, in Fresno,
Tulare, and Kern Counties.
79

Table 2.5 Measures of Association Between Breast Cancer and
Exposure to Selected Pesticides Based on Residential
Address at the Time of Diagnosis Using Linked PUR and
Land-Use Data for 1974 –2012, in Fresno, Tulare, and Kern
Counties, by Geocode Certainty of the Residential Address
at Diagnosis.
80

Table 2.6 Measures of Association Between Breast Cancer and
Exposure to Selected Pesticides at Ages 20 –39 Based on
Residential Address at the Time of Diagnosis Using Linked
PUR and Land- Use Data for 1974 –2012, in Fresno, Tulare,
and Kern Counties.
81

Table 2.7 Measures of Association Between Breast Cancer and
Exposure to Selected Pesticides Based on Residential
Address at the Time of Diagnosis Among Women in Fresno,
Tulare, and Kern Counties Using Linked PUR and Land-Use
Data for 1974 –2012 Compared to a Random Selection of
Tax-Assessor Parcels.
82

Table 3.1 Comparison of Selected Characteristics in Breast Cancer
Cases (Diagnosed in 2007 –2008) and Population-Based
Controls (2001–2011), in Fresno, Tulare and Kern Counties.
124

Table 3.2 Comparison of Occupational and Residential Characteristics
and Self -Reported Pesticide Use Among Breast Cancer
Cases (Diagnosed in 2007 –2008) and Population-Based
Controls (2001–2011), in Fresno, Tulare and Kern Counties.
126

vii

List of Tables, Continued

Table 3.3 Measures of Association Between Breast Cancer and
Ambient Exposure to Selected Pesticides Based on
Residential and Occupational Address Histories Among
Women in Fresno, Tulare, and Kern Counties Using Linked
PUR data for 1974–2011.
127

Table 3.4 Measures of Association Between Breast Cancer and
Exposure to Organochlorine Pesticides and Chlorpyrifos at
Ages 20–39 and at Ages 40– 59 Among Women in Fresno,
Tulare, and Kern Counties using linked PUR data for 1974 –
2011.
128

Table 3.5 Comparison of Self-Reported Ever Lived on or Worked on a
Farm and Ambient Pesticide Exposure to Organochlorines,
Chlorpyrifos, Diazinon, or 1,3 -Dichloropropene Using
Linked PUR data from 1974–2011.
130

Table 3.6 Comparison of Residential Histories (1974 to Year of
Diagnosis or Interview) in Breast Cancer Cases (Diagnosed
in 2007 –2008) and Population -Based Controls (2001 –
2011), in Fresno, Tulare and Kern Counties.
131

Table 3.7 Measures of Association Between Breast Cancer and
Ambient Exposure to Selected Pesticides Among Women in
Fresno, Tulare and Kern Counties Using Linked PUR Data
for 1974 –2011, When Exposure is Based on Address at
Diagnosis or Interview Only or Complete Residential
Histories.
132

Table 3.8a Comparison of Length Resided in Fresno, Tulare, or Kern
Counties and Length Resided in California in Breast Cancer
Cases (Diagnosed in 2007 –2008) and Population -Based
Controls (2001–2011), by Age and SES.
133

Table 3.8b Comparison of Breast Cancer Cases (Diagnosed in 2007 –
2008) and Population -Based Controls (2001 –2011) by
Length Resided in Fresno, Tulare, or Kern Counties, and by
Age and SES.
134

viii

List of Tables, Continued

Table 3.8c Comparison of Breast Cancer Cases (Diagnosed in 2007 –
2008) and Population -Based Controls (2001 –2011) by
Length Resided in California, and by Age and SES.
135

Table 3.9 Comparison of Selected Characteristics in N on-Hispanic
White Women Ages 55 –74 Diagnosed with Invasive Breast
Cancer in 2007 –2008 Enrolled as Participants in the Case -
Control Study from Fresno, Tulare, and Kern Counties and
Those in the Cancer Registry of Central California (CCRC).
136

Table 4.1 Measures of Association Between Breast Cancer and
Exposure to Selected Pesticides Among Women in Fresno,
Tulare and Kern Counties Using Linked PUR data for 1974–
2011, All Geocode Certainty Levels Combined.
177

Table 4.2 Measures of Association Between Breast Cancer and
Exposure to Selected Pesticides Among Women in Fresno,
Tulare and Kern Counties Using Linked PUR Data for
1974–2011, Stratified by the Proportion of Low (Red)
Geocode Certainty Locations in the Residential Histories.
177

Table 4.3 Measures of Association Between Breast Cancer and
Exposure to Selected Pesticides Among Women in Fresno,
Tulare and Kern Counties Using Linked PUR Data for
1974–2011, Stratified by the Proportion of Low (Red)
Geocode Certainty Locations in the Residential Histories,
Excluding Subjects Exposed Only at Low Geocode Certainty
Locations in Their Residential Histories.
181

Table 4.4 Measures of Association Between Breast Cancer and
Exposure to Selected Pesticides Among Women in Fresno,
Tulare and Kern Counties Using Linked PUR Data for
1974–2011 for All Geocode Certainty Levels Combined,
Including vs. Excluding Subjects Exposed Only at Low
Geocode Certainty Locations in Their Residential Histories.
181

ix

List of Tables, Continued

Table 4.5 Measures of Association Between Breast Cancer and
Exposure to Selected Pesticides Among Women in Fresno,
Tulare, and Kern Counties Using Linked PUR Data for
1974–2011 for All Geocode Certainty Levels Combined,
Including vs. Excluding Subjects Unexposed Only at Low
Geocode Certainty Locations in Their Residential Histories.
183

x

List of Figures

Figure 1.1 Map of California Highlighting the Central Valley and
the Agriculturally Productive Counties of Fresno, Tulare
and Kern that are the Regional Focus of the Studies in
This Dissertation.
27

Figure 2.1

Map of App roximate Residential Locations at the Time
of Diagnosis for Women with Breast Cancer and Other
Types of Cancer as Controls in California’s Central
Valley, 1988–2012.
83

Figure 2.2 Map of Approximate Residential Locations at the Time
of Diagnosis for W omen with Breast Cancer in 1988 –
2009 and Matched to a Random Selection of 10,000 Tax
Assessor Parcel Centroids in California’s Central Valley.
84

Figure 3.1 Power Calculation for Pilot Case-Control Study of Breast
Cancer Risk and Exposure to Ambient Pesticides
Assuming a Sample Size of 150 Cases and 150 Controls.
105

Figure 3.2 Trends in Pesticide Use in Fresno, Tulare, and Kern
Counties, 1974–2011.
137

Figure 3.3 Map of Approximate Residential Locations at the Time
of Diagnosis for Breast Cancer Cases (2007–2008) and
at the Time of Interview for Population -Based Controls
(2001–2011), in Fresno, Tulare and Kern Counties.
138

Figure 4.1 A ZCTA in Fresno County With the Centroid Marked
With a P oint and 500 -Meter Circular Buffer Around it.
The Green Topography I ndicates Agricultural Areas and
the Orange S hapes Represent Fields Where Chlorpyrifos
was Applied.
157

Figure 4.2 Trends in Use of a Selected Group of Organochlorines
Pesticides and Chlorpyrifos in Fresno, Tulare, and Kern
Counties, 1974–2011.

174

xi

List of Figures, Continued

Figure 4.3 Comparison of Levels of High (Green) and Low (Red)
Geocode Certainty at H istorical Residential Addresses
(1974 to Year of Diagnosis or Interview) in Breast Cancer
Cases (Diagnosed in 2007 –2008) and Population-Based
Controls (2001 –2011), in Fresno, Tulare and Kern
Counties.
175

Figure 4.4a

Comparison of Levels of High (Green) and Low (Red)
Geocode Certainty at Historical Residential Addresses
(1974 to Year of Diagnosis or Interview) in Breast Cancer
Cases (Diagnosed in 2007 –2008) and Population-Based
Controls (2001 –2011) in Fresno, Tulare an d Kern
Counties, Stratified by Ever Exposed to Organochlorines.
176

Figure 4.4b

Comparison of Levels of High (Green) and Low (Red)
Geocode Certainty at Historical Residential Addresses
(1974 to Year of Diagnosis or Interview) in Breast Cancer
Cases (Diagnosed in 2007 –2008) and Population-Based
Controls (2001 –2011) in Fresno, Tulare and Kern
Counties, Stratified by Ever Exposed to Chlorpyrifos.
176

Figure 4.5a

Comparison of Levels of High (Green) and Low (Red)
Geocode Certainty at Historical Residential Addresses
(1974 to Year of Diagnosis or Interview) in Breast Cancer
Cases and Population -Based Controls, in Fresno, Tulare
and Kern Counties, Stratified by Ever Exposed to
Organochlorines and Boxed Around Subjects With 1 0%
or More of Their Residential Histories Having Low
Geocode Certainty Locations.
178

Figure 4.5b

Comparison of Levels of High (Green) and Low (Red)
Geocode Certainty at Historical Residential Addresses
(1974 to Year of Diagnosis or Interview) in Breast Cancer
Cases and Population -Based Controls, in Fresno, Tulare
and Kern Counties, Stratified by Ever Exposed to
Chlorpyrifos and Boxed Around Subjects With 10% or
More o f Their Residential Histories Having Low
Geocode Certainty Locations.
178

xii

List of Figures, Continued

Figure 4.6a Comparison of Levels of High (Green) and Low (Red)
Geocode Certainty at Historical Residential Addresses
(1974 to Year of Diagnosis or Interview) in Breast Cancer
Cases and Population -Based Controls, in Fresno, Tulare
and Kern Counties, Shaded for Person -Years Exposed to
Organochlorines and Boxed Around Subjects With 10%
or More of Their Residential Histories Having Low
Geocode Certainty Locations.
179

Figure 4.6b Comparison of Levels of High (Green) and Low (Red)
Geocode Certainty at Historical Residential Addresses
(1974 to Year of Diagnosis or Interview) in Breast Cancer
Cases and Population -Based Controls, in Fresno, Tulare
and Kern Counties, Shaded for Person -Years Exposed to
Chlorpyrifos and Boxed Around Subjects With 10% or
More of Their Residential Histories Having Low
Geocode Certainty Locations.
179

Figure 4.7a

Comparison of Levels of High (Green) and Low (Red)
Geocode Certainty at Historical Residential Addresses
(1974 to Year of Diagnosis or Interview) in Breast Cancer
Cases and Population -Based Controls, in Fresno, Tulare
and Kern Counties, Shaded for Person -Years Exposed to
Organochlorines, Boxed Around Subjects With 10% or
More of Their Residential Histories Having Low
Geocode Certainty Locations, and Shaded and Boxed
Around Subjects Exposed Only at Low Geocode
Certainty Locations in Their Residential Histories.
180

Figure 4.7b

Comparison of Levels of High (Green) and Low (Red)
Geocode Certainty at Historical Residential Addresses
(1974 to Year of Diagnosis or Interview) in Breast Cancer
Cases and Population -Based Controls, in Fresno, Tulare
and Kern Counties, Shaded for Person -Years Exposed to
Chlorpyrifos, Boxed Around Subjects With 10% or More
of Their Residential Histories Having Low Geocode
Certainty Locations, and Shaded and Boxed Around
Subjects Exposed Only at Low Geocode Certainty
Locations in Their Residential Histories.
180

xiii

List of Figures, Continued

Figure 4.8a Comparison of Levels of High (Green) and Low (Red)
Geocode Certainty at Historical Residential Addresses
(1974 to Year of Diagnosis or Interview) in Breast Cancer
Cases and Population-Based Controls, in Fresno, Tulare,
and Kern Counties, Shaded for Person -Years Exposed to
Organochlorines, and Shaded and Boxed Around
Subjects Unexposed at Low Geocode Certainty Locations
in Their Residential Histories.
182

Figure 4.8b Comparison of Levels of High (Green) and Low (Red)
Geocode Certainty at Historical Residential Addresses
(1974 to Year of Diagnosis or Interview) in Breast Cancer
Cases and Population-Based Controls, in Fresno, Tulare,
and Kern Counties, Shaded for Person -Years Exposed to
Chlorpyrifos, and Shaded and Boxed Around Subjects
Unexposed at Low Geocode Certainty Locations in Their
Residential Histories.
182

xiv

Abbreviations

BMI

Body Mass Index (kg/m
2
)

CCR

California Cancer Registry

CRCC

Cancer Registry of Central California

DDT

Dichlorodiphenyltrichloroethane

DDE

Diphenyldichloroethylene

ER

Estrogen Receptor

GIS

Geographic Information Systems

GPS

Global Positioning System

PLSS Public Land-use Survey System

PUR

California Pesticide Usage Reporting System

xv

Abstract

There is strong evidence that a variety of pesticides affect hormone
metabolism. Exposure to these chemicals may be involved in the development of
breast cancer, but previous studies have had conflicting results. Current
methodology is severely lacking for assessing cumulative exposures that amass
over a lifetime. The objectives of this research were to assess breast cancer risk
from ambient pesticide exposure using a Geographical Information Systems
(GIS)-based approach to examine historical effects in the most agriculturally
productive region in the U.S. , where pesticide drift from neighboring application
presents a major source of exposure.
We conducted registry -based and pilot case-control studies employing a
GIS-based exposure assessment to see if a comprehensive exposure model
revealed true risk for breast cancer in postmenopausal women from specific,
hormone-related pesticides . We evaluated risk from a selected group of
organochlorine pesticides with biologically -plausible links to breast cancer and
three other pesticides detected in ambient air monitoring at lev els of concern to
human health (chlorpyrifos, diazinon, and 1,3 -dichloropropene). We observed no
association between breast cancer and exposure to any of the pesticides in the
registry-based study; however, the exposure metric likely underestimated risk by
comparing cases to wo men with other cancer diagnoses with limited information
on potential confounders and relying on exposures occurring only at address of
diagnosis.
xvi

The pilot case -control study incorporated self-reported residential and
occupational address timelines to show how historical exposure information
reduced the non-differential bias in estimates obtained from only address at time
of diagnosis. We observed no association between breast cancer risk and exposure
to the group of organochlorine s, but found a three -fold increased risk of breast
cancer for exposure to the organophosphate chlorpyrifos af ter adjusting for
exposure to other pesticides including organochlorines (OR = 3.22, 95% CI: 1.38,
7.53). Organophosphate pesticides have rarely been evaluated in studies of breast
cancer risk and additional research is needed to confirm this finding.
The GIS-based approach relied on geocoding methods to identify where
people lived or worked in relation to historic pesticide applications. We evaluated
the “certainty” of the geocoded location s and the potential to impact exposure
classification. In the registry-based study, we adjusted for geocode certainty in the
analyses since there was only one location per person ; however, in the pilot study
we developed novel methods to assess geocode certainty associated with mul tiple
historical locations per subject. The certainty of the historical geocoded locations
can impact the estimates of relative risk and the magnitude and direct ion of
change will depend on the type of pesticide and how far in the past the exposure
and location occurred. As GIS -based methods are employed with increasing
frequency in environmental studies, researchers need to take into account the
certainty of the geocoded locations.
We explored whether or not breast cancer risk was associated with specific
pesticides that had strong biological plausibility or have been measured in the
ambient air at levels deemed of concern to public health. Future directions will
xvii

explore ways to improve our GIS -based approach for evaluating exposure to
pesticides in epidemiologic studies.
xviii

Chapter 1

1. Exposure to Pesticides and Breast Cancer Risk

1.1 Introduction
Breast cancer is the most common cancer among American women. Breast
cancer mortality is decreasing but the num bers of new cases being diagnosed has
remained steady (1). Investigating other environmental and genetic factors that
contribute to breast cancer risk can potentially lead to a better understanding of
the mechanisms involved in breast cancer etiology and contribute to new
prevention strategies. Chemicals that persist in the env ironment, bioaccumulate
in humans and have biologically plausible links to breast cancer should be
evaluated as possible risk factors. One such factor is pesticide exposure.
Known risk factors can explain only 40% of the variation in breast cancer
risk in the U.S. (2). Lifetime exposure to estrogens is a strong risk factor for breast
cancer (3-5). Several established breast cancer risk factors are related to exposure
to endogenous or exogenous hormones, including early menarche, late
menopause, parity, postmenopausal hormone therapy and postmenopausal
obesity (3,6-8). Estrogens act as promoters of cell growth and precursors for
carcinogenic metabolites (9,10), raising concerns about the possible role of the
cumulative impacts of exogenous xenoestrogens such as p esticides or other
industrial compounds that have estrogenic or hormone disruptive effects on
animals and humans.
1

Known risk factors for breast cancer have mainly been established through
epidemiologic studies collecting information on prior exposures through self -
reported interview, as is the case for family history, number of pregnancies,
alcohol use and physical inactivity. Breast cancer risk was mainly linked to
exposure to endogenous through prospective studies (11,12) and to exogenous
estrogens through clinical trials of h ormone replacement therapy and cancer
treatment (13-15), where the levels of hormones were directly measured or
controlled. These methods however, are inadequate to assess lifetime exposure to
pesticides since most people are unaware of exposures to pesticides that may
come from ambient sources such as through the air, water, soil or food, and it
would be highly une thical to suggest any kind of study involving intentional
exposure to pesticides. Epidemiologic studies aimed at addressing the risk of
breast cancer from pesticide exposure must therefore develop new methods that
are suitably designed to detect risk from chronic, low -level exposure to
environmental chemicals throughout a person’s lifetime.

1.2 Pesticides as Endocrine Disruptors
As part of the Federal Food, Drug, and Cosmetic Act of 1996, the U.S.
Environmental Protection Agency evaluated chemicals that may a ct similar to
estrogen and have a possible effect on the human endocrine system. These
chemicals were “endocrine disrupting” if they could bind to the estrogen receptor
(ER) in animal models, activate ER -alpha transcription, or produce an estrogen-
like response through in vivo assays at development, puberty or reproduction
stages of animal models (16). Organochlorine pesticides are among the chemicals
2

considered endocrine disrupting, with the ability to interfere with the metabolism
of hormones through various mechanisms.
Organochlorine pesticides affect hormone function ing by mimicking
estrogen or affecting enzyme systems involved in hormone metabolism (17,18).
Since lifetime estrogen exposure is a key factor in breast cancer development
(19,20), exposure to organochlorines may be plausibly involved in breast cancer
development. The mechanism is thought to be an increase in cellular proliferation
in the breast, which in turn increases the chances of genetic errors occurring
during cell division (21).
Persistent organochlorine pesticides bioaccumulate in fat and by -products
are measurable in breast tissue and breast milk (22-24). Organochlorine
pesticides such as dichlorodiphenyltrichloroethane ( DDT),

dieldrin, endosulfan,
methoxychlor and toxaphene have chemical structure similarities to estrogen
(25,26). As a result they interact with estrogen receptors (27), increase the
estrogen-dependent transcription of target genes and modulate metabolic
enzymes (28). In toxicological studies done in vitro, exposure to organochlorines
show dose-dependent estrogenic effects (29) and stronger effects with exposure to
multiple organochlorines (17,30). Even though organochlorine pesticides have
weaker estrogenic activity than endogenous estrogens, the prolonged exposure to
combinations of organochlorines may have a cumulative impact even at low
concentrations, inducing breast cell proliferation by upregulating the expression
of certain genes (31).
Breast cancer susceptibility depends on the levels of circulating carcinogens
in the breast. Many organochlorines are known or suspected human carcinogens
3

(32). In breast cancer patients, higher concentrations of organochlorines have
been found in tumor tissue compared to surrounding tumor -free tissue (33). It is
plausible that the carcinogenic potential of organochlorines is related to the rate of
their activation or elimination by metabolic enzymes. Cytochrome P450 e nzymes
metabolize estrogens and xenobiotics such as organochlorines or other synt hetic
substances that are usually toxic to living cells and, in the process, produce
reactive oxygen species that can react with deoxyribonucleic acid ( DNA).
Organochlorines have been shown to interact with P450 metabolic enzymes by
binding to their promoter region s (34) and increasing the production of reactive
species that contribute to carcinogenesis (35). Organochlorines promote cell
proliferation in breast cells in vitro, at doses similar to pesticide levels found in
humans (36,37).

1.3 Previous Epidemiologic Studies of Breast Cancer Risk from
Pesticide Exposure
Although the estrogenic properties of orga nochlorine pesticides have been
widely recognized, results from previous epidemiologic studies of the impact of
pesticide exposure on breast cancer risk have been conflicting. A case-control
study by Teitelbaum et al. 2007 conducted in New York found an increased risk of
breast cancer with self-reported pesticide use where the sum of lifetime
applications was categorized as quintiles (OR = 1.39, 95% CI: 1.15, 1.68 for quintile
2–5 vs. the referent 1
st
quintile), but did not observe a dose-response relationship
with increasing quintiles of lifetime pesticide application (Table 1.1) (38).
Associations were observed for breast cancer and use of lawn and garden
4

pesticides (OR = 1.34, 95% CI: 1.11, 1.63), but not for use of nuisance -pest
insecticides (OR = 1.07, 95% CI: 0.80, 1.42). Another case -control study
conducted by Farooq et al. 2010 in New York found no associations for outdoor
use of lawn pesticides (OR = 1.30, 95% CI: 0.79, 2.14 ) or for indoor use of
insecticides to control for ants or cockroaches (OR = 1.13, 95% CI: 0.75, 1.72 ),
however controls were recruited from hospital surgery patients and may have
higher exposures than the general population if exposure to pesticides is also
related to benign breast disease and other non -breast related chronic conditions
(39). Neither of these studies were able to assess exposure to particular pesticides
of interest that may be most relevant to breast cancer etiology , but instead
grouped together broad categories of pesticide use by pest type or by location such
as indoor or outdoor use.
Self-reported residential pesticide use may be an inappropriate method for
measuring prior exposure to specific types of pesticides since most people are not
likely to be able to recall specific products or chemicals that they used unless they
worked in agriculture or directly in or around pesticide application. The grouping
of many different types of pesticides may misclassify the true exposure among
both cases and controls, leadi ng to non -differential misclassification biasing the
estimates towards the null. Stratification in these two studies based on the type of
pest targeted by the pesticide and who applied the product (self vs. professional or
other) usually resulted in low nu mbers of exposed subjects. For example, only
3.4% of cases and 2.8% of controls reported that they applied lawn insecticides
themselves in the study by Teitelbaum et al. (O R = 1.56, 95% CI: 1.01, 2.43) and
only 2.5% of cases and 0.9% of controls applied outdoor insecticides themselves in
5

the study by Farooq et al. (OR = 2.68, 95% CI: 0.72, 9.92) (Table 1.1) . Low
exposure prevalence could also lead to estimates that are biased towards the null
from too few people being exposed and wide confidence intervals indicating that
the study was underpowered to detect an effect if one truly existed.
A third study to base pesticide exposure on self -reported use at home was
conducted by El-Zaemey et al. in Australia and found no association for household
use (OR = 1.10, 95% CI: 0.86, 1.37) (40). Only large, general categories of self -
application in home, on yard and on pets were examined, as well as if they had
“treatment for white ants/termites,” which could potentially include applications
conducted by professionals. Occupational pesticide exposure was evaluated using
a job exposure matrix to identify potential exposures from self -reported job
histories; however, the prevalence of exposure was very low, totaling only 1.7% of
cases and 2.2% of controls (OR = 0.77, 95% CI: 0.45, 1.32) . El-Zaemey et al. also
explored self -reported noticing of pesticide spray drift near residences in
agricultural areas (41). Although the prevalence of exp osure was not particularly
high (15.5% of cases and 11.6% of controls), they did find an association between
breast cancer and noticing pesticide drift (OR = 1.43, 95% CI: 1.15, 1.78) . This
finding however, was completely confounded by participants’ belief in whether or
not pesticides cause breast cancer. There was a statistically significant association
between breast cancer and noticing pesticide drift for women who believed that
pesticides cause breast cancer (OR = 1.47, 95% CI: 1.15, 1.87), but not among
women who did not (OR = 0.94, 95% CI: 0.51, 1.74).
The Agricultural Health Study, a prospective cohort of pesticide applicators
in Iowa and North Carolina, evaluated the link between breast cancer and
6

pesticide exposure among the pesticide applicators’ wives (42). Cumulative
lifetime pesticide exposure was determined from questionnaires completed by the
wives themselves regarding personal use as well as from indirec t exposure via
occupational use reported by their applicator husbands. Approximately 11.0% of
wives reported frequently washing work clothing worn by their husbands. The
wives who frequently washed their husbands’ work clothing were no more likely to
develop breast cancer than wives who rarely washed their husbands’ work clothing
(RR = 1.4, 95% CI: 0.8, 2.7). No associations were observed for self-reported use
by the applicators’ wives for any specific pesticide, but the wives had an increased
risk of breast cancer when their husbands occupationally applied captan (RR =
2.7, 95% CI: 1.7, 4.3), aldrin (RR = 1.9, 95% CI: 1.3, 2.7), chlordane (RR = 1.7,
95%CI: 1.2, 2.5), dieldrin (RR = 2.0, 95% CI: 1.1, 3.3), heptachlor (RR = 1.6, 95%
CI: 1.1, 2.4), lindane (RR = 1.7, 95% CI: 1.1, 2.5), and 2,4,5 -
trichlorophenoxypropionic acid (RR = 2.0, 95% CI: 1.2, 3.2) (43). The role of
chance however, cannot be ruled out since forty -seven different pesticides were
included in the analysis without accounting for multiple comparisons . Pesticide-
specific results were also inconsistent between the wives’ and husbands’ reported
use, as well as between the two locations of the study. T he prevalence of exposure
for the majority of individual pesticides was sufficient among use reported by the
applicators themselves, but perhaps too low (<5%) to detect a difference in risk for
use reported by the wives themselves.
The method of assessing pesticide exposure among wives of applicators via
their husbands’ occupational use may not accurately reflect actual exposure by the
women. Among the wives in the cohort, 55.2% of breast cancer cases and 48.4% of
7

non-cases lived within 100 yards from areas of pesticide application. There was a
70% increase in risk for those living within 100 yards compared to over 300 yards,
although this finding was not statistically si gnificant (RR = 1.7, 95% CI: 1.0, 2.9).
Exposure to pesticides via the ambient air in close proximity to commercial
pesticide application either from pesticide drift or contamination of the soil, water
or food may represent a significant source of exposure for these wives of
applicators, and warrants further investigation.
Only two studies have attempted to examine breast cancer risk in female
agricultural workers (Table 1.2). A recent case -control study in Canada found a
slight non-significant increased risk of postmenopausal breast cancer for
agricultural workers compared workers in other occupations (OR = 1.25, 95% CI:
0.63, 2.47 for agriculture/plants) (44). The prevalence of exposure for the study
was only 3.7% for cases and 2.0% for controls, which may be too low to detect a n
association even if one existed. A nested case- control study of female Hispanic
agricultural workers in California linked employment records indicating dates of
work on a specific crop type to corresponding state -mandated reported
applications of pesticides to that crop and summed the pounds of active
ingredients applied during the time of employment as a proxy measure of
pesticide exposure (45). Mills et al. showed a modest, though not statistically
significant, increased in risk of breast cancer among the highest quartile of
pesticide use compared to the lowest for all chemicals combined (OR = 1.41, 95%
CI: 0.66, 3.02) . Six specific crop types were also analyzed separately, but only
mushrooms were associated with an increased risk of breast cancer (OR = 6.02,
95% CI: 2.01, 18.0) , however the finding was based on low numbers of exposed
8

individuals (5.5% cases and 0.9% controls). The prevalence of exposure was 25.0%
or more for grapes, vegetable crops, and all crop types combined, but may not be
sufficient for the other crop categories where fewer than 10.0% were exposed.
A nested case -control study among breast cancer screening participants in
Greece found an increased number of malignant changes detected in
mammograms among women who reported working in greenhouses for at least 10
years compared to women who never worked in agriculture (18 exposed vs. 12
unexposed women, χ
2
= 1.45, P = 0.228), however the finding was not statistically
significant. There were statistically-significant increases in risk for six of eight
other kinds of abnormal breast lesions, including ductal hyperplasia (OR = 1.87,
95% CI: 1.1 0, 3.13) fibrocystic and ductal hy perplasia (OR = 1.85, 95% CI: 1.3 0,
2.60), fibroadenoma OR = 4.85, 95% CI: 1.4 0, 16.70), inflammatory mastitis (OR
= 2.21, 95% CI: 1.20 , 4.00), and gross cystic disease (OR = 1.44, 95% CI: 1.10 ,
2.00), suggesting that the overall risk of future breast cancer may be higher
among women working i n greenhouses where pesticides were sprayed for several
hours daily (46).
Most studies in female farm workers have analyzed proportionate mortality
among different categories of occupations , yielding mixed results. Some
proportionate mortality studies have suggested an increased risk of breast cancer
among women who work in farming compared to other occupations (47,48), while
others show ed no association (49-51). One study did not find a higher than
expected rate of breast cancer mortality compared to other causes of death among
farmers across several states (50). These proportionate mortality studies were
designed as ecological studies that used very broad exposure categories such as
9

“farming” occupation and did not collect detailed information on individual -level
risk factor s so they were unable to control for confounding by other established
breast cancer risk factors . Results from these kinds of occupational studies we re
also subject to biases resulting from healthy worker populations , especially if the
comparison group consisted of people in the general source population .
Furthermore, associations observed for breast cancer mortality may not be a
reflection of the factors that affect breast cancer etiology.
The International Agen cy for Research on Cancer groups occupational
exposures from applying or spraying pesticides as “probable human carcinogens”
(category 2A) (32), however, very few women work directly in the application of
pesticides. A study of proportionate mortality in a Florida cohort of pesticide
applicators found no increased risk of death from breast cancer among women
(SMR = 0.76, 95% CI: 0.20, 1.94) (52). Female farmworkers have twice the risk of
acute pesticide-related illness than their male counterparts and the most common
cause is exposure to pesticide drift (53). A case-control study in North Carolina
found that women who worked in agricultural fields during or soon after pesticide
application had an 80% increase in risk of breast cancer (OR = 1.8, 95% CI: 1.1,
2.8) (Table 1.2) (54). In the same study, w omen who reported potential exposure
to pesticides from using pesticide application equipment or from mixing, loading
or spills had a modest though not statistically significant increased risk of bre ast
cancer (OR = 1.2, 95% CI: 0.6, 2.2). Although only 7% of female farmworkers in
the North Carolina study had applied pesticides themselves, women who did not
wear protective clothing while applying pesticides were twice as likely to develop
breast cancer as women who reported that they did (OR = 2.0, 95% CI: 1.0, 4.3).
10

Studies involving women who work directly with pesticides may be unable to
detect an association because of the relatively few women working in these
occupations and for typically shorter durations of exposure than their male
counterparts, and because the women in these occupations tend to be younger and
healthier than the general population . Both of these occupatio nal studies were
underpowered to detect associations in women applicators for any pesticide, let
alone specific pesticides of interest or groups of pesticides such as
organochlorines.
1.3.1 Previous Studies Using Biospecimens
The vast majority of studies to date assessing breast cancer risk from
exposure to pesticides have examined levels of metabolites found in the body by
measuring biomarkers such blood or adipose tissue samples. The primary
chemical of interest in these studies has been the persistent organoch lorine
pesticide dichlorodiphenyltrichloroethane ( DDT) and its main metabolite
diphenyldichloroethylene (DDE). In 1993, Wolff et al. reported the findings from a
nested case -control study of over 14,000 women participating in the Women’s
Health Study cohort between 1985 and 1991 and found a statistically-significant
fourfold increase in breast cancer risk for those in the 90
th
percentile of serum
DDE concentrations (19.1 ng/mL) compared to those in the 10
th
percentile (2.0
ng/mL) (P = 0.031) (55). The many studies that tried to replicate these findings by
analyzing levels off DDE in biospecimens , however, have yielded mixed results .
Some failed to confirm the findings or found modest associations that were not
statistically significant (23,56-73).
11

There are several significant challenges to assessing the amount of pesticide
exposure accumulated over a person’s lifetime by methods of measuring pesticide
biomarkers. First, the collection of biospecimens for studies of cancer risk is
usually done close to the time of cancer diagnosis and may not be an accurate
estimate of prior exposure given the long latency period of a typical cancer
diagnosis (74). Second, the determination of cumulative exposure must take into
consideration that the levels of organochlorine compounds in the body are highly
correlated with the timing and duration of exposure as well as individual
metabolism and body mass index that may also change over a person’s life time
(57,75,76). Finally, the levels of organochlorine metabolites in the body will
decrease over time but the rate of elimination is substantially affected by changes
in body fat and lactation, two factors that also influence breast cancer risk (77).
The most recent study of serum organochlorine pesticide measures used a
cross-sectional approach and found positive associations for prostate cancer but
not for breast cancer (78). A review by Calle et al. concluded that there was not
enough evidence to support an association between organochlorine pesticides and
breast cancer based on the biospecimen studies to date (79), while another review
suggested more research was needed (19). A more recent review included
published articles from across the globe and found that studies conducted in
developing countries collectively supported an association between serum DDT
and breast cancer risk; whereas those in developed countries did not, and that this
may be due to the detection of generally higher exposure levels in studies from
developing countries than those reported in studies from developed countries
(80). Another study found that women whose serum DDT levels were measured at
12

two different time periods had a three -fold increased risk of breast cancer for the
highest quartile compared to the lowest and a statistically significant increasing
trend (OR = 3.6, 95% CI: 1.1, 12.2, P = 0.02) (81). These findings suggest that
studies that consider multiple measurements or the cumulative impact over a
person’s lifetime may lead to a more accurate estimate of ri sk than those that rely
on a single exposure point in time.
There is also evidence that early lifetime exposures to pesticides may
increase risk of breast cancer. A prospective study of women found that women
under 14 years of age when first exposed to DDT had a dose response increase in
breast cancer risk compared to the lowest s erum level (OR = 2.8, 95% CI: 1.1, 6.8
and OR = 5.4, 95% CI: 1.7, 17.1) (82). Reviews of the scientific literature suggest
that exposures prior to the first pregnancy when breast tissue is not ful ly
differentiated may be a critical period for the breast and a time of particular
vulnerability to endogenous and exogenous hormones and hormone -like
chemicals (83,84). These early in life time periods of exposure are often difficult to
assess in epidemiologic stud ies because of the cost and feasibility of a large -scale
follow-up study beginning before these possible critical windows of exposure and
lasting until after a possible breast cancer diagnosis decades later.
1.3.2 Proxy Methods to Estimate Pesticide Exposure
Ecologic studies that are usually exploratory in nature have sought to
examine the risk of breast cancer from exposure to agricultural areas where
pesticides are utilized. Two studies that used regression analyses to look at breast
cancer incidence in areas of high pesticide application found no evidence of
elevated breast cancer at the census block level or ward level (85,86), while a third
13

study using county -level pesticide use information in California observed an
increased risk of breast cancer for the highest quartile of total pounds of exposure
to methoxychlor (RR = 1.18, 95% CI: 1.03, 1.35) and t oxaphene (RR = 1.16, 95%
CI: 1.01, 1.34) (Table 1.3) (87). These kinds of regression analyses using
aggregated data have limited interpretations regarding risk to individuals since
the data is subject to bias resulting from heterogeneity within the group-level data
and lacks information on individual -level exposure to pesticides as well as
potential confounders.
Some of the mos t recent studies have utilized advancements in
Geographical Information Systems (GIS) software to examine proximity to
pesticide application as a proxy measure of exposure. These methods indirectly
estimate exposure to pesticide application and do not rely on subject recall of
prior exposures or the collection of costly, invasive biospecimens. A GIS- based
approach in a nested case -control study in Long Island, New York found an
increased risk of breast cancer for women living within one mile of a hazardous
waste site that contained organochlorine pesticides (OR = 2.9, 95% CI: 1.1, 7.2)
but no association for living on or near previous agricul tural land (OR = 1.5, 95%
CI: 0.8, 2.9) (Table 1.4) (88). GIS software was used to measure the distance from
each subject’s residential location in 1980 to hazardous waste site locations
obtained from county records. Aerial photography taken in 1947 and 1950 was
used to determine prior farmland locat ions, however this method was unable to
distinguish different crop s and therefore wa s unable to identify locations where
specific pesticides of interest were used historically. Another study conducted on
Prince Edward Island, Canada also used a GIS -approach to link residential postal
14

code to the density of fungicide applied in the nearest census area in a case-
control study of mammogram screening participants , but found no increased risk
of breast cancer (OR = 0.70, 95% CI: 0.46, 1.17) (89). Exposure was based on
whether greater than 10% of the land within the census area was treated with
fungicides procured from agricultural census data conducted every ten years.
Neither of these studies had very large prevalence of exposure or sample
sizes, and thus may be underpowered to detect an y effect. In the New York Study,
only 11.4% of cases and 5.3% of controls lived within one mile of a hazardous
waste site and 19.0% of cases and 15.7% of controls lived on or near previous
agricultural land. In the Prince Edward Island study, only 14.1% of cases and
18.4% of controls lived in a census area where >10% of the land was treated with
fungicides. In addition, both studies assumed that the participants lived at one
location their entire lives, which may underestimate the true exposure and bias
results towards the null.
A study in Cape Cod, Massachusetts collected complete residential histories
from cases and controls and assessed their proximity to historical locations of
pesticide spraying identified through government reports and aerial photography
in 1951, 1971, 1984 and 1990 to estimate lifetime exposure dating from 1948 when
DDT was first used there (90). Associations were observed for the highest
exposure group compared to the lowest for proximity to aerial pesticide
applications at cranberry bogs (OR = 1.8, 95% CI: 0.7, 4.5) and for tree pests (OR
= 1.2, 95% CI: 0.7, 1.8), but their findings were not statistically signific ant and no
associations were observed for ground applications on wetlands (OR = 0.4, 95%
CI: 0.1, 1.15) or for other agriculture (OR = 0.8, 95% CI: 0.3, 2.3) (91). The
15

prevalence of exposure ranged from fairly low for ground applications on wetlands
(5.3% of cases and 6.0% of controls) to more moderate for aerial applications for
tree pests (16.3% of cases and 13.8% of controls).
A GIS- based approach was used to assess breast cancer risk in the
California Teachers’ Study cohort based on the density of reported pesticides
application within a half -mile of participants’ residential addresses (92). The
highest category of pesticide use density compared to the lowest resulted in no
increase in the risk of breast cancer f or a group of four pesticides considered
mammary carcinogens (HR = 1.15, 95% CI: 0.90, 1.48), a group of 34 endocrine
disrupting pesticides (HR = 1.03, 95% CI: 0.86, 1.25), or a group of three
organochlorine pesticides (HR = 0.99, 95% CI: 0.63, 1.55). Only 6.0% of the
cohort was exposed to the group of organochlorine pesticides defined by living in
areas of pesticide use density ≥1 pound/mile
2
, but 17.0% were exposed to the
group of mammary carcinogenic pesticides and 36.0% were exposed to the group
of endocrine disrupting pesticides. Unlike the two previous studies using location
to historical pesticide use as a proxy for exposure, the study conducted in
California was able to examine specific pesticide chemicals of interest and group
them based on etiologic relevance, assessing a limited number of meaningful
comparisons. The hazard risk estimates however, were based on residential
address only at the time of enrollment in the cohort, thus limiting the potential to
examine risk over a lifetime.
Only one previous study conducted in Cape Cod, Massachusetts used a GIS-
based approach to examine proximity of residential histories collected from cases
and controls to locations of pesticide spraying identified through government
16

reports and aerial photography beginning in 1948 when DDT was first used in the
area (91). Associations were observed for the highest exposure group compared to
the lowest for proximity to aerial pesticide spraying at cranberry bogs (OR = 1.8,
95% CI: 0.7, 4.5) and to aerial spraying for tree pests (OR = 1.2, 95% CI: 0.7, 1.8),
but the 95% confidence intervals include the null and therefore the role of chance
cannot be ruled out in these findings. No associations were observed for exposure
to ground applications of other agriculture (OR = 0.8, 95% CI: 0.3, 2.3) or for
ground applications on wetlands (OR = 0.4, 95% CI: 0.1, 1.15). Exposure
categories in the study were grouped by land -use type, which may combine
chemicals of different toxicological effects, and the prevalence of exposure was
rather low, ranging from 5.3% to 16.3% depending on the exposure category.

1.4 Future Direction in Studies Assessing Pesticide Exposure in
Breast Cancer Risk
There is a strong biologically plausible link between endocrine- disrupting
chemicals such as certain pesticides and breast cancer , yet the epidemiologic
studies to date are unable to confirm this. It is possible however, that the lack of
association observed in st udies of breast cancer risk from pesticide exposure
arises because the method of estimating exposure to pesticides over a person’s
lifetime has been highly inaccurate, thus reporting null effects when there could
actually be a significant association. Many of the studies to date have been
conducted in areas such as Long Island, where the level of exposure to specific
pesticides of interest may be too low to detect an effect. In addition, most studies
were not designed to examine the impacts of pesticide exposures that may
17

accumulate over a lifetime or to look at critical time periods that may be most
pertinent for breast cancer. Without sufficiently high exposures or consideration
of cumulative doses spa nning decades prior to a cancer diagnosis, there is
potential for non- differential misclassification of pesticide exposure to bias the
estimates of relative risk towards the null.
Future studies involving the measurement of biomarkers and future breast
cancer risk are highly costly and would require long -term tracking of several
pesticide residues in addition to monitoring individual variation in metabolism
and changes to potential confounding factors such as body fat over time. Even if
pesticide exposure levels in the population of study were high enough, it would be
cost prohibitive to conduct the large -scale, comprehensive biological monitoring
study needed to determine lifetime exposure levels. Thus, new methods are
needed to adequately address whether chronic exposure to pesticides over a
person’s lifetime leads to an increased risk of breast cancer in postmenopausal
women. These methods should be feasible, economical and capable of detecting
historical exposure to specific pesticides with known estrogenic effects.
18

Table 1.1. Studies of Breast Cancer Risk and Pesticide Exposure Based on Self-Reported Exposure to Pesticides.
Reference Location
State Ranking
Pesticide Use
in Agriculture
(93)
Number
of
Subjects
Method for
Determining
Pesticide Exposure
Prevalence of Exposure and
Estimates of Relative Risk
(95% CI)
Concerns Regarding the Study
Design
El-Zaemey,
2014
Australia --- 1,205 cases
and 1,789
controls
Self-reported
pesticide use in
home, lawn and pets,
plus job matrix
identified exposure
from occupations in
mailed questionnaire.
11.5% cases and 11.8% controls used
pesticides at home OR = 1.10, 95%
CI: 0.86, 1.37; and 1.7% cases and
2.2% controls worked with
pesticides, OR = 0.77, 95% CI: 0.45,
1.32.
Specific pesticides of interest were
not examined and general pest
categories were grouped for
analyses.
El-Zaemey,
2013
Australia --- 1,743 cases
and 1,169
controls
Self-reported notice
of pesticide spray
drift near residences
was obtained from
mailed questionnaire.
15.5% cases and 11.6% controls
noticed spray drift, OR = 1.43, 95%
CI: 1.15, 1.78 but stratified on belief
that pesticides cause breast cancer
OR = 1.47, 95% CI: 1.15, 1.87 for
believers and OR = 0.94, 95% CI:
0.51, 1.74 for non-believers.
The association between self-
reported noticing of pesticide drift
appears confounded by belief in
whether or not pesticides cause
breast cancer.
Farooq,
2010
New
York
26 447 cases
and 758
controls
Women were
interviewed before
surgery for breast
cancer (cases) or
benign breast disease
or other non-breast
related reasons
(controls).
44.7% cases and 45.3% controls used
pesticides for ants/cockroaches
indoors, OR = 1.13, 95% CI: 0.75,
1.72; 21.7% cases and 18.2% controls
used outdoor insecticides, OR =
1.30, 95% CI: 0.79, 2.14; and 2.5%
cases and 0.9% controls applied
outdoor insecticides themselves, OR
= 2.68, 95% CI: 0.72, 9.92.
Controls were recruited from
hospital patients and may not
represent exposures found in the
general population. Specific
pesticides of interest were not
examined and years of pesticide
use were not obtained for
calculating lifetime exposure.
19

Table 1.1. Continued:
Reference Location
State Ranking
Pesticide Use
in Agriculture
(93)
Number
of
Subjects
Method for
Determining
Pesticide Exposure
Prevalence of Exposure and
Estimates of Relative Risk
(95% CI)
Concerns Regarding the Study
Design
Teitelbaum,
2007
Long
Island,
New
York
26 1,505 cases
and 1,553
controls
Self-reported
pesticide use in
home, lawn and pets
was categorized into
exposure quintiles.
84.7% cases and 74.2% controls ever
used any residential pesticide where
lifetime applications were
categorized into quintiles, OR =
1.39, 95% CI: 1.15, 1.68 for quintiles
2–5 vs. referent quintile 1; 93.3%
cases and 92.5% controls ever used
insecticides, OR = 1.07, 95% CI:
0.80, 1.42; 83.3% cases and 79.3%
controls ever used lawn/garden
pesticides, OR = 1.34, 95% CI: 1.11,
1.63; and 3.4% cases and 2.8%
controls ever applied lawn
insecticides themselves, OR = 1.56,
95% CI: 1.01, 2.43.
Specific pesticides of interest were
not examined and general pest
categories were grouped for
analyses.
Engel, 2005 Iowa and
North
Carolina
2 and 8 309 cases
and 30,145
non-cases
Lifetime pesticide use
obtained via self-
administered survey
sent to pesticide
applicators and their
wives.
50.8% cases and 55.9% non-cases
ever used any pesticide, RR = 0.8,
95% CI: 0.7, 1.1; 10.7% cases and
11.1% controls washed their
husbands work clothing >20
days/year, RR = 1.4, 95% CI: 0.8,
2.7; and 55.2% cases and 48.4%
non-cases lived <100 yards from
pesticide application, RR = 1.7, 95%
CI: 1.0, 2.9.
Inaccurate recall of personal
pesticide use may lead to non-
differential exposure
misclassification biasing the
estimates towards the null.
Prevalence of exposure to specific
pesticides too low to detect
difference in risk.

20

Table 1.2. Studies of Breast Cancer Risk and Pesticide Exposure in Agricultural Workers.
Reference Location
State Ranking
Pesticide Use
in Agriculture
(93)
Number
of
Subjects
Method for
Determining
Pesticide Exposure
Prevalence of Exposure and
Estimates of Relative Risk
(95% CI)
Concerns Regarding the Study
Design
Brophy,
2012
Canada --- 1,005
cases and
1,146
controls
Exposure was based
on occupational
farming category
obtained from
interview.
3.7% cases and 2.0% controls
worked in agriculture/plants sector,
OR = 1.25, 95% CI: 0.63, 2.47.
The risk estimates are based on
broad exposure and occupational
categories, not specific pesticides
plausibly linked to breast cancer.
Prevalence of exposure to specific
pesticides too low to detect
difference in risk.
Mills, 2005 California 26 128 cases
and 640
controls
Union employment
records in Hispanic
agricultural workers
were linked with state
pesticide use reports
(1974–1999). Cases
were identified
through the cancer
registry between
1988–2001.
25.8% cases and 25.0% controls
exposed to all chemicals combined,
OR = 1.41, 95% CI: 0.66, 3.02 for
highest quartile compared to lowest;
48.4% cases and 43.4% controls
exposed to grapes, OR = 1.22, 95%
CI: 0.84, 1.79; 29.7% cases and
35.2% controls exposed to
vegetables, OR = 0.78, 95% CI: 0.52,
1.18; 7.0% cases and 7.2% controls
exposed to strawberries, OR = 0.98,
95% CI: 0.45, 2.05; 5.5% cases and
8.9% controls exposed to citrus, OR
= 0.59, 95% CI: 0.26, 1.33; 5.5%
cases and 0.9% controls exposed to
mushrooms, OR = 6.02, 95% CI:
2.01, 18.0; and 4.7% cases and 3.6%
controls exposed to horticulture, OR
= 1.09, 95% CI: 0.41, 2.93.
Analyses were based on union
records of employment history,
adjusted for age. Individual-level
information on other factors was
not available.
Dolapsakis,
2001
Island of
Crete,
Greece
--- 1,053
women
in the
cohort
Women participants
of a breast cancer
screening program in
1988–1992 were
asked about farm
work and had taken
mammograms every
1–2 years until 1998.
Compared to never working in
agriculture, 49.0% of the cohort has
worked in greenhouses for at least
10 years. Malignant changes were
found in 18 exposed and 12
unexposed women’s mammograms,
χ
2
= 1.45, P = 0.228.
These results are preliminary
findings. Analyses were stratified
by age, but no other confounders
were considered. Working in
greenhouses was used as a proxy
for exposure to pesticides, but
exposures to specific chemicals of
interest were not evaluated.

21

Table 1.2. Continued:
Reference Location
State Ranking
Pesticide Use
in Agriculture
(93)
Number
of
Subjects
Method for
Determining
Pesticide Exposure
Prevalence of Exposure and
Estimates of Relative Risk
(95% CI)
Concerns Regarding the Study
Design
Duell, 2000 North
Carolina
8 259 cases
and 291
controls
Women who reported
farming in the
Carolina Breast
Cancer Study were
interviewed and
exposure was based
on recall of historical
farm-specific
exposures.
13.9% cases and 12.0% controls
applied pesticides to crops, OR = 1.3,
95% CI: 0.8, 2.3; 39.4% cases and
32.6% controls worked in fields soon
after pesticide application, OR = 1.8,
95% CI: 1.1, 2.8; 8.5% cases and
5.5% controls did not wear
protective clothing to apply
pesticides, OR = 2.0, 95% CI: 1.0,
4.3; and 16.2% cases and 16.2%
controls were exposed to pesticides
from spills, mixing, or using
equipment, OR = 1.2, 95% CI: 0.6,
2.2.
Few women work directly with
pesticides so not enough power to
study these occupational
exposures in detail. Studies in the
general population means that
results will be applicable to the
largest proportion of breast cancer
cases.

22

Table 1.3. Ecologic or Regression Studies of Breast Cancer Risk and Pesticide Exposure.
Reference Location
State Ranking
Pesticide Use
in Agriculture
(93)
Number
of
Subjects
Method for
Determining
Pesticide Exposure
Prevalence of Exposure and
Estimates of Relative Risk
(95% CI)
Concerns Regarding the Study
Design
Mills, 2006 California 1 23,513
cases
County-level
pesticide use reports
from 1970–1988 were
compared to county
rates of breast cancer
in Latina women,
based on residence at
diagnosis between
1988–1999.
Prevalence of pesticide use was not
available at the county level.
Comparing the highest quartile of
pesticide use to the lowest, RR =
1.18, 95% CI: 1.03, 1.35 for
methoxychlor and RR = 1.16, 95%
CI: 1.01, 1.34 for toxaphene.
Regression analysis was used to
examine aggregated county-level
pesticide data among women with
breast cancer.
Reynolds,
2005
California 1 176,302
cases
State pesticide use
reports in 1990–1997
were linked to census
block group at the
time of diagnosis
between 1988–1997
among women with
breast cancer.
Exposure was based
on density of census
block groups having
≥1 lb/mi
2
application
to groups of
pesticides.
8.0% of census block groups
exposed to group of 3
organochlorine pesticides, RR =
0.98, 95% CI: 0.92, 1.04 for the
highest exposed group; 17.0% of
census block groups exposed to
group of 4 mammary carcinogen
pesticides, RR = 1.01, 95% CI: 0.97,
1.04 for the highest exposed group;
and 37.0% of census block groups
exposed to group of 34
xenoestrogens, RR = 0.99, 95% CI:
0.82, 1.20 for the highest exposed
group.
This study was ecologic in nature
and did not estimate individual-
level risks. Aggregated pesticide
use data at the census block group
level was compared among
women with breast cancer.
Muir, 2004 England --- County-
level
breast
cancer
rates for
152 rural
and 223
urban
ward
areas
Rates of breast cancer
at zip code level from
cancer registry
(1989–1991) was
spatially linked to
ward-level pesticide
use in 1991. Wards
were categorized into
quartiles of pesticide
use for 4 different
pesticides (kg/km
2
).
Prevalence of pesticide use was not
available at the ward level. There
were significantly positive slope
values for 3 of 4 pesticides analyzed
in one county but not for another
county.
Regression analysis was used to
spatially analyze aggregated ward-
level pesticide use data and
county-level breast cancer rates.
23

Table 1.4. Studies of Breast Cancer Risk and Pesticide Exposure Based on GIS-Based Proxy Measures of Exposure.
Reference Location
State Ranking
Pesticide Use
in Agriculture
(93)
Number
of
Subjects
Method for
Determining
Pesticide Exposure
Prevalence of Exposure and
Estimates of Relative Risk
(95% CI)
Concerns Regarding the Study
Design
Ashley-
Martin,
2012
Prince
Edward
Island,
Canada
--- 207 cases
and 617
controls
Residential postal
code in 1999–2002
was linked to nearest
census area and
considered exposed if
>10% of the land in
the census area was
treated with
fungicides according
to 1991 agricultural
records.
14.1% cases and 18.4% controls live
in a census area that has >10% of
land treated with fungicides, OR =
0.72, 95% CI: 0.46, 1.12.
The study was not able to look at
specific types of fungicides used
and did not collect information on
other sources of pesticide
exposure. Exposure was based on
postal code at the time of
enrollment and aggregated data
on application of fungicides.
O’Leary,
2004
Long
Island,
New York
26 105 cases
and 210
controls
Proximity of 1980
residential location to
hazardous waste sites
prior to diagnosis
(1980–1992). Aerial
photos in 1947 and
1950 were used to
determine prior
agricultural land use.
11.4% cases and 5.3% controls live
within one mile of hazardous waste
site containing pesticides, OR = 2.9,
95% CI: 1.1, 7.2; and 19.0% cases
and 15.7% controls living on or near
previous agricultural land, OR = 1.5,
95% CI: 0.8, 2.9.
The study was not conducted in a
highly agricultural region so
prevalence of exposure may be too
low to detect any difference in
risk. Historical records of actual
pesticide use and crop locations
were not available.
Reynolds,
2004
California 1 1,552
cases
among
cohort of
114,835
women
Amount of pesticide
applied within ½-
mile buffer around
diagnosis address.
Address at time of
diagnosis (1996–
1999) was linked to
state pesticide use
data from 1993–1995.
17.0% of cohort exposed to group of
4 mammary carcinogen pesticides,
HR = 1.15, 95% CI: 0.9, 1.5 for
highest quartile of use; 6.0% of
cohort exposed to group of 3
organochlorine pesticides, HR =
0.99, 95% CI: 0.63, 1.55 for highest
quartile of use; and 36.0% of cohort
exposed to group of 34 endocrine
disrupting pesticides, HR = 1.03,
95% CI: 0.9, 1.3 for highest quartile
of use.
Exposure for this study was based
on only the address around the
time of diagnosis.
24

Table 1.4. Continued:
Reference Location
State Ranking
Pesticide Use
in Agriculture
(93)
Number
of
Subjects
Method for
Determining
Pesticide Exposure
Prevalence of Exposure and
Estimates of Relative Risk
(95% CI)
Concerns Regarding the Study
Design
Brody,
2004
Cape
Cod,
Massach
usetts
45 1,165
cases and
1,006
controls
Exposure was based
on residential history
obtained through
interview and
proximity to known
historical pesticide
application locations
gathered from
government agencies
and aerial
photographs taken in
1951, 1971, 1984 and
1990.
9.6% cases and 9.6% controls
exposed to aerial applications at
cranberry bogs, OR = 1.8, 95% CI:
0.7, 4.5 for highest exposed group;
16.3% cases and 13.8% controls
exposed to aerial applications for
tree pests, OR = 1.2, 95% CI: 0.7, 1.8
for highest exposed group; 5.3%
cases and 6.0% controls exposed to
ground applications on wetlands,
OR = 0.4, 95% CI: 0.1, 1.5 for
highest exposed group; and 9.4%
cases and 9.6% controls exposed to
ground applications for other
agriculture, OR = 0.8, 95% CI: 0.3,
2.3 for highest exposed group.
The study was not conducted in a
highly agricultural region so
prevalence of exposure may be too
low to detect any difference in
risk.

25

1.5 New and Improved Methods for Pesticide Exposure Assessment
The focus of this dissertation was to show that a highly accurate method
for estimating historical exposures to pesticides in an area of intense agriculture
and commercial pesticide use reduced the non -differential misclassification to
reveal a true risk for postmenopausal breast cancer. The new approach utilized
GIS technology to evaluate ambient exposure to historical pesticide application.
An individual’s proximity to agricultural cropland that was targeted by reported
pesticide application w as used as a proxy measure for exposure to specific
chemicals of biologic relevance for breast cancer.
Most of the previous studies of pesticide exposure and breast cancer have
been conducted in areas where the level of exposure to specific pesticides of
interest may be too low in the population to detect a difference in risk, even if one
existed. The method described here assess ed pesticide exposure in California’s
Central Valley, the largest agricultural region in the nation with the highest
potential for pesticide exposure (Figure 1.1). In 2010, the use of pesticides for
agricultural purposes in California exceeded 170 million pounds of active
ingredients, about a quarter of all pesticides used in the U.S. (94). The Central
Valley counties of Fresno, Tulare , and Kern are the top three ranked agricultural
counties in California for total pounds of pesticide active ingredients report ed.
There has been more opportunity for residents of these counties than elsewhere
in the U.S. to be exposed to ambient pesticides from neighboring agricultural
application through aerial drift, field runoff, or through contamination of local
drinking water sources (95-98).
26

Figure 1.1. Map of California Highlighting the Central Valley and the
Agriculturally Productive Counties of Fresno, Tulare , and Kern that are the
Regional Focus of the Studies in This Dissertation.

27

1.5.1 Pesticide Use Reporting
Pesticides are chemicals or mixtures of chemicals used to control, kill,
repel or attract pests, or to regulate plant growth, and include insecticides,
herbicides, rodenticides, molluscicides, fungicides, repellents, disinfectants and
sanitizers. Those who work with or around pesticides may be at risk of exposure
to pesticides, including pesticide handlers, field workers, office workers or others
nearby exposed to pesticide application drift or residue. People who live or work
near agricultural fields may b e exposed to pesticides application through the air,
water, or soil or via residues in food or drinking water. People may also be
exposed when pesticides are used in or around their home, on their pet or in their
neighborhood. There is evidence that people who live close to commercial
agricultural areas have higher concentrations of pesticides in household dust and
soil (99).
The U.S. Environmental Protection Agency requires pesticides to be
registered and develops the standards for enforcement carried out by individual
states. California has some of the strictest pesticide regulations in the world. The
California Department of Pesticide Regulation works with local county
agricultural commissioners to evaluate the safety of pesticides for registration, to
train professional pesticide handlers, to investigate pesticide -related illnesses,
and to monitor air, water and produce for residues (100). Beginning in 1970,
California Department of Pesticide Regulation required pest control operators to
report all pesticides applications and farmers to report all “restricted -use”
pesticide applications. “Restricted -use” pesticides require applicators of such
28

pesticides to be certified. In 1990, California expanded the reporting
requirements to include full reporting of all pesticides used in agriculture
applications (excluding livestock). Over 2.5 million records of Pesticide Use
Reporting (PUR) are logged each year. Data on pesticide use reporting since 1974
has been made available to the public and can be accessed via the Department of
Pesticide Regulation’s web -based California Pesticide Information Portal
(http://calpip.cdpr.ca.gov/). PUR data includes information on the active and
inactive ingredients, the location, amount, frequency and timing of use, the
location and crop information that the pesticide will be used on, and the method
of application (air, ground, etc.). The reported PUR data is linked to the Public
Land S urvey System (PLSS), a grid that parcels the United States into
approximately one square mile sections.
1.5.2 Land Use Surveys
To more precisely pinpoint the location of pesticide application on specific
crops at a finer resolution than the squar e-mile PLSS sec tions, PUR data wa s
combined with land-use survey information. Land-use data contains information
expressed as polygon shapes indicating various crop types such as fields,
vineyards and orchards, plus the crop acreage contained within the polygons.
California’s Department of Water Resources, Division of Planning and Local
Assistance maintains an extensive, statewide set of digital land -use and crop
cover surveys by county from 1976 to 2012 or more current (101). Each of the
counties is surveyed at approximately 6 to 10-year intervals by field observations
taken at one time during the growing season and the resulting maps made
29

available to the public. The maps for the years 1999 –2012 were digitized and
accessible via the California Department of Water Resources web site
(http://www.water.ca.gov/landwateruse/lusrvymain.cfm). Historical electronic
maps of land use and crop type were constructed from the most recen t digitally
available land use surveys for all California counties (1986 –2012) and manually
digitized for the counties of Fresno, Tulare, and Kern using the earliest available
paper maps (1977–1985) (102,103).
Since specific pesticides are applied to certain types of crops, Rull and Ritz
developed methods to match land -use data and PUR data by crop types to
determine where the pesticides are likely applied within the PLSS section (104).
Each land-use polygon was thus linked with a specific pesticide and its density,
which is the total pou nds of active ingredients applied to the corresponding
polygon during the year reported by the PUR data. Nuckols et al. found that
California PUR data combined with land- use surveys showed 88 –98% correct
identification of a specific pesticide for a given c rop location (105) and reduced
non-differential misclassification of exposure compared to GIS -based approaches
based only on historical pesticide records (104,106).
1.5.3 Satellite Imagery in Pesticide Use Data
One of the limitations of relying on land-use surveys to specify the location
of crops is that the 6 to 10-year intervals in between surveys does not allow for
the identification of land use where there is yearly crop rotation or where
multiple crops are grown on the same land. The utilization of satellite image data
has been proposed as a way to define land- use bounda ries and characterize
30

agricultural land (107). Various land cover types differ in the way that they reflect
light as captured in satellite images, and these “spectral signatures” can be used
to differentiate bare soil, water, urban areas and vegetation areas in the images.
Landsat image data is now freely available via the US Geological Survey website
(http://landsat.usgs.gov). Lands at imag es are collected every 8 to 16 days,
providing information on how various agricultural crops change throughout the
year. Higher resolution images that show smaller features such as buildings and
trees are taken only once a year and can be combined with Landsat images to
refine the images indicating different types of land cover.
Maxwell et al. developed methods to classify the spectral signatures for
several types of field crops in California by comparing Landsat images taken over
the growing season (108). A feasibility study was conducted using Landsat
images in combination with pesticide application data from California PUR to
identify the various crop locations within the square- mile PLSS sections where
the pesticide paraquat was likely app lied (109). In order to utilize these methods
in epidemiologic studies, a spectral signature library of all the different crop types
for every year in the growing region would need to be developed, as well as
methods to differentiate crops that have similar spectral signatures. Further
research is also needed to test and validate these methods for other pesticides of
interest. For an epidemiologic study that intends to assess exposure to
commercial pesticide use via proximity to hundreds or thousands of geographic
locations spanning several years or decades, the process would be a substantially
large undertaking that may become feasible as the technology develops and
31

becomes more automated (110). Maxwell et al. noted that for certain crops and
regional areas, the resolution of Landsat images requires incorporating aerial
photography to identify the actual boundaries, which would add yet another layer
to resolve and automate before these methods can be readily implemented in an
epidemiologic study.
1.5.4 Individual-Level Lifetime Pesticide Exposure
Proximity to ambient pesticide application in this dissertation was used as
a proxy measure for an individual’s actual exposure to specific pesticides at the
time they lived or worked at a particular location. Each location in a person’s
residential and occupational history was geocoded to yield latitude and longitude
spatial coordinates that refer to a point, line segment or a polygon shape on a
map using ArcView software (ESRI, Redlands, CA). Geocoding was the process
we used to identify a location’s geog raphy, however there was still some variation
even with our best estimate of where the location wa s. We first used a process of
geocoding in batches to identify the spatial coordinates (i.e. latitude and
longitude) that were readily identified with the add ress information provided.
Manual geocode correction methods developed by Goldberg et al. were
subsequently used to pinpoint a more precise geospatial location by using
additional information from web -based interactive maps and satellite imagery
(111).
A 500- meter buffer was then created around each pair of spatial
coordinates in order to identify which pesticides were applied within the area of
the buffer . The use of location proximity to crops where pesticides are applied
32

relies on the assumption that the distance from the location to the crops is
correlated with pesticide levels found in homes which results in exposure to
individuals in those homes. This buffer distance was chosen based on studies that
found measurable concentrations of pesticides drifted from commercial pesticide
application (107,112,113) and were detected in indoor and outdoor air of
neighboring homes as well as in household dust, clothing and food (114). A study
conducted in northern and central California found that pesticide residues
measured in household c arpet dust was correlated with use of agricultural
pesticides within the previous year at 500 meter and 1,250 meter buffer distances
(115). The same study found that occupational and home pesticide use only
explained 4 –28% of the va riability in pesticide concentrations in these homes.
Another study conducted in Iowa found that herbicide concentrations in carpets
dust increased with crop acreage within 750 meters of residences (116). A buffer
distance of 500 meters therefore may be conservative based on the findings from
the two studies of carpet dust measurements of pesticides, but has been most
frequently used in previously published epidemiologic studies of residential or
workplace proximity to commercial pesticide application (102,103,117).
An individual’s ex posure to specific pesticide s was determined by
intersecting their geographical location to the land -use and PUR combined data.
For each year that an individual lived or worked at a particular location, the
amount of pesticide exposure during that year was summed for the areas of the
land-use polygons that fe ll within the buffer, and the corresponding densities of
active ingredients applied. Pesticide- specific historical averages were then
33

calculated by taking the total density in the buffer divided by the number of years
of exposure.
1.5.5 Missing Information
Methods were developed to deal with missing information in order to
reduce misclassification of exposure and potential bias. Missing values in the
pesticide exposure assessment data may result from the following:
1) Incomplete PUR data
2) Annual PUR data and 6 to 10-year periodic land -use survey data do not
match
3) Poor participant recall of prior residential and occupational histories
4) Lack of PUR or land-use information outside of California
Since PUR data was collected through self -reporting of pesticide
application, there may be missing information in the data as a result of
incomplete reporting, inaccuracies or human entry error. After excluding PUR
records with obviously incomplete or incorrect entries, approximately 18.0
million records out of a total 48.3 million PUR records entered between 1974 and
2012 have what appeared to be valid information for the type of pesticide, pounds
applied, or area, but did not include information on where the pesticide was
applied, i.e. the linked PLSS spatial location was mis sing. This means that the
pesticide applications for 37.3% of PUR data entries with incomplete PLSS
coordinates could not be combined with land -use data, and exposure to many
pesticide applications was underestimated. Consequently, study subjects were
more likely to be misclassified as unexposed than was true. Assuming that the
34

potential misclassification resulting from incomplete PUR data is not related case
or control status, the bias would be directed towards the null.
Another limitation of PUR data is that we only have information on
restricted pesticide use since 1974 and for all pesticide application since 1990. In
determining a person’s lifetime exposure to ambient pesticide application,
exposures occurring before 1974 are either ignored or assumed t o have the same
exposure during the earlier timeframe as occurred in 1974, the earliest year we
have data for. We examined these two scenarios and both produced estimates
that were nearly identical so we decided to report only those exposures for which
we have complete PUR and land -use data available. We will examine if there are
any differences in the missing information for exposures prior to 1974 between
cases and controls.
Our pesticide exposure model incorporated annual PUR data with land-
use data from the closest year available from county su rveys that were conducted
every 6 to 10 years. It is possible that PUR data and land- use crop information
may not perfectly match up for the same PLSS unit due to the lag time between
land-use surveys. A tier ed approach was created to match the most
comprehensive PUR and land -use data given various levels of geocoding
certainty. When land -use and PUR data matched exactly, the pesticide was
assumed to be applied in the land -use polygon of the corresponding crop type;
when PUR data did not match by crop type but other crops were identified in the
PLSS section, the pesticide was assumed to be applied to the location where the
35

existing crops were reported; and when no crops were identified in the PLSS
section, the pesticide was assumed to be applied across the entire PLSS section.
Another source of potential missing exposure information is inaccurate
reporting, lack of recall of residential histories, or locations that are outside of
California. We will examine those missing address es at diagnosis by other factors
to see if they are different in other ways that might be related to their diagnosis or
to their potential for exposure to pesticides. Exposure based on residential
histories may include gaps of missing data within people’s lifetimes. Participants
may recall their own address histories to varying degrees of detail or include
places that were unable to be geocoded or that were outside of California and
thus do not have corresponding information on neighboring pesticide
application. We will determine the person -years of missing information for those
with incomplete address histories and compare those with substantial gaps in
their histories by disease status.

1.6 Previous Studies of Pesticide Exposure and Risk of Other
Diseases
Several studies have been conducted using these GIS -based methods that
combine land-use data and historical pesticide records to create a detailed model
of historical pesticide application in the counties of Fresno, Tulare, and Kern
(Table 1.5) . Compared to metho ds used in other epidemiologic studies, the
strength of these GIS -based methods include: 1) the prevalence of exposure for
individual pesticides of interest was high enough to be powered to detec t an
36

association if one existed , 2) the stud ies focused on pesticides related to the
outcome based on scientific literature, thereby reducing the need to control for
large numbers of comparisons, and 3) the studies were able to examine exposures
that accumulate over decades or during historic exposure windows. While these
methods have not been validated against biomarkers of exposure , they do
correctly identify high serum DDE levels and persistent organochlorine pesticide
exposure with 87% specificity (118).
Previous studies assessing pesticide exposure on the risk of Parkinson’s
disease and one study of pesticide exposure on the risk of prostate cancer show
substantially increased risk for only those pesticides with biologic relevance.
Costello et al. found that individual’s exposed via their residential history to both
the herbicide paraquat and fungicide maneb had a 75% increased risk of
Parkinson’s disease (OR = 1.75, 95% CI: 1.13, 2.73) and that the results were
stronger for persons with younger-onset (≤60 years old) Parkinson’s disease (OR
= 5.07, 95% CIL 1.75, 14.71) (103). Wang et al. examined risk of Parkinson’s
disease and associations with paraquat, maneb and ziram from residential as well
as occupational histories, observing larger risk estimates for work place
exposures and when exposed at both residential and occupational locations (102).
In another study, Wang et al. evaluated risk of Parkinson’s disease and exposure
to 36 kinds of organophosphate pesticides, observing a d ose-response
relationship with exposure to multiple organophosphates at residences and
workplaces ( OR = 1.98, 95% CI: 1.32, 2.97 for exposure to 1 to 7 different
37

organophosphate pesticides and OR = 3.37, 95% CI: 1.90, 5.95 for exposure to 8
to 14 different organophosphate pesticides) (119).
Cockburn et al. assessed risk of prostate cancer from specific pesticides of
interest associated via residential histories, while controlling for self -reported
home pesticide use and occupational pesticide exposure (117). The GIS -based
methods showed prostate cancer associations for methyl bromide (OR = 1.62,
95% CI: 1.02, 2.59) and a group of eight organochlorine pesticides with known
hormone-related effects (OR = 1.64, 95% CI: 1.02, 2.63). Statistically significant
dose-response relationships were observed for a selected group of organochlorine
pesticides (ORlow vs. none = 1.25, 95% CI: 0.75, 2.08 and ORhigh vs. none = 2.03 95%
CI: 1.17, 3.52, P for trend = 0.04) and for the fungicide captan (OR low vs. none =
0.68, 95% CI: 0.34, 1.36 and OR high vs. none = 1.74, 95% CI: 1.01, 3.13, P for trend =
0.04). Previous epidemiologic studies examining the risk of prostate cancer from
pesticide exposure had found null effects for many of the same organochlorine
pesticides that may also be relevant for breast cancer etiology.
Since these studies were conducted, the GIS -based method described here
has been expanded and improved. The previous published studies incorporated
PUR data beginning in 1974 through 1999 and land- use surveys for the three
Central Valley counties of Fresno, Tulare, and Kern but for the studies pres ented
here the GIS -based method utilized PUR data through 2012 and all digitally
available lan d-use surveys for every county within the state of California. The
method is capable of estimating exposure to specific pesticides of interest in the
most agricul turally productive region in the nation . It is a substantial
38

improvement over methods such as self -reported exposure or those that are
ecologic by design that were used in previous studies of pesticide exposure on the
risk of breast cancer , and can account for a longer period of time than has been
done previously to capture exposures that may accumulate over a person’s
lifetime.

39

Table 1.5. Previous Studies Assessing Disease Risk and Exposure to Pesticides Based on GIS-Methods That Utilize
California Pesticide Use Reporting and Land-Use Data.
Reference Outcome
Number of
Subjects
Prevalence of Exposure and Estimates of Relative Risk
Pesticides Cases
%
Controls
%
OR (95% CI)
Costello, 2009 Parkinson’s
disease
368 cases and
341 controls
Paraquat and maneb 24.0 14.0

OR = 1.75, 95% CI: 1.13, 2.73,
OR = 1.36, 95% CI: 0.83, 2.23 for
those aged >60 years and OR =
5.07, 95% CI: 1.75, 14.71 for those
aged ≤60 years
Wang, 2011 Parkinson’s
disease
362 cases and
341 controls
Exposed at both residences and workplaces to:
Paraquat
Maneb
Ziram

Residentially to paraquat, maneb and ziram
Occupationally to paraquat, maneb and ziram

47.5
11.1
15.2

14.6
12.7

32.6
5.0
5.3

8.2
5.3

OR = 1.50, 95% CI: 1.03, 2.18
OR = 2.26, 95% CI: 1.22, 4.20
OR = 1.50, 95% CI: 1.03, 2.18

OR = 1.86, 95%CI: 1.09, 3.18
OR = 3.09, 95% CI: 1.69, 5.64
Cockburn, 2011 Prostate
cancer
173 cases and
162 controls
Methyl bromide
Captan
Group of 8 organochlorines
50.3
33.5
54.9
43.2
31.5
47.5
OR = 1.62, 95% CI: 1.02, 2.59
OR = 1.20, 95% CI: 0.74, 1.96
OR = 1.64, 95% CI: 1.02, 2.63
Wang, 2014 Parkinson’s
disease
357 cases and
752 controls
Exposed at residences only to:
1-7 organophosphates
Exposed at workplaces only to:
1-7 organophosphates
Exposed at residences and workplaces to:
1-7 organophosphates
8-14 organophosphates

20.7

12.9

22.4
10.4

20.3

10.0

19.0
4.5

OR = 1.63, 95% CI: 1.09, 2.16

OR = 2.16, 95% CI: 1.34, 3.48

OR = 1.98, 95% CI: 1.32, 2.97
OR = 3.37, 95% CI: 1.90, 5.95
40

Chapter 1 References

1. Siegel R, Naishadham D, Jemal A. Cancer statistics, 2012. CA: Cancer J
Clin. 2012;62(1):10–29.
2. Madigan MP, Ziegler RG, Benichou J, et al. Proportion of breast cancer
cases in the United States explained by well -established risk factors. J
Natl Cancer Inst. 1995;87(22):1681–1685.
3. Colditz GA. Relationship between estrogen levels, use of hormone
replacement therapy, and breast cancer. J Natl Cancer Inst.
1998;90(11):814–823.
4. Key T, Appleby P, Barnes I, et al. Endogenous sex hormones and breast
cancer in postmenopausal women: reanalysis of nine prospective studies.
J Natl Cancer Inst. 2002;94(8):606–616.
5. Key T, Pike MC. The role of oestrogens and progestagens in the
epidemiology and prevention of breast cancer. Eur J Cancer Prev.
1988;24(1):29–43.
6. Althuis MD, Fergenbaum JH, Garcia-Closas M, et al. Etiology of hormone
receptor-defined breast cancer: a systematic review of the literature.
Cancer Epidemiol Biomarkers Prev. 2004;13(10):1558–1568.
7. Ma H, Henderson KD, Sullivan -Halley J, et al. Pregnancy -related factors
and the risk of breast carcinoma in situ and invasive breast cancer among
postmenopausal women in the California Teachers Study cohort. Breast
Cancer Res. 2010;12(3):R35.
8. Tamimi RM, Hankinson SE, Chen WY, et al. Combined estrogen and
testosterone use and risk of breast cancer in postmenopausal women.
Arch Intern Med. 2006;166(14):1483–1489.
9. Pike MC, Spicer DV, Dahmoush L, et al. Estrogens, progestogens, normal
breast cell proliferation, and breast cancer risk. Epidemiol Rev.
1993;15(1):17–35.
10. Liehr JG. Dual role of oestrogens as hormones and pro- carcinogens:
tumour initiation by metabolic activation of oestrogens. Eur J Cancer
Prev. 1997;6(1):3–10.
11. Kaaks R, Rinaldi S, Key T, et al. Postmenopausal serum androgens,
oestrogens and breast cancer risk: the European prospective investigation
into cancer and nutrition. Endocr Relat Cancer. 2005;12(4):1071–1082.
41

12. Onland-Moret NC, Kaaks R, van Noord PAH, et al. Urinary endogenous
sex hormone levels and the risk of postmenopausal breast cancer. Br J
Cancer. 2003;88(9):1394–1399.
13. Chlebowski RT, Kuller LH, Prentice RL, et al. Breast cancer after use of
estrogen plus progestin in postmenopausal women. N. Engl. J. Med.
2009;360(6):573–587.
14. Rossouw JE, Anderson GL, Prentice RL, et al. Risks and benefits of
estrogen plus progestin in healthy postmenopausal w omen - Principal
results from the Women's Health Initiative randomized controlled trial.
JAMA. 2002;288(3):321–333.
15. Clarke M, Collins R, Davies C, et al. Tamoxifen for early breast cancer: an
overview of the randomised trials. Lancet. 1998;351(9114):1451–1467.
16. Endocrine Disruptor Screening Program , Office of Chemical Safety and
Pollution Protection. Weight-of-Evidence: Evaluating Results of EDSP
Tier 1 Screening to Identify the Need for Tier 2 Testing. Washington, DC:
U.S. Environmental Protection Agency; 2011. Federal Register (EPA-HQ-
OPPT-2010-0877-0021).
17. Soto AM, Chung KL, Sonnenschein C. The pesticides endosulfan,
toxaphene, and dieldrin have estrogenic effects on human estrogen -
sensitive cells. Environ Res. 1994;102:380–383.
18. Høyer AP, Grandjean P, Jørgensen T, et al. Organochlorine exposure and
risk of breast cancer. Lancet. 1998;352(9143):1816–1820.
19. Snedeker SM. Pesticides and breast cancer risk: a review of DDT, DDE,
and dieldrin. Environ Res. 2001;109(Suppl 1):35–47.
20. Fuhrman BJ, Schairer C, Gail MH, et al. Estrogen metabolism and risk of
breast cancer in postmenopausal women. J Natl Cancer Inst.
2012;104(4):326–339.
21. Henderson BE, Feigelson HS. Hormonal carcinogenesis. Carcinogenesis.
2000;21(3):427–433.
22. Phillips KP, Foster WG, Leiss W, et al. Assessing and managing risks
arising from exposure to endocrine- active chemicals. J Toxicol Environ
Health B Crit Rev. 2008;11(3-4):351–372.
23. Falck FJ, Ricci AJ, Wolff MS, et al. Pesticides and polychlorinated
biphenyl residues in human breast lipids and their relation to breast
cancer. Arch Environ Health. 1992;47(2):143–146.
42

24. Rogan WJ, Gladen BC, McKinney JD, et al. Polychlorinated biphen yls
(PCBs) and dichlorodiphenyl dichloroethene (DDE) in human milk:
effects on growth, morbidity, and duration of lactation. Am J Public
Health. 1987;77(10):1294–1297.
25. Kaushik PP, Kaushik GG. An assessment of structure and toxicity
correlation in orga nochlorine pesticides. J Hazard Mater. 2007;143(1 -
2):102–111.
26. Mukherjee S, Koner BC, Ray S, et al. Environmental contaminants in
pathogenesis of breast cancer. Indian J Exp Biol. 2006;44(8):597–617.
27. Briz V, Molina -Molina J -M, Sánchez- Redondo S, e t al. Differential
estrogenic effects of the persistent organochlorine pesticides dieldrin,
endosulfan, and lindane in primary neuronal cultures. Toxicol Sci.
2011;120(2):413–427.
28. Rattenborg T, Gjermandsen I, Bonefeld -Jørgensen E. Inhibition of E2 -
induced expression of BRCA1 by persistent organochlorines. Breast
Cancer Res. 2002;4(6):R12.
29. Lemaire G, Mnif W, Mauvais P, et al. Activation of alpha - and beta-
estrogen receptors by persistent pesticides in reporter cell lines. Life Sci.
2006;79(12):1160–1169.
30. Kortenkamp A. Breast cancer, oestrogens and environmental pollutants: a
re-evaluation from a mixture perspective. Int J Androl. 2006;29(1):193–
198.
31. Valerón PF, Pestano JJ, Luzardo OP, et al. Differential effects exerted on
human mammary epithelial cells by environmentally relevant
organochlorine pesticides either individually or in combination. Chem
Biol Interact. 2009;180(3):485–491.
32. International Agency for Research on Cancer. IARC Monographs on the
Evaluation of Carcinogenic Risk of Chemicals to Humans. Occupational
exposures in insecticide application, and some pesticides. 1999;53.
(http://monographs.iarc.fr/ENG/Monographs/vol53/volume53.pdf).
(Accessed May 31, 2013).
33. Iscan MM, Coban TT, Cok II, et al. The organochlorine pesticide residues
and antioxidant enzyme activities in human breast tumors: is there any
association? Breast Cancer Res Treat. 2002;72(2):173–182.
34. Linko PP, Yeowell HNH, Gasiewicz TAT, et al. Induction of cytochrome P -
450 isozymes by hexachlorobenzene in rats and aromatic hydrocarbon
(Ah)-responsive mice. J Biochem Toxicol. 1986;1(2):95–107.
43

35. Cassidy RAR, Natarajan SS, Vaughan GMG. The link between the
insecticide heptachlor epoxide, estradiol, and breast cancer. Breast
Cancer Res Treat. 2005;90(1):55–64.
36. García MA, Peña D, Álvarez L, et al. Hexachlorobenzene induces cell
proliferation and IGF -I signaling pathway in an estrogen receptor alpha-
dependent manner in MCF -7 breast cancer cell line. Toxicol Lett.
2010;192(2):195–205.
37. Coumoul X, Diry M, Robillot C, et al. Differential regulation of
cytochrome P450 1A1 and 1B1 by a combination of dioxin and pesticides
in the breast tumor cell line MCF-7. Cancer. 2001;61(10):3942–3948.
38. Teitelbaum SL, Gammon MD, Britton JA, et al. Reported residential
pesticide use and breast cancer risk on Long Island, New York. Am J Ind
Med. 2007;165(6):643–651.
39. Farooq U, Joshi M, Nookala V, et al. Self -reported exposure to pesticides
in residential settings and risk of breast cancer: a case -control study.
Environ Health. 2010;9(30):1–9.
40. El-Zaemey S, Heyworth J, Glass DC, et al. Household and occupational
exposure to pesticides and risk of breast cancer. Int J Environ Health
Res. 2014;24(2):91–102.
41. El-Zaemey S, Heyworth J, Fritschi L. Noticing pesticide spray drift from
agricultural pesticide application areas and breast cancer: a case -control
study. Aust N Z J Public Health. 2013;37(6):547–555.
42. Engel LS, Hill DA, Hoppin JA, et a l. Pesticide use and breast cancer risk
among farmers' wives in the Agricultural Health Study. Am J Ind Med.
2005;161(2):121–135.
43. Weichenthal S, Moase C, Chan P. A review of pesticide exposure and
cancer incidence in the Agricultural Health Study cohort. Environ Res.
2010;118(8):1117–1125.
44. Brophy JT, Keith MM, Watterson A, et al. Breast cancer risk in relation to
occupations with exposure to carcinogens and endocrine disruptors: a
Canadian case–control study. Environ Health. 2012;11:87.
45. Mills PKP, Yang RR. Breast cancer risk in Hispanic agricultural workers
in California. Int J Occup Environ Health. 2005;11(2):123–131.
46. Dolapsakis G, Vlachonikolis IG, Varveris C, et al. Mammographic findings
and occupational exposure to pesticides currently in use on Crete. Eur. J.
Cancer. 2001;37(12):1531–1536.
44

47. Band PR, Le ND, Fang R, et al. Identification of occupational cancer ris ks
in British Columbia. A population -based case -control study of 995
incident breast cancer cases by menopausal status, controlling for
confounding factors. J. Occup. Environ. Med. 2000;42(3):284–310.
48. Gardner KM, Ou Shu X, Jin F, et al. Occupations and breast cancer risk
among Chinese women in urban Shanghai. Am J Ind Med.
2002;42(4):296–308.
49. Rubin CH, Burnett CA, Halperin WE, et al. Occupation as a risk identifier
for breast cancer. Am J Public Health. 1993;83(9):1311–1315.
50. Blair A, Dosemeci M, Heineman EF. Cancer and other causes of death
among male and female farmers from twenty -three states. Am J Ind Med.
1993;23(5):729–742.
51. Morton WE. Major differences in breast cancer risks among occupations.
J. Occup. Environ. Med. 1995;37:328–335.
52. Fleming LE, Bean JA, Rudolph M, et al. Mortality in a cohort of licensed
pesticide applicators in Florida. Occup Environ Med. 1999;56(1):14–21.
53. Kasner EJ, Keralis JM, Mehler L, et al. Gender differences in acute
pesticide-related illnesses and injuries among farmworkers in the United
States, 1998-2007. Am J Ind Med. 2012;55(7):571–583.
54. Duell EJ, Millikan RC, Savitz DA, et al. A population -based case-control
study of farming and breast cancer in North Carolina. Epidemiology.
2000;11(5):523–531.
55. Wolff MS, Toniolo PG, Lee EW, et al. Blood levels of organochlorine
residues and risk of breast cancer. J Natl Cancer Inst. 1993;85(8):648–
652.
56. Wolff MS, Zeleniuch -Jacquotte A, Dubin N, et al. Risk of breast cancer
and organochlorine exposure. Cancer Epidemiol Biomarkers Prev.
2000;9(3):271–277.
57. Lopez-Carrillo L, Torres -Sanchez L, Moline J, et al. Breast -feeding and
serum p,p'DDT levels among Mexican women of childbearing age: a pilot
study. Environ Res. 2001;87:131–135.
58. Romieu I, Hernandez -Avila M, Lazcano -Ponce E, et al. Breast cancer,
lactation history, and serum organochlorines. Am J Ind Med.
2000;152(4):363–370.
45

59. Gammon MD, Wolff MS, Neugut AI, et al. Environmental toxins and
breast cancer on Long Island. II. organochlorine compound levels in
blood. Cancer Epidemiol Biomarkers Prev. 2002;11:686–697.
60. Krieger N, Wolff MS, Hiatt RA, et al. Breast cancer and serum
organochlorines: a prospective study among white, black, and Asian
women. J Natl Cancer Inst. 1994;86(8):589–599.
61. Bagga D, Anders KH, Wang H -J, et al. Organochlorine pesticide content
of breast adipose tissue from women with breast cancer and control
subjects. Journal Natl Cancer Inst. 2000;92(9):750–753.
62. Charlier C, Albert A, Herman P, et a l. Breast cancer and serum
organochlorine residues. Occup Environ Med. 2003;60(5):348–351.
63. Cocco P, Kazerouni N, Zahm SH. Cancer mortality and environmental
exposure to DDE in the United States. Environ Res. 2000;108(1):1–4.
64. Demers A, Ayotte P, B risson J, et al. Risk and aggressiveness of breast
cancer in relation to plasma organochlorine concentrations. Cancer
Epidemiol Biomarkers Prev. 2000;9(2):161–166.
65. Dorgan JF, Brock JW, Rothman N, et al. Serum organochlorine pesticides
and PCBs and breast cancer risk: results from a prospective analysis
(USA). Cancer Causes Control. 1999;10(1):1–11.
66. Hunter DJ, Hankinson SE, Laden F, et al. Plasma organochlorine levels
and the risk of breast cancer. N. Engl. J. Med. 1997;337(18):1253–1258.
67. Laden F, Hankinson SE, Wolff MS, et al. Plasma organochlorine levels
and the risk of breast cancer: an extended follow -up in the Nurses' Health
Study. Int. J. Cancer. 2001;91(4):568–574.
68. Millikan R, DeVoto E, Duell EJ, et al. Dichlorodiphenyldichloroethen e,
polychlorinated biphenyls, and breast cancer among African -American
and white women in North Carolina. Cancer Epidemiol Biomarkers Prev.
2000;9(11):1233–1240.
69. van't Veer P, Lobbezoo IE, Martín -Moreno JM, et al. DDT (dicophane)
and postmenopausal br east cancer in Europe: case -control study. BMJ.
1997;315(7100):81–85.
70. Zheng T, Holford TR, Mayne ST, et al. DDE and DDT in breast adipose
tissue and risk of female breast cancer. Am J Ind Med. 1999;150(5):453–
458.
46

71. Stellman SD, Djordjevic MV, Bri tton JA, et al. Breast cancer risk in
relation to adipose concentrations of organochlorine pesticides and
polychlorinated biphenyls in Long Island, New York. Cancer Epidemiol
Biomarkers Prev. 2000;9(11):1241–1249.
72. Moysich KB, Ambrosone CB, Vena JE, et al. Environmental
organochlorine exposure and postmenopausal breast cancer risk. Cancer
Epidemiol Biomarkers Prev. 1998;7:181–188.
73. Moysich KB, Ambrosone CB, Mendola P, et al. Exposures associated with
serum organochlorine levels among postmenopausal women from
Western New York state. Am J Ind Med. 2002;41(2):102–110.
74. Verner M-A, Bachelet D, McDougall R, et al. A case study addressing the
reliability of polychlorinated biphenyl levels measured at the time of
breast cancer diagnosis in representing early-life exposure. Cancer
Epidemiol Biomarkers Prev. 2011;20(2):281–286.
75. Wolff MS, Britton JA, Teitlebaum SL, et al. Improving organochlorine
biomarker models for cancer research. Cancer Epidemiol Biomarkers
Prev. 2005;14(9):2224–2236.
76. Pelletier CC, Doucet EE, Imbeault PP, et al. Associations between weight
loss-induced changes in plasma organochlorine concentrations, serum
T(3) concentration, and resting metabolic rate. Toxicol. Sci.
2002;67(1):46–51.
77. Vo TT, Gladen BC, Cooper GS, e t al. Dichlorodiphenyldichloroethane and
polychlorinated biphenyls: intraindividual changes, correlations, and
predictors in healthy women from the southeastern United States. Cancer
Epidemiol Biomarkers Prev. 2008;17(10):2729–2736.
78. Xu X, Dailey AB, Talbott EO, et al. Associations of serum concentrations
of organochlorine pesticides with breast cancer and prostate cancer in
U.S. adults. Environ Res. 2010;118:60–66.
79. Calle EE, Frumkin H, Henley SJ, et al. Organochlorines and breast cancer
risk. CA: Cancer J Clin. 2002;52:301–309.
80. Shakeel MK, George PS, Jose J, et al. Pesticides and breast cancer risk: a
comparison between developed and developing countries. Asian Pac J
Cancer Prev. 2010;11(10):173–180.
81. Høyer AP, Jørgensen T, Grandjean P, et al. Repeated measurements of
organochlorine exposure and breast cancer risk (Denmark). Cancer
Causes Control. 2000;11(2):177–184.
47

82. Cohn BA, Wolff MS, Cirillo PM, et al. DDT and breast cancer in young
women: new d ata on the significance of age at exposure. Environ Res.
2007;115(10):1406–1414.
83. Ruder EH, Dorgan JF, Kranz S, et al. Examining breast cancer growth
and lifestyle risk factors: early life, childhood, and adolescence. Clin
Breast Cancer. 2008;8(4):334–342.
84. Okasha M, McCarron P, Gunnell D, et al. Exposures in childhood,
adolescence and early adulthood and breast cancer risk: a systematic
review of the literature. Breast Cancer Res Treat. 2003;78(2):223–276.
85. Reynolds P, Hurley SE, Gunier RB, et al. Residential proximity to
agricultural pesticide use and incidence of breast cancer in California,
1988–1997. Environ Res. 2005;113(8):993–1000.
86. Muir K, Rattanamongkolgul S, Smallman- Raynor M, et al. Breast cancer
incidence and its possible spatia l association with pesticide application in
two counties of England. Annu Rev Public Health. 2004;118(7):513–520.
87. Mills PK, Yang R. Regression analysis of pesticide use and breast cancer
incidence in California Latinas. J Environ Health. 2006;68(6):15–14.
88. O'Leary ESE, Vena JEJ, Freudenheim JLJ, et al. Pesticide exposure and
risk of breast cancer: a nested case -control study of residentially stable
women living on Long Island. Environ Res. 2004;94(2):134–144.
89. Ashley-Martin J, VanLeeuwen J, Cribb A, et al. Breast cancer risk,
fungicide exposure and CYP1A1*2A gene- environment interactions in a
province-wide case control study in Prince Edward Island, Canada. Int J
Environ Health Res Public Health. 2012;9(12):1846–1858.
90. Brody JG, Vorhees DJ, Melly SJ, et al. Using GIS and historical records to
reconstruct residential exposure to large- scale pesticide application. J
Expo Anal Environ Epidemiol. 2002;12(1):64–80.
91. Brody JG, Aschengrau A, McKelvey W, et al. Breast cancer risk and
historical e xposure to pesticides from wide- area applications assessed
with GIS. Environ Res. 2004;112(8):889–897.
92. Reynolds P, Hurley S, Goldberg DE, et al. Regional var iations in breast
cancer among California teachers. Epidemiology. 2004;15(6):746–754.
93. United States Department of Agriculture, Economic Research Service.
Agricultural Productivity in the U.S. National Tables 1948 –2011.
(http://www.ers.usda.gov/data-products/agricultural-productivity-in-
the-us.aspx#.Up6dwr--WXo). (Accessed December 3, 2013).
48

94. California Department of Pesticide Regulation. Summary of Pesticide Use
Report Data–2010. Sacramento, CA: California Environmental Protection
Agency; 2011. (http://www.cdpr.ca.gov/docs/pur/pur10rep/10sum.htm#
Development). (Accessed June 11, 2013).
95. Luo Y, Zhang M. Spatially distributed pesticide exposure assessment in
the Central Valley, California, USA. Envir Pollut. 2010;158(5):1629–1637.
96. Pfleeger TGT, Olszyk DD, Burdick CAC, et al. Using a geographic
information system to identify areas with potential for off-target pesticide
exposure. Environ Toxicol Chem. 2006;25(8):2250–2259.
97. Brouwer DH, Brouwer EJ, van Hemmen JJ. Estimation of long -term
exposure to pesticides. Am J Ind Med. 1994;25(4):573–588.
98. Russell HH, Jackson RJ, Spath DP, et al . Chemical Contamination of
California Drinking Water. West J Med. 1987;147(5):615–622.
99. Simcox NJ, Fenske RA, Wolz SA, et al. Pesticides in household dust and
soil: exposure pathways for children of agricultural families. Environ Res.
1995;103(12):1126–1134.
100. California Department of Pesticide Regulation. A Guide to Pesticide
Regulation in California. Sacramento, CA : California Environmental
Protection Agency ; 2011. (http://www.cdpr.ca.gov/docs/pressrls/dpr
guide.htm). (Accessed May 31, 2013).
101. California Department of Water Resources . California Land & Water Use.
Land Use Data. (http://www.water.ca.gov/landwateruse/lusrvymain .
cfm). (Accessed July 12, 2013).
102. Wang AA, Costello SS, Cockburn MM, et al. Parki nson's disease risk from
ambient exposure to pesticides. Eur J Epidemiol. 2011;26(7):547–555.
103. Costello S, Cockburn M, Bronstein J, et al. Parkinson's disease and
residential exposure to maneb and paraquat from agricultural
applications in the central valley of California. Am J Ind Med.
2009;169(8):919–926.
104. Rull RP, Ritz B. Historical pesticide exposure in California using pesticide
use reports and land-use surveys: an assessment of misclassification error
and bias. Environ Res. 2003;111(13):1582–1589.
105. Nuckols JR, Gunier RB, Riggs P, et al. Linkage of the California Pesticide
Use Reporting Database with spatial land use data for exposure
assessment. Environ Res. 2007;115(5):684–689.
49

106. Marusek JC, Cockburn MG, Mills PK, e t al. Control selection and
pesticide exposure assessment via GIS in prostate cancer studies. Am J
Prev Med. 2006;30(2):S109–S116.
107. Gunier RB, Harnly ME, Reynolds P, et al. Agricultural pesticide use in
California: Pesticide prioritization, use densit ies, and population
distributions for a childhood cancer study. Environ Res.
2006;109(10):1071–1078.
108. Maxwell SK, Airola M, Nuckols JR . Using Landsat satellite data to
support pesticide exposure assessment in California. Int J Health Geogr.
2010;9(46):1–14.
109. Maxwell SK. Downscaling pesticide use data to the crop field level in
California using Landsat satellite imagery: paraquat case study. Remote
Sensing. 2011;3(12):1805–1816.
110. Maxwell SK. Generating land cover boundaries from remotely sensed data
using object -based image analysis: overview and epidemiological
application. Spat Spatiotemporal Epidemiol. 2010;1(4):231–237.
111. Goldberg DW, Wilson JP, Knoblock CA, et al. An effective and efficient
approach for manually improving geocoded data. Int J Health Geogr.
2008;7:60.
112. Woods N, Craig IP, Dorr G, et al. Spray drift of pesticides arising from
aerial application in cotton. J Environ Qual. 2001;30(3):697–701.
113. Cox C. Pesticide drift: indiscriminately from the skies. Journal of
Pesticide Reform [electronic article]. 1995;15:1 –7. (http://sunridge.net/
assets/pdf/pesticide_drift.pdf).
114. Bradman A, Schwartz JM, Fenster L, et al. Factors predicting
organochlorine pesticide levels in pregnant Latina women living in a
United States agricultural area. J Expo Sci Environ Epidemiol.
2007;17:388–399.
115. Gunier RB, Ward MH, Airola M, et al. Determinants of agricultural
pesticide concentrations in carpet dust. Environ Res. 2011;119(7):970 –
976.
116. Ward MH, Lubin J, Giglierano J, et al. Pr oximity to crops and residential
exposure to agricultural herbicides in Iowa. Environ Res.
2006;114(6):893–897.
117. Cockburn M, Mills P, Zhang X, et al. Prostate cancer and ambient
pesticide exposure in agriculturally intensive areas in California. Am J Ind
Med. 2011;173(11):1280–1288.
50

118. Ritz B, Costello S. Geographic model and biomarker -derived measures of
pesticide exposure and Parkinson's disease. Ann N Y Acad Sci.
2006;1076(1):378–387.
119. Wang A, Cockburn M, Ly TT, et al. The association between ambient
exposure to organophosphates and Parkinson's disease risk. Occup
Environ Med. 2014;71(4):275–281.

51

Chapter 2

2. Registry-Based Case-Control Study of Breast Cancer Risk from
Ambient Exposure to Pesticides

2.1 Abstract
Since lifetime estrogen exposure is a key factor in breast cancer
development, exposure to pesticides that are suspected to mimic estrogen and
affect hormone metabolism may be plausibly involved in breast cancer
development. Breast cancer cases were identified from a population -based cancer
registry in 1988– 2012 ( n = 9,540), and compared to two kinds of controls,
registry-based patients diagnosed in 1988–2012 with cancers at sites other than
the breast ( n = 22,106) and a random selection of tax assessor parcel -based
controls (n = 10,000), which may be more representative of population -based
controls than people with other cancer diagnoses. Compared to no exposure,
there was no significant difference in risk of breast cancer among those exposed
to a selected group of organochlorine pesticides (OR = 1.01, 95% CI: 0.96, 1.07 )
or to any of the pesticides detecting in local air monitoring (chlorpyrifos,
diazinon, or 1,3-dichloropropene). No increased risk was observed after including
10-year or 15 -year latency periods for any pesticide. No associations were found
for specific estrogen receptor breast cancer subtypes; however, different
pesticides exhibit varying affinities for the estrogen receptor and some pesticides
have also shown anti -estrogenic effects in vitro. In this study, pesticide exposure
was based on the residential address at the time of diagnosis, which assumes that
52

this is a relevant period at risk for breast cancer. Using only the address at th e
time of diagnosis for both cases and controls may have led to non -differential
misclassification of exposure and the attenuated risk estimates.

2.2 Introduction
Known risk factors for breast cancer only account for 40% of breast cancer
cases in the U. S. (1). Several established breast cancer risk factors are related to
exposure to endogenous or exogenous hormones, including early menarche, late
menopause, parity, postmenopausal hormone therapy and postmenopausal
obesity (2-5). Lifetime exposure to estrogens is a strong risk factor for breast
cancer (3,6,7). Estrogens act as prom oters of cell growth and precursors for
carcinogenic metabolites (8,9), raising concerns about the possible role of
exogenous xenoestrogens such as pesticides (10).
Many pesticides, particularly organochlorines, are endocrine -disrupting
chemicals that affect hormone functi oning by mimicking estrogen or affecting
enzyme systems involved in hormone metabolism (11). Persistent
organochlorines bioaccumulate in fat and by -products are measurabl e in breast
tissue and breast milk (12-14). Organochlorines such as DDT, dieldrin,
endosulfan, methoxychlor and toxaphene have chemical structure similarities to
estrogen (15,16). As a result they interact with estrogen recept ors (17), increase
the estrogen -dependent transcription of target genes and modulate metabolic
enzymes (18). In toxicological studies done in vitro, exposure to organochlorines
show dose-dependent estrogenic effects (19) and stronger effects with exposure
to multiple organochlorines (11,20). Even though organochlorines have weaker
53

estrogenic activity than endogenous estrogens, the prolonged exposure to
combinations of organochlorines may have a cumulative impact even at low
concentrations, inducing breast cell proliferation (21).
Breast cancer susceptibility depends on the levels o f circulating
carcinogens in the breast . Many organochlorines are known or suspected human
carcinogens (22-24). Organochlorines have been shown to interact with P450
metabolic enzymes by binding to their promoter regions (25,26) and increasing
the production of reactive species that contribute to carcinogenesis (27,28). In
breast cancer patients, higher concentrations of organochlorines have also been
found in tumor tissue compared to surrounding tumor-free tissue (29).
Although the estrogenic properties of organochlorines have been widely
recognized, results from previous epidemiologic studies of the impact of pesticide
exposure on breast cancer risk have been conflicting (30-38). A case-control
study conducted in Long Island, NY by Gammon et al. showed slightly elevated
risk of breast cancer for women with the highest levels of organochlorines DDE
and dieldrin compared to the lowest levels, although these findings were not
statistically significant (OR = 1.20, 95% CI: 0.76, 1.90 and OR = 1.37, 95% CI:
0.69, 2.72, respectively) (39). Another case -control study conducted by
Teitelbaum et al. in Long Island found modest breast cancer associations with
lifetime self-reported residential pesticide use (OR = 1.39, 95% CI: 1.15, 1.68) (31)
and a New York study conducted by O’Leary et al. found elevated breast cancer
risk among women residing nea r hazardous waste sites containing
organochlorine pesticides (OR = 2.80, 95% CI: 1.10, 7.10) (35).
54

California is the most agricultural state in the nation where one quarter of
all commercial pesticides in the U.S. are applied (40). The Central Valley counties
of Fresno, Tulare , and Kern are the top three ranked agricultural counties in
California for total pounds of pesticide active ingredients reported. In contrast to
more urban areas like Long Island (31), residents of intense agricultural regions
are more likely to be exposed to ambient pesticides from neighboring agricultural
application through aerial drift, field runoff, or through contamination of local
drinking water sources (41-44). Previous studies in the Central Valley have found
statistically significant risk of breast cancer using an ecologic regression approach
for the organochlorine pesticides methoxychlor and toxaphene (RR = 1.18, 95%
CI: 1.03, 1.35 and RR = 1.16, 95% CI: 1.01, 1.34 respectively) (45), as well as a
case-control study for chlordane (OR = 3.85, 95% CI: 1.22, 12.20 ) (46). These
findings mos t closely resemble results from the Agricultural Health Study, a
cohort of pesticide applicators in Iowa and North Carolina, in which wives of
applicators had elevated risks of breast cancer when their husbands
professionally applied the pesticides chlorda ne (RR = 1.70, 95% CI: 1.20, 2.50),
dieldrin (RR = 2.00, 95% CI: 1.10, 3.30), heptachlor (RR = 1.60, 95% CI: 1.10),
lindane (RR = 1.70, 95% CI: 1.10, 2.50), and toxaphene (RR = 1.40, 95% CI: 0.90,
2.10) (47).
It is possible that the conflicting findings of previous studies of breast
cancer risk and pesticide exposure arises from two main methodological
problems. The first is that the majority of studies have been conducted in areas
where the level of exposure to pesticides may be too low to detect an effect and
too few exposed persons leads to wide confidence intervals from being
55

underpowered, subsequently reporting null findings. Secondly, the methods of
measuring exposure to pesticides has been highly inaccurate and reported no
effect when there could truly be an effect (48). Several studies relied on self -
reported pesticide exposure where the level of exposure may be too infrequent or
failed to find an association because exposure misclassification biased the
estimates towards the null (31,47). Other studies considered only a small window
of time to evaluate exposure (49), grouped many pesticides together that may not
be relevant for breast cancer etiology (35,37), or were ecological studies by design
(33,36,45).
A Geographical Information Systems (GIS) -based approach in a region of
high pesticide use has been shown to reduce exposure misclassification in a
previous study involving prostate cancer (30). This GIS-based model has shown
high exposure accuracy and specificity compared to previous studies showing null
effects for hormone-related pesticides that are also likely to be relevant for breast
cancer. The present study will assess bre ast cancer risk and ambient lifetime
pesticide exposure using these GIS -based methods in a registry -based case -
control study to observe associations using a detailed exposure model that is
designed to reduce exposure misclassification as compared with estimates based
on less comprehensive measures used in previous studies.

2.3 Materials and Methods
2.3.1 Selection of cases and controls
Cases and controls were residents at the time of diagnosis in the counties
of Fresno, Tulare and Kern, identified from California’s population-based Cancer
56

Registry of the Central California (CRCC) with histologically confirmed primary
invasive cancer diagnoses occurring from 1988 to 2012 (n = 31,646), which
includes all currently available data. The registry includes all cancers except non-
melanoma skin and in- situ cervical cancers . The demographic and diagnostic
variables used in these analyses, including age, gender, race/ethnicity, birthplace,
residential address, tumor stage and estrogen receptor (ER) subtype were
collected by the CRCC from patient medical records at the time of diagnosis (50).
Race/ethnicity was identified by the patient or assumptions made by the patient’s
appearance, ethnicity of the parents, birth place, surname or maiden name.
Neighborhood socioeconomic status (SES) was derived from 1990 U.S. census
data at the block group level based on the patient’s address at the time of
diagnosis and categorized into a quintile score ranging from low SES (1) to high
SES (5) by Yost et al. (51).
For this registry -based study, cases were women diagnosed with breast
cancer and controls were women with cancer diagnoses at sites other than the
breast. We considered several control groups, including controls that consisted of
all other cancer diagnoses, controls that excluded lung cancer diagnoses, since
lung cancer may plausibly be related to inhalation of ambient pesticide exposure,
and controls that excluded ovarian, endometrial (uterine) and other female
genital cancers, since these cancers are asso ciated with hormone balance and
may also be associated with pesticides that mimic estrogen.
Cases and controls were included in the study if they were residents in
California’s Central Valley counties of Fresno, Tulare, and Kern, and were
between the ages of 55 and 74 years (so as to include only postmenopausal breast
57

cancer, which is more likely of hormone- related origin). The study protocol was
approved by the Office f or the Protection of Research Subjects at the University
of Southern California.
Cases were also compared to a control group consisting of tax assessor
parcels randomly selected from the same geographic area as the population -
based cancer registry cases , where cases and tax assessor parcel controls were
from the entire Central Valley regio n including the counties of Fresno, Kern,
Kings, Madera, Mariposa, Merced, Stanislaus, Tulare and Tuolumne. Each case
was individually matched to one of the 10,000 randomly selected tax assessor
parcels, weighted for the proportion of census-level age and race demographics of
the case (Figure 2.1).
2.3.2 Ambient pesticide exposure assessment
Individual pesticide exposure was determined from proximity to
commercial pesticide application as a proxy measure (52,53). Since 1970, farmers
and pest control operators in California have been required to report their
applications of restricted -use pesticides to local agricultural commissioners, and
applications of all pesticides starting in 1990 (10). California pesticide usage
reporting (PUR) data starting from 1974 is made publically available from the
California Department of Pesticide Regulation. PUR data includes the active
ingredients, amounts applied, types of crop and acreage, application methods,
dates of use, and locations identified as Public Land -use Survey System (PLSS)
sections, which are equivalent to one square mile. In addition, each county
conducts land- use surveys every 6 t o 10 years, documenting the locations of
agricultural crops and types of land- use such as fields, vineyards and orchards.
58

The data are available from California’s Department of Water Resources, Division
of Planning and Local Assistance for the years 1976 t o 20 12 (54). Historical
electronic maps of land use and crop type were constructed from the most recent
digitally available land use surveys fo r all Californ ia counties (1986 –2012) and
manually digitized for the counties of Fresno, Tulare , and Kern using the earliest
available paper maps for the years 1977–1985 (52,55).
PUR data was linked with land- use survey data by PLSS section to more
precisely identify where the pesticide was applied (48). When land-use and PUR
data matched exactly, the pesticide was assumed to be applied in the land- use
polygon of the corresponding crop type; when PUR data did not match by crop
type but other crops were identified in the PLSS section, the pesticide was
assumed to be applied to the location where the existing crops were reported; and
when no crops were identified in the PLSS section, the pesticide was assumed to
be applied across the entire PLSS section.
2.3.2.1 Geocoding
Residential address at the time of diagnosis was used to measure exposure
to ambient commercial pesticide application, assuming that individuals resided
at that address over the course of their lifetimes. The residential addresses were
then geocoded to identify the corresponding geographic locat ion coordinates (i.e.
latitude and longitude). Locations that were geocoded to a complete and valid
street address (i.e. centroid of a tax assessor parcel) were considered to have
“high” geocode certainty whereas locations that could not be geocoded to a
complete and valid street address were considered to have “low” geocode
certainty.
59

2.3.2.2 Pesticide exposure estimates
Exposure to a given pesticide was determined by the density of the
pesticide applied within a 500- meter buffer around each residential location,
weighted for the number of years exposed to that pesticide (48). This buffer
distance was chosen based on studies of pesticide drift that found measu rable
concentrations of pesticides in homes near commercial application sites (56,57).
Ambient pesticide expos ure for each year and location was identified and an
estimated exposure history was calculated for each individual from birth year to
the year of diagnosis. An individual’s exposure to a particular pesticide of interest
was determined by averaging the cumulative total number of pounds of pesticide
applied within the buffer surrounding the best estimated geocoded location
divided by the total number of years exposed at that location. Individuals missing
information on the residential address at the time of diagnosis were excluded
from the estimated exposure history.
The GIS -based pesticide exposure model incorporated data from 1974 –
2012, years with complete pesticide reporting and land -use information. We
examined exposure to a group of organochlorine insecticides most likely
associated with breast cancer carcinogenesis was selected, including
aldrin/dieldrin (21,58-63), chlordane (19,46,60,64-66), dicofol (67), endosulfan
(62,65,68,69), endrin (70), heptachlor (27,71,72), hexachlorobenzene
(28,60,73,74), lindane (also known as gamma -hexachlorocyclohexane)
(59,75,76), methyoxychlor (77,78), and toxaphene (19,62,66,79,80). An
individual was considered exposed to the group of organochlorines if they were
exposed to at least one of the selected organochlorine insecticides during the time
60

period of interest. In order to evaluate the specificity of the pesticide exposure
assessment method, three chemicals that have not been linked to breast cancer
but were detected in ambient air monitoring conducted in Fresno County at levels
of concern to human health were also examined (81).
2.3.3 Statistical Analyses
In the registry -based case-control study, unconditional logistic regression
was used to estimate lifetime ambient pesticide exposure on the risk of breast
cancer in postmenopausal women. For the group of organochlorines and each of
the three other pesticides of interest, odds ratios (OR) and 95% confidence
intervals (CI) were determined from binary o utcome measures of exposed
compared to unexposed groups where an individual was considered exposed if
the average pesticide exposure over the time period from 1974 until the address
of diagnosis was not null . Among those exposed, the total average pounds of
pesticide applied per area within the 500 -meter buffer of each location was often
heavily skewed, therefore “low” and “high” exposure cut -points were created
based on the median average total pounds of pesticides exposed among the
exposed control group subjects. Analyses were adjusted for age (continuous),
ethnicity (non -Hispanic white, Hispanic, or other ), SES (low to high quintiles) ,
and geocode quality (geocoded to a complete and valid street address or not) .
Stage of the cancer at diagnosis (localized, regional or distant/missing) and
birthplace (USA or other) were included in the model if the beta est imates
changed by >10% after the addition of either variable.
Since cancer development can often have a long latency period before
diagnosis and the time frame of which is unknown, exposures 10 years before
61

diagnosis and 15 years before diagnosis were excluded to examine any changes in
association due to latency. Sensitivity analyses were conducted to examine the
potential impacts of including certain other types of cancer diagnoses that may
possibly share a common etiology with breast cancer and are associ ated with
hormone-related pesticides among the group of control subjects, including lung
cancer diagnoses as well as ovarian, uterine, and other reproductive cancer
diagnoses.
Breast cancer -pesticide associations were also analyzed among breast
cancer est rogen receptor (ER) subtypes to evaluate whether there was effect
modification for ER( -) compared to ER(+) tumors. ER(+) breast cancers tend to
be correlated with endogenous estrogen and hormone- driven factors and
therefore may respond differently than ER( -) breast cancers to exogenous
estrogens such as organochlorine pesticides (2,21,82). The analysis examining ER
subtypes included breast cancer as well as other cancer controls diagnosed
between 1991 and 2012, the time period when this information was available in
the registry.
In addition to adjusting for geocode certainty in our analyses, we also
stratified by geocode certainty to evaluate the potential impact of geocode
certainty on the estimates of relative risk, comparing high to low geocode
certainty.
Younger age at the time of exposure has been suggested as a period of
particular vulnerability to exposure from endogenous and ex ogenous hormone -
like chemicals. Since exposures prior to the first pregnancy when the breast tissue
is not fully differentiated has been proposed as a critical time period of exposure
62

opportunity (83,84), we examined exposures to pesticides at ages 20 –39 years.
All analyses were conducted using SAS 9.3 (SAS Institute, Inc., Cary, North
Carolina).
Conditional logistic regression was used to analyze the registry -based
breast cancer cases that were individually matched to randomly selected tax
assessor parcels defined as control s, weighted for the proportion of census -level
age and race demographics.

2.4 Results
Breast cancer cases were more likely to be younger, non -Hispanic white
and to be diagnosed with localized carcinoma compared to other types of cancer
in the CRCC (Table 2.1). Prevalence of exposure to specific pesticides ranged from
22% for 1,3 -dichloropropene to 65% for the selected group of organochlorines
among both cases and other cancer diagnosed controls. There were no significant
differences in risk of breast cancer for exposed versus unexposed based on
residential address at the time of diagnosis for the selected group of
organochlorines (OR = 1.01, 95% CI: 0.95, 1.07) or any of the three pesticides not
linked to breast cancer but detected in monitoring studies (Table 2.2). Null
associations were consistent when comparisons were made using the control
group consisting of all other cancer diagnoses, as well as control groups excluded
lung cancer diagnoses or ovarian, uterine and other reproductive cancer
diagnoses.
No statistically significant dose -response relationships were observed for
those in the high and low exposure categories (based on the median level among
63

the controls for total pounds of pesticide applied within the buffer surrounding
each location) compared to those unexposed to the group of organochlorines
(OR low = 1.03, 95% CI: 0.96, 1.10, OR high = 0.99, 95% CI: 0.93, 1.06, P = 0.74), or
any of the three other pesticides (Table 2.3). After accounting for 10 -year or 15 -
year latency periods, there were no associations for exposed versus unexposed
and for comparisons of low and high exposure categories compared to
unexposed. As a result, there did not appear to be any trend in risk with
increasing exposure levels after considering these latency periods. This was true
for all pesticides we considered.
2.4.1 Estrogen receptor breast cancer subtypes
When breast cancer cases were stratified by their ER subtypes, no
increased risk was observed for ER(+) breast cancer cases compared to controls
for all pesticides we considered , but there was an increased risk of ER( -) breast
cancer from exposure to 1,3 -dichloropropene (OR = 1.16, 95% CI: 1.03, 1.32)
(Table 2.4). There were positive associations among ER(-) breast cancer cases for
exposure to low levels of the organochlorines (OR = 1.17, 95% CI: 1.03, 1.35) and
low levels of 1,3-dichloropropene (OR = 1.26, 95% CI: 1.07, 1.49), but no increase
in risk for those exposed to high levels of these same chemicals, and no apparent
trend in risk with increasing exposure levels for any of the pesticides examined.
2.4.2 Geocode certainty
The majority of addresses at diagnosis were able to be matched by the
geocoding methods to a complete and valid address (high geocode certainty), and
a similar proportion was observed for both cases and controls (96.0% and 95.7%
64

respectively). When analyses were stratified by geocode certainty (Table 2.5), we
observed no association between breast cancer risk and all pesticides considered.
2.4.3 Young adult ages at exposure
When we examined ages 20 –39 at exposure, we observed an association
for exposure to chlorpyrifos among women exposed in their 20’s and 30’s
compared to those unexposed at the se ages (OR = 1.32, 95% CI: 1.01, 1.72),
however, the confidence intervals were wide and the finding wa s based on low
prevalence of exposure (6.2% of cases and 5.1% of contr ols) (Table 2.6). There
was no increase in risk for exposure to chlorpyrifos at ages 20’s or 30’s (OR =
0.90, 95% CI: 079, 1.02), and we did not observe an association for overall
exposure to chlorpyrifos at ages 20 –39 (OR = 0.94, 95% CI: 0.84, 1.06). We
found no associations for exposure to any other pesticide at ages 20–39.
2.4.4 Breast cancer cases compared to tax assessor parcel controls
There were statistically significant differences in risk of breast cancer
compared to the tax assessor parcels defined as controls for all of the selected
pesticides of interest, including organochorines ( OR = 1.65, 95% CI: 1.58, 1.74)
(Table 2.7). (Note that comparison with the other data in this Chapter is not
appropriate since the area from which the cases arose is not comparable and
different pesticides were selected for comparisons with the group of
organochlorines). For all pesticides considered, the magnitude of the estimate for
low exposure indicated increased risk of breast cancer, while the magnitude of
the estimate for high exposure was protective. For example, compared to those
unexposed, those with low exposure to organochlorines had a two- fold increased
risk of breast cancer (OR = 2.56, 95% CI: 2.42, 2.70) while those with high
65

exposure to organochlorines were inversely associated with breast cancer (OR =
0.74, 95% CI: 0.69, 0.79), and there was no statistically significant increasing
trend in risk for exposure to organochlorines (P = 0.69).

2.5 Discussion
Overall, no associations were found between breast cancer risk and the
selected group of organochlorines or any of the three pesticides that have not
been previously associated with breast cancer but were detected in local air
monitoring, even after taking 10 - and 15 -year latency periods into account. We
also did not observe associations between breast cance r risk and exposure to the
same pesticides at ages 20–39. Previous studies have suggested critical periods of
development such as puberty (32), or prior to the first pregnancy (83,84) may be
times of particular vulnerability to hormone -related chemicals such as
organochlorines. We were unable to examine exposures during these specific age
ranges or to account for exposures at the actual residences of subjects during
these time periods. Further research is needed to explore whether exposure to
pesticides at younger ages is associated with increased breast cancer risk.
In this registry -based study that used other cancer diagnoses as controls,
there is potential for selection bias towards the null if the control group included
those cancers that share a common etiology with breast cancer and are also
associated with exposure to the same pesticides. We attempted to account for this
bias by excluding lung, bronchus, ovarian, uterine and other female reproductive
cancers and did not see any effect on our odds ratios, however , we cannot be
66

completely sure that we eliminated all bias from other possible cancers that also
share underlying associations with breast cancer and exposure to pesticides.
Pesticide exposure in this study was based on the address at the time of
diagnosis to estimate a person’s entire lifetime exposure. We were not able to
evaluate cumulative exposures over known residential histories or to account for
residential mobility. In this highly agricultural region, we are more likely to miss
opportunities for pesticide exposure in the absence of all prior residential
information. Exposure for cases and controls based only on the address at the
time of diagnosis is likely to result in non-differential misclassification that would
bias our estimates towards the null and might account for the lack of associations
observed in this study.
2.5.1 Estrogen receptor breast cancer subtypes
When considering ER breast cancer subtype, no associations were
observed for ER(+) breast cancer for any pesticide considered here. There was a
16% increase in risk of ER( -) breast cancer from exposure to 1,3 -dichloropropene
(OR = 1.16, 95% CI: 1.- 3, 1.32) and positive associations among ER( -) breast
cancer and women exposed to low levels of organochlorines or 1,3 -
dichloropropene, but no increased risk for exposure to high levels of these same
chemicals and no evidence of dose response relationships, which does not
support an overall association between ER( -) breast cancer risk and these
chemicals. Based on in -vitro studies, ER(+) breast cancers are more likely to be
associated with hormone-related risk factors, however, not all studies have found
a difference in estimates of relative risk when stratified by subtype (74). The
organochlorines selected in these analyses have been shown to affect hormone
67

signaling, but the mechanisms may vary for different pesticides or for exposures
to combinations of pesticides. In studies done in vitro, aldrin, chlordane, dieldrin,
DDT, endosulfan, endrin, heptachlor, lindane, methoxychlor and toxaphene have
endocrine-disrupting effects, but different estrogenic affinities for ER(alpha) and
ER(beta) (17,19). Aldrin, dieldrin, endrin, endosulfan and methoxychlor were
also shown to b e weak estrogen receptor antagonists (19), which may be why we
did not observe any differential effects on the risk of ER(+) breast cancer
subtypes.
2.5.2 Geocode certainty
When we stratified by geocode certainty, the associations for all pesticides
were null regardless of whether the locations were of high or low geocode
certainty, however very few locations were of low geocode certainty (4.0% of
cases and 4.3% of controls). Geocoding methods used by cancer registries usually
process the address information in batches after the initial cancer reporting data
is received from hospitals or other facilities. Although efforts have been made to
standardize the geocoding process for interpolating location data collected by
cancer registries into spatial coordinates (85), there is variation in all levels of
geocoded locations depending on the level of detail included in the input dataset
as well as the ability of geocoding methods to pinpoint a location (86,87).
Although inaccuracies in the geocoded locations could have impacted the
association between breast cancer risk, the majority of locations of both cases and
controls were able to be geocoded to high geocode certainty (96.0% of cases and
95.7% of controls) and we controlled for geocode certainty as a potential
confounder in the analyses.
68

2.5.3 Breast cancer cases compared to tax assessor parcel controls
In an attempt to compare our registry -based breast cancer cases to
population-based controls that were not chosen from other cancer diagnoses in
the registry, we examined pesticide exposure in a random selection of tax
assessor parcels. We hoped that these tax assessor parcels would be more
representative of typical exposures found in the underlying Central Valley
population (based on proximity to pesticide application within a 500 -meter
buffer around each location) . We observed lower prevalence of exposure for the
pesticides of interest and each of the control pesticides compared to our breast
cancer cases. Without information on individual -level confounders for each tax
assessor parcel, we attempted to control for confounding by indivi dually
matching tax assessor parcels to each case, weighted for the proportion of census-
level age and race demographics. Other confounders such as SES were not
evaluated and may significantly differ between our cases and the randomly
selected parcels, biasing our estimates in any direction. For example, in the
absence of controlling for SES if the randomly selected tax assessor parcels were
more likely to be in areas of low SES that are also more likely to be rural parcels
near agricultural pesticide appli cation, then we might expect our estimates to be
biased towards the null; whereas, if the parcels were more likely to be in areas of
high SES that were urban neighborhoods not near pesticide use, then we might
expect our estimates to be biased away from the null.
We did not restrict the selection of tax assessor parcels to include only
residential types of properties and this could have substantially affected whether
the random selection of tax assessor parcels was representative of a random
69

selection of p ersons living in the area. This study was conducted in an
agriculturally intensive region and if the selected tax assessor parcels were more
likely to be in the mountain regions where few people reside, we would expect
fewer parcel-defined controls to be exposed than we might find in the underlying
population and our estimates would be biased away from the null. A map of the
randomly selected tax assessor parcel centroids from California’s Central Valley
shows that many centroids cluster around town centers but there are also many
centroids scattered in mountainous areas and near freeways that are less likely to
be residences near commercial farming areas than the areas from which our cases
arose (Figure 2.2). A more accurate comparison group for this study would be to
use population-based controls derived from the same geographic area.
2.5.4 Strengths
This study was able to examine exposure to specific pesticides thought to
mimic hormones in a large number of breast cancer cases from a population-
based registry in an agriculturally intensive area. The GIS -based approach
utilized detailed pesticide repor ting and land- use surveys to determine an
individual’s proximity to specific pesticides of interest, rather than assign
individual exposure categories based on subject recall (31,47,88), or aggregated
geographical and pesticide data (33,36,45,46).
2.5.5 Limitations
In general, registry -based studies have limited data on potential
confounders. Many breast can cer risk factors, including use of hormone
replacement therapy, age at first pregnancy and weight, are associated with
higher education, income and SES. Urban areas with high SES tend to have
70

higher rates of breast cancer, while rural areas with high pesticide exposure have
lower rates of other established breast cancer risk factors (49). W ithout
controlling for these confounding factors associated with residential location that
favor lower rates of breast cancer in rural agricultural areas, the associations
between pesticides and breast cancer risk are likely to be biased towards the null.
In addition, SES influences behavioral risk factors for breast cancer such as
tobacco smoking, alcohol intake and physical inactivity, and these were not
accounted for in this study. Inability to control for known or unknown
confounders can affect estimate s of relative risk in unknown ways, either
attenuating the risks or leading to spurious results.
Breast cancer also has a well -established screening program and is more
likely to be diagnosed in women of high SES and at earlier stages than the other
types of cancer diagnoses used as controls in this study. When we adjusted for
SES and stage of the cancer at the time of diagnosis however, we found no change
in our estimates. The registry SES variable is derived from 1990 census -block
level data at the resid ence of diagnosis and may not be accurate compared to
individual measures of SES such as education and income gathered directly from
patients, however this neighborhood measure of SES was developed for breast
cancer incidence in California and shows that S ES is related to breast cancer
incidence. A recent study found that the SES variable by Yost et al. used in our
analyses adequately captures socioeconomic gradients in breast cancer incidence
(89).
Our GIS-based exposure assessment for pesticides is based on California
PUR data, whic h is the most complete pesticide reporting system in the nation;
71

however, approximately 37% of PUR data is incomplete and could not be
incorporated into our model. Another limitation of PUR data is that we only have
information on restricted pesticide use since 1974 and for all pesticide application
since 1990. In the analyses we ignored exposures occurring prior to 1974, but we
also examined exposures occurring before 1974 under the assumption that
pesticide applications before 1974 were the same as those applied during the year
1974 and this produced nearly identical estimates to those that ignored pesticide
exposure prior to 1974. We observed that there were fewer pesticide applications
reported in 1974, the first year PUR data is available, compared to o ther years,
and this is possibly due to poor compliance and record keeping in the early years
of the state -mandated reporting system. Assuming that incomplete pesticide
reporting data is not differential by cancer type, exposure misclassification would
be expected to bias our odds ratios towards the null.
Lack of listed residential address at the time of diagnosis may also miss
identification of pesticide exposure. In the registry, 1.6% of breast cancer cases
and 2.3% of other types of cancer diagnosed controls were missing address
information. We compared those with missing addresses to those with known
addresses by demographics to see if those missing were different by certain
characteristics and whether these differences were similar for cases and controls
(Table 2.1). A similar proportion of non -Hispanic white and Hispanic subjects
were missing addresses as had addresses listed, and more subjects with other or
unknown race were missing addresses than were known. Compared to controls,
fewer Hispanic cases and more cases of other ethnicity were missing addressees.
Subjects in the lowest SES quintile were more likely to be missing addresses, and
72

this was true for both cases and controls. M ore controls in the lowest SES and
more cases in the highest SES had missing addresses.
Although the CRCC rigorously reviews and cleans its data, we cannot rule
out misclassification of outcome based on error recorded in the registry for tumor
or demographic characteristics. This error is not likely to differ by cancer type so
misclassification would be non -differential and bias the estimates towards the
null.
2.5.6 Conclusions and future work
Recent advancements in GIS -based methodologies have made it possible
to explore historical exposures by proxy methods utilizing state pesticide
reporting data and individual addresses. Our GIS-based approach was able to
evaluate specific pesticides thought to be important in breast cancer etiology with
over thirty years of documented pesticide use history. The next step would be to
employ this model in a case -control study comparing breast cancer cases to
population-based controls from the same highly agricultural region where the
prevalence of exposure to specific pesticides of interest is sufficient to be able to
detect an association if one truly exists. This type of detailed and specific model
of pesticide exposure, combined with actual residential and occupation address
histories, would allow us to examine lifetime expo sure on the risk of breast
cancer and to assess critical time periods of exposure that may be most pertinent
to the development of breast cancer decades later.
In this non -contact study, we utilized demographic and diagnostic
information from a large numbe r of cancer patients collected for a population -
based cancer registry. Cancer registries are a valuable source of data for research
73

that aims to assess causes and treatments of cancer, however, the level of patient
detail included in them is typically not sufficient for conducting epidemiologic
studies without further patient contact or linkage to other existing datasets. In
the registry -based study presented here for example, we would ideally want to
collect residential histories on all registry participan ts. Future efforts could be
made to collect this information from patients in the cancer registry through
follow-up interviews by registry staff or by linking cancer registry data to
residential history information collected by credit reporting agencies. Linking to
existing data such as that collected by credit reporting agencies is more cost -
effective than trying to collect this information through interviews and reduces
the potential for selection bias resulting from patients who do not respond or who
are unwilling or unable to provide the information. Given that two goals of
population-based cancer registries are cancer prevention and research, efforts
should be made to improve upon the information collected by registries as well as
develop new mechanisms to expand health data while adhering to patient privacy
laws. Cancer patients, as well as the general public, have much to gain from the
identification of new cancer prevention measures and new developments in
medical research that are possible with improving the capacity of the cancer
registry data.
74

Table 2.1. Comparison of Characteristics of Breast Cancer Cases and Other Types of Cancer
Controls in Fresno, Tulare, and Kern Counties, 1988–2012.

Address at Diagnosis

Missing Address at Diagnosis
Characteristic
Cases

Controls

Cases Controls
(n = 9,391)

(n = 21,604)

(n = 149)

(n = 502)
No. % No. %

No. % No. %
Age at Diagnosis, years
55 – 59 2,339 24.9 4,360 20.2 42 28.2 109 21.7
60 – 64 2,366 25.2 5,133 23.8 44 29.5 135 26.9
65 – 69 2,446 26.1 5,985 27.7 31 20.8 122 24.3
70 – 74 2,240 23.9 6,126 28.4 32 21.5 136 27.1
Stage
Localized 5,885 62.7 6,966 32.2 75 50.3 149 29.7
Regional 2,828 30.1 4,543 21.0 29 19.5 66 13.2
Distant or Missing 678 7.2 10,095 46.7 45 30.2 287 57.2
Birthplace
United States 5,418 57.7 12,881 59.6 47 31.5 208 41.4
Other 3,964 42.2 8,660 40.1 102 68.5 293 58.4
Missing 9 0.1 63 0.3 0 0.0 1 0.2
Race/Ethnicity
Non-Hispanic White 7,149 76.1 15,437 71.5 95 63.8 306 61.0
Hispanic 1,512 16.1 4,157 19.2 19 12.8 97 19.3
Other or Unknown 730 7.8 2,010 9.3 35 23.5 99 19.7
SES
1 Lowest Quintile 3,003 32.0 8,442 39.1 62 41.6 227 45.2
2 2,640 28.1 5,947 27.5 39 26.2 140 27.9
3 1,894 20.2 3,998 18.5 30 20.1 81 16.1
4 1,463 15.6 2,531 11.7 13 8.7 45 9.0
5 Highest Quintile 391 4.2 686 3.2 5 3.4 9 1.8

75

Table 2.2. Prevalence of Exposure and Measures of Association for Women Diagnosed with Breast Cancer Compared to Several Types of Cancer
Registry Control Groups, Including All Women Diagnosed at Sites Other Than the Breast, All Diagnoses Except Lung Cancers, and All Diagnoses
Except Ovarian, Uterine and Other Reproductive Cancers, Fresno, Tulare, and Kern Counties, 1988–2012.

Cases
(n = 9,540)
All Controls
(n = 22,106)
Controls, Excluding Lung Cancers
(n = 17,250)
Controls, Excluding Ovarian, Uterine,
and Reproductive Cancers
(n = 18,762)
Exposure Type No. % No. % OR* 95% CI No. % OR* 95% CI No. % OR* 95% CI
Missing† 149 1.6 502 2.3

437 2.5

442 2.4

Organochlorines‡

Unexposed 3,272 34.8 7,471 34.6 1.00 --- 5,682 33.8 1.00 ---

6,411 35.0 1.00 ---
Exposed 6,119 65.2 14,133 65.4 1.01 0.95, 1.07 11,131 66.2 1.00 0.94, 1.06

11,909 65.0 1.03 0.97, 1.10
Chlorpyrifos

Unexposed 4,322 46.0 10,129 46.9 1.00 --- 7,724 45.9 1.00 ---

8,650 47.2 1.00 ---
Exposed 5,069 54.0 11,475 53.1 1.04 0.99, 1.10 9,089 54.1 1.02 0.96, 1.08

9,670 52.8 1.05 1.00, 1.12
Diazinon

Unexposed 3,571 38.0 8,327 38.5 1.00 --- 6,305 37.5 1.00 ---

7,106 38.8 1.00 ---
Exposed 5,820 62.0 13,277 61.5 1.03 0.97, 1.08 10,508 62.5 1.00 0.94, 1.06

11,214 61.2 1.03 0.97, 1.09
1,3-Dichloropropene

Unexposed 7,310 77.8 16,770 77.6 1.00 --- 12,920 76.9 1.00 ---

14,223 77.6 1.00 ---
Exposed 2,081 22.2 4,834 22.4 1.03 0.96, 1.09 3,893 23.2 1.00 0.94, 1.07 4,097 22.4 1.03 0.96, 1.10
* Adjusted for age, race, SES, birthplace, cancer stage, and geocode certainty
† Missing observations had no exposure data for the period 1974–2012
‡ Aldrin, chlordane, dicofol, dieldrin, endosulfan, lindane, methoxychlor, and toxaphene
76

Table 2.3. Measures of Association between Breast Cancer and Exposure to Selected Pesticides Based on Residential Address at the Time of
Diagnosis Using Linked PUR and Land-Use Data for 1974–2012, in Fresno, Tulare, and Kern Counties, Assuming 10-Year or 15-Year Latency
Periods or No Latency Period.
Exposure Type
10-Year Latency 15-Year Latency
Cases
(n = 9,540)
Controls
(n = 22,106)

Cases
(n = 9,540)
Controls
(n = 22,106)

Cases
(n = 9,540)
Controls
(n = 22,106)

No. % No. % OR* 95% CI No. % No. % OR* 95% CI No. % No. % OR* 95% CI
Missing† 149 1.6 502 2.3

149 1.6 502 2.3

149 1.6 502 2.3

Organochlorines‡

Unexposed 3,272 34.8 7,471 34.6 1.00 --- 3,579 38.1 8,106 37.5 1.00 --- 4,001 42.6 9,239 42.8 1.00 ---
Exposed 6,119 65.2 14,133 65.4 1.01 0.95, 1.07 5,812 61.9 13,498 62.5 1.00 0.94, 1.05 5,390 57.4 12,365 57.2 1.02 0.96, 1.08
Low 3,091 32.9 7,066 32.7 1.03 0.96, 1.10 2,909 31.0 6,749 31.2 1.00 0.93, 1.06 2,731 29.1 6,182 28.6 1.03 0.97, 1.10
High 3,028 32.2 7,067 32.7 0.99 0.93, 1.06 2,903 30.9 6,749 31.2 1.00 0.93, 1.06 2,659 28.3 6,183 28.6 1.00 0.94, 1.07

Trend P = 0.74

Trend P = 0.91

Trend P = 0.85
Chlorpyrifos
Unexposed 4,322 46.0 10,129 46.9 1.00 --- 5,400 57.5 12,609 58.4 1.00 --- 6,387 68.0 14,791 68.5 1.00 ---
Exposed 5,069 54.0 11,475 53.1 1.04 0.99, 1.10 3,991 42.5 8,995 41.6 1.02 0.97, 1.08 3,004 32.0 6,813 31.5 0.99 0.94, 1.05
Low 2,515 26.8 5,737 26.6 1.05 0.98, 1.12 1,982 21.1 4,497 20.8 1.03 0.97, 1.11 1,489 15.9 3,406 15.8 1.00 0.93, 1.08
High 2,554 27.2 5,738 26.6 1.04 0.97, 1.10 2,009 21.4 4,498 20.8 1.00 0.94, 1.08 1,515 16.1 3,407 15.8 0.99 0.92, 1.06

Trend P = 0.25

Trend P = 0.74

Trend P = 0.73
Diazinon

Unexposed 3,571 38.0 8,327 38.5 1.00 --- 3,796 40.4 8,820 40.8 1.00 --- 4,251 45.3 9,976 46.2 1.00 ---
Exposed 5,820 62.0 13,277 61.5 1.03 0.97, 1.08 5,595 59.6 12,784 59.2 1.02 0.97, 1.08 5,140 54.7 11,628 53.8 1.01 0.96, 1.07
Low 2,840 30.2 6,638 30.7 1.04 0.98, 1.11 2,730 29.1 6,392 29.6 1.03 0.97, 1.10 2,505 26.7 5,814 26.9 1.03 0.96, 1.10
High 2,980 31.7 6,639 30.7 1.02 0.95, 1.08 2,865 30.5 6,392 29.6 1.01 0.95, 1.08 2,635 28.1 5,814 26.9 1.00 0.94, 1.07

Trend P = 0.62

Trend P = 0.71

Trend P = 0.90

77

Table 2.3. Continued:
Exposure Type
10-Year Latency 15-Year Latency
Cases
(n = 9,540)
Controls
(n = 22,106)

Cases
(n = 9,540)
Controls
(n = 22,106)

Cases
(n = 9,540)
Controls
(n = 22,106)

No. % No. % OR* 95% CI No. % No. % OR* 95% CI No. % No. % OR* 95% CI
1,3-Dichloropropene
Unexposed 7,310 77.8 16,770 77.6 1.00 --- 7,890 84.0 18,143 84.0 1.00 --- 8,309 88.5 18,992 87.9 1.00 ---
Exposed 2,081 22.2 4,834 22.4 1.03 0.96, 1.09 1,501 16.0 3,461 16.0 1.01 0.94, 1.08 1,082 11.5 2,612 12.1 0.96 0.88, 1.04
Low 1,039 11.1 2,417 11.2 1.03 0.95, 1.13 719 7.7 1,730 8.0 0.98 0.89, 1.08 508 5.4 1,306 6.1 0.93 0.83, 1.04
High 1,042 11.1 2,417 11.2 1.02 0.94, 1.11 782 8.3 1,731 8.0 1.04 0.94, 1.14 574 6.1 1,306 6.1 0.99 0.88, 1.10

Trend P = 0.52

Trend P = 0.62

Trend P = 0.48
* Adjusted for age, race, SES, birthplace, cancer stage, and geocode certainty
† Missing observations had no exposure data for the period 1974–2012
‡ Aldrin, chlordane, dicofol, dieldrin, endosulfan, lindane, methoxychlor, and toxaphene
78

Table 2.4. Measures of Association for Estrogen Receptor Breast Cancer Subtypes and Exposure to Selected Pesticides Based on Residential
Address at the Time of Diagnosis Using Linked PUR and Land-Use Data for 1974–2012, in Fresno, Tulare, and Kern Counties.*

No.
Controls
ER (-) Cases ER (+) Cases ER (Status Unknown) Cases
No. OR† 95% CI No. OR† 95% CI No. OR† 95% CI
(n = 19,622) (n = 1,475) (n = 5,798) (n = 1,252)
Missing‡ 414 14 62 45
Organochlorines§
Unexposed 6,526 456 1.00 --- 2,002 1.00 --- 430 1.00 ---
Exposed 12,682 1,005 1.13 1.00, 1.27 3,734 0.98 0.91, 1.05 777 0.99 0.87, 1.12
Low 6,341 515 1.17 1.03, 1.35 1,909 1.01 0.93, 1.09 353 0.91 0.78, 1.05
High 6,341 490 1.09 0.95, 1.25 1,825 0.95 0.88, 1.03 424 1.06 0.92, 1.23
Trend P = 0.27 Trend P = 0.19 Trend P = 0.41
Chlorpyrifos
Unexposed 8,680 645 1.00 --- 2,549 1.00 --- 545 1.00 ---
Exposed 10,528 816 1.02 0.91, 1.14 3,187 1.04 0.987, 1.11 662 1.06 0.94, 1.19
Low 5,264 417 1.05 0.92, 1.20 1,558 1.03 0.95, 1.12 311 1.00 0.87, 1.16
High 5,264 399 0.98 0.86, 1.12 1,629 1.05 0.97, 1.13 351 1.12 0.97, 1.29
Trend P = 0.86 Trend P = 0.23 Trend P = 0.16
Diazinon
Unexposed 7,161 539 1.00 --- 2,052 1.00 --- 490 1.00 ---
Exposed 12,047 922 0.99 0.88, 1.11 3,684 1.06 1.00, 1.14 717 0.93 0.83, 1.06
Low 6,023 447 0.99 0.87, 1.13 1,772 1.07 0.99, 1.16 374 0.99 0.86, 1.14
High 6,024 475 0.99 0.86, 1.13 1,912 1.06 0.98, 1.15 343 0.88 0.76, 1.02
Trend P = 0.83 Trend P = 0.13 Trend P = 0.10
1,3-Dichloropropene
Unexposed 14,783 1,085 1.00 --- 2,002 1.00 --- 430 1.00 ---
Exposed 4,425 376 1.16 1.03, 1.32 3,734 1.00 0.93, 1.08 777 0.92 0.80, 1.07
Low 2,212 204 1.26 1.07, 1.49 1,909 0.96 0.86, 1.06 353 1.02 0.84, 1.23
High 2,213 172 1.07 0.90, 1.27 1,825 1.05 0.95, 1.16 424 0.83 0.67, 1.02
Trend P = 0.10 Trend P = 0.59 Trend P = 0.12
* Only breast cancer cases and other cancer controls diagnosed between 1991 and 2012 were included in these analyses since these were the years estrogen receptor status was reported
by the cancer registry.
† Adjusted for age, race, SES, birthplace, cancer stage, and geocode certainty
‡ Missing observations had no exposure data for the period 1974–2012
§ Aldrin, chlordane, dicofol, dieldrin, endosulfan, lindane, methoxychlor, and toxaphene
79

Table 2.5. Measures of Association Between Breast Cancer and Exposure to Selected Pesticides Based on Residential Address at the Time of
Diagnosis Using Linked PUR and Land-Use Data for 1974–2012, in Fresno, Tulare, and Kern Counties, by Geocode Certainty of the Residential
Address at Diagnosis.*
Exposure Type
High Geocode Certainty Low Geocode Certainty
Cases
(n = 8,896)
Controls
(n = 20,324)

Cases
(n = 369)
Controls
(n = 913)

No. % No. % OR† 95% CI No. % No. % OR† 95% CI
Organochlorines‡
Unexposed 3,000 33.7 6,788 33.4 1.00 --- 190 51.5 452 49.5 1.00 ---
Exposed 5,896 66.3 13,536 66.6 1.01 0.96, 1.07 179 48.5 461 50.5 0.92 0.69, 1.22
Chlorpyrifos
Unexposed 4,036 45.4 9,374 46.1 1.00 --- 205 55.6 510 55.9 1.00 ---
Exposed 4,860 54.6 10,950 53.9 1.04 0.99, 1.10 164 44.4 403 44.1 1.00 0.76, 1.33
Diazinon
Unexposed 3,306 37.2 7,630 37.5 1.00 --- 183 49.6 457 50.0 1.00 ---
Exposed 5,590 62.8 12,694 62.5 1.03 0.97, 1.09 186 50.4 456 50.0 1.00 0.76, 1.32
1,3-Dichloropropene
Unexposed 6,907 77.6 15,716 77.3 1.00 --- 297 80.5 735 80.5 1.00 ---
Exposed 1,989 22.4 4,608 22.7 1.03 0.96, 1.10 72 19.5 178 19.5 0.96 0.68, 1.35
* Missing observations had no exposure data for the period 1974–2012 due to missing location information (149 cases and 502 controls) or
inability to be geocoded (126 cases and 367 controls).
† Adjusted for age, race, SES, birthplace and cancer stage
‡ Aldrin, chlordane, dicofol, dieldrin, endosulfan, lindane, methoxychlor, and toxaphene

80

Table 2.6. Measures of Association Between Breast Cancer and Exposure to Selected Pesticides at
Ages 20–39 Based on Residential Address at the Time of Diagnosis Using Linked PUR and Land-
Use Data for 1974–2012, in Fresno, Tulare, and Kern Counties.
Exposure Type
Cases
(n = 1,987)

Controls
(n = 4,089)

OR* 95% CI
No. % No. %
Missing† 39 2.0 123 3.0

Organochlorines‡

Ages 20–39
Unexposed ages 20–39 661 33.9 1,381 34.8 1.00 ---
Exposed ages 20–39 1,287 66.1 2,585 65.2 1.00 0.89, 1.14

Unexposed at ages 20's or 30's 661 33.9 1,381 34.8 1.00 ---
Exposed at ages 20's or 30's 492 25.3 951 24.0 1.07 0.91, 1.25
Exposed at ages 20's & 30's 795 40.8 1,634 41.2 0.97 0.84, 1.11

Chlorpyrifos

Ages 20–39
Unexposed ages 20–39 1,117 57.3 2,250 56.7 1.00 ---
Exposed ages 20–39 831 42.7 1,716 43.3 0.94 0.84, 1.06

Unexposed at ages 20's or 30's 1,117 57.3 2,250 56.7 1.00 ---
Exposed at ages 20's or 30's 710 36.5 1,515 38.2 0.90 0.79, 1.02
Exposed at ages 20's & 30's 121 6.2 201 5.1 1.32 1.01, 1.72

Diazinon

Ages 20–39
Unexposed ages 20–39 677 34.8 1,444 36.4 1.00 ---
Exposed ages 20–39 1,271 65.3 2,522 63.6 1.01 0.89, 1.14

Unexposed at ages 20's or 30's 677 34.8 1,444 36.4 1.00 ---
Exposed at ages 20's or 30's 516 26.5 1,035 26.1 1.01 0.87, 1.18
Exposed at ages 20's & 30's 755 38.8 1,487 37.5 1.00 0.87, 1.15

1,3-Dichloropropene

Ages 20–39
Unexposed ages 20–39 1,628 83.6 3,338 84.2 1.00 ---
Exposed ages 20–39 320 16.4 628 15.8 1.06 0.90, 1.24

Unexposed at ages 20's or 30's 1,628 83.6 3,338 84.2 1.00 ---
Exposed at ages 20's or 30's 300 15.4 568 14.3 1.09 0.93, 1.29
Exposed at ages 20's & 30's 20 1.0 60 1.5 0.71 0.41, 1.21

* Adjusted for age, race, SES, birthplace, cancer stage, and geocode certainty
† Missing observations had no exposure data for the period 1974–2012
‡ Aldrin, chlordane, dicofol, dieldrin, endosulfan, lindane, methoxychlor, and toxaphene

81

Table 2.7. Measures of Association Between Breast Cancer and Exposure to Selected Pesticides
Based on Residential Address at the Time of Diagnosis Among Women in Fresno, Tulare, and
Kern Counties Using Linked PUR and Land-Use Data for 1974–2012 Compared to a Random
Selection of Tax-Assessor Parcels.*
Exposure Type
Cases
(n = 14,569)

Parcel
Controls
(n = 14,569)

Crude
OR
95% CI
No. % No. %
Organochlorines†

Unexposed 5,285 36.3 7,037 48.3 1.00 ---
Exposed 9,284 63.7 7,532 51.7 1.65 158, 1.74
Low 7,219 49.6 3,766 25.9 2.56 2.42, 2.70
High 2,065 14.2 3,766 25.9 0.74 0.69, 0.79
Trend P = 0.69
Simazine

Unexposed 7,918 54.4 9,134 62.7 1.00 ---
Exposed 6,651 45.7 5,435 37.3 1.41 1.35, 1.48
Low 4,783 32.8 2,716 18.6 2.04 1.93, 2.16
High 1,868 12.8 2,719 18.7 0.80 0.74, 9.85
Trend P < 0.01
Maneb

Unexposed 10,740 73.7 11,470 78.7 1.00 ---
Exposed 3,829 26.3 3,099 21.3 1.32 1.25, 1.39
Low 2,631 18.1 1,549 10.6 1.83 1.70, 1.96
High 1,198 8.2 1,550 10.6 0.82 0.76, 0.89
Trend P < 0.01
Paraquat dichloride

Unexposed 4,916 33.7 6,739 46.3 1.00 ---
Exposed 9,653 66.3 7,830 53.7 1.71 1.63, 1.79
Low 7,401 50.8 3,915 26.9 2.58 2.44, 2.73
High 2,252 15.5 3,915 26.9 0.79 0.74, 0.84
Trend P = 0.21
* Breast cancer cases included all those with invasive breast cancer diagnosed in the California
Central Valley region (Fresno, Kern, Kings, Madera, Merced, Stanislaus, Tulare, and Tuolumne
Counties) individually matched to a random selection of tax-assessor parcels from the Central
Valley, weighted for the proportion of census-level age and race demographics of the case.
† Aldrin, chlordane, dicofol, dieldrin, endosulfan, lindane, methoxychlor, and toxaphene
82

Figure 2.1. Map of Approximate Residential Locations* at the Time of Diagnosis
for Women with Breast Cancer and Other Types of Cancer as Controls in Fresno,
Tulare, and Kern Counties in California, 1988–2012.

* Mapped points have been masked to randomly selected distances within 500
meters from the geocoded location at the time of diagnosis or interview to protect
participants’ privacy.

83

Figure 2.2. Map of Approximate Residential Locations* at the Time of Diagnosis
for Women with Breast Cancer in 1988–2009 and a Random Selection of 10,000
Tax Assessor Parcel Centroids in California’s Central Valley.†

*Mapped points have been masked to randomly selected distances within 500
meters from the geocoded location at the time of diagnosis or interview to protect
participants’ privacy.
† California’s Central Valley includes the counties of Fresno, Kern, Kings,
Madera, Mariposa, Merced, Stanislaus, Tulare and Tuolumne.

84

Chapter 2 References

1. Madigan MP, Ziegler RG, Benichou J, et al. Proportion of breast cancer
cases in the United States explained by well -established risk factors. J Natl
Cancer Inst. 1995;87(22):1681–1685.
2. Althuis MD, Fergenbaum JH, Garcia -Closas M, et al. Etiology of horm one
receptor-defined breast cancer: a systematic review of the literature.
Cancer Epidemiol Biomarkers Prev. 2004;13(10):1558–1568.
3. Colditz GA. Epidemiology of breast -cancer. F indings from the Nurses '
Health Study. Cancer. 1993;71(4):1480–1489.
4. Ma H, Henderson KD, Sullivan -Halley J, et al. Pregnancy -related factors
and the risk of breast carcinoma in situ and invasive breast cancer among
postmenopausal women in the California Teachers Study cohort. Breast
Cancer Res. 2010;12(3):R35.
5. Tamimi RM, H ankinson SE, Chen WY, et al. Combined estrogen and
testosterone use and risk of breast cancer in postmenopausal women. Arch
Intern Med. 2006;166(14):1483–1489.
6. Key T, Appleby P, Barnes I, et al. Endogenous sex hormones and breast
cancer in postmenopausal women: reanalysis of nine prospective studies. J
Natl Cancer Inst. 2002;94(8):606–616.
7. Key T, Pike MC. The role of oestrogens and progestagens in the
epidemiology and prevention of breast cancer. Eur J Cancer Prev.
1988;24(1):29–43.
8. Pike MC, Spicer DV, Dahmoush L, et al. Estrogens, progestogens, normal
breast cell proliferation, and breast cancer risk. Epidemiol Rev.
1993;15(1):17–35.
9. Liehr JG. Dual role of oestrogens as hormones and pro- carcinogens:
tumour initiation by metabolic activation of oestrogens. Eur J Cancer Prev.
1997;6(1):3–10.
10. California Department of Pesticide Regulation. A Guide to Pesticide
Regulation in California. Sacramento, CA: California Environmental
Protection Agency; 2011. (http://www.cdpr.ca.gov/docs/pressrls/dprguide.
htm). (Accessed May 31, 2013).
11. Soto AM, Chung KL, Sonnenschein C. The pesticides endosulfan,
toxaphene, and dieldrin have estrogenic effects on human estrogen -
sensitive cells. Environ Res. 1994;102:380–383.
85

12. Phillips KP, Foster WG, Leiss W, et al. Assessing and managing risks
arising from exposure to endocrine -active chemicals. J Toxicol Environ
Health B Crit Rev. 2008;11(3-4):351–372.
13. Rogan WJ, Gladen BC, McKinney JD, et al. Polychlorinated biphenyls
(PCBs) and dichlorodiphenyl dichloroethene (DDE) in human milk: effects
on growth, morbidity, and duration of lactation. Am J Public Health.
1987;77(10):1294–1297.
14. Falck FJ, Ricci AJ, Wolff MS, et al. Pesticides and polychlorinated biphenyl
residues in human breast lipids and their relation to breast cancer. Arch
Environ Health. 1992;47(2):143–146.
15. Kaushik PP, Kaushik GG. An assessment of structure and toxicity
correlation in organochlorine pesticides. J Hazard Mater. 2007;143(1 -
2):102–111.
16. Mukherjee S, Koner BC, Ray S, et al. Environmental contaminants in
pathogenesis of breast cancer. Indian J Exp Biol. 2006;44(8):597–617.
17. Briz V, Molina- Molina J -M, Sánchez- Redondo S, et al. Differential
estrogenic effects of the persistent organochlorine pesticides dieldrin,
endosulfan, and lindane in primary neuronal cultures. Toxicol Sci.
2011;120(2):413–427.
18. Rattenborg T, Gjermandsen I, Bonefeld -Jørgensen E. Inhibition of E2 -
induced expression of BRCA1 by persistent organochlorines. Breast Cancer
Res. 2002;4(6):R12.
19. Lemaire G, Mnif W, Mauvais P, et al. Activation of alpha - and beta -
estrogen receptors by persistent pesticides in reporter cell lines. Life Sci.
2006;79(12):1160–1169.
20. Kortenkamp A. Breast cancer, oestrogens and environmental pollutants: a
re-evaluation from a mixture perspective. Int J Androl. 2006;29(1):193 –
198.
21. Valerón PF, Pestano JJ, Luzardo OP, et al. Differential effects exerted on
human mammar y epithelial cells by environmentally relevant
organochlorine pesticides either individually or in combination. Chem Biol
Interact. 2009;180(3):485–491.
22. International Agency for Research on Cancer. IARC Monographs on the
Evaluation of Carcinogenic Risk of Chemicals to Humans. Agents
classified by the IARC monographs. 2010;1-100. (http://monographs.iarc.
fr/ENG/Classification/ClassificationsAlphaOrder.pdf). (Accessed May 29,
2013).
86

23. International Agency for Research on Cancer. IARC Monographs on the
Evaluation of Carcinogenic Risk of Chemicals to Humans. Occupational
exposures in insecticide application, and some pesticides. 1999;53.
(http://monographs.iarc.fr/ENG/Monographs/vol53/volume53.pdf).
(Accessed May 31, 2013).
24. Rudel RA, Attfield KR, Sch ifano JN, et al. Chemicals causing mammary
gland tumors in animals signal new directions for epidemiology, chemicals
testing, and risk assessment for breast cancer prevention. Cancer.
2007;109(S12):2635–2666.
25. Linko PP, Yeowell HNH, Gasiewicz TAT, et a l. Induction of cytochrome P -
450 isozymes by hexachlorobenzene in rats and aromatic hydrocarbon
(Ah)-responsive mice. J Biochem Toxicol. 1986;1(2):95–107.
26. Coumoul X, Diry M, Robillot C, et al. Differential regulation of cytochrome
P450 1A1 and 1B1 by a combination of dioxin and pesticides in the breast
tumor cell line MCF-7. Cancer. 2001;61(10):3942–3948.
27. Cassidy RAR, Natarajan SS, Vaughan GMG. The link between the
insecticide heptachlor epoxide, estradiol, and breast cancer. Breast Cancer
Res Treat. 2005;90(1):55–64.
28. García MA, Peña D, Álvarez L, et al. Hexachlorobenzene induces cell
proliferation and IGF -I signaling pathway in an estrogen receptor alpha -
dependent manner in MCF -7 breast cancer cell line. Toxicol Lett.
2010;192(2):195–205.
29. Iscan MM, Coban TT, Cok II, et al. The organochlorine pesticide residues
and antioxidant enzyme activities in human breast tumors: is there any
association? Breast Cancer Res Treat. 2002;72(2):173–182.
30. Cockburn M, Mills P, Zhang X, et al. Prostate ca ncer and ambient pesticide
exposure in agriculturally intensive areas in California. Am J Ind Med.
2011;173(11):1280–1288.
31. Teitelbaum SL, Gammon MD, Britton JA, et al. Reported residential
pesticide use and breast cancer risk on Long Island, New York. Am J Ind
Med. 2007;165(6):643–651.
32. Cohn BA, Wolff MS, Cirillo PM, et al. DDT and breast cancer in young
women: new data on the significance of age at exposure. Environ Res.
2007;115(10):1406–1414.
33. Reynolds P, Hurley SE, Gunier RB, et al. Residen tial proximity to
agricultural pesticide use and incidence of breast cancer in California,
1988–1997. Environ Res. 2005;113(8):993–1000.
87

34. Reynolds P, Hurley SE, Goldberg DE, et al. Residential proximity to
agricultural pesticide use and incidence of breast cancer in the California
Teachers Study cohort. Environ Res. 2004;96(2):13–13.
35. O'Leary ESE, Vena JEJ, Freudenheim JLJ, et al. Pesticide exposure and
risk of breast cancer: a nested case -control study of residentially stable
women living on Long Island. Environ Res. 2004;94(2):134–144.
36. Muir K, Rattanamongkolgul S, Smallman- Raynor M, et al. Breast cancer
incidence and its possible spatial association with pesticide application in
two counties of England. Annu Rev Public Health. 2004;118(7):513–520.
37. Brody JG, Aschengrau A, McKelvey W, et al. Breast cancer risk and
historical exposure to pesticides from wide -area applications assessed with
GIS. Environ Res [electronic article]. 2004;112(8):889–897.
38. Brody JG, Vorhees DJ, Melly SJ, et al. Using GIS and historical records to
reconstruct residential exposure to large-scale pesticide application. J Expo
Anal Environ Epidemiol. 2002;12(1):64–80.
39. Gammon MD, Wolff MS, Neugut AI, et al. Environmental toxins and breast
cancer on Long Island. II. organochlorine compound levels in blood.
Cancer Epidemiol Biomarkers Prev. 2002;11:686–697.
40. California Department of Pesticide Regulation. Summary of Pesticide Use
Report Data–2010. Sacramento, CA: California Environmental Protection
Agency; 2011. (http://www.cdpr.ca.gov/docs/pur/pur10rep/10sum.htm#
Development). (Accessed June 11, 2013).
41. Luo Y, Zhang M. Spatially distributed pesticide exposure assessment in the
Central Valley, California, USA. Envir Pollut. 2010;158(5):1629–1637.
42. Pfleeger TGT, Olszyk DD, Burdick CAC, et al. Using a geographic
information system to identify areas with potential for off -target pesticide
exposure. Environ Toxicol Chem. 2006;25(8):2250–2259.
43. Brouwer DH, EJ B, van Hemmen JJ. Estimation of long- term exposure to
pesticides. Am J Ind Med. 1994;25(4):573–588.
44. Russell HH, Jackson RJ, Spath DP, et al. Chemical contamination of
California drinking water. West J Med. 1987;147(5):615.
45. Mills PK, Yang R. Regression analysis of pesticide use and breast cancer
incidence in California Latinas. J Environ Health. 2006;68(6):15–14.
46. Mills PKP, Yang RR. Breast cancer risk in Hispanic agricultural workers in
California. Int J Occup Environ Health. 2005;11(2):123–131.
88

47. Engel LS, Hill DA, Hoppin JA, et al. Pesticide use and breast cancer risk
among farmers' wives in the Agricultural Health Study. Am J Ind Med.
2005;161(2):121–135.
48. Rull RP, Ritz B. Historical pesticide exposure in California using pesticide
use reports and land- use surveys: an assess ment of misclassification error
and bias. Environ Res. 2003;111(13):1582–1589.
49. Reynolds P, Hurley S, Goldberg DE, et al. Regional variations in breast
cancer among california teachers. Epidemiology. 2004;15(6):746–754.
50. California Cancer Registry. Cancer Reporting in California: Standards for
Automated Reporting. California Cancer Reporting System Standards for
2012 Volume II. Sacramento, CA: California Department of Public Health;
2012. (http://www.ccrcal.org/DSQC_Pubs/V2- 2012/Vol_II_2012.pdf).
(Accessed July 20, 2013).
51. Yost K, Perkins C, Cohen R, et al. Socioeconomic status and breast cancer
incidence in California for different race/ethnic groups. Cancer Causes
Control. 2001;12(8):703–711.
52. Costello S, Cockburn M, Bronstein J, et al. Parki nson's disease and
residential exposure to maneb and paraquat from agricultural applications
in the central valley of California. Am J Ind Med. 2009;169(8):919–926.
53. Marusek JC, Cockburn MG, Mills PK, et al. Control selection and pesticide
exposure assessment via GIS in prostate cancer studies. Am J Prev Med.
2006;30(2):S109–S116.
54. California Department of Water Resources. California Land & Water Use.
Land Use Data. (http://www.water.ca.gov/landwateruse/lusrvymain.cfm).
(Accessed July 12, 2013).
55. Wang AA, Costello SS, Cockburn MM, et al. Parkinson's disease risk from
ambient exposure to pesticides. Eur J Epidemiol. 2011;26(7):547–555.
56. Woods N, Craig IP, Dorr G, et al. Spray drift of pesticides arising from
aerial application in cotton. J Environ Qual. 2001;30(3):697–701.
57. Cox C. Pesticide drift: indiscriminately from the skies. Journal of Pesticide
Reform [electronic article]. 1995;15:1 –7. (http://sunridge.net/assets/pdf/
pesticide_drift.pdf)
58. Agency for Toxic Substances and Disease Registry. Toxicological Profile for
Aldrin/Dieldrin. Atlanta, GA: U.S. Department of Health and Human
Services; 2002. (http://www.atsdr.cdc.gov/toxprofiles/tp.asp?id=317&tid=
56). (Accessed September 22, 2013).
89

59. Xu X, Dailey AB, Talbott EO, et al. Associations of serum concentrations of
organochlorine pesticides with breast cancer and prostate cancer in U.S.
adults. Environ Res. 2010;118:60–66.
60. Muñoz-de-Toro M, Durando M, Beldoménico PM, et al. Estrogenic
microenvironment generated by organochlorine residues in adipose
mammary tissue modulates biomarker expression in ERalpha -positive
breast carcinomas. Breast Cancer Res. 2006;8(4):R47.
61. Høyer AP, Jørgensen T, Grandjean P, et al. Repeated measurements of
organochlorine exposure and breast cancer risk (Denmark). Cancer Causes
Control. 2000;11(2):177–184.
62. Arnold SF, Klotz DM, Collins BM, et al. Synergistic activation of estrogen
receptor with combinations of environmental chemicals. Science.
1996;272(5267):1489–1492.
63. Høyer AP, Grandjean P, Jørgensen T, et al. Organochlorine exposure and
risk of breast cancer. Lancet. 1998;352(9143):1816–1820.
64. Agency for Toxic Substances and Disease Registry. Toxicological Profile for
Chlordane. Atlanta, GA: U.S. Department of Health and Human Services;
1994. (http://www.atsdr.cdc.gov/toxprofiles/tp.asp?id= 355&tid= 62).
(Accessed September 22, 2013).
65. Cossette LJ, Gaumond I, Martinoli MG. Combi ned effect of xenoestrogens
and growth factors in two estrogen -responsive cell lines. Endocrine.
2002;18(3):303–308.
66. Chen S, Zhou D, Yang C, et al. Modulation of aromatase expression in
human breast tissue. J. Steroid Biochem Mol Biol. 2001;79(1-5):35–40.
67. Agency for Toxic Substances and Disease Registry. Toxicological Profile for
DDT, DDE, and DDD. Atlanta, GA: U.S. Department of Health and Human
Services; 2002. (http://www.atsdr.cdc.gov/toxprofiles/tp35.pdf).
(Accessed October 20, 2013).
68. Agency for Toxic Substances and Disease Registry. Draft Toxicological
Profile for Endosulfan. Atlanta, GA: U.S. Department of Health and
Human Services; 2013. (http://www.atsdr.cdc.gov/toxprofiles/tp.asp?id=6
09&tid=113). (Accessed September 30, 2013).
69. Andersen HR, Vinggaard AM, Rasmussen TH, et al. Effects of currently
used pesticides in assays for estrogenicity, androgenicity, and aromatase
activity in vitro. Toxicol Appl Pharmacol. 2002;179(1):1–12.
90

70. Agency for Toxic Substances and Disease Registry. Toxicological Profile for
Endrin. Atlanta, GA: U.S. Department of Health and Human Services;
1996. (http://www.atsdr.cdc.gov/toxprofiles/tp.asp?id= 617&tid= 114).
(Accessed September 22, 2013).
71. International Agency for Research on Cancer. IARC Monographs on the
Evaluation of Carcinogenic Risk of Chemicals to Humans. Clordane and
Heptachlor. 1991;53:115 –177. ( http://monographs.iarc.fr/ENG/Monograp
hs/vol53/mono53-8.pdf). (Accessed May 29, 2013).
72. Agency for Toxic Substances and Disease Registry. Toxicological Profile for
Heptachlor and Heptachlor Epoxide. Atlanta, GA: U.S. Department of
Health and Human Services; 2007. (http://www.atsdr.cdc.gov/toxprofiles/
tp.asp?id=746&tid=135). (Accessed September 22, 2013).
73. Bradman A, Whitaker D, Quirós L, et al. Pesticides and their metabolites in
the homes and urine of farmworker children living in the Salinas Valley,
CA. J Expo Sci Environ Epidemiol. 2007;17(4):331–349.
74. Charlier C, Albert A, Herman P, et al. Breast cancer and serum
organochlorine residues. Occup Environ Med. 2003;60(5):348–351.
75. Wong PS, Matsumura F. Promotion of breast cancer by beta -
hexachlorocyclohexane in MCF10AT1 cells and MMTV -neu mice. BMC
Cancer. 2007;7:130.
76. Akkina JE, Reif JS, Keefe TJ, et al. Age at natural men opause and exposure
to organochlorine pesticides in Hispanic women. J Toxicol Environ Health
A. 2004;67(18):1407–1422.
77. Agency for Toxic Substances and Disease Registry. Toxicological Profile for
Methoxychlor. Atlanta, GA: U.S. Department of Health and Human
Services; 2002. (http://www.atsdr.cdc.gov/toxprofiles/tp.asp?id=778&tid=
151). (Accessed September 22, 2013).
78. Thomas P, Dong J. Binding and activation of the seven -transmembrane
estrogen receptor GPR30 by environmental estrogens: a potential nov el
mechanism of endocrine disruption. J. Steroid Biochem Mol Biol.
2006;102(1-5):175–179.
79. Agency for Toxic Substances and Disease Registry. Draft Toxicological
Profile for Toxaphene. Atlanta, GA: U.S. Department of Health and Human
Services; 2010. (http://www.atsdr.cdc.gov/toxprofiles/tp.asp?id=548&tid=
99). (Accessed September 22, 2013).
80. Arcaro KF, Yang Y, Vakharia DD, et al. Toxaphene is antiestrogenic in a
human breast -cancer cell assay. J Toxicol Environ Health A.
2000;59(3):197–210.
91

81. Wofford P, Segawa R, Schreider J, et al. Community air monitoring for
pesticides. Part 3: using health- based screening levels to evaluate results
collected for a year. Environ Monit Assess. 2013;186(3):1355–1370.
82. St-Hilaire S, Mandal R, Commendador A, et al. Estrogen receptor positive
breast cancers and their association with environmental factors. Int J
Health Geogr. 2011;10:32.
83. Ruder EH, Dorgan JF, Kranz S, et al. Examining breast cancer growth and
lifestyle risk factors: early life, childhood, and a dolescence. Clin Breast
Cancer. 2008;8(4):334–342.
84. Okasha M, McCarron P, Gunnell D, et al. Exposures in childhood,
adolescence and early adulthood and breast cancer risk: a systematic
review of the literature. Breast Cancer Res Treat. 2003;78(2):223–276.
85. Goldberg DW. A Geocoding Best Practices Guide. North American
Association of Central Cancer Registries, Inc; 2008.
(http://www.naaccr.org/LinkClick.aspx?fileticket=ZKekM8k_IQ0%3d&ta
bid=239&mid=699). (Accessed September 22, 2013).
86. Goldberg DW, Wilson JP, Knoblock CA. From Text to Geographic
Coordinates: The Current State of Geocoding. URISA Journal.
2007;19:33–47.
87. Cayo MR, Talbot TO. Positional error in automated geocoding of
residential addresses. Int J Health Geogr. 2003;2(1):10.
88. Duell EJ, Millikan RC, Savitz DA, et al. A population -based case -control
study of farming and breast cancer in North Carolina. Epidemiology.
2000;11(5):523–531.
89. Yu M, Tatalovich Z, Gibson JT, et al. Using a composite index of
socioeconomic status to i nvestigate health disparities while protecting the
confidentiality of cancer registry data. Cancer Causes Control.
2013;25(1):81–92.

92

Chapter 3

3. Case-Control Study of Breast Cancer Risk from Ambient Exposure
to Pesticides

3.1 Abstract
We conducted a pilot study to evaluate the risks of postmenopausal breast
cancer associated with historical pesticide exposure in California’s Central Valley,
the most agriculturally productive and diverse region in the United States.
Residential and occupati onal histories were linked to commercial pesticide
reports and land use data to determine exposure to specific chemicals using a
Geographic Information Systems (GIS) approach. Cases ( n = 155) were recruited
from a population- based cancer registry and contr ols ( n = 150) were obtained
from tax assessor and Medicare list mailings. There was no association between
breast cancer risk and exposure to a selected group of organochlorine pesticides
thought to have endocrine -disrupting potential, however we observed a more
than three -fold increase in risk of breast cancer for exposure to the
organophosphate chlorpyrifos after adjusting for exposure to other pesticides
including organochlorines (OR = 3.22, 95% CI: 1.3 8, 7.53). Organophosphate
pesticides have rarely been evaluated in studies of breast cancer risk and
additional research is needed to confirm this finding and to understand the
underlying mechanisms given that chlorpyrifos has been detected during local air
monitoring at levels of concern for residents of the Central Valley.
93

3.2 Introduction
Certain types of pesticides are considered to be endocrine-disrupting
chemicals that affect hormone functioning by mimicking estrogen, interacting
with estrogen receptors and affecting enzyme systems involved in hormone
metabolism (1). Since lifetime estrogen exposure is a key factor in breast cancer
development, exposure to pesticides may be plausibly involved in breast cancer
development (2-5). Although the estrogenic properties of some p esticides such as
organochlorines have been established (6-11), results from previous studies that
assessed the impact of pesticide exposure on breast cancer risk are conflicting
with some studies showing a positive association (12-15), while others are null
(16-20).

Studies in which pesticide exposure was self -reported by pest type or
location (indoors or outdoors), or based on occupation categories, may have
misclassified exposure due to having to group together pesticides with varying
toxicological effects and subsequently reported null effects when there could be
effects for specific chemicals (12,16,17). Other studies based on measurements of
pesticide metabolites in serum samples taken near the time of cancer diagnosis
may not have accurately reflected prior exposures most relevant for breast cancer
etiology given the questionable validity of a one- time biomarker measure for long
term exposure (21). A recent review concluded that too few epidemiologic studies
to date evaluated pesticide -specific exposures on the risk of breast cancer and
there was need for methods capable of assessing lifetime exposure as well as
exposures at relevant ages (22).
94

We conducted a pilot case -control study of breast cancer risk from
exposure to pesticides using a Geographical Information Systems (GIS) –based
method that combines state pesticide use reports, land use data, and geocoded
residential and occupational histories (23). Participants were recruited from the
California Central Valley coun ties of Fresno, Tulare, and Kern the three top
ranked counties for agricultural density and commercial pesticide use in the U.S.
(24). In highly agricultural regions, pesticide drift from neighboring application
sites presents a major source of exposure (25-28). We evaluated a group of
organochlorine insecticides with known estrogenic effects that are most likely
related to breast carcinogenesis (aldrin, chlordane, dicofol, dieldrin, endosulfan,
methoxychlor, and toxaphene) (1,7,29-31), as well as three commonly used
agricultural chemicals that have not previously been linked to breast cancer but
were detected in ambient air monitoring conducted in Fresno County at levels of
concern to human health (chlorpyrifos, diazinon, and 1,3-dichloropropene) (32).

3.3 Materials and Methods
3.3.1 Selection, recruitment and data collection of cases
Cases were recruited from among women with histologically confirmed
breast cancer diagnosed in 2007–2008 in the counties of Fresno, Tulare, or Kern
from the Cancer Registry of Central California (CRCC). Cases selected were aged
55 to 74 years (over 55 in order to include only postmenopausal breast cancer,
which is more likel y of hormone -related origin, and under 75 in order to avoid
interviewing people who are less likely to recall prior events), and of non-
Hispanic white ethnicity. From 2011 –2013, cases were recruited by telephone (n
95

= 369). In attempting to recruit participants, up to 20 telephone calls were made
at various times of the morning, afternoon and evening with at least 10 messages
left. Women selected from the CRCC whose contact phone numbers were not
valid were sent to tracing services, which uses a combination of free and paid
search e ngines to obtain a phone number from other known identifying
information including name, address and date of birth. Over one quarter of the
telephone numbers received from CRCC were sent to tracing services ( n = 116),
with 64.7% resulting in a valid contact number.
Among the 3 28 we attempted to recruit , 10 were deceased and 4 too ill,
123 refused to participate, and we were unable to contact 32 . To match control
selection criteria (below), cases were excluded if they did not live in California for
at least five years prior to diagnosis (n = 0), if they reported Hispanic ethnicity (n
= 2) or if they had been diagnosed with ovarian, uterine, other female
reproductive cancers or Parkinson’s disease prior to their diagnosis of breast
cancer (n = 2). A total of 155 participants with breast cancer enrolled in the study
and completed interviews.
3.3.2 Selection, recruitment and data collection of controls
Controls were obtained from a study of Parkinson’s disease conducted in
the same geographic area from 2001 –2011. Details are described elsewhere (33-
36). Controls were eligible for the Parkinson’s disease study if they lived in
California for at least five years prior to study, were at least 35 years old and
resided in Fresno, Tulare, and Kern counties and did not have Parkinson’s
disease. Initially, controls were recruited from Medicare listings, but this method
of contact was discontinued in compliance with the Health Insurance Portability
96

and Accountability Act (HIPAA) of 2003. Recruitment continued with mailings to
a random selection of tax assessor parcels using Internet searches and marketing
companies to identify contact information. A total of 1,212 potential participants
were recruited for the Parkinson’s disease study through mailings and telephone
screenings and of these, 457 were ineligible (409 were too young, 44 were too ill
and 4 did not reside in o ne of the three counties). Among those eligible, 409
declined to participate, became too ill or moved out of the study area, lea ving 346
enrolled in the Parkinson’s disease study and among these 169 were female (36).
In 2009, another recruitment strategy for the Parkinson’s disease was
implemented where trained staff enrolled participants in person during home
visits to randomly selected neighborhood households. Among the 4,756 residents
screened during home visits, 3,515 individuals were ineligible (88.0% were too
young) and 634 declined to participate, le aving 607 enrolled in the Parkins on’s
disease study and among these 342 were female.
From the controls enrolled in the Parkinson’s disease study (169 + 342),
participants were selected as controls in this study if they were postmenopausal
women aged 55 –74 years of non-Hispanic white ethnicity ( n = 208). After
excluding women who had been diagnosed with breast cancer ( n = 20), ovarian,
uterine or other female reproductive cancers ( n = 9), and women who had opted
to complete a shortened questionnaire without lifetime residential and
occupational histories ( n = 29), there were 150 participants included as controls
in these analyses.

97

3.3.3 Source of exposure data
All controls (n = 150) and the majority of cases ( n = 111) were interviewed
over the telephone, with an additional 44 cases opting to complete a mailed
questionnaire with follow -up by telephone to clarify or complete responses. All
study participants were mailed a timeline to complete their historical residen tial
and occupational workplace information (addresses and dates) prior to their
telephone interviews. Telephone interviews were conducted to obtain
demographic and risk factor information such as age (in years), employment
status at the time of diagnosis for cases or interview for controls (unemployed,
employed, retired, and disabled), ever lived on a farm (yes or no), ever worked on
a farm (yes or no), education (in years), weight (pounds) and height (feet and
inches) at the time of diagnosis or interview were used to calculate body mass
index (kg/m
2
), age at menarche , age at menopause (note that we did not
distinguish between natural and surgical menopause) , number of births
(including st illbirth), oral contraceptive use (in years), menopausal hormone
therapy use (in years) by type (estrogen only, progesterone only, estrogen plus
progesterone, or a mixture of treatments), ever smoked (current, former, and
never), ever consumed alcohol at least once a week (yes or no), and vigorous
physical activity (defined as the number of hours of strenuous or moderate
activity per week). Neighborhood socioeconomic status (SES) was based on the
residential address at the time of diagnosis for cases or at the time of interview
for controls using income and occupation infor mation obtained from the 1990
U.S. census data at the block group level and categorized into a quintile score by
Yost et al. (37).
98

3.3.4 Historical and age-specific pesticide exposure assessment
Historical and age -specific pesticide exposures were determined from our
GIS-based method that combines state -reported pesticide use data, land use
surveys, and geocoded addresses to provide estimates of pesticide exposure
within a 500 -meter buffer around residential and occupational locations. These
methods are described in detail elsewhere (23,36,38).
3.3.4.1 Pesticide usage reporting
Briefly, California farmers and pest control operators have been required
to report pesticide applications to local agricultural commissioners starting in
1970 for restricted -use pesticides and in 1990 for all pesticides (39). California
maintains an extensive pesticide usage reporting system (P UR) with public data
available from 1974 to 2012. Each PUR record documents the name of the
pesticide’s active ingredient, the pounds applied, the crop and acreage of the field
to which it was applied, the application method, and the date and location. The
application location is identified in Public Land -use Survey System (PLSS)
sections, which equates to approximately one square mile.
3.3.4.2 Land-use maps
In addition to the PUR, California’s Department of Water Resources,
Division of Planning and Local Assistance, maintains an extensive, statewide set
of land-use and crop cover surveys by county from 1976 to 2011 (40). Each county
is surveyed at 6 to 10-year intervals, documenting crop areas and types of land-
use such as fields, vineyards and orchards. Historical electronic maps of land use
and crop type were constructed from the most recent digitally available land use
surveys for all California counties (1986 –2011) and manually digitized for the
99

counties of Fresno, Tulare, and Kern using the earliest available paper maps for
the years 1977–1985 (35,36).
California land -use survey and PUR data were matched by crop type to
indicate where the pesticides were most likely applied within the PLSS section
using an algorithm previously developed and validated (23,41). When land -use
and PUR data matched exactly, the pesticide was assumed to be applied in the
land-use polygon of the corresponding crop type; when PUR data did not match
by crop type but other crops were identified in the PLSS section, the pesticide was
assumed to be applied to the location where the existing crops were reported; and
when no crops were identified in the PLSS section, the pestici de was assumed to
be applied across the entire PLSS section.
3.3.4.3 Geocoding
All historical residential and workplace addresses were geocoded using
Texas A&M GeoServices (available at http://geoservices.tamu.edu) and manually
resolved using methods developed by Goldberg et al. to more precisely identify a
point location (referenced by its latitude and longitude coordinates) using
additional information provided by the participants such as the cross streets and
landmarks (42,43). We noted the level of “certainty” of each geocoded location
and considered addresses to have high geocode certainty if geocoded to the
centroid of a building, parcel, nearest parcel, street or street intersection.
Addresses geocoded to the centroid of a zip code, city, county or state, or those
that were unable to be geocoded were considered to have low geocode certainty.

100

3.3.4.4 Pesticide exposure estimates
Historical and age -specific ambient exposure to specific pesticides of
interest were calculated by summing the annual density (total pounds of a
pesticide’s active ingredients applied per acre) of applied pesticide within a 500-
meter buffer around each residential and workplace location (23). This buffer
distance was chosen based on studies that found measurable concentrations of
pesticides drifted from commercial pesticide application (44,45), and were
detectable in household dust of neighboring homes (46-48).
3.3.5 Other measures of pesticide exposure
Self-reported pesticide exposure information was obtained during the
telephone interviews. Participants were asked if they ever personally applied
pesticides (yes or no) inside their homes, outdoors in their yards or gardens, on
their pets, and whether they had ever hired a professional to spray or fumigate, as
well as whether they had ever worked on a farm or with pesticides or fertilizers.
Participants were also asked details about their personal pesticide use such as
frequency of application, but too few had reported regular use of once a month or
more often for stratified analyses by frequency (7.8% of cases and 12.7% of
controls reported regular use of pesticides inside their homes and 5.8% of cases
and 7.3% of controls reported regular use of pesticides in their yards or gardens).
In order to identify occupational pesticide exposure based on self -reported data,
other studies have used job exposure matrices (49-51), but only 26 cases and 43
controls reported farming so occupational exposure was based on self -reported
“ever” or “never” worked on a farm or worked with pesticides.

101

3.3.6 Statistical analyses
Unconditional logistic regression was used to estimate ambient pesticide
exposure on risk of breast cancer in postmenopausal women. Odds ratios (OR)
and 95% confidence intervals (CI) were calculated for study participants exposed
to specific pesticides com pared to those not exposed. An individual was
considered exposed to a particular pesticide (or any one of the pesticides in the
group of organochlorine pesticides) when the pounds per acre of applied
pesticide within the buffer area was greater than zero during the period from
1974 until the year of diagnosis for cases and the year of interview for controls.
We chose 1974 as the start of our exposure assessment so as to include all years
with complete pesticide information recorded by the state. Other studies have
also defined exposure by a cut -point at the median observed for the total pounds
of applied pesticide within the buffer area among the control subjects (33,52), but
this pilot study did not have a sufficient number s of exposed. In order to account
for time between exposure and the development of the disease, we conducted
sensitivity analyses excluding 10 years and 20 years prior to diagnosis.
Since young adulthood (defined as 20– 39) may be a critical time of
vulnerability to endocrine- disrupting chemicals, we also examined exposure t o
organochlorines and other pesticides found to be associated with breast cancer at
ages 20 –39 as well as at ages 40 –59 to assess relevant ages at the time of
exposure. An individual was considered exposed at these ages if the density of
applied pesticide within the buffer area was not null during the specific age
period, mutually adjusted for the other age period of exposure.
All a nalyses were adjusted for established breast cancer risk factors
102

including age (continuous), SES (quintiles 1 lowest to 5 highest), body mass index
(<25, 25 –29, and ≥30 kg/m2), age at menarche (<12, 12, >12 years), age at
menopause (<45, 45 –54, ≥55), number of births (0, 1, 2, and ≥3), o ral
contraceptive use (never, 1 –5 years, ≥5 years), menopausal hormone therapy use
(never used, estrogen only, progesterone only, estrogen plus progesterone, or a
mixture of treatments), and ever consume alcohol at least weekly (yes or no) , as
well as the number of years lived in Fresno, Tulare, and Kern counties during the
exposure assessment tim e period (continuous). Other factors, including the year
of diagnosis, education, ever smoked, vigorous physical activity, employment
status, ever lived or worked on a farm, ever personally applied pesticides inside
their home, outdoors in their yard or ga rden, or on their pets, and ever hired a
professional to spray or fumigate were evaluated as potential confounders and
were included in the final models if they changed the estimates by >10%. All
analyses were conducted using SAS, version 9.3 software (SAS Institute, Inc.,
Cary, North Carolina).
For the years residential histories had missing location data due to
incomplete recall of addresses or addresses that could not be found , we imputed
exposures using the average exposures during all years with data for each person
(53). Gaps in workplace histories where women did not report addresses because
they were unemployed, at home caring for children, retired or disabled were
imputed with the average exposures from the participant’s residential exposure
history (assuming that they most likely resided at home during typi cal work
hours). Pesticide exposure could not be identified for addresses outside of
California since the exposure model includes only California PUR data ; therefore,
103

these locations were considered unexposed. Since California has the highest
agricultural productivity anywhere in the nation, people living outside of
California can reasonably be considered unexposed. We assessed the influence of
missing data due to recall and locations outside of California by examining the
change in our estimates of relativ e risk after excluding missing data and
conducted sensitivity analyses that included only women who lived in California
for at least 30 years between 1974 and their year of diagnosis or interview.
Ambient lifetime exposure to pesticides was determined from the
residential histories of the cases and controls obtained through interviews, and
separately, under the assumption that the participant lived at the address at the
time of diagnosis or interview throughout her entire lifetime. The differences in
the prevalence of exposure and the estimates of relative risk obtained from
lifetime reported histories compared to only the address at diagnosis or interview
were used to estimate the impact of exposure misclassification in the absence of
complete historical location information.
The institutional review boards at the California Health and Human
Services Agency and the University of Southern California approved the study
protocol for cases participating in this study. The institutional review board at the
University of California, Los Angeles approved the study protocol for controls
used in this study. Informed consent was obtained for all participants.
3.3.7 Power calculation
With a sample size of 150 cases and 150 controls, we are powered at 82%
to detect an effect size of 2.0 assuming at least 30% prevalence of pesticide
104

exposure, using Quanto 1.2.4, May 2009 available at
http://biostats.usc.edu\software.
Figure 3.1: Power Calculation for Pilot Case -Control Study of Breast Cancer Risk
and Exposure to Ambient Pesticides Assuming a Sample Size of 150 Cases and
150 Controls.
Prevalence of Exposure
Effect Size 15% 20% 25% 30%
1.25 0.11 0.13 0.14 0.15
1.50 0.27 0.32 0.36 0.38
1.75 0.47 0.55 0.60 0.64
2.00 0.67 0.75 0.79 0.82
2.25 0.81 0.87 0.91 0.92
2.50 0.90 0.94 0.96 0.97

3.4 Results
3.4.1 Breast cancer risk factors
Cases and controls appeared similar in terms of established breast cancer risk
factors such as age, SES, education, body mass index, age at menarche, age at
menopause, number of births, menopausal hormone therapy use, and vigorous
physical activity (Table 3.1). Cases were more likely to have used oral
contraceptives for 5 years or longer (OR = 1.17, 95% CI: 0.70, 1.97). Cases were
half as likely to be current s mokers (OR = 0.45, 95% CI: 0.20, 1.03), but were
more likely to consume alcohol at least weekly compared to controls (OR = 1.74,
95% CI: 1.09, 2.79). Cases were more likely to have lived in Fresno, Tul are, and
Kern Counties at least 30 years since the start of our exposure assessment in 1974
and to have lived in California at least 30 years (OR = 1.59, 95% CI: 1.01, 2.51 and
OR = 2.52, 95% CI: 1.28, 4.97, respectively).
105

Employment status differed between cases and controls (Table 3.2). Cases
were less than half as likely to be retired as controls (OR = 2.02, 95% CI: 1.23,
3.31) and were more likely to be employed or unemployed (OR = 0.77, 95% CI:
0.35, 1.70). Cases were also half as likely to have ever lived and worked on a farm
(OR = 0.44, 95% CI: 0.23, 0.83) or to have ever worked with pesticides compared
to controls, however only 3 cases and 7 controls reported working with pesticides
(OR = 0.40, 95% CI: 0.10, 1.56 ). Cases and controls were equa lly as likely to
report applying pesticides in their homes or on their pets, and to have hired a
professional to spray or fumigate; however, cases were significantly more likely to
have applied pesticides in their yards or gardens than controls (OR = 1.75, 95%
CI: 1.08, 2.82).
3.4.2 Ambient pesticide exposure
When assessing ambient pesticide exposure, the prevalence of exposure at
residences and workplaces for the selected group of organochlorines,
chlorpyrifos, and diazinon were over 40% among cases and controls (Table 3.3).
After adjusting for established breast cancer risk factors, the number of years
lived in Fresno, Tulare, and Kern counties, vigorous physical activity (since its
inclusion changed the estimates by 19.1%), and exposure to the other pesticides,
there was a more than three- fold increase in breast cancer risk for those exposed
to chlorpyrifos at both residences and workplaces compared to women not
exposed at either location (OR = 3.22, 95% CI: 1.38, 7.53). Associations more
moderate in magnitude were observed between breast cancer and exposure to
organochlorines and to diazinon, however after adjusting for exposures to other
pesticides, in particular chlorpyrifos, the associations were null (OR = 0.98, 95%
106

CI: 0.42, 2.28 and OR = 0.81, 95% CI: 0.35 , 1.84, respectively). There was no
increased breast cancer risk for exposure to 1,3 -dichloropropene. Results that
excluded exposures occurring 10 years or 20 years prior to diagnosis or interview
did not qualitatively change the risk estimates. Risk estimates were slightly
attenuated for exposure to chlorpyrifos (OR 10 year = 2.78, 95% CI: 1.20, 6.43 and
OR 20 year = 3.13, 95%CI: 1.30, 7.52) and remained close to the null for exposure to
organochlorines, diazinon, and 1,3 -dichloropropene, after accounting for the 10 -
and 20-year latency periods.
3.4.3 Pesticide exposure at relevant ages
When we considered specific age periods of exposure, there was no
association between breast cancer risk and exposur e to organochlorines at ages
20–39 or at ages 40–59, after adjusting for breast cancer covariates and exposure
to the other pesticides (OR = 0.88, 95% CI: 0.44, 1.75 and OR = 0.89, 95% CI:
0.45, 1.78, respectively) (Table 3.4). Expo sure to chlorpyrifos at ages 20 –39 was
not associated with increased breast cancer risk (OR = 0.96, 95% CI: 0.45, 2.08),
but women exposed to chlorpyrifos at ages 40 –59 had more than three times the
odds of breast cancer (OR = 3.31, 95% CI: 1.60, 6.85). Compared to unexposed
women, women exposed to chlorpyrifos in either their 40’s or 50’s were 2.8 times
as likely to develop breast cancer (OR = 2.82, 95% CI: 1.28, 6.23) and women
exposed to chlorpyrifos at both ages 40’s and 50’s were 4.2 times as likely to
develop breast cancer (OR = 4.18, 95% CI: 1.74, 10.05).
3.4.4 Self-reported pesticide exposure vs. GIS-based method
Among those who reported that they had never lived on a farm, 43.9% of
cases and 40.0% of controls were exposed to one of the selected pesticides at
107

their residences according to our GIS-based exposure method. Among those who
self-reported that they had never worked on a farm , 71.6% of cases and 54.7% of
controls were exposed to one of the selected pesticides at their workplaces.
3.4.5 Exposure assessment based on address at diagnosis or
interview vs. exposure based on address histories
The associations between breast cancer risk and exposure to
organochlorines, chlorpyrifos, diazinon, and 1,3 -dichloropropene were null when
the exposure assessment was based on a single residential address at the time of
diagnosis for cases and the time of interview for controls, after adjusting for
breast cancer covariates and mutually adjusting for the other pesticides (Table
3.7). When the exposure assessment was based on entire residential histories, the
associations remained null for exposure to organochorines, diazinon, and 1,3 -
dichloropropene, while a more than two -fold increased risk between breast
cancer and exposure to chlorpyrifos was observed after adjusting for breast
cancer covariates and exposure to the other pesticides (OR = 2.27, 95% CI: 1.07 ,
4.80).

3.5 Discussion
This population -based study examine d historical and chemical -specific
effects of hormone-related pesticides that are plausibly related to breast cancer in
a region of intense agricultural production. We did not observe an association
between breast cancer risk and exposure to a selected grou p of organochlorines
after adjustment for co -exposures to other commonly applied pesticides , but we
did observe a more than three- fold increased risk of breast cancer with exposure
108

to chlorpyrifos, one of 3 pesticides detected in air monitoring studies at levels of
concern to public health (OR = 3.22, 95% CI: 1.38, 7.53). Although the confidence
intervals were relatively wide given the small sample size, we found an even
stronger association for exposure to chlorpyrifos at ages 40 –59 (OR = 3.31, 95%
CI: 1.60, 6.85) and a dose response relationship for exposure at ages 40’s or 50’s
and exposure at both ages 40’s and 50’s (P = 0.001).
3.5.1 Rationale and biological plausibility
The majority of epidemiologic studies involving pesticides and breast
cancer risk have focused on organochlorine s, but we found an increased risk of
breast cancer with exposure to the organophosphate chlorpyrifos that became
stronger after adjusting for exposure to other pesticides including
organochlorines. Conversely, the association between breast cancer risk and
exposure to organochlorines became null after adjusting for exposure to the other
pesticides, specifically adjustment for chlorpyrifos was important (OR organochlorines
= 0. 79, 95% CI: 0.36, 1.73 for exposure at both residences and workplaces,
adjusting for covariates and exposure to chlorpyrifos). Previous studies
examining exposure to organochlorines have not considered other kinds of
pesticides such as chlorpyrifos that may be driving the breast cancer associations
because of high correlations between pesticides.
When assessing ambient pesticide exposure, it is usually difficult to
distinguish the effects of specific chemicals because applications may be
correlated and people can be exposed to multiple chemicals (as is the case wi th
other kinds of toxic air pollutants). For example, we examined exposure among
the controls in our study and found that 54.0% were exposed to both the
109

organochlorines and chlorpyrifos, 24.7% were unexposed to either, and only
21.3% were exposed to one bu t not the other. Although most of the research to
date has focused on organochlorines, these findings suggest that exposure to
organochlorines may be proxies for exposure to chlorpyrifos, or other pesticides
that have not been well studied but may affect breast cancer risk.
In vitro studies show that chlorpyrifos is weakly estrogenic (7,9), anti -
androgenic (54,55), and as an arylhydrocarbon receptor agonist can also affect
hormone pathways (56) or inhibit adrenal steroidogenesis (57). One study found
that chlorpyrifos induced proliferation of estrogen -dependent breast cancer cells
in vitro but not estrogen -independent breast cancer cells (58). In addition,
chlorpyrifos has been shown to irreversibly inhibit cytochrome P450 enzymes
involved estrogen and testosterone metabolism (59,60). Since cytochrome P450
enzymes activate or detoxify many xenobiotics and chemical carcinogens, it is
possible that chlorpyrifos could affect breast cancer risk through non -hormonal
mechanisms. More toxicological research is needed to understand its mechanistic
potential with regards to breast cancer r isk. The findings from this study also
support the 2012 regulations that further restrict the agricultural use of
chlorpyrifos in order to reduce the potential for exposure through pesticide drift
(61).
Although chlorpyrifos was the most commonly used pesticide inside
homes prior to its ban for residential use in 2000 by the U.S . EPA, previous
studies of home pesticide use grouped pesticides by pest type and were unable to
examine chemical-specific effects since chlorpyrifos was used on a variety of pests
(12,16). Chlorpyrifos is also widely used in agriculture and its use has been
110

correlated with measurements found in household dust (26). It is a
cholinesterase inhibitor that affects the human nervous system and was included
in an ecological study of breast cancer rates and pesticide use density by census
blocks, but no associations were observed for a group of 20 organophosphate and
carbamate pesticides that may or may not share similar toxicological mechanisms
(OR = 0.99, 95% CI: 0.77, 1.28) (18). In the Agricultural Health Study of pesticide
applicators and their wives, a modest association was observed between breast
cancer among wives who reported using chlorpyrifos at home (OR = 1.40, 95%
CI: 0.90, 2.40), however t he 95% CI included the null and the prevalence of self -
reported use was very low (5.3% of cases and 3.9% of non-cases) (17).
Although we did not observe an increased risk of breast cancer with
exposure to organochlorines overall or at ages 20 –39, we cannot rule out the
possibility that exposures at younger age periods might be relevant. Exposures to
organochlorine pesticides prior to the first pregnancy when breast tissue is not
fully differentiated may be a critical period of vulnerability to endogenous
hormones and exogenous hormone -like chemicals (6,8). One prospective study
found that women exposed to the organochlorine DDT before puberty had a
dose-response increase in risk of breast cancer with increasing serum levels of
DDT, suggesting that exposures during breast growth and development may be
important (5). We were unable to examine exposures prior to the first pregnancy
or at puber ty since we did not have information on age at pregnancy for the
control subjects in this pilot study and our exposure assessment was not able to
capture younger ages at exposure. Future research is needed to explore these
younger age periods that may be important for breast cancer risk.
111

Although chlorpyrifos and diazinon are b oth organophosphate pesticides
and as many women were exposed to diazinon as to chlorpyrifos in our study, we
did not find an association between diazinon and breast cancer risk after
adjusting for exposure to chlorpyrifos . In vitro studies indicate that chlorpyrifos
is cytotoxic at far lower concentrations than diazinon (62).
We also did not observe an association between breast cancer and 1,3 -
dichloropropene, however the prevalence of exposure in our study was relatively
low (14.2% of cases and 13.3% of controls exposed at both residences and
workplaces). In an ecologic study, Reynolds et al. included 1,3 -dichloropropene
along with 15 other pesticides of various chemical classes that were “probably or
likely human carcinogens” and found no association between breast cancer rates
and pesticide use density at the census tract level (OR = 1.00, 95% CI: 0.96, 1.04)
(18). The pesticide 1,3-dichloropropene is a respiratory carcinogen, but its role as
a breast carcinogen is unknown (63-65).
3.5.2 Advantages to our GIS-based method
Our GIS -based ambient exposure method has several advantages over
previous methods used to assess breast cancer risk from pesticide exposure
including that it allows us to examine spe cific pesticides of interest, to assess
historical exposures spanning three decades as well as age -specific exposures,
and to reduce the potential for recall bias . Unlike GIS -based ecologic or
regression studies, our method evaluated individual residential and occupational
histories, while obtaining and controlling for established breast cancer risk
factors. One previous ecologic study found no association comparing participants’
current residential zip codes with agricultural census data reports of areas t reated
112

with any fungicides (OR = 0.72, 95% CI: 0.46, 1.12) (66). Two previous regression
analyses also found no associations between breast cancer risk and pesticide use
at the census block group or ward level (14,18), while another study comparing
county-level pesticide use found a statistically significant increased risk of breast
cancer for two types of organochlorines, methoxychlor and toxaphene (RR = 1.18,
95% CI: 1.03, 1.35 and RR = 1.16, 95% CI: 1.01, 1.34, respectively) (13). In all of
these prior studies, exposure was based on aggregated pesticide use data and
proximity to address at the time of cancer diagnosis, with limited information on
potential confounding variables. Our risk estimates based on a single residential
address at diagnosis or interview were null for all pesticides considered in these
analyses, including exposure to chlorpyrifos (OR = 1.21, 95% CI: 0.56 , 2.61), and
may underestimate actual risk since they do not account for exposures occurring
at workplaces (Table 3.7) . Only one previous study conducted in Cape Cod,
Massachusetts collected residential histories to assess GIS -based proximity to
pesticide applications and found no associations; however, the p revalence of
pesticide exposure was much lower than observed in our study and exposures
were grouped by land-use type instead of by specific pesticide of interest (20).
We based exposure on reported address histories rather than self-reported
pesticide use, which reduces the potential for bias from differential recall of prior
environmental exposures by cases and controls since participants were most
likely not aware of specific pesticides historical applied in the vicinity of their
residences or workplaces . A study conducted in Australia found that the
association between breast cancer risk and self -reported “noticing of pesticide
spray drift ” was strongly confounded by participants’ belief in whether or not
113

pesticides cause d breast cancer (OR = 1.47, 95% CI: 1.15, 1.87 among believers
and OR = 0.94, 95% CI: 0.51, 1.74 among non -believers) (67). Self -reported
pesticide exposure can also lead to non -differential misclassification from the
inability of both cases and controls to recall prior pesticide use that may have
occurred decades ago. In addition, cases and controls may be unable to
remember or do not know specific details about their pesticide use such as the
type of chemical applied, leading to broad exposure categories such as “indoor
insecticides” or “lawn pesticides.” These categories assume that all chemicals
would have similar effects on risk of breast cancer, leading to bias of the
estimates towards the null. In addition, self -reported exposures do not account
for exposures that people may not be aware of but are routinely exposed to, which
is often the case with pesticide drift. Among participants who had never lived on
a farm, 40% or more were exposed to one of the selected pesticides in close
proximity to their residential locations according to our model and over 50%
were exposed via their occupational locations (Table 3.5). If our study had been
based on these self -reported measures of having ever lived or worked on a farm,
our estimates would have been biased towards the null since cases were more
likely than controls to report not living or working on a farm but were found to be
exposed to the selected pesticides via our GIS-based approach.
We also looked at the potential impacts of migration in our study by
comparing those who resided in Fresno, Tulare, and Kern Counties for the
majority of the exposure assessment (30 or more years) to those who moved
during this time (lived <30 years in one of the three counties) by age (55 –59, 60–
64, 65 –69, or 70 –74) and SES (quintiles 1 lowest to 5 highest) . We did not
114

observe any differences in the distribution of age or SES between those who lived
in Fresno, Tulare, and Kern Counties compared to those who moved ( P age = 0.76
and P SES = 0.56), and these were not differential among cases and controls
(Tables 3.8a and 3.8b). We also did no t find any differences in age or SES among
residents who lived in California for the majority of the exposure assessment (30
or more years) compared to those who moved out of California during this time
(lived <30 years in California) ( P age = 0.55 and P SES = 0.52), and again this was
not differential by disease status (Tables 3.8a and 3.8c). These results suggest
that participants who migrated during the time frame of our exposure assessment
did not differ from participants who did not migrate by these demo graphic
factors or by disease status, and migration is not likely to have differentially
impacted our overall findings.
3.5.3 Limitations to our GIS-based method
Although we were able to construct residential and occupational histories
and link them to state reports of pesticide application, we were limited to data
collected since 1974 for restricted use pesticides and 1990 for all pesticides. The
GIS-based method used in this study thus was able to account for 28 –36 years of
exposure, but exposures at even younger ages such as puberty may be relevant.
Our exposure assessment was also based on ever vs. never exposed, which meant
that women exposed to a lot of pesticides were grouped with women exposed to
very little at their residential and occupational locations. We may have
misclassified women with exposure that did not reach a level high enough to
impact breast cancer etiology, thus inflating our estimates of relative risk.
115

Additional research is needed to evaluate detectable threshold limits related to
pesticide exposure on risk of breast cancer.
Another limitation to our GIS-based method is that pesticide exposure was
underestimated due to incomplete PUR data. Approximately 37% of California
PUR data lacked information on where the pesticide was applied, meaning that
more participants were likely to be exposed than our method was able to identify.
We assumed that the potential misclassification resulting from incomplete PUR
data was not related to case or control status and therefore the bias would be
directed towards the null.
Our GIS -based method relied on land- use data to ident ify the specific
location of crops, however land -use surveys are conducted every 6 to 10- year
intervals and may not indicate land use where the crops change more frequently
or where multiple crops are grown on the same land (10). The use of satellite
imagery has been proposed as a way to characterize agricultural land since these
images are collected every 8 to 16 days (68); however, in order to use such
methods in an epidemiologic study an image library that identifies all of the
different crop types from their spectral signatures would need to be developed as
well as methods to distinguish different crops that appear similar in the images.
Further research is also needed to validat e such methods using satellite images to
assess historical pesticide applications spanning decades, and this may become
more feasible in the future as the technology develops and becomes more
automated.

116

3.5.4 Selection bias
Since participants were unlikely to be aware of their exposure to pesticides
in the ambient air via their location histories, our GIS -based method reduces the
potential for selection bias from differential participation as a result of concerns
about exposures in the environment. The response rate was low but typical for
telephone recruitment methods used for cases (42.0% of cases), as well as
mailing and home visit methods used for controls. We compared the cases that
participated in our pilot study to other women diagnosed with breast cance r
during the same time period in the CRCC that lived in Fresno, Tulare, and Kern
Counties as well as those that lived in all counties in the CRCC and found them to
be similar in terms of age, SES, and estrogen and progesterone receptor subtypes
(Table 3.9). Cases in our pilot study were slightly more likely to have their cancer
stage defined as localized rather than to distant compared to non -participating
women with breast cancer in the CRCC. Nothing was known about the non -
responding controls.
In this pil ot study however, selection bias may be a concern since cases
were recruited from a population- based cancer registry while controls were
obtained from another population-based study in the same location and
overlapping in time but still may not be representative of the base population that
gave rise to the breast cancer cases. The validity of the findings depends on how
well the controls represent the distribution of exposure in the same population at
risk to become breast cancer cases in this study (69,70). Therefore, it’s important
that the selection of controls in the study not be affected by their exposure status.
Controls were more likely to have lived or worked on a farm than cases and as a
117

result would be expected to have higher likelihood of exposure to pesticides near
their residences or workplaces, thus biasing our estimates towards the null; and
yet, we still observed a strong association to one of the pesticides of interest.
Cases, on the other hand, were more likely to have lived in Fresno, Tulare, or
Kern counties longer than controls and therefore may have had more opportunity
to be exposed to ambient pesticides in the region. We adjusted the a nalyses for
the number of years lived in Fresno, Tulare, or Kern counties to account for this
potential confounding factor (accounting for only a 5.4% change in the estimate),
but still observed chemical-specific effects.
Many of the established risk facto rs for breast cancer, including SES,
education, body mass index, number of births, and use of hormone replacement
therapy were not associated with breast cancer risk in this study population,
which may indicate bias in the way the controls were selected. For example, we
did not observe an association with body mass index as we might have expected
and this could have been related to our method of recruiting controls through
mailings and home visits. It is possible that the controls selected for our study
were more likely to be at home and have higher body mass indices than non -
participating women in the base population. We observed that controls were
more likely to be retired than our cases, and although controls reported similar
vigorous physical activity as cases at the time of interview, they may have reduced
activity compared to before they retired or perhaps they may consume more high
calorie meals at restaurants as a result of retirement and thus have higher body
mass indices than the underlying populati on. Therefore, it is possible that
selection bias could have altered the association between breast cancer and
118

pesticide exposure if established breast cancer risk factors that were affected by
the selection of controls were also associated with pesticide exposure. None of the
breast cancer risk factors adjusted for in our model however, were associated
with pesticide exposure, including body mass index (OR = 1.10, 95% CI: 0.81,
1.48 for exposure to chlorpyrifos). Adjustment cannot be used to control for
selection bias except under certain circumstances as discussed in Rothman and
Greenland (69), but it would be better to avoid selection bias at the onset of the
study. An improved study design using population- based controls selected to
represent the base population of women in the region is needed to confirm these
findings, however despite the limitations in the design of our pilot study we
observed a strong association for chlorpyrifos, but not for exposure to the other
pesticides.
Other study designs may be considered, but each one has its own strengths
and drawbacks as well. For example, one way to obtain a representative sample of
controls from the base population might be to recruit neighborhood controls via
several strategies including recruitment through home visits or at communities
centers such as senior citizen centers, churches, or heal thcare screening
facilities), or by obtaining contact information from commercial databases
available for marketing purposes. During recruitment however, you would ideally
want to find out information on the underlying individuals you are trying to
recruit regardless of eligibility, particularly for certain risk factors for the disease
and information to evaluate their exposure, and this is likely to be challenging
and may not be feasible or cost -effective. Another study design might be to
conduct a nested case-control study using an existing cohort such as the
119

California Teacher’s Study that began in 1995 where complete residential and
workplace histories would have to be collected on the participants. Selection
issues may be of concern when using this coho rt of educated women, particularly
with regards to covariate information on breast cancer risk factors that differ in
this population compared to other women in California.
3.5.5 Misclassification of covariates
In addition to selection bias, our measurement of the covariates could
have led to information bias. The breast cancer covariates for this study were
obtained from questions developed for a study of Parkinson’s disease and may be
inadequate measures of the adjusted risk factors, leading to residual
confounding. For example, our evaluation of body mass index based on
participant recall at the time of diagnosis or interview could have led to an
underestimate of body mass index in our cases. Our measurement for body mass
index was based on height and weight at the time of diagnosis for cases, which
may not represent what it was before diagnosis since cases may have lost weight
right before or after being diagnosed with breast cancer. Furthermore, body mass
index at the time of diagnosis or interview may not be as etiologically relevant for
breast cancer risk as body mass index at earlier ages or other kinds of weight
measures such as waist -to-hip ratio, which were not measured or controlled for.
These possibly misclassified covariates could have confounded the association
between breast cancer and pesticide exposure if they were also associated with
pesticide exposure, but again, these established breast cancer risk factors were
not associated with pesticide exposure in our study.

120

3.5.6 Misclassification of pesticide exposure
Since pesticide exposure was determined from residential and
occupational histories, missing data in the lifetime exposure calculations may
have misclassified exposure status. Missing exposures resulted from incomplete
recall of address histories , addresses that could not be found, or locations that
were outside of California. Excluding the subjects with substantial missing
information in their residential histories did not qualitatively change our
estimates. (There were 3 cases and 4 controls wit h more than one third of their
residential timelines missing from the start of our exposure assessment until the
year of diagnosis or interv iew, and only 5 controls with 1 –5 years missing). The
majority of participants had all residential locations within California during the
timeframe of interest (82.0% of cases and 72.7% of controls) (Table 3.6). Cases
were significantly more likely to have spent at least 30 years in California during
the exposure assess ment time frame than controls (P = 0.02), however f ew
participants reported that their residences located outside of California were on
farms (3 cases and 5 controls). If more controls than cases were misclassified as
unexposed as a result of living outside of California when they were actually
exposed, we would have underestimated exposure in the controls and inflated our
risk estimates. We restricting analyses to include only those who had lived in
California for at least 30 years during our exposures assessment time frame (n =
141 cases and n = 120 controls), and found that our estimates increased for
exposure to chlorpyrifos at both residences and workplaces (OR = 3. 98, 95% CI:
1.48, 10.72), while estimates remained null for exposure to the other pesticides
(data not shown).
121

Our GIS-based method is also dependent on the certainty of the geocoded
location information. The geocode certainty of the historical addresses was
similar for cases and controls for the period of our exposure assessment from
1974 until the year of diagnosis or interview. There were 61.9% of cases and
62.7% of controls whose entire residential address histories had high geocode
certainty, and another 13.5% of cases and 18.7% of controls with 85% of their
residential histories having high geocode certainty (1 –5 years having po or
geocode certainty). Workplace addresses tended to be recalled with less detail
than residential addresses. There were 43.2% of cases and 44.7% of controls
whose entire occupational address histories had high geocode certainty and
another 13.5% of cases and 16.1% of controls with 85% of their occupational
histories having high geocode certainty. Few participants had address histories
with 50% or more of the addresses having low geocode certainty. There were
7.7% of cases and 3.3% of controls with 50% or m ore of their residential
addresses having low geocode certainty and 21.9% of cases and 16.7% of controls
with 50% or more occupational addresses having low geocode certainty. This
suggests that the certainty of the geocoding is not likely to account for di fference
in the estimated effects.
3.5.7 Conclusions
The GIS -based approach presented here likely reduced exposure
misclassification compared to estimates based on self -report or those based on ly
on address at the time of diagnosis. Estimating adult cancer risks associated with
prior cumulative exposures present methodological challenges for epidemiologic
studies as it requires accurate measurements over lifetime and decades. We have
122

constructed a more comprehensive exposure assessment using historical da ta
than has been done previously, however, a larger study is needed to confirm the
chemical-specific associations we report and to examine different levels of
exposure as well as potential dose-response relationships.
This study suggests that pesticides other tha n organochlorines, such as
chlorpyrifos, may be important for breast cancer risk and that additional research
is needed to improve etiologically -relevant measures of exposure in order to
protect people who are exposed unwittingly to these chemicals in the ambient air.
Future studies of interest should be designed to evaluate lifetime exposures
beginning when the study subjects are young in age (pre- puberty) until they
reach the age of menopause, to assess exposure at various ages for specific
pesticides of interest (exposures at menarche as well as prior to first pregnancy),
and to take into consideration the distribution of pesticide us age over time
(Figure 3.2).
123

Table 3.1. Comparison of Selected Characteristics in Breast Cancer Cases (Diagnosed in 2007–
2008) and Population-Based Controls (2001–2011), in Fresno, Tulare, and Kern Counties.
Characteristic
Cases

Controls

(n = 155)

(n = 150)

P Value*
No. % No. %

Year of diagnosis or interview

2001–2007 73 47.1

68 45.3

0.76

2008–2011 82 52.9

82 54.7

Age, years

55–59 40 25.8

34 22.7

0.86

60–64 40 25.8

37 24.7

65–69 43 27.7

43 28.7

70–74 32 20.7

36 24.0

Neighborhood SES

1 Lowest quintile 29 18.7

21 14.0

0.56

2 41 26.5

34 22.7

3 35 22.6

41 27.3

4 41 26.5

47 31.3

5 Highest quintile 9 5.8

7 4.7

Education, years

<12 8 5.2

7 4.7

0.97

12 33 21.3

31 20.7

>12 114 73.6

112 74.7

Body mass index (kg/m
2
)†

<25 54 34.8

36 27.3

0.33

25–30 47 30.3

41 31.1

≥30 54 34.8

55 41.7

Unknown/missing 0

18

Age at menarche, years

<12 35 22.6

33 22.0

0.99

12 38 24.5

37 24.8

>12 82 52.9

79 53.0

Unknown/missing 0

1

Age at menopause, years

<45 39 25.2

47 32.9

0.28

45–54 81 52.3

63 44.1

≥55 35 22.6

33 23.1

Unknown/missing 0

7

Number of births

0 16 10.3

19 12.7

0.76

1 25 16.1

19 12.7

2 47 30.3

49 32.7

≥3 67 43.2

63 42.0

124

Table 3.1. Continued:
Characteristic
Cases

Controls

(n = 155)

(n = 150)

P Value*
No. % No. %

Oral contraceptive use

No 59 39.1

54 36.0

0.02

1–5 years 24 15.9

43 28.7

≥5 years 68 45.0

53 35.3

Unknown/missing 4

0

Menopausal hormone therapy use

None 61 39.4

49 33.1

0.11

Estrogen only 59 38.1

58 39.2

Progesterone only 3 1.9 3 2.0

Estrogen + progesterone 25 16.1

19 12.8

Mixture of treatments 7 4.5

19 12.8

Unknown/missing 0

2

Ever smoked

Never smoked 90 58.1

77 51.3

0.15

Quit smoking 55 35.5

54 36.0

Currently smoked 10 6.5

19 12.7

Ever consumed alcohol at least once a week

No 56 36.1

66 49.6

0.02

Yes 99 63.9

67 50.4

Unknown/missing 0

17

Vigorous physical activity, hours per week

None 45 30.2

36 24.3

0.43

1–6 38 25.5

38 25.7

7–14 31 20.8 42 28.4

≥14 35 23.5

32 21.6

Unknown/missing 6

2

Number of years lived in Fresno, Tulare, and
Kern Counties‡

<30 56 36.1

71 47.3

0.05

≥30 99 63.9

79 52.7

Number of years lived in California‡

<30 14 9.0

30 20.0

0.01

≥30 141 91.0 120 80.0

* Pearson's χ2 test on n-1 degrees of freedom.
† Body mass index was calculated from height (feet and inches) and weight (pounds) at the time of diagnosis for cases and
at the time of interview for controls. Height ranged 4’9”–6’0” for cases and 4’9.5”–5’10” for controls and weight ranged
90–247 pounds for cases and 102–305 pounds for controls.
‡ Years lived was calculated from residential histories that included addresses recorded from 1974 until the year of
diagnosis or interview. Missing addresses in residential histories were imputed with the nearest known address; however,
excluding these missing addresses did not change proportions.
125

Table 3.2. Comparison of Occupational and Residential Characteristics and Self-Reported
Pesticide Use Among Breast Cancer Cases (Diagnosed in 2007–2008) and Population-Based
Controls (2001–2011), in Fresno, Tulare, and Kern Counties.
Characteristic
Cases

Controls

(n = 155)

(n = 150)

P Value*
No. % No. %

Employment status†

Employed 72 46.5

51 34.0

0.01

Unemployed 22 14.2

12 8.0

Retired 56 36.1 80 53.3

Disabled 5 3.2

7 4.7

Ever lived or worked on a farm

No 84 54.2

71 47.3

0.06
Lived on a farm only 45 29.0 36 24.0

Worked on a farm only 8 5.2 8 5.3

Lived and worked on a farm 18 11.6

35 23.3

Ever applied pesticides at home

No 31 20.0

31 20.7

0.89

Yes 124 80.0

119 79.3

Ever applied pesticides in yard or garden

No 45 29.4

62 42.2

0.02

Yes 108 70.6

85 57.8

Unknown/missing 2

3

Ever applied pesticides on pets

No 68 45.3

61 42.1

0.57

Yes 82 54.7

84 57.9

Unknown/missing 5

5

Ever hired professional to spray or fumigate
in or around home

No 37 24.5

34 25.6

0.84

Yes 114 75.5 99 74.4

Unknown/missing 4

17

* Pearson's χ2 test on n-1 degrees of freedom.
† Employment defined at the time of diagnosis for cases and at the time of interview for controls.

126

Table 3.3. Measures of Association Between Breast Cancer and Ambient Exposure to Selected Pesticides Based on Residential and Occupational
Address Histories Among Women in Fresno, Tulare, and Kern Counties Using Linked PUR data for 1974–2011.
Exposure Type
Cases
(n = 155)

Controls
(n = 150)
Adjusted
OR*

Adjusted
OR†

No. % No. % 95% CI 95% CI
Organochlorines‡

Unexposed 32 20.7 47 31.3 1.00 --- 1.00 ---
Exposed at residences only 16 10.3 11 7.3 1.26 0.43, 3.72 0.76 0.24, 2.44
Exposed at workplaces only 15 9.7 15 10.0 1.21 0.43, 3.35 0.89 0.28, 2.80
Exposed at both residences and workplaces 92 59.4 77 51.3 1.41 0.72, 2.73 0.98 0.42, 2.28

Chlorpyrifos

Unexposed 35 22.6 59 39.3 1.00 --- 1.00 ---
Exposed at residences only 17 11.0 9 6.0 3.91 1.29, 11.85 4.58 1.42, 14.80
Exposed at workplaces only 21 13.6 18 12.0 2.90 1.12, 7.54 3.70 1.31, 10.51
Exposed at both residences and workplaces 82 52.9 64 42.7 2.27 1.18, 4.38 3.22 1.38, 7.53

Diazinon

Unexposed 37 23.9 45 30.0 1.00 --- 1.00 ---
Exposed at residences only 8 5.2 9 6.0 1.06 0.32, 3.50 0.73 0.21, 2.56
Exposed at workplaces only 18 11.6 17 11.3 1.23 0.47, 3.26 0.86 0.30, 2.52
Exposed at both residences and workplaces 92 59.4 79 52.7 1.30 0.69, 2.45 0.81 0.35, 1.84

1,3-Dichloropropene

Unexposed 112 72.3 109 72.7 1.00 --- 1.00 ---
Exposed at residences only 8 5.2 6 4.0 1.40 0.38, 5.14 1.06 0.27, 4.17
Exposed at workplaces only 13 8.4 15 10.0 0.72 0.28, 1.86 0.47 0.17, 1.27
Exposed at both residences and workplaces 22 14.2 20 13.3 0.86 0.39, 1.91 0.58 0.25, 1.37
* Adjusted for age, SES (quintiles), body mass index (<25, 25–29, and ≥30 kg/m2), age at menarche (<12, 12, >12 years), age at menopause (<45, 45–54, ≥55), number of births (0, 1,
2, and ≥3), oral contraceptive use (never, 1–5 years, ≥5 years), menopausal hormone therapy use (never used, estrogen only, progesterone only, estrogen plus progesterone, or a
mixture of treatments), ever consumed alcohol at least once a week (yes or no), vigorous physical activity (0, 1–6, 7–13, or ≥14 hours per week), and number of years lived in Fresno,
Tulare, or Kern counties.
† Adjusted for the above covariates plus exposure at residences and/or workplaces to the other pesticides in the table (yes or no).
‡ Aldrin, chlordane, dicofol, dieldrin, endosulfan, lindane, methoxychlor, and toxaphene 127

Table 3.4. Measures of Association Between Breast Cancer and Exposure to Organochlorine Pesticides and Chlorpyrifos at Ages 20–39 and at Ages
40–59 Among Women in Fresno, Tulare, and Kern Counties using linked PUR data for 1974–2011.
Exposure Type
Cases
(n = 155)

Controls
(n = 150)
Adjusted
OR†

Adjusted
OR‡
95% CI P Value*
No. % No. % 95% CI
Organochlorines§

Ages 20–39
Unexposed ages 20–39 75 48.4 91 60.7 1.00 --- 1.00 --- 0.71
Exposed ages 20–39 80 51.6 59 39.3 1.15 0.61, 2.18 0.88 0.44, 1.75

Unexposed at ages 20's or 30's 75 48.4 91 60.7 1.00 --- 1.00 --- 0.96
Exposed at ages 20's or 30's 59 38.1 44 29.3 1.09 0.56, 2.16 0.84 0.41, 1.73
Exposed at ages 20's & 30's 21 13.6 15 10.0 1.51 0.58, 3.92 1.06 0.38, 2.92

Ages 40–59
Unexposed ages 40–59 65 41.9 74 49.3 1.00 --- 1.00 --- 0.75
Exposed ages 40–59 90 58.1 76 50.7 1.10 0.62, 1.99 0.89 0.45, 1.78

Unexposed at ages 40's or 50's 65 41.9 74 49.3 1.00 --- 1.00 --- 0.78
Exposed at ages 40's or 50's 44 28.4

37 24.7

1.13 0.58, 2.21 0.89 0.43, 1.85

Exposed at ages 40's & 50's 46 29.7

39 26.0

1.02 0.49, 2.11 0.89 0.37, 2.16

Chlorpyrifos
Ages 20–39
Unexposed ages 20–39 117 75.5 119 79.3 1.00 --- 1.00 --- 0.92
Exposed ages 20–39 38 24.5 31 20.7 0.83 0.39, 1.75 0.96 0.45, 2.08

Unexposed at ages 20's or 30's 117 75.5 119 79.3 1.00 --- 1.00 --- 0.67
Exposed at ages 20's or 30's 33 21.3 30 20.0

0.72 0.34, 1.55 0.85 0.39, 1.86
Exposed at ages 20's & 30's 5 3.2 1 0.7

4.46 0.40, 50.22 5.15 0.44, 60.03

128

Table 3.4. Continued:
Exposure Type
Cases
(n = 155)

Controls
(n = 150)

Adjusted
OR†

Adjusted
OR‡
95% CI P Value*
No. % No. % 95% CI
Chlorpyrifos

Ages 40–59
Unexposed ages 40–59 54 34.8 77 51.3 1.00 --- 1.00 --- 0.001
Exposed ages 40–59 101 65.2 73 48.7 2.52 1.39, 4.59 3.31 1.60, 6.85

Unexposed at ages 40's or 50's 54 34.8 77 51.3 1.00 --- 1.00 --- 0.001
Exposed at ages 40's or 50's 38 24.5

35 23.3 2.31 1.12, 4.76 2.82 1.28, 6.23
Exposed at ages 40's & 50's 63 40.7 38 25.3 2.72 1.36, 5.45 4.18 1.74, 10.05
* Pearson's χ2 test on n-1 degrees of freedom.
† Adjusted for age, SES (quintiles), body mass index (<25, 25–29, and ≥30 kg/m2), age at menarche (<12, 12, >12 years), age at menopause (<45, 45–54, ≥55), number of births (0, 1,
2, and ≥3), oral contraceptive use (never, 1–5 years, ≥5 years), menopausal hormone therapy use (never used, estrogen only, progesterone only, estrogen plus progesterone, or a
mixture of treatments), ever consumed alcohol at least once a week (yes or no), vigorous physical activity (0, 1–6, 7–13, or ≥14 hours per week), and number of years lived in Fresno,
Tulare, or Kern counties.
‡ Adjusted for the above covariates plus exposure at the other age period and ever exposed to the other pesticides in the table (yes or no).
§ Aldrin, chlordane, dicofol, dieldrin, endosulfan, lindane, methoxychlor, and toxaphene
129

Table 3.5. Comparison of Self-Reported Ever Lived on or Worked on a Farm and Ambient
Pesticide Exposure to Organochlorines, Chlorpyrifos, Diazinon, or 1,3-Dichloropropene Using
Linked PUR data from 1974–2011.
Ambient and self-reported exposure
Cases

Controls
(n = 155)

(n = 150)

P Value*
No. % No. %
Residential ambient exposure

Unexposed

Never lived on a farm 24 15.5

19 12.7

0.04

Lived on a farm 5 3.2

18 12.0

Exposed

Never lived on a farm 68 43.9

60 40.0

Lived on a farm 58 37.4

53 35.3

Occupational ambient exposure

Unexposed

Never worked on a farm 18 11.6

25 16.7

0.02

Worked on a farm 4 2.6

9 6.0

Exposed

Never worked on a farm 111 71.6

82 54.7

Worked on a farm 22 14.2 34 60.7

* Pearson's χ2 test on n-1 degrees of freedom.

130

Table 3.6. Comparison of Residential Histories† (1974 to Year of Diagnosis or Interview) in Breast
Cancer Cases (Diagnosed in 2007–2008) and Population-Based Controls (2001–2011), in Fresno,
Tulare, and Kern Counties.
Characteristic
Cases

Controls
P Value* (n = 155)

(n = 150)
No. % No. %
Years Lived on a Farm

0 124 80.0

124 82.7
0.83

1–9 16 10.3

14 9.3

≥10 15 9.7

12 8.0

Years Drinking Water Source was a Private Well

0 104 67.1

90 60.0 0.36

1–9 20 12.9

27 18.0

≥10 31 20.0

33 22.0

Years Lived in Fresno, Tulare, and Kern Counties

<10 13 8.4

12 8.0
0.11

10–19 10 6.5

20 13.3

20–29 33 21.3

39 26.0

≥30 99 63.9

79 52.7

Years Lived in California

<20 1 0.7

4 2.7
0.02

20–29 13 8.4

26 17.3

≥30 141 91.0

120 80.0

Years Lived in California but not in Fresno,
Tulare, or Kern Counties

<10 125 80.7

115 76.7
0.21

10–19 14 9.0

23 15.3

≥20 16 10.3

12 8.0

Years Lived Outside of California

0 127 81.9

109 72.7
0.13

1–9 23 14.8

31 20.7

≥10 5 3.2 10 6.7

* P value from Pearson’s χ2 tests for case-control differences in frequency distribution of
categorical variables. Fisher's exact test was used when the expected value for each cell was less
than 5.
† Residential histories included addresses recorded from 1974 to year of diagnosis or interview.
Missing addresses in residential histories were imputed with the nearest known address;
however, excluding these missing addresses did not change the estimates.

131

Table 3.7. Measures of Association Between Breast Cancer and Ambient Exposure to Selected Pesticides Among Women in Fresno, Tulare, and
Kern Counties Using Linked PUR Data for 1974–2011, When Exposure is Based on Address at Diagnosis Only or Complete Residential Histories.

Diagnosis Address Only Complete Residential Histories
Exposure Type
Cases
(n = 155)
Controls
(n = 150) Adjusted
OR*

Adjusted
OR†

Cases
(n = 155)
Controls
(n = 150) Adjusted
OR*

Adjusted
OR†

No. % No. % 95% CI 95% CI No. % No. % 95% CI 95% CI
Organochlorines‡

Unexposed 52 33.6 65 43.3 1.00 --- 1.00 ---

47 30.3 62 41.3 1.00 --- 1.00 ---
Exposed 103 66.5 85 56.7 1.09 0.62, 1.92 0.93 0.45, 1.91

108 69.7 88 58.7 1.32 0.74, 2.34 1.05 0.51, 2.18

Chlorpyrifos

Unexposed 63 40.7 79 52.7 1.00 --- 1.00 ---

56 36.1 77 51.3 1.00 --- 1.00 ---

Exposed 92 59.4 71 47.3 1.23 0.70, 2.14 1.21 0.56, 2.61

99 63.9 73 48.7 1.84 1.05, 3.24 2.27 1.07, 4.80

Diazinon

Unexposed 50 32.3 58 38.7 1.00 --- 1.00 ---

55 35.5 62 41.3 1.00 --- 1.00 ---

Exposed 105 67.7 92 61.3 1.06 0.60, 1.87 0.90 0.43, 1.86

100 64.5 88 58.7 1.19 0.68, 2.09 0.77 0.37, 1.63

1,3-Dichloropropene

Unexposed 125 80.7 126 84.0 1.00 --- 1.00 ---

125 80.7 124 82.7 1.00 --- 1.00 ---
Exposed 30 19.4 24 16.0 1.53 0.72, 3.26 1.50 0.65, 3.47 30 19.4 26 17.3 1.02 0.51, 2.03 0.76 0.36, 1.62
* Adjusted for age, SES (quintiles), body mass index (<25, 25–29, and ≥30 kg/m2), age at menarche (<12, 12, >12 years), age at menopause (<45, 45–54, ≥55), number of births (0, 1,
2, and ≥3), oral contraceptive use (never, 1–5 years, ≥5 years), menopausal hormone therapy use (never used, estrogen only, progesterone only, estrogen plus progesterone, or a
mixture of treatments), ever consumed alcohol at least once a week (yes or no), vigorous physical activity (0, 1–6, 7–13, or ≥14 hours per week), and number of years lived in Fresno,
Tulare, or Kern counties.
† Adjusted for the above covariates plus exposure at residences and/or workplaces to the other pesticides in the table (yes or no).
‡ Aldrin, chlordane, dicofol, dieldrin, endosulfan, lindane, methoxychlor, and toxaphene

132

Table 3.8a. Comparison of Length Resided in Fresno, Tulare, or Kern Counties and Length Resided in California in Breast Cancer Cases
(Diagnosed in 2007–2008) and Population-Based Controls (2001–2011), by Age and SES.
Characteristic
Lived in Fresno, Tulare, or
Kern Counties†
P Value*
Lived in California†
P Value*
≥30 Years

<30 Years

≥30 Years

<30 Years
(n = 178)

(n = 127)

(n = 261)

(n = 44)
No. % No. % No. % No. %
Age

55–59 43 24.2

31 24.4
0.76

66 25.3

8 18.2
0.55

60–64 44 24.7

33 26.0

67 25.7

10 22.7

65–69 54 30.3

32 25.2

73 28.0

13 29.6

70–74 37 20.8

31 24.4

55 21.1

13 29.6

SES

1 Lowest Quintile 30 16.9

20 15.8
0.56

44 16.9

6 13.6
0.52

2 38 21.4

37 29.1

63 24.1

12 27.3

3 44 24.7

32 25.2

61 23.4

15 34.1

4 56 31.5

32 25.2

78 29.9

10 22.7

5 Highest Quintile 10 5.6

6 4.7

15 5.8

1 2.3

* P value from Pearson’s χ2 tests for differences in frequency distribution of categorical variables. Fisher's exact test was used when the expected
value for each cell was less than 5.
† Missing addresses in residential histories were imputed with the nearest known address; however, excluding these missing addresses did not
change the estimates.
133

Table 3.8b. Comparison of Breast Cancer Cases (Diagnosed in 2007–2008) and Population-Based Controls (2001–2011) by Length Resided in
Fresno, Tulare, or Kern Counties, and by Age and SES.
Characteristic
Lived in Fresno, Tulare, or
Kern Counties ≥30 Years†
P Value*

Lived in Fresno, Tulare, or
Kern Counties <30 Years†
P Value*
Cases

Controls

Cases

Controls
(n = 99)

(n = 79)

(n =56)

(n = 71)
No. % No. % No. % No. %
Age

55–59 24 24.2

19 24.1
0.82

1 28.6

15 21.1
0.23

60–64 23 23.2

21 26.6

17 30.4

16 22.5

65–69 29 29.3

25 31.7

14 25.0

18 25.4

70–74 23 23.2

14 17.7

9 16.1

22 31.0

SES

1 Lowest Quintile 18 18.2

12 15.2
0.51

11 19.6

9 12.7
0.23

2 25 25.3

13 16.5

16 28.6

21 29.6

3 23 23.2

21 26.6

12 21.4

20 28.2

4 29 29.3

27 34.2

12 21.4

20 28.2

5 Highest Quintile 4 4.0

6 7.6

5 8.9

1 1.4

* P value from Pearson’s χ2 tests for differences in frequency distribution of categorical variables. Fisher's exact test was used when the expected
value for each cell was less than 5.
† Missing addresses in residential histories were imputed with the nearest known address; however, excluding these missing addresses did not
change the estimates.
134

Table 3.8c. Comparison of Breast Cancer Cases (Diagnosed in 2007–2008) and Population-Based Controls (2001–2011) by Length Resided in
California, and by Age and SES.
Characteristic
Lived in California ≥30
Years†
P Value*

Lived in California <30
Years†
P Value*
Cases

Controls

Cases

Controls
(n = 141)

(n = 120)

(n = 14)

(n = 30)
No. % No. % No. % No. %
Age

55–59 38 27.0

28 23.3
0.91

1 28.6

6 20
0.31

60–64 35 24.8

32 26.7

17 30.4

5 16.7

65–69 38 27.0

35 29.2

14 25.0

8 26.7

70–74 30 21.3

25 20.8

9 16.1

11 36.7

SES

1 Lowest Quintile 27 19.2

17 14.2
0.74

2 14.3

4 13.3
0.50

2 36 25.5

27 22.5

5 35.7

7 23.3

3 31 22.0

30 25.0

4 28,6

11 36.7

4 39 28.0

39 32.5

2 14.3

8 26.7

5 Highest Quintile 8 5.7

7 5.8

1 7.1

0 0.0

* P value from Pearson’s χ2 tests for differences in frequency distribution of categorical variables. Fisher's exact test was used when the expected
value for each cell was less than 5.
† Missing addresses in residential histories were imputed with the nearest known address; however, excluding these missing addresses did not
change the estimates.

135

Table 3.9. Comparison of Selected Characteristics in Non-Hispanic White Women Ages 55–74
Diagnosed with Invasive Breast Cancer in 2007–2008 Enrolled as Participants in the Case-
Control Study from Fresno, Tulare, and Kern Counties and Those in the Cancer Registry of
Central California (CCRC).
Characteristic
Case-Control Study
Participants,
Fresno, Tulare, and
Kern Counties
CRCC,
Fresno, Tulare,
and Kern Counties

CRCC,
All Central Valley
Counties*
(n = 155)

(n = 566)

(n = 963)
No. % No. %

No. %
Age

55–59 40 25.8

144 25.4

250 26.0

60–64 40 25.8

155 27.4

268 27.8

65–69 43 27.7

149 26.3

248 25.8

70–74 32 20.7

118 20.9

197 20.5
Neighborhood SES

1 Lowest quintile 29 18.7

131 23.1

200 20.8

2 41 26.5

157 27.7

271 28.1

3 35 22.6

134 23.7

289 30.0

4 41 26.5

112 19.8

171 17.8

5 Highest quintile 9 5.8

32 5.7

32 3.3
Stage

Localized 109 70.3

375 66.3

620 64.4

Regional 44 28.4

158 27.9

277 28.8

Distant or missing 2 1.3

33 5.8

66 6.9
Estrogen Receptor Subtype

ER (+) 110 75.3

433 76.5

721 74.9

ER (-) 29 19.9

98 17.3

169 17.6

Missing 7 4.8

35 6.2

73 7.6
Progesterone Receptor
Subtype

PR (+) 81 55.5

346 61.1

577 59.9

PR (-) 54 37.0

177 31.3

302 31.4
Missing 11 7.5 43 7.6

84 8.7
* All counties of the CRCC include Fresno, Kern, Kings, Madera, Mariposa, Merced, Stanislaus,
Tulare, and Tuolumne.

136

Figure 3.2. Trends in Pesticide Use in Fresno, Tulare, and Kern Counties, 1974–2011.

0
1,000,000
2,000,000
3,000,000
4,000,000
5,000,000
6,000,000
7,000,000
1974-1979 1980-1984 1985-1989 1990-1994 1995-1999 2000-2004 2005-2011
Pounds of Pesticide
Applied
Year
Organochlorines
Chlorpyrifos
Diazinon
137

Figure 3.3. Map of Approximate Residential Locations* at the Time of Diagnosis
for Breast Cancer Cases (2007–2008) and at the Time of Interview for
Population-Based Controls (2001–2011), in Fresno, Tulare, and Kern Counties.

*Mapped points have been masked to randomly selected distances within 500
meters from the geocoded location at the time of diagnosis or interview to protect
participants’ privacy.

138

Acknowledgments

Support for this study was provided by the National Institute of
Environmental Health Sciences grant through the Southern California
Environmental Health Sciences Center.

139

Chapter 3 References

1. Soto AM, Chung KL, Sonnenschein C. The pesticides endosulfan,
toxaphene, and dieldrin have estrogenic effects on human estrogen -
sensitive cells. Environ Res. 1994;102:380–383.
2. Key T, Appleby P, Barnes I, et al. Endogenous sex hormones and breast
cancer in postmenopausal women: reanalysis of nine prospective studies. J
Natl Cancer Inst. 2002;94(8):606–616.
3. Colditz GA. Epidemiology of breast -cancer. F indings from the Nurses '
Health Study. Cancer. 1993;71(4):1480–1489.
4. Key T, Pike MC. The role of oestrogens and progestagens in the
epidemiology and prevention of breast cancer. Eur J Cancer Prev.
1988;24(1):29–43.
5. Cohn BA, Wolff MS, Cirillo PM, et al. DDT and breast cancer in young
women: new data on the significance of age at exposure. Environ Res.
2007;115(10):1406–1414.
6. Ruder EH, Dorgan JF, Kranz S, et al. Examining breast cancer growth and
lifestyle risk fac tors: early life, childhood, and adolescence. Clin Breast
Cancer. 2008;8(4):334–342.
7. Kojima H, Katsura E, Takeuchi S, et al. Screening for estrogen and
androgen receptor activities in 200 pesticides by in vitro reporter gene
assays using Chinese hamste r ovary cells. Environ Res. 2003;112(5):524 –
531.
8. Okasha M, McCarron P, Gunnell D, et al. Exposures in childhood,
adolescence and early adulthood and breast cancer risk: a systematic
review of the literature. Breast Cancer Res Treat. 2003;78(2):223–276.
9. Andersen HR, Vinggaard AM, Rasmussen TH, et al. Effects of currently
used pesticides in assays for estrogenicity, androgenicity, and aromatase
activity in vitro. Toxicol Appl Pharmacol. 2002;179(1):1–12.
10. Gunier RB, Harnly ME, Reynolds P, et al. Agricultural pesticide use in
California: Pesticide prioritization, use densities, and population
distributions for a childhood cancer study. Environ Res.
2006;109(10):1071–1078.
11. Sohoni P, Sumpter JP. Several environmental oestrogens are also anti -
androgens. J Endocrinol. 1998;158(3):327–339.
140

12. Teitelbaum SL, Gammon MD, Britton JA, et al. Reported residential
pesticide use and breast cancer risk on Long Island, New York. Am J Ind
Med. 2007;165(6):643–651.
13. Mills PK, Yang R. Regression analysis o f pesticide use and breast cancer
incidence in California Latinas. J Environ Health. 2006;68(6):15–14.
14. Muir K, Rattanamongkolgul S, Smallman- Raynor M, et al. Breast cancer
incidence and its possible spatial association with pesticide application in
two counties of England. Annu Rev Public Health. 2004;118(7):513–520.
15. O'Leary ESE, Vena JEJ, Freudenheim JLJ, et al. Pesticide exposure and
risk of breast cancer: a nested case -control study of residentially stable
women living on Long Island. Environ Res. 2004;94(2):134–144.
16. Farooq U, Joshi M, Nookala V, et al. Self -reported exposure to pesticides in
residential settings and risk of breast cancer: a case -control study. Environ
Health. 2010;9(30):1–9.
17. Engel LS, Hill DA, Hoppin JA, et al. Pestic ide use and breast cancer risk
among farmers' wives in the Agricultural Health Study. Am J Ind Med.
2005;161(2):121–135.
18. Reynolds P, Hurley SE, Gunier RB, et al. Residential proximity to
agricultural pesticide use and incidence of breast cancer in Cal ifornia,
1988–1997. Environ Res. 2005;113(8):993–1000.
19. Reynolds P, Hurley S, Goldberg DE, et al. Regional variations in breast
cancer among California teachers. Epidemiology. 2004;15(6):746–754.
20. Brody JG, Aschengrau A, McKelvey W, et al. Breast cancer risk and
historical exposure to pesticides from wide -area applications assessed with
GIS. Environ Res. 2004;112(8):889–897.
21. Verner M-A, Bachelet D, McDougall R, et al. A case study addressing the
reliability of polychlorinated biphenyl levels measured at the time of breast
cancer diagnosis in representing early -life exposure. Cancer Epidemiol
Biomarkers Prev. 2011;20(2):281–286.
22. Alavanja MCR, Ross MK, Bonner MR. Increased cancer burden among
pesticide applicators and others due to pesticide exposure. CA: Cancer J
Clin. 2013;63(2):120–142.
23. Rull RP, Ritz B. Historical pesticide exposure in California using pesticide
use reports and land- use surveys: an assessment of misclassification error
and bias. Environ Res. 2003;111(13):1582–1589.
141

24. California Department of Pesticide Regulation. A Guide to Pesticide
Regulation in California. Sacramento, CA: California Environmental
Protection Agency; 2011. (http://www.cdpr.ca.gov/docs/pressrls/dprguide.
htm). (Accessed May 31, 2013).
25. Quirós-Alcalá L, Bradman A, Nishioka M, et al. Pesticides in house dust
from urban and farmworker households in California: an observational
measurement study. Environ Health. 2011;10(1):19.
26. Harnly ME, Bradman A, Nishioka M, et al. Pesticides in dust from homes
in an agricultural area. Environ Sci Technol. 2009;43(23):8767–8774.
27. Fenske RA, Chensheng L, Barr D, et al. Children's exposure to chlorpyrifos
and parathion in an agricultural community in central Washington State.
Environ Res. 2002;110(5):549–553.
28. Simcox NJ, Fenske RA, Wolz SA, et al. Pesticides in household dust and
soil: exposure pathways for children of agricultural families. Environ Res.
1995;103(12):1126–1134.
29. Lemaire G, Mnif W, Mauvais P, et al. Activation of alpha - and beta -
estrogen receptors by persistent pesticides in reporter cell lines. Life Sci.
2006;79(12):1160–1169.
30. Coumoul X, Diry M, Robillot C, et al. Differential regulation of cytochrome
P450 1A1 and 1B1 by a combination of dioxin and pesticides in the breast
tumor cell line MCF-7. Cancer. 2001;61(10):3942–3948.
31. Valerón PF, Pestano JJ, Luzardo OP, et al. Differential effects exerted on
human mammary epithelial cells by environmentally relevant
organochlorine pesticides either individually or in combination. Chem Biol
Interact. 2009;180(3):485–491.
32. Wofford P, Segawa R, Schreider J, et al. Community air monitoring for
pesticides. Part 3: using health- based screening levels to evaluate results
collected for a year. Environ Monit Assess. 2013;186(3):1355–1370.
33. Wang A, Cockburn M, Ly TT, et al. The association between ambient
exposure to organophosphates and Parkinson's disease risk. Occup
Environ Med. 2014;71(4):275–281.
34. Narayan S, Liew Z, Paul K, et al. Household organophosphorus pesticide
use and Parkinson's disease. Int J Epidemiol. 2013;42(5):1476–1485.
35. Wang AA, Costello SS, Cockburn MM, et al. Parkinson's disease risk from
ambient exposure to pesticides. Eur J Epidemiol. 2011;26(7):547–555.
142

36. Costello S, Cockburn M, Bronstein J, et al. Parki nson's disease and
residential exposure to maneb and paraquat from agricultural applications
in the central valley of California. Am J Ind Med. 2009;169(8):919–926.
37. Yost K, Perkins C, Cohen R, et al. Socioeconomic status and breast cancer
incidence in California for different race/ethnic groups. Cancer Causes
Control. 2001;12(8):703–711.
38. Marusek JC, Cockburn MG, Mills PK, et al. Control selection and pesticide
exposure assessment via GIS in prostate cancer studies. Am J Prev Med.
2006;30(2):S109–S116.
39. California Department of Pesticide Regulation. Summary of Pesticide Use
Report Data –2010. Sacramento, CA: California Environmental Protection
Agency; 2011. (http://www.cdpr.ca.gov/docs/pur/pur10rep/10sum.htm#
Development). (Accessed June 11, 2013)
40. California Department of Water Resources. California Land & Water Use.
Land Use Data. (http://www.water.ca.gov/landwateruse/lusrvymain.cfm).
(Accessed July 12, 2013).
41. Ritz B, Costello S. Geographic model and biomarker -derived measures of
pesticide exposure and Parkinson's disease. Ann N Y Acad Sci.
2006;1076(1):378–387.
42. Goldberg D. The USC WebGIS Open Source Geocoding Platform. USC GIS
Research Laboratory Technical Report No. 11; 2009.
(http://spatial.usc.edu/wp-content/uploads/2014/03/gislabtr111.pdf).
(Accessed June 17, 2014).
43. Goldberg DW, Wilson JP, Knoblock CA, et al. An effective and efficient
approach for manually improving geocoded data. Int J Health Geogr.
2008;7:60.
44. Woods N, Craig IP, Dorr G, et al. Spray d rift of pesticides arising from
aerial application in cotton. J Environ Qual. 2001;30(3):697–701.
45. Cox C. Pesticide drift: indiscriminately from the skies. Journal of Pesticide
Reform [electronic article]. 1995;15:1 –7. (http://sunridge.net/assets/pdf/
pesticide_drift.pdf).
46. Gunier RB, Ward MH, Airola M, et al. Determinants of agricultural
pesticide concentrations in carpet dust. Environ Res. 2011;119(7):970–976.
47. Ward MH, Lubin J, Giglierano J, et al. Proximity to crops and residential
exposure t o agricultural herbicides in Iowa. Environ Res.
2006;114(6):893–897.
143

48. Bradman A, Whitaker D, Quirós L, et al. Pesticides and their metabolites in
the homes and urine of farmworker children living in the Salinas Valley,
CA. J Expo Sci Environ Epidemiol. 2007;17(4):331–349.
49. Young HA. Use of a crop and job specific exposure matrix for estimating
cumulative exposure to triazine herbicides among females in a case- control
study in the Central Valley of California. Occup Environ Med.
2004;61(11):945–951.
50. Coronado GD, Thompson B, Strong L, et al. Agricultural task and exposure
to organophosphate pesticides among farmworkers. Environ Res.
2003;112(2):142–147.
51. London L, Myers JE. Use of a crop and job specific exposure matrix for
retrospective asses sment of long -term exposure in studies of chronic
neurotoxic effects of agrichemicals. Occup Environ Med. 1998;55(3):194–
201.
52. Cockburn M, Mills P, Zhang X, et al. Prostate cancer and ambient pesticide
exposure in agriculturally intensive areas in California. Am J Ind Med.
2011;173(11):1280–1288.
53. Weinberg CR, Moledor ES, Umbach DM, et al. Imputation for exposure
histories with gaps, under an excess relative risk model. Epidemiology.
1996;7(5):490.
54. Viswanath G, Chatterjee S, Dabral S, et al. Jo urnal of Steroid Biochemistry
and Molecular Biology. J. Steroid Biochem Mol Biol. 2010;120(1):22–29.
55. Kang HG, Jeong SH, Cho JH, et al. Chlropyrifos -methyl shows anti -
androgenic activity without estrogenic activity in rats. Toxicology.
2004;199(2-3):219–230.
56. Medjakovic S, Zoechling A, Gerster P, et al. Effect of nonpersistent
pesticides on estrogen receptor, androgen receptor, and aryl hydrocarbon
receptor. Environ Toxicol. 2014;29(10):1201–1216.
57. Mandal TK, Das NS. Testicular gametogenic and steroidogenic activities in
chlorpyrifos insecticide -treated rats: a correlation study with testicular
oxidative stress and role of antioxidant enzyme defence systems in
Sprague-Dawley rats. Andrologia. 2011;44(2):102–115.
58. Ventura C, Núñez M, Miret N, et al. Differential mechanisms of action are
involved in chlorpyrifos effects in estrogen -dependent or -independent
breast cancer cells exposed to low or high concentrations of the pesticide.
Toxicol Lett. 2012;213(2):184–193.
144

59. Usmani KA. Inhibition of the human liver microsomal and human
cytochrome P450 1A2 and 3A4 metabolism of estradiol by deployment -
related and other chemicals. Drug Metab Dispos. 2006;34(9):1606–1614.
60. Usmani KA, Rose RL, Hodgson E . Inhibition and activation of the human
liver microsomal and human cytochrome P450 3A4 metabolism of
testosterone by deployment -related chemicals. Drug Metab Dispos.
2003;31(4):384–391.
61. Office of Pesticide Programs. Pesticide news story: new use restrictions on
insecticide chlorpyrifos address bystander risk from spray drift; EPA’s
partial response to chlorpyrifos petition denies claims. Washington, DC:
U.S. Environmental Protection Agency; 2012. (http://www.epa.gov/
oppfead1/cb/csb_page/updates/2012/chlorpyrifos.html). (Updated
October 10, 2012. Accessed July 5, 2014).
62. Oostingh GJ, Wichmann G, Schmittner M, et al. The cytotoxic effects of the
organophosphates chlorpyrifos and diazinon differ from their
immunomodulating effects. J Immunotoxicol. 2009;6(2):136–145.
63. Agency for Toxic Substances and Disease Registry. ToxGuide for
Dichloropropenes. Atlanta, GA: U.S. Department of Health and Human
Services; 2011. (http://www.atsdr.cdc.gov/toxguides/toxguide -40.pdf).
(Accessed June 25, 2014).
64. Agency for Toxic Substances and Disease Registry. Priority Data Needs for
Dichloropropenes. Atlanta, GA: U.S. Department of Health and Human
Services; 2009. (http://www.atsdr.cdc.gov/pdns/pdfs/pdn_doc_40.pdf).
(Accessed June 25, 2014).
65. Agency for Toxic Substances and Disease Registry. Toxicological Profile for
Dichloropropenes. Atlanta, GA: U.S. Department of Health and Human
Services; 2008. (http://www.atsdr.cdc.gov/toxprofiles/tp40.pdf).
(Accessed June 25, 2014).
66. Ashley-Martin J, VanLeeuwen J, Cribb A, et al. Breast cancer risk,
fungicide exposure and CYP1A1*2A gene- environment interactions in a
province-wide case control study in Prince Edward Island, Canada. Int J
Environ Health Res Public Health. 2012;9(12):1846–1858.
67. El-Zaemey S, Heyworth J, Fritschi L. Noticing pesticide spray drift from
agricultural pesticide application areas and breast cancer: a case -control
study. Aust N Z J Public Health. 2013;37(6):547–555.
68. Maxwell SK. Generating land cover boundaries from remotely sensed data
using object -based image analysis: overview and epidemiological
application. Spat Spatiotemporal Epidemiol. 2010;1(4):231–237.
145

69. Rothman KJ, Greenland S, Lash TL. Modern Epidemiology. 3rd ed.
Philadelphia, PA: Lippincott, Williams & Wilkins; 2008.
70. Wacholder S, McLaughlin JK, Silverman DT, et al. Selection of controls in
case-control studies. I. Principles. Am J Epidemiol. 1992;135(9):1019 –
1028.

146

Chapter 4

4. Impact of Geocoding Certainty on the Pesticide Exposure
Assessment in a Case-Control Study of Breast Cancer Risk

4.1 Abstract
Although environmental epidemiologic studies frequently employ
Geographic Information Systems (GIS) -based methods to assess exposure to
contaminants using threshold distance or proximity to study subjects’ residences,
it is not common to evaluate the certa inty of the geocode locations or the
potential for misclassification in the exposure- disease associations. We assessed
the geocode certainty over historical time periods in a case -control study of
pesticide exposure on breast cancer risk and evaluated the potential impact on
the estimates of relative risk for exposure to selected pesticides of interest,
chlorpyrifos and a group of organochlorines. Compared to risk estimate that
included all locations, the odds ratio for exposure to organochlorines was
attenuated after excluding participants who were only exposed at locations with
low geocode certainty (centroids of zip code, city, state or unknown), but
unchanged for chlorpyrifos. Subjects were more likely to be exposed to
organochlorines earlier in their residential histories when they were also more
likely to reside at locations with low geocode certainty, whereas they were more
likely to be exposed to chlorpyrifos later in their residential histories when they
were also more likely to reside at locations with high geocode certainty. Our
analyses demonstrate that geocode certainty can affect the validity of GIS -based
147

exposure assessments, especially for historical exposures and locations that occur
further back in time.

4.2 Introduction
Geographic Information Systems ( GIS)-based methods used for
determining exposures like those described in the registry and pilot case- control
studies are often used to investigate spatial -temporal associations between
environmental contaminants and human health outcomes. Linking individuals to
sources of environmental exposures relies on locating subjects over time and
space and using GIS to create the spatial data that defines a subject’s exposure via
proximity to, intersection with the surface, or distance to where the exposure of
interest is (1). Accurately locating individuals is imperative in such studies since it
is the basis for correctly assigning the exposure that is used to assess the
exposure-outcome relationship.
Residential addresses are commonly used as proxy measures of the
subjects’ locations, and must first be geocoded to convert the address into it’s
corresponding spatial coordinates (i.e. latitude and longitude) to be employed in
the analysis. The quality of the geocoded locations depends on how detailed the
information was recalled or reported, in addition to how well these addresses
were resolved to their corresponding spatial data points by the geocoding
methods. We will use the term ‘geocode quality’ to mean how well the geocoding
method captures the true location defined by two aspects, 1) the positional
accuracy of the spatial location (how closely the geocoded coordinates match the
true location coordinates) and 2) the level of certainty determined by the
148

completeness of a given address and the ability of the geocoding methods to
resolve that address (ranging from a spatial point placed in the centroid of a
rooftop to the centroid of a country or unknown if no address information is
recalled or rep orted). The validity of the spatial -temporal findings, that is the
degree to which the results of the study are free from systematic error, depends
on the confidence one has in the quality of the geocoded data, but epidemiologic
studies rarely mention geocode quality, let alone assess the variation in geocode
quality on the estimated exposure-disease associations (2).
4.2.1 Positional accuracy of geocoded data
The process of geocoding is based on several assumptions: 1) all addresses
in the address range exist and have a position along the street segment, 2) the tax
assessor parcels corresponding to the addresses are the same size, 3) the centroid
of the parcel is most appropriate spatial point location, and 4) the input dataset
contains all information to identify an address to the parcel level (3). Several
studies have sought to quantify the amount of positional errors associated with
the accuracy of the geocode based on these assumptions, comparing addresses
geocoded using a geo -referenced street file to “gold standard” locations
determined by Global Positioning System (GPS) satellite receivers or aerial
imagery. Cayo et al. found average positional errors of 58 meters in urban and
614 meters in rural areas, and maximum errors of 299 meters in urban and 3,567
meters in rural areas (4). Ward et al. reported average er rors of 77 meters in
urban and 210 in rural areas, with maximum errors of 687 meters in urban and
1,731 meters in rural areas (5). Jones et al. found an average of 6 6 meters in
urban and 44 meters in rural areas, with maximum errors of 4,694 meters in
149

urban and 15,171 meters in rural areas (6). Bonner et al. examined historical
addresses and found average errors of 96 meters in urban and 129 meters in rural
areas, with maximum errors of 1,551 meters in urban and 2,552 meters in rural
areas (7).
These findings suggest that the positional error of the geocoding methods
may be substantial and that exposure assessments based on short distances will
call for more precise locations with h igher levels of positional accuracy to
produce valid risk estimates (at least as valid as can be estimated based on the
spatial resolution of the exposure data). In addition, locations in rural areas may
be more likely to be misclassified than locations in urban areas if the distance of
the exposure assessment is not sufficient to account for such positional variability
in the measured geocoded location compared to the true location. Rural areas are
also more prone to geocoding errors of larger distances si nce they tend to have
longer street segments with fewer intersections and houses that are less evenly
spaced compared to dense street networks typically found in cities where streets
tend to be aligned in grid patterns, and this reduces the ability of the geocoding
method to interpolate a spatial location based on the assumptions of the
geocoding process previously mentioned (4).
The studies that assess the positional accuracy of geocoded locations reli ed
on the availability of “gold standard” locations and excluded any addresses that
could not be geocoded to the tax assessor parcel level (to meet the fourth
assumption of the geocoding process). In reality, automated geocoding methods
using geo -referenced street files are routinely used in epidemiologic studies
because they are more feasible and efficient than using resource- intensive
150

methods such as GPS or aerial imagery. In addition, such “gold standard”
methods are typically not available for historical locations. A web -based
interactive GIS-approach is another method that can be used that does not rely
on “gold standard” locations that has been shown to improve the geocode quality
of the spatial data in a cost effective manner by allowing users to manu ally
correct previously geocoded data to pinpoint more precise locations using
additional information provided in the address such as cross streets or landmarks
(8). GIS -based epidemiologic studies are employing such manual corr ection
methods to the geocoding process with increasing frequency (9-12).
4.2.2 Level of certainty of geocoded data
In addition to positional accuracy, error resulting from what we define
here as the level of geocode certainty can also affect the geocode quality (3). The
level of geocode certainty depends on the completeness of the address
information and how well it can be resolved to the level of tax assessor parcel
coordinates. For example, the address data may contain information that is
missing, extraneous, or misspelled, include locations or roads that no longer exist
or were developed after the reference map, or describe street intersections or
landmarks instead of complete street numbers and names, so that the geocoding
method resolves the addresses to the geo graphic coordinates corresponding to
various levels of geocode certainty. The levels of geocode certainty can range from
a spatial point placed in the centroid of a parcel (using commercial geocoding
methods) or a rooftop (using manual resolution, GPS, or aerial imagery methods)
to the centroid of a country or unknown if no address information can be
151

matched to the geographical reference dataset. We will consider the centroid of a
rooftop to have the highest geocode certainty.
Previous studies have reported simply the rate of successful matches (the
percentage of addresses geocoded) for the addresses input to the geocoding
process (13-15), but there is no standard for an acceptable geocoding match rate
(14) and studies typically do not include details concerning the various levels of
geocode certainty for all of the addresses geocoded. Matching rates using
commercially available software range from 44% to 84% (16), depending on the
criteria used to match (street interpolatio n, parcel centroid, or address points for
example), suggesting that there can be wide variation in the matching criteria and
the level of geocode certainty included or not included in the match. The distance
between the highest level of geocode certainty ( centroid of a rooftop) and other
levels of geocode certainty can vary depending on the level of certainty that the
address is able to be resolved to by the geocoding method, as well as whether the
landscape is rural or urban.
Bow et al. compared two levels of geocode certainty, the centroid of the zip
code and the actual street address, in order to assess the distance errors
associated with using spatially aggregated units such as postal codes as proxies
for residential addresses in an urban setting and found an average distance of 146
meters, with a maximum distance error of 4,482 meters between the two point
locations (17). Healy et al. also assessed geocoding distance errors between the
centroid of the postal code and the residential street address and found median
errors up to 109 meters in urban and 1,363 meters in rural areas (18). These
studies suggest that the distance errors between various level s of geocode
152

certainty may be large and caution the use of certain levels of geocode certainty
such as zip codes when the distance for defining exposure is small relative to the
magnitude of the potential errors. Healy et al. observed that the distances
between postal code and street address locations was most problematic for
locations in rural areas and advised against using postal codes to represent
locations for threshold distances or density buffers as small as 500 meters (18).
Methods that simply restrict analyses to only those locations with high geocode
certainty however, can reduce the statistical power of the analysis or even lead to
selection bias (15,19).
4.2.3 Evaluating geocode quality in exposure assessments
Several epidemiologic studies have sought to examine the impact of
geocode quality on the exposure assessment for various environmental expos ures
including air pollutants associated with distance to high traffic roadways,
proximity to agricultural crops, and drinking water contamination. Whitsel et al.
observed non-differential misclassification for exposure to air pollutants within a
100-meter distance to the nearest highway that biased a hypothetical association
towards the null (20), while Zand bergen, P. found geocoding errors tended to
overestimate proximity to major roads at distances up to 250 meters, introducing
substantial bias in the estimates of exposure to traffic -related air pollution (21).
Ward et al. assessed crop occurrence within several different buffer distances of
residential locations and found that geocoding errors were likely to misclassify
exposure if based on proximity with a 100 -meter buffer, but not for buffer
distances greater than 250 meters (5). Jones et al. observed attenuation of the
odds ratios for crops within 250 and 500-meter buffers of residences, particularly
153

for residences in rural areas and when crop prevalence was low (<50%) (6).
Vieira et al. found that estimates were not affected by geocode quality in an
assessment of drinking water contamination since addresses were typically
geocoded to the correct street which tended to share the same water supply (22).
These studies highlight that the direction a nd magnitude of the bias
resulting from misclassification of exposure due to poor geocode quality depends
on the distance associated with the positional error or level of geocode certainty
as well as the spatial refinement of the exposure assessment that m ay be
sufficient to account for variation in geocode quality as indicated in the study of
drinking water contamination, or insufficient when exposure was defined by
small distances (<500 meters) as observed in the studies of air pollution and
proximity to field crops.
Ideally, an assessment of the variation in geocode quality locations would
evaluate the exact relationship between the exposure, the true location, and the
geocoded location for every person -location in the analysis. We do not always
have information on the true location however, and labor -intensive methods to
improve the positional accuracy may not be feasible. Methods are needed for
evaluating the impacts of geocode quality that are simple and can be easily
incorporated into epidemiologic studies as part of routine assessment (3). W e
propose evaluating the geocode quality by estimating the level of geocode
certainty obtained from the geocoded output. Readily available information can
be found from the output of the geocoding method that assigns a level of geocode
certainty for every geocoded location, and this can be used as a guideline to
154

indicate how confident one can be that the geocoded location represents the true
location.
4.2.4 Evaluating geocode certainty in a pilot study of pesticide
exposure on the risk of breast cancer
The purpose of this study was to evaluate the geocode quality in a pilot
case-control study of breast cancer risk and exposure to pesticides in a highly
agricultural region (Chapter 3), where exposure was based on proximity of
participants’ residential address timelines within a 500 -meter buffer to historic
records of commercial pesticide application. This buffer distance was chosen
based on studies that found measurable concentrations of pesticides drifted fr om
commercial application (23-25), and were detectable in household dust of nearby
homes (26). We assessed geocode quality by first ensuring that the geocoded
locations were as close as possible to the true locations even though “gold
standard” locations for the historical residential addresses were not availab le by
employing manual resolution methods to improve the quality of all the geocoded
locations (note that previous studies have typically used manual resolution
methods only for those addresses that do not readily match using automated
geocoding methods but here we used it for all addresses) and then evaluating the
level of geocode certainty provided in the output of the geocoding method.
The aims of this study were to assess the level of geocode certainty over
historical time periods among cases and controls, to consider possible biases that
may be introduced by using poor geocode certainty data, and to evaluate the
potential impact of geocode certainty on the estimates of relative risk between
breast cancer and exposure to selected pesticides of interest, chlorpyrifos and a
155

group of organochlorines. We accounted for multiple addresses and geocode
certainty across participants’ historical timelines, examining first the overall
proportion of various levels of geocode certainty in participants’ timelines and
second, the year -specific exposures that occurred along historical timelines with
corresponding year-specific levels of geocode certainty to identify the participants
that were most likely misclassified as a result of their geocode certainty.
4.2.5 Illustration of misclassification of pesticide exposure from
inaccurate geocode location information
Figure 4.1 illustrates the potential for exposure misclassification when
exposure is based on locations that have various levels of geocode certainty in an
epidemiological study of pesticide exposure where exposure is defined by a
greater than zero value for the total pounds of pesticides applied within a 500-
meter buffer around each location. The figure outlines a zip code tabulation area
(ZCTA) in a highly agricultural region of Fresno County with the centroid of the
ZCTA identified by a point and a 500-meter buffer drawn around it. Areas shaded
gray indicate densely populated places where presumably many of the residents
live, such as near the town of Selma. Green land -use polygons indicate
agricultural areas where pesticides are applied, and areas shaded orange denote
where the pesticide of interest is applied, for exa mple, chlorpyrifos. In this
example, if a study participant’s location is geocoded to the centroid of the ZCTA
then they would not be considered exposed to chlorpyrifos. There are many areas
within the ZCTA however where someone might live and be exposed to
applications of chlorpyrifos. Therefore, using the centroid of the ZCTA to identify
a participant’s exposure could have underestimated (or overestimated ) their
156

exposure by incorrectly identifying their actual location , leading to
misclassification of a person’s true exposure to pesticides. On the other hand, if a
participant resided in a metropolitan zip code where pesticides were not applied
anywhere in the zip code then we would correctly identify them as unexposed
regardless of level of geocode certainty (up to the ZCTA centroid level in our
example), but simply excluding those persons would reduce the statistical power
of the association and may lead to selection bias in the study. Thus, not
considering the certainty of the geocoded locations, when these locations are
used as the basis for determining exposure in an epidemiologic study, could
result in an estimate of relative risk that has a different magnitude or even a
different direction than the one observed.
Figure 4.1. A ZCTA in Fresno County With the Centroid Marked With a Point and
500-Meter Circular Buffer Around it. The Green Topography Indicates
Agricultural Areas and the Orange Shapes Represent Fields Where Chlorpyrifos
was Applied.

Centroid of ZCTA plus
500-meter buffer
Densely populated area
Agricultural land-use
Chlorpyrifos application

157

4.3 Materials and Methods
4.3.1 Case and Control Recruitment
As described previously (Chapter 3), c ases were identified from women
diagnosed with histologically confirmed breast cancer in 2007 –2008 in the
counties of Fresno, Tulare, or Kern from the Cancer Registry of Central California
(CRCC). Controls were obtaine d from a study of Parkinson’s disease being
conducted in the same geographic area from 2001 –2011 (27,28). Controls were
eligible for the Parkinson’s disease study if they lived in California for at least five
years prior to the study, were at least 35 years old and resided in Fresno, Tulare,
or Kern C ounties and did not have Parkinson’s disease. To match control
selection criteria, cases were excluded if they did not live in California for at least
five years prior to diagnosis . Cases and controls were included if they were aged
55 to 74 years (so as to include only postmenopausal breast cancer, which is more
likely of hormone-related origin), non-Hispanic white ethnicity, and had not been
diagnosed with ovarian, uterine, other female reproductive cancers . A total of 155
participants with breast cancer and 150 controls enrolled in the study and
completed interviews.
4.3.2 Source of exposure data
All controls (n = 150) and the majority of cases ( n = 111) were interviewed
over the telephone, with an additional 44 cases opting to complete a mailed
questionnaire with follow -up by telephone to clarify or complete responses. All
study participants were mailed a timeline to complete their historical residential
information (addresses and dates) prior to their telephone interviews.
158

4.3.3 Geocode certainty
Residential historical addresses and dates lived at each address were
collected from all study participants, geocoded using Texas A& M GeoServices
(available at http://geoservices.tamu.edu), and manually resolved using web -
based interactive geocoding methods developed by Goldberg et al. to more
precisely identify a point location (referenced by its latitude and longitude
coordinates) us ing additional information provided by the participants such as
the cross streets and landmarks (8,29). We noted the level of geocode certainty
for each location provided in the geocoding output and categorized the levels of
certainty to include the centroid of a building (center of a visible residential
rooftop), tax assessor parcel, street segment, street intersection, zip code
tabulation area, city, or state, or noted that the location was unable to be
geocoded. For simplicity in our analysis that involved multiple locations (with
corresponding geocode certainties) occurring over decades, we chose to stratify
geocode certainty into binary cut points, with the centroids of a building, tax
assessor parcel, street segment, and street intersection considered to have “high
geocode certainty” and those geocoded to the centroids of a zip code tabulation
area, city, or state or that were unable to be geocoded considered to have “low
geocode certainty.”
4.3.4 Historical geocode certainty data
One novelty of our case -control design was that we obtained historical
data on residence and, as a result, need to consider the geocode certainty
associated with each of the historical point locations. Note that we used
geocoding methods that referenced current geographical data even for historical
159

locations, which also adds some additional uncertainty in the precision of the
geocoded location estimates. In order to visualize historical geocoded data, we
created a location -year grid that included one row for each study participant’s
residential timeline based on where they lived each year from the time they were
born until the year of diagnosis or interview, and assigned a level of geocode
certainty for each location -year. Since location dates were listed by year and in
order to account for residential mobility within a given year, we made the
assumption that all subjects remained at their primary listed locations for any
given year for the duration of that year. Since levels of geocode certainty can vary
over time, we evaluated various proportions at which a person’s timeline was a
mixture of high and low levels of geocode certainty.
4.3.5 Historical pesticide exposure assessment
The geocoded locations were used as the basis for dete rmining exposure to
nearby commercial pesticide application using a GIS -based approach that
combines California pesticide data and land use maps to produce estimates of
pesticide exposure within a 500 -meter buffer around each location . Details are
described elsewhere (30-33). Briefly, California maintains Pesticide Use
Reporting (PUR) data from 1974 for all restricted -use pesticides and from 1990
for all pesticides applied in commercial agriculture. Each PUR record includes
the name of the pesticide, the pounds applied, the date and method of
application, an d the location of application as referenced by the Public Land
Survey System (PLSS), which divides the land into a grid of roughly one square
mile units. In addition to the PUR, California’s Department of Water Resources,
Division of Planning and Local Ass istance, maintains an extensive, statewide set
160

of land- use and crop cover sur veys by county from 1976 to 2012 (34). Each
county is surveyed at 6 to 10-year intervals, documenting crop areas and types of
land-use such as fields, vineyards and orchards. Land-use maps and PUR data
were matched by crop type to indicate where the pesticides were most likely
applied within the PLSS section using an algorithm previously developed and
validated (23,24,35).
Historical exposure to ambient pesticides was calculated by summing the
annual density (total pounds of a pesticide’s active ingredients applied per acre)
of applied pesticide within a 500 -meter buffer around each residential location
(35). Residential addresses that were missing because of poor participant recall
(only 3 cases and 4 controls had more than one third of their residential histories
missing) or that were located outside of California were considered “unexposed to
pesticides.” Since California has the highest agricultural productivity anywhere in
the nation, there is a low probability that someone living outside of California
would have been exposed to commercial agricultural pesticides in the ambient
air.
4.3.6 Assessing geocode certainty plus exposure over time
We compared the geocode certainty histories between cases and controls
with a timeline beginning in 1974 , so as to include all years with complete
pesticide information recorded by the state , until the year of diagnosis for the
cases or the year of interview for the controls. First, we evaluated the geocode
certainty histories stratified by ever exposed to the selected pesticides over this
time period, where an individual was considered ever exposed if they were
exposed at any point in their residential timeline. We chose to assess a selected
161

group of organochlorine pesticides and chlorpyrifos since historical records
indicate very different time frames for exposure (Figure 4.2). Organochlorines
were more heavily applied during the early years of our exposure assessment
(prior to bans on usage for several kinds), while chlorpyrifos was applied during
later years in the exposure assessment.
We then assessed what impact the geocode certainty had on the estimates
of relative risk when exposure was defined as ever exposed to the pesticides of
interest over the study subject’s residential timeline. Since levels of geocode
certainty can vary over time in our historical exposure assessment, we evaluated
various proportions at which a study subject ’s timeline was a mixture of high and
low levels of geocode certainty over the total number of years the participant was
at risk in th e study, including having 5%, 10%, 15%, 20%, and 25% low geocode
certainty, and at which point these made a difference on the estimates of relative
risk. We investigated several cut points in the data and chose to dichotomize the
geocode certainty level at the 10% proportion of low geocode certainty due to
sparseness of the cell counts among those with any proportion of low geocode
certainty in their timelines. We placed the location -year grids showing the
proportion of low and high geocode certainty timeli nes side by side for cases and
controls in order to visually observe whether the proportion and pattern of
geocode certainty was similar for cases and controls.
Secondly, we evaluated the impact of low and high levels of geocode
certainty on the estimates of relative risk by refining our definition of exposure to
identify each person-year in the subjects’ timelines that they were exposed to the
162

pesticides of interest and whether or not that exposure occurred at a residence of
low or high geocode certainty.
4.3.7 Statistical analyses
Unconditional logistic regression was used to estimate ambient pesticide
exposure on risk of breast cancer in postmenopausal women. Odds ratios (OR)
and 95% confidence intervals (CI) were calculated for study participants exposed
to specific pesticides compared to those not exposed. An individual was
considered exposed if the density of applied pesticide within the buffer area was
not null during the period from 1974 until the year of diagnosis for cases and the
year of interview for controls. Formal tests for interaction were conducted to
assess statistical differences in associations by geocode certainty by adding a
multiplicative term in the model with the main exposure for the proportion at
which a person’s timeline was a mixture of levels of geocode certainty (geocode
certainty*exposure), and evaluating the Wald χ
2
test for significance. All analyses
were conducted using SAS, version 9.3 software (SAS Institute, Inc., Cary, North
Carolina), and P values are 2 sided.

4.4 Results
4.4.1 Geocode certainty
Figure 4.3 shows side -by-side timelines of location-year geocode certainty
for cases and controls where cells with high geocode certainty are shaded in green
and those with low geocode accuracy are shaded in red. The majority of cases and
controls had all residential locations in their timeline geocoded to high geocode
certainty (61.9% of cases and 62.7% of controls) . Another 13.5% of cases and
163

18.7% of controls had only 1 to 4 years in their timelines with locations with low
geocode certainty and the rest of the locations with high geocode certainty, while
24.5% of cases and 18.7% of controls had 5 or more years in their timelines with
locations of low geocode certainty. Locations of low geocode certainty were more
likely to occur earlier in the timelines rather than in more recent years. It also
appeared that more cases than controls reported residential addresses that
resolved to low geocode certainty, with 7.7% of cases and 3.3% of controls having
over half of their timeline locations with low geocode certainty.
4.4.2 Historical geocode certainty and pesticide exposure assessment
When we stratified by exposure status defined by ever exposed at any
point in the residential timeline, cases ever exposed to organochlorines were
more likely to have low geocode certainty than controls ever exposed to
organochlorines (Figures 4.4a and 4.4b). Cases ever exposed to chlorpyrifos were
also more likely to have l ow geocode certainty than controls ever exposed to
chlorpyrifos, but to a lesser degree.
There was increased risk of breast cancer for exposure to organochlorines
(OR = 1.62, 95% CI: 1.10, 2.60) and exposure to chlorpyrifos (OR = 1.87, 95% CI:
1.18, 2.95), based on inclusion of all types of geocode certainty (Table 4.1). When
stratified by the proportion of 10% low geocode certainty locations in a person’s
timeline, those with less than 10% low geocode certainty locations in their
timelines had attenuated risk estimates and those with 10% or more low geocode
certainty locations in their timelines had inflated risk estimates compared to the
risk estimates that included all levels of geocode certainty (Table 4.2). Compared
to unexposed, those exposed to organochlorines with 10% or more low geocode
164

certainty locations in their timelines had a three -fold increased risk of breast
cancer (OR = 3.01, 95% CI: 0.99, 9.14) , whereas those exposed to
organochlorines with less than 10% low geocode certainty locations in t heir
timelines showed a much more moderate increase in magnitude (OR = 1.3 2, 95%
CI: 0.77, 2.26). Those exposed to chlorpyrifos with 10% or more low geocode
certainty locations in their timelines compared to unexposed had a more than
two-fold increase in r isk of breast cancer (OR = 2.36, 95% CI: 0.89, 6.20) , while
those with less than 10% low geocode certainty locations in their timelines were
1.7 times as likely to develop breast cancer (OR = 1.67, 95% CI: 0.98, 2.83). The
differences between the magnitudes of relative risk between the strata of 10%
proportion of low geocode certainty locations appeared more pronounced for
exposure to organochlorines than for chlorpyrifos, however all the 95%
confidence intervals contained the null value and therefore we cannot rule out the
role of chance in these findings. Although the magnitude of the risk by geocode
certainty strata appear to differ, the tests for interaction were not statistically
significant (P = 0.19 for exposure to organochlorines and P = 0.54 for exposure to
chlorpyrifos).
4.4.3 Assessing geocode certainty and exposure over time
We observed that stratifying by a 10% proportion of low geocode certainty
locations in the timelines resulted in study participants who had a mixture of
both low and high geocode certainty locations in their residential histories
(Figures 4.5a and 4.5b). S ince these participants could have possibly been
exposed at their locations of high geocode certainty as well as their locations of
low geocode certainty during their timelines, we examined our refined definition
165

of exposure that considered exposure at eac h location -year in the subjects’
residential histories. Among those participants that have 10% or more low
geocode certainty locations in their timelines, 61.7% of cases and 61.1% of
controls were actually exposed during the years when they reported locati ons of
high geocode certainty in their timelines (Figures 4.6a and 4.6b).
Examining participants who were only exposed at locations in their
timelines with low geocode certainty, there were 12 cases and 3 controls exposed
to organochlorines and 1 case and 2 controls exposed to chlorpyrifos (Figures
4.7a and 4.7b). After excluding those participants who were only exposed at
locations in their timelines with low geocode certainty, we observed that the
magnitude of risk among those with 10% or more low certain ty geocode locations
in their timelines decreased for exposure to organochlorines ( OR = 2.42, 95% CI:
0.77, 7.55) and increased slightly for exposure to chlorpyrifos (OR = 2.52, 95%
CI: 0.94, 6.70) (Table 4.3). Although the magnitudes of relative risk seem ed to
differ by geocode certainty strata, the 95% confidence intervals contained the null
value and the tests for interaction were not statistically significant ( P = 0.35 for
exposure to organochlorines and P = 0.47 for exposure to chlorpyrifos).
Compared to estimates of relative risk based on exposures at all locations
regardless of geocode certainty level, when we excluded those participants who
were exposed only at low geocode certainty locations in their timelines, the
estimate of relative risk was attenuated for exposure to organochlorines (OR =
1.49, 95%CI: 0.92, 2.40) and the resulting 95% confidence interval contained the
null value. The estimate for exposure to chlorpyrifos on the other hand was
166

nearly the same as the estimate that included locations of all geocode certainty
levels (OR = 1.90, 95% CI: 1.20, 3.01) (Table 4.4).

4.5 Discussion
The results of this study suggest that the certainty of the geocoded location
information that is the basis for determining pesticide exposure may impact the
estimates of relative risk in the association with breast cancer and that the
magnitude of the change (and possibly the direction) depends on the type of
pesticide as well as how far in the past the exposure and the location occurred.
We observed that compared t o the relative risk estimates that included all
locations regardless of geocode certainty level, the estimates that excluded
subjects based on an overall proportion of geocode certainty in their histories
(that had 10% or more locations with low geocode ce rtainty in their timelines)
were attenuated (Table 4.2). Since the overall proportion of geocode certainty for
the historical residential address timeline included individuals who were exposed
at both low and high geocode certainty locations, we analyzed individual location-
year exposures and excluded subjects that were only exposed at their low geocode
certainty locations throughout their timelines, resulting in an attenuated estimate
for exposure organochlorines and essentially no change in the estimate f or
exposure to chlorpyrifos. Subjects were more likely to be exposed to
organochlorines earlier in their timelines when they were also more likely to
reside at locations with low geocode certainty (Figure 4.3a), whereas subjects
were more likely to be exposed to chlorpyrifos later in their timelines (more
167

recent years) when they were more likely to reside at locations with high geocode
certainty (Figure 4.3b).
4.5.1 Misclassification of pesticide exposure
When the exposure to nearby pesticide applications is based on less
detailed location information identified by having low geocode certainty, a
potential source of exposure misclassification can result from incorrectly
identifying a person’s true exposure to specific pesticides used, resulting in
changes to t he magnitude and possibly the direction of the estimates. Since the
exposure assessment in this study utilized a 500 -meter buffer to estimate
exposure to pesticide applications in highly agricultural areas that are mainly
rural, we assumed that low geocode certainty locations (geocoded to the centroid
of a zip code, city, or state or t hat were unable to be geocoded) were unreliable
and likely to be misclassified for exposure. We therefore consider ed the estimates
of relative risk that excluded subjects who were exposed only at low geocode
certainty locations in their timelines to be less biased than those that included
exposures at all locations regardless of geocode certainty, however our confidence
intervals are wide given the small sample size of the stud y (Table 4.4). More
participants were exposed to organochlorines only at their locations of low
geocode certainty than to chlorpyrifos (Figures 4.6a and 4.6b, Table 4.4), and this
resulted in an inflated risk estimate based on all levels of geocode certain ty
combined that was particularly pronounced for organochlorines.
On the other hand, the estimates could also underestimate the true risk by
missing those truly exposed but whose locations were resolved to low geocode
certainty if there was no exposure, such as the centroid of a zip code placed where
168

there is a post office in the most commercially dense area of the zip code. Over
twice as many cases as controls had 50% or more of the locations in their timeline
with low geocode certainty (7.7% of cases and 3.3% of controls), and it is possible
that these individuals would have been exposed if we had more precise location
information for them. We assessed the impact of excluding those who were
unexposed at their low geocode quality locations (and who were not also exposed
at any of their high geocode quality locations within their timeline) (Figures 4.8a
and 4.8b). Excluding those that were unexposed at their low geocode quality
locations attenuated the odds ratios for exposure to the organochlorines (OR =
1.32, 95% CI: 0.79, 2.24) and for exposure to chlorpyrifos (OR = 1.73, 95% CI:
1.04, 2.89) (Table 4.5).
4.5.2 Advantages to our methods for evaluating geocode certainty
In this study we assessed the level of geocode certainty over historical time
periods and evaluated the potential for misclassification resulting from low
geocode certainty locations on the estimates of relative risk. Two previous ly
published studies of pesticide exposure on the risk of breast cancer using GIS -
based methods neither mentioned nor estimated geocode certainty at all (36,37).
Several studies of pesticide exposure on risk of other diseases evaluated geocode
quality by comparing the proportions of “high” geocode certainty locations
(defined as centroids of a street segment, street intersectio n, parcel, or building )
resulting after manual resolution between cases and controls , and noting that
similar proportions between cases and controls suggested that geocode certainty
would not likely account for differences in the observed estimates (30-33). One
previously published GIS -based study that also used historical residential
169

addresses to assess ambient pesticide application on the risk of breast cancer
noted the geocode certainty of the locations, but did not account for geocode
certainty in the analyses (38). In that study by Brody et al., the majority of
addresses (83.0%) were geocoded to a rooftop, parcel or nearest parcel or cluster
of rooftops ( n = 3,794), the remaining addresses identified by cross streets or
landmarks were then manually resolved to the nearest rooftop, nearest parcel, or
midpoint of the street (12.9%), wh ile those that could not be geocoded by those
criteria were excluded from the analyses (4.3%). None of these studies directly
measured the impact of historical geocode certainty on the exposure- outcome
associations nor conducted any analyses to evaluate th e degree to which
misclassification might affect the true exposure prevalence.
The strengths of the method s used in this study is that it is we estimated
misclassification due to geocode certainty accounting for all addresses resolved to
various levels of geocode certainty, instead of simply excluding those that could
not be geocoded to the level of street segment or higher. Two studies that
evaluated the positional accuracy of the geocoded locatio ns in exposure
assessments of crop locations within buffer distances as proxy measures of
pesticide exposure included only those addresses with GPS coordinates (5,6),
which are resource -intensive spatial d ata to obtain with staff often driving to
residences to identify or verify locations, and not feasible for studies like ours
that are based on historical residential addresses. Ward et al. also noted that the
relatively small impact of geocode accuracy on their exposure assessment
resulted because the majority (68 –85%) of homes were within 500 meters of
corn or soybean fields (5), and therefore areas with lower prevalence of exposure
170

or more diverse agriculture like that found in California would be expected to
have more misclassification of exposure.
4.5.3 Limitations to our methods for evaluating geocode certainty
Although previously published studies a ssessing pesticide exposure on
disease risk have employed the use of a geocoded point plus buffer distance as
was done in this study (30-33), use of this method is limiting because it does not
describe the variability in the area where the exposure occurs. A limitation of this
method of assessing the geocode certainty in a spatial -temporal association is
that it does not account for the geographical area, or shape, of the underlying
location. In order to evaluate the potential for misclassification from the certainty
of the geocoded locations, we need to understand the full spatial extent of
geocoded locations of where the study subjects are located as well as that of the
exposure we are measuring. Methods developed by Rull and Ritz to link PUR
data to land- use data can be used to identify the spatial extent of pesticide
exposure for a variety of geographical shapes in California (39), depending on the
level of geocode certainty. Geocoding methods also currently exist that can
identify standard shapes from zip codes or tax assessor parcels, but methods are
still needed to create geographical shapes for other levels of geocode certainty
such as street segment or street intersection that were identified in the output of
the manual resolution geocoding method used in this study. For studies that
evaluate historical addresses, geocoding methods that use historical reference
datasets could be used to provide historical geographical shapes that would
further increase the accuracy of the historical locations at various levels of
geocode certainty.
171

Another limitation is that these methods are based on the assumption that
a person’s location is where they spend all of their time. Most spatial
epidemiologic studies rely on residential addresses to identify subjects’ locations
and do not typically account for time -activity patterns. This would require actual
time and motion studies with real -time measurement that is data and labor
intensive and may become more feasible as the technology develops for
conducting such studies. Technologies are being developed for detecting
pollutants in real -time and coupled with location -identifying devices such as
mobile phones that people typically have with them, means that it may become
possible to identify individual -level exposure estimates of people’s day -to-day
activities (40). Such methods , however, are probably cost prohibitive for
assessments of long- term exposures and are unable to account for historical
time-activity patterns or exposures that occurred prior to the use of these
technologies.
Another way of looking at geocode certainty besides the methods
described here would be to compare exposure prevalence among those with
entire residential histories geocoded to the highes t geocode certainty level
(centroid of a building) to the exposure prevalence of those same individuals
under the assumption that we only knew their zip code of residence where the
corresponding levels of geocode certainty are the centroids of the zip code
tabulation areas. This would be similar to methods used by Healy et al. (18) and
Bow et al. (17) but it would not only assess the potential distance err ors
associated with differences between the building -centroid and zip -code-
tabulation-area-centroid point locations, but also examine the effect of those
172

distance errors with respect to pesticide exposure classification as defined by
exposure to pesticides within a 500-meter buffer around each location.
4.5.4 Conclusions
Our findings show that the level of geocode certainty can affect the validity
of the GIS -based pesticide exposure assessment, particularly for historical
exposures and locations that occurred further back in time. As the technology and
techniques for the geocoding process improve, addresses will be able to be
geocoded to a higher level of accuracy and with more precision, and yet, there will
inherently remain non -random variation in the level of certainty to which these
geocoded points represent a given location. Studies that have evaluated geocode
certainty indicate that errors are not typically random and can be substantial. It
is also common to use spatially aggregated units such as zip codes to represent
the locations of subjects when more precise address information is not available,
however this level of geocode certainty may not be sufficient for determining
exposure based on distances as small as 500 meters . Evaluating the geocode
certainty using methods such as those described here in future epidemiologic
studies will allow researchers to determine if the potential for misclassification at
various levels is beyond the exposure assessment threshold or buffer distance
where it has potential to jeopardize the integrity of the association between
exposure and health effects.
173

Figure 4.2. Trends in Use of a Selected Group of Organochlorines Pesticides and Chlorpyrifos in Fresno, Tulare, and Kern
Counties, 1974–2011.

0
1,000,000
2,000,000
3,000,000
4,000,000
5,000,000
6,000,000
7,000,000
1974-1979 1980-1984 1985-1989 1990-1994 1995-1999 2000-2004 2005-2011
Pounds of Pesticide
Applied
Year
Organochlorines
Chlorpyrifos
174

Figure 4.3. Comparison of Levels of High (Green) and Low (R ed) Geocode Certainty at Historical Residential Addresses (1974 to Year of Diagnosis
or Interview) in Breast Cancer Cases (Diagnosed in 2007 –2008) and P opulation-Based Controls (2001 –2011), in Fresno, Tulare , and Kern
Counties.
Cases Controls

175

Exposed
Unexposed
Exposed
Unexposed
Exposed
Unexposed
Exposed
Unexposed
Figure 4.4a. Comparison of Levels of High (Green) and Low (Red) Geocode Certainty at Historical
Residential Addresses (1974 to Year of Diagnosis or Interview) in Breast Cancer Cases (Diagnosed
in 2007 –2008) and Population -Based Controls (2001 –2011) in Fresno, Tulare, and Kern
Counties, Stratified by Ever Exposed to Organochlorines.
Cases Controls

Figure 4.4b. Comparison of L evels of High
(Green) and Low (Red) Geocode Certainty at Historical Residential Addresses (1974 to Year of
Diagnosis or Interview) in Breast Cancer Cases (Diagnosed in 2007–2008) and Population-Based
Controls (2001 –2011) in Fresno, Tulare, and Kern Counties, Stratified by Ever E xposed to
Chlorpyrifos.
Cases Controls

176

Table 4.1. Measures of Association Between Breast Cancer and Exposure to Selected Pesticides Among W omen in Fresno, Tulare, and Kern
Counties Using Linked PUR data for 1974–2011, All Geocode Certainty Levels Combined.
Exposure Type
Cases
(n = 155)
Controls
(n = 150)

No. % No. % OR 95% CI
Organochlorines

Unexposed 47 30.3

62 41.3

1.00 ---

Exposed 108 69.7

88 58.7

1.62 1.01, 2.60

Chlorpyrifos

Unexposed 56 36.1

77 51.3

1.00 ---

Exposed 99 63.9

73 48.7

1.87 1.18, 2.95

Table 4.2. Measures of Association Between Breast Cancer and Exposure to Selected Pesticides Among Women in Fresno, Tulare, and Kern
Counties Using Linked PUR Data for 1974 –2011, Stratified by the Proportion of Low (Red) G eocode Certainty Locations in the Residential
Histories.

Proportion of Low (Red) Geocode Certainty <10 % Proportion of Low (Red) Geocode Certainty ≥10%
Exposure Type
Cases
(n = 108)
Controls
(n = 114)

Cases
(n = 47)
Controls
(n = 36)

No. % No. % OR 95% CI

No. % No. % OR 95% CI
Organochlorines

Unexposed 41 38.0

51 44.7

1.00 ---

6 12.8

11 30.7

1.00 ---

Exposed 67 62.0

63 55.3

1.32 0.77, 2.26

41 87.2

25 69.4

3.01 0.99, 9.14

Chlorpyrifos

Unexposed 46 42.6

63 55.3

1.00 ---

10 21.3

14 38.9

1.00 ---

Exposed 62 57.4

51 44.7

1.67 0.98, 2.83

37 78.7

22 61.1

2.36 0.89, 6.20

177

Exposed
Unexposed
Exposed
Unexposed
Exposed
Unexposed
Exposed
Unexposed
Figure 4.5a. Comparison of Levels of High (Green) and Low (Red) Geocode Certainty at Historical
Residential Addresses (1974 to Year of Diagnosis or Interview) in Breast Cancer Cases and
Population-Based Controls, in Fresno, Tulare, and Kern Counties, Stratified by Ever E xposed to
Organochlorines and Boxed Around Subjects With 10% or Mo re of Their Residential Histories
Having Low Geocode Certainty Locations.
Cases Controls

Figure 4.5b. Comparison of Levels of High (Green) and Low (Red) Geocode Certainty at Historical
Residential Addresses (1974 to Year of Diagnosis or Interview) in Breast Cancer Cases and
Population-Based Controls, in Fresno, Tulare , and Kern Counties, Stratified by Ever Exposed to
Chlorpyrifos and Boxed Around Subjects With 10% or More of Their Residential Histories
Having Low Geocode Certainty Locations.
Cases Controls

178

Exposed
Unexposed
Exposed
Unexposed
Exposed
Unexposed
Exposed
Unexposed
Figure 4.6a. Comparison of Levels of High (Green) and Low (Red) Geocode Certainty at Historical
Residential Addresses (1974 to Year of Diagnosis or Interview) in Breast Cancer Cases and
Population-Based Controls, in Fresno, Tulare, and Kern Counties, Shaded for Person -Years
Exposed to Organochlorines and B oxed Around Subjects With 10% or More of Their
Residential Histories Having Low Geocode Certainty Locations.
Cases Controls

Figure 4.6b. Comparison of L evels of High (Green) and Low (Red) Geocode Certainty at
Historical Residential Addresses (1974 to Year of Diagnosis or Interview) in Breast Cancer Cases
and Population-Based Controls, in Fresno, Tulare, and Kern Counties, Shaded for Person -Years
Exposed to Chlorpyrifos and B oxed Around Subjects With 10% or More of Their Residential
Histories Having Low Geocode Certainty Locations.
Cases Controls

179

Exposed
Unexposed
Exposed
Unexposed
Exposed
Unexposed
Exposed
Unexposed
Figure 4.7a. Comparison of Levels of High (Green) and Low (Red) Geocode Certainty at Historical
Residential Addresses (1974 to Year of Diagnosis or Interview) in Breast Cancer Cases and
Population-Based Controls, in Fresno, Tulare, and Kern Counties, Shaded for Person -Years
Exposed to Organochlorines, B oxed Around Subjects Wit h 10% or More of Their Residential
Histories Having Low G eocode Certainty Locations, and Shaded and Boxed Around Subjects
Exposed Only at Low Geocode Certainty Locations in Their Residential Histories.
Cases Controls

Figure 4.7b. Comparison of Levels of High (Green) and Low (Red) Geocode Certainty at Historical
Residential Addresses (1974 to Year of Diagnosis or Interview) in Breast Cancer Cases and
Population-Based Controls, in Fresno, Tulare, and Kern Counties, Shaded for Person -Years
Exposed to Chlorpyrifos, B oxed Around Subjects With 10% or Mor e of Their Residential
Histories Having Low G eocode Certainty Locations, and Shaded and Boxed Around Subjects
Exposed Only at Low Geocode Certainty Locations in Their Residential Histories.
Cases Controls

180

Table 4.3. Measures of Association Between Breast Cancer and E xposure to Selected Pesticides Among W omen in Fresno, Tulare, and Kern
Counties Using Linked PUR Data for 1974 –2011, Stratified by the Proportion of Low (Red) G eocode Certainty Locations in the Residential
Histories, Excluding Subjects Exposed Only at Low Geocode Certainty Locations in Their Residential Histories.

Proportion of Low (Red) Geocode Certainty <10 % Proportion of Low (Red) Geocode Certainty ≥10%*
Exposure Type Cases

Controls

Cases

Controls

No. % No. % OR 95% CI

No. % No. % OR 95% CI
Organochlorines 108

114

35

33

Unexposed 41 38.0

51 44.7

1.00 ---

6 17.1

11 33.3

1.00 ---

Exposed 67 62.0

63 55.3

1.32 0.77, 2.26

29 82.9

22 66.7

2.42 0.77, 7.55

Chlorpyrifos
108

114

46

34

Unexposed 46 42.6

63 55.3

1.00 ---

10 21.7

14 41.2

1.00 ---

Exposed 62 57.4

51 44.7

1.67 0.98, 2.83

36 78.3

20 58.8

2.52 0.94, 6.70
* Excluding 12 cases and 3 controls in the analysis for organochlorines and 1 case and 2 controls in the analysis for chlorpyrifos because these subjects were only exposed at low geocode
certainty locations in their residential histories.

Table 4.4. Measures of Association Between Breast Cancer and E xposure to Selected Pesticides Among W omen in Fresno, Tulare, and Kern
Counties Using Linked PUR Data for 1974 –2011 for All Geocode Certainty Levels Combined, Including vs. E xcluding Subjects Exposed Only at
Low Geocode Certainty Locations in Their Residential Histories.

All Geocode Certainty Levels Combined

Excluding Subjects Exposed Only At Low Geocode
Certainty Locations in Their Residential Histories*
Exposure Type Cases

Controls

Cases

Controls

No. % No. % OR 95% CI

No. % No. % OR 95% CI
Organochlorines 155

150

143

147

Unexposed 47 30.3

62 41.3

1.00 ---

47 32.9

62 42.2

1.00 ---

Exposed 108 69.7

88 58.7

1.62 1.01, 2.60

96 67.1

85 57.8

1.49 0.92, 2.40

Chlorpyrifos
155

150

154

148

Unexposed 56 36.1

77 51.3

1.00 ---

56 36.4

77 52.0

1.00 ---

Exposed 99 63.9

73 48.7

1.87 1.18, 2.95

98 63.6

71 48.0

1.90 1.20, 3.01
* Excluding 12 cases and 3 controls in the analysis for organochlorines and 1 case and 2 controls in the analysis for chlorpyrifos because these subjects were only exposed at low geocode
certainty locations in their residential histories.
181

Exposed
Unexposed
Exposed
Unexposed
Exposed
Unexposed
Exposed
Unexposed
Figure 4.8a. Comparison of Levels of High (Green) and Low (Red) Geocode Certainty at Historical
Residential Addresses (1974 to Y ear of Diagnosis or Interview) in Breast Cancer Cases and
Population-Based Controls, in Fresno, Tulare, and Kern Counties, Shaded for Person -Years
Exposed to Organochlorines, and Shaded and Boxed Around Subjects Unexposed at Low
Geocode Certainty Locations in Their Residential Histories.
Cases Controls

Figure 4.8b. Comparison of Levels of High (Green) and Low (Red) Geocode Certainty at
Historical Residential Addresses (1974 to Year of Diagnosis or Interview) in Breast Cancer Cases
and Population-Based Controls, in Fresno, Tulare, and Kern Counties, Shaded for Person -Years
Exposed to Chlorpyrifos, and Shaded and Boxed Around Subjects Unexposed at Low Geocode
Certainty Locations in Their Residential Histories.
Cases Controls

182

Table 4.5. Measures of Association Between Breast Cancer and E xposure to Selected Pesticides Among W omen in Fresno, Tulare, and Kern
Counties Using Linked PUR Data for 1974–2011 for All Geocode Certainty Levels Combined, Including vs. Excluding Subjects Unexposed Only at
Low Geocode Certainty Locations in Their Residential Histories.

All Geocode Certainty Levels Combined

Excluding Subjects Unexposed Only At Low Geocode
Certainty Locations in Their Residential Histories*
Exposure Type Cases

Controls

Cases

Controls

No. % No. % OR 95% CI

No. % No. % OR 95% CI
Organochlorines 155

150

146

129

Unexposed 47 30.3

62 41.3

1.00 ---

38 26.0

41 31.8

1.00 ---

Exposed 108 69.7

88 58.7

1.62 1.01, 2.60

108 74.0

88 68.2

1.32 0.79, 2.24

Chlorpyrifos
155

150

139

124

Unexposed 56 36.1

77 51.3

1.00 ---

40 28.8

51 41.1

1.00 ---

Exposed 99 63.9

73 48.7

1.87 1.18, 2.95

99 71.2

73 58.9

1.73 1.04, 2.89
* Excluding 9 cases and 21 controls in the analysis for organochlorines and 1 6 cases and 2 6 controls in the analysis for chlorpyrifos because these subjects were unexposed at low
geocode certainty locations in their residential histories.

183

Chapter 4 References

1. Nuckols JR, Ward MH, Jarup L. Using geographic information s ystems for
exposure assessment in environmental epidemiology studies. Environ Res.
2004;112(9):1007–1015.
2. Goldberg DW, Wilson JP, Knoblock CA. From Text to Geographic
Coordinates: The Current State of Geocoding. URISA Journal.
2007;19:33–47.
3. Jacquez GM. Spatial and Spatio -temporal Epidemiology. Spatial and
Spatio-temporal Epidemiology. 2012;3(1):7–16.
4. Cayo MR, Talbot TO. Positional error in automated geocoding of
residential addresses. Int J Health Geogr. 2003;2(1):10.
5. Ward MH, Nuckols JR, Giglierano J, et al. Positional accuracy of two
methods of geocoding. Epidemiology. 2005;16(4):542–547.
6. Jones RR, DellaValle CT, Flory AR, et al. Accuracy of residential geocoding
in the Agricultural Health Study. Int J Health Geogr. 2014;13:37.
7. Bonner MR, Han D, Nie J, et al. Positional accuracy of geocoded addresses
in epidemiologic research. Epidemiology. 2003;14(4):408–412.
8. Goldberg DW, Wilson JP, Knoblock CA, et al. An effective and efficient
approach for manually impro ving geocoded data. Int J Health Geogr.
2008;7:60.
9. Mazumdar S, Rushton G, Smith BJ, et al. Geocoding accuracy and the
recovery of relationships between environmental exposures and health. Int
J Health Geogr. 2008;7:13.
10. Zandbergen PA, Green JW. Error and bias in determining exposure
potential of children at school locations using proximity -based GIS
techniques. Environ Res. 2007;115(9):1363–1370.
11. Strickland MJ, Siffel C, Gardner BR, et al. Quantifying geocode location
error using GIS methods. Environ Health. 2007;6(1):10.
12. Lovasi GS, Weiss JC, Hoskins R, et al. Comparing a single -stage geocoding
method to a multi -stage geocoding method: how much and where do they
disagree? Int J Health Geogr. 2007;6:12.
184

13. Zhan FB, Brender JD, DE Lima I, et al. Match rate and positional accuracy
of two geocoding methods for epidemiologic research. Annu Rev Public
Health. 2006;16(11):842–849.
14. Zandbergen PA. A comparison of address point, parcel and street
geocoding techniques. Comput Environ Urban Syst. 2008;32(3):214–232.
15. Gilboa SM, Mendola P, Olshan AF, et al. Comparison of residential
geocoding methods in population -based study of air quality and birth
defects. Environ Res. 2006;101(2):256–262.
16. Krieger N, Waterman P, Lemieux K, et al. On the wrong side of the tracts?
Evaluating the accuracy of geocoding in public health research. Am J Public
Health. 2001;91(7):1114–1116.
17. Bow CJD, Waters NM, Faris PD, et al. Accuracy of city postal code
coordinates as a proxy for location of residence. Int J Health Geogr.
2004;3(1):5.
18. Healy MA, Gilliland JA. Quantifying the magnitude of environmental
exposure misclassification when using imprecise address proxies in public
health research. Spat Spatiotemporal Epidemiol. 2012;3(1):55–67.
19. Rushton G, Armstrong MP, Gittler J, et al. Geocoding in cancer research.
Am J Prev Med. 2006;30(2):S16–S24.
20. Whitsel EA, Quibrera PM, Smith RL, et al. Accuracy of commercial
geocoding: assessment and implications. Epidemiol Perspect Innov.
2006;3:8.
21. Zandbergen PA. Influence of geocoding quality on environmental exposure
assessment of children living near high traffic roads. BMC Public Health.
2007;7:37.
22. Vieira VM, Howard GJ, Gallagher LG, et al. Geocoding rural addresses in a
community contaminated by PFOA: a comparison of methods. Environ
Health. 2010;9:18.
23. Woods N, Craig IP, Dorr G, et al. Spray drift of pesticides arising from
aerial application in cotton. J Environ Qual. 2001;30(3):697–701.
24. Cox C. Pesticide drift: indiscriminately from the skies. Journal of Pesticide
Reform [electronic article]. 1995;15:1 –7. (http://sunridge.net/assets/pdf/
pesticide_drift.pdf).
185

25. Gunier RB, Harnly ME, Reynolds P, et al. Agricultural pesticide use in
California: Pesticide prioritization, use densities, and population
distributions for a childhood cancer study. Environ Res.
2006;109(10):1071–1078.
26. Bradman A, Schwartz JM, Fenster L, et al. Factors predicting
organochlorine pesticide levels in pregnant Latina women living in a
United States agricultural area. J Expo Sci Environ Epidemiol.
2007;17:388–399.
27. Narayan S, Liew Z, Paul K, et al. Househo ld organophosphorus pesticide
use and Parkinson's disease. Int J Epidemiol. 2013;42(5):1476–1485.
28. Armes MN, Liew Z, Wang A, et al. Residential pesticide usage in older
adults residing in Central California. Int J Environ Res Public Health.
2011;8(8):3114–3133.
29. Goldberg D. The USC WebGIS Open Source Geocoding Platform. USC GIS
Research Laboratory Technical Report No. 11 ; 2009.
(http://spatial.usc.edu/wp-content/uploads/2014/03/gislabtr111.pdf).
(Accessed June 17, 2014).
30. Wang A, Cockburn M, Ly TT, et al. The association between ambient
exposure to organophosphates and Parkinson's disease risk. Occup
Environ Med. 2014;71(4):275–281.
31. Cockburn M, Mills P, Zhang X, et al. Prostate cancer and ambient pesticide
exposure in agriculturally intensive areas in California. Am J Epidemiol.
2011;173(11):1280–1288.
32. Wang AA, Costello SS, Cockburn MM, et al. Parkinson's disease risk from
ambient exposure to pesticides. Eur J Epidemiol. 2011;26(7):547–555.
33. Costello S, Cockburn M, Bronstein J, et al. Parkinson's disease and
residential exposure to maneb and paraquat from agricultural applications
in the central valley of California. Am J Epidemiol. 2009;169(8):919–926.
34. California Department of Water Resources. California Land & Water Use.
Land Use Data. (http://www.water.ca.gov/landwateruse/lusrvymain.cfm).
(Accessed July 12, 2013).
35. Rull RP, Ritz B. Historical pesticide exposure in California using pesticide
use reports and land- use surveys: an assessment of misclassification error
and bias. Environ Res. 2003;111(13):1582–1589.
36. O'Leary ESE, Vena JEJ, Freudenheim JLJ, et al. Pesticide exposure and
risk of breast cancer: a nested cas e-control study of residentially stable
women living on Long Island. Environ Res. 2004;94(2):134–144.
186

37. Reynolds P, Hurley SE, Goldberg DE, et al. Residential proximity to
agricultural pesticide use and incidence of breast cancer in the California
Teachers Study cohort. Environ Res. 2004;96(2):13–13.
38. Brody JG, Aschengrau A, McKelvey W, et al. Breast cancer risk and
historical exposure to pesticides from wide -area applications assessed with
GIS. Environ Res. 2004;112(8):889–897.
39. Rull RP, Gunier R, Behren Von J, et al. Residential proximity to
agricultural pesticide applications and childhood acute lymphoblastic
leukemia. Environ Res. 2009;109(7):891–899.
40. Jacquez GM, Meliker JR, Rommel RR. Exposure reconstruction using
space-time information technology. In: Geocoding Health Data - The Use
of Geographic Codes in Cancer Prevention and Control, Research, and
Practice. Boca Raton, FL: Elsevier; 2008.

187

Chapter 5

5. Summary and Discussion of Findings
There are strong biologically plausible links between certain endocrine-
disrupting pesticides such as organochlorines and breast cancer, yet the results of
epidemiologic studies to date have been mixed. Previous studies have been
hampered by inadequate met hods that either relied on biomarkers without
consideration of cumulative doses spanning decades prior to diagnosis, used self-
reported exposures that were unable to examine chemical -specific associations,
or were conducted in areas where the prevalence of exposure may be too low to
detect an effect if one truly existed . Without sufficiently high exposures or
consideration of cumulative doses spanning decades prior to a cancer diagnosis,
there is potential for non -differential misclassification of pesticide exposure to
bias the estimates of relative risk towards the null. We assessed chronic exposure
to specific pesticides of interest using a GIS-based approach that combined
geocoded address es with reported pesticide application records and land -use
data to explore the relationship between ambient pesticide exposure and breast
cancer risk.
First, a registry-based case control study using a GIS- based approach was
conducted to see if a comprehensive exposure model reduced non -differential
misclassification to reveal a true risk for breast cancer from specific, hormone -
related pesticides. Compared to no exposure, there was no significant difference
in risk of breast cancer among those exposed to any of the selected pesticides of
interest. We observed that the use of a detailed exposure model provided
188

substantial reduction in exposure misclassification as compared with estimates
based on less comprehensive measures used in previous studies of breast cancer
however, the study lacked the ability to examine lifetime exposure since estimates
were based only on addresses at the time of diagnosis . In addition, the registry
provided very limited information on other potential confounding factors and
controls were people diagnosed with other cancers that may not be represen tative
of the base population.
Secondly, we aimed to quantify how historical pesticide exposure
information reduced the non -differential bias in exposure estimates obtained
from only the a ddress at the time of diagnosis from the registry -based study. We
conducted a pilot case-control study using breast cancer cases recruited from the
cancer registry and population -based controls obtained from a study of
Parkinson’s disease conducted in the same region during the same time period.
Exposure was determined vi a our GIS -based approach that incorporated self -
reported residential and occupational histories. In the pilot study, risk estimates
based on only address at diagnosis or interview were attenuated compared to
estimates based on complete address histories. For exposure based on complete
residential and occupational histories, we found that exposure to organochlorines
was not associated with increased risk of breast cancer after adjusting for
exposure to chlorpyrifos, but chlorpyrifos was strongly associated w ith increased
risk of breast cancer after adjusting for exposure to other pesticides including
organochlorines. Although organochlorines have been the most well studied
pesticides in terms of breast cancer risk, exposures to different pesticides are
likely to be highly correlated (as is true of other air toxic chemicals) and it may be
189

that organochlorines serve as proxy measures for other pesticides such as
chlorpyrifos that are responsible for increased risk of breast cancer. A study with
a larger sample s ize is needed to confirm these chemical -specific findings and to
compare different levels of exposure as well as examine potential dose- response
relationships.
Lastly, we aimed to evaluate the certainty of the geocoded location
information used as the basi s for our GIS -based exposure assessment and its
potential to impact exposure classification. Studies that have evaluated geocode
certainty have shown that errors associated with inaccurate locations can be
substantial and are not randomly distributed but a re typically worse for rural
locations where proximity to agricultural pesticides is most likely to occur. In the
registry-based study, we were able to adjust for geocode certainty (obtained from
registry data) in the analyses since there was only one loca tion per person,
however in the pilot case control study exposure was based on address histories
where each participant had typically several addresses and each address
corresponded to a level of geocode certainty. For the pilot study, we developed
novel methods for assessing the geocode certainty (obtained from the geocoding
process output) associated with multiple historical locations recorded per study
subject. Since our pesticide exposure assessment relied on proximity to pesticides
within a 500-meter buffer from each location, we considered locations having low
geocode certainty to be those that were most likely beyond this distance threshold
(centroid of the zip code, city, state or unable to be geocoded), and, as a result,
most likely to misclassify exposure.
190

After analyzing geocode certainty at each location -year, we excluded
subjects who were only exposed to pesticides at their low geocode certainty
locations in their historical timelines and observed risk estimates were attenuated
for exposure to organochlorines but relatively unchanged for exposure to
chlorpyrifos. Locations with low geocode certainty were more likely to occur
further in the past when subjects were also more likely to be exposed to
organochlorines, whereas exposure to chlorpyrifos occurred more recently. Thus
we determined that the certainty of the geocoded locations that are the basis for a
GIS-based exposure assessment can impact the estimates of relative risk in the
pesticide exposure- disease association, and the magnitude of the change will
depend on the trends in pesticide usage over time as well as how far in the past
the exposure and the location occurred. The potential for misclassification
resulting from uncertainty in the geocoded locations is able to affect the risk
estimates, particularly when exposure is based on a relatively small buffer
distance and the potential errors associated with lower geocode certainties is
likely to be greater than this distance (500 meters in our case). Our results
highlight the need to account for geocode certainty in studies that employ GIS -
based methods to evaluate disease risk from spatial temporal exposures in the
environment.

5.1 Future Directions
5.1.1 Improving the GIS-based exposure assessment
In our pilot study, we were limited by an insufficient sample size to be able
to examine different levels of pesticide exposure or potential dose -response
191

relationships. Exposure was defined as ever exposed to the pesticides of interest
as determined by whether or not those pesticides were applied to crops within a
500-meter buffer surrounding the participants’ locations, but future directions
will seek to explore new methods for asses sing various quantities of exposure to
ambient pesticides using our existing GIS -based approach. A next step could be
to examine the distance from a person’s location to the nearest crop (where the
pesticides of interest are applied) within the 500 -meter b uffer so that those
exposed to pesticides very close to their location (<100 meters, for example)
could be compared to those who are exposed to pesticides applied on crops that
are 100’s of meters away. This measure of distance to nearest crop (pounds of
pesticide applied per meter distance) could be used to measure different levels of
exposure for assessing a potential dose-response relationship.
In addition to considering levels of different pesticide exposure, we could
improve upon our GIS-based approach by incorporating meteorological data and
wind patterns at various locations to consider whether our distance to crops
where pesticides are applied should be uniform in every direction or vary
according to typical weather or wind directions.
We also could improve upon our GIS-based approach for future studies by
evaluating the potential for non -differential misclassification resulting from
incomplete California PUR data. Roughly 37% of PUR data did not include
information on where the pesticide was applied, and, assuming that this missing
data was not related to case or control status, would likely lead to
misclassification biasing the estimates towards the null. We could evaluate this
non-located PUR data by pesticide type to determine whether the potential for
192

misclassification differed for different pesticides. In order to do this, we would
need to assess the proportion of PUR data, where a specific pesticide application
was reported but no information on its location, quantity (total pounds applied),
or crop area was listed in the record, by pesticide type for each year of PUR data
(1974–2012). Decisions would need to be made throughout this assessment as to
whether the information given in the PUR record was sufficient to include in our
GIS-based approach.
5.1.2 Alternative approaches to collecting control subjects for case-
control studies
In chapters 2 and 3, we described methods for conducting two case -
control studies where incident cases were obtained from a population -based
cancer registry and compared to several different types of controls. The validity of
the case-control studies’ findings depend on whether or not the controls
represent the distribution of exposure in the same population at risk to become
cases in the study (1,2) (and would be reported to the cancer registry in the same
region if they developed breast cancer) . The registry -based case -control study
highlighted the limitations of using controls that had other cancer diagnoses as
well as controls that were a random selection of tax as sessor parcels instead of
population-based controls. The pilot case -control study, on the other hand,
utilized population -based controls, but these were a convenient sample of
participants recruited for another study of Parkinson’s disease conducted in the
same region during the same time period. An improved study design for the pilot
study would be to use population-based controls randomly selected from disease-
free women of the same source population.
193

We can evaluate how well our cases and controls repre sent the base
population and the potential for selection bias if we know or have collected
information on certain factors that affect participation in our study (1). For
example, for the cases in our pilot case- control study, we compared those that
participated in our study to other breast cancer patients in the cancer registry and
found that they were similar in terms of age and SES. As for the contr ols in our
pilot study, unfortunately, we did not have demographic information on the non-
responding residents that were initially recruited for the Parkinson’s disease
study to be able to determine if the control participants of our pilot study
represented the same population that gave rise to the breast cancer cases. In our
future directions, we would like to conduct sensitivity analyses attempting to
adjust for potential selection bias using methods such as those by Thompson et
al. that are applicable when covariate information on factors that affect selection
is unknown (3).
Participation in epidemiologic studies in general has been declining,
particularly among controls participating in population- based case -control
studies (4). Population-based controls are often challenging to obtain si nce they
consist of healthy individuals who may not be interested in participating and
recruitment methods are typically time consuming and expensive. In addition to
low response, the controls that do participate may not be comparable to the cases
in terms of recruitment or recall of exposure and covariate information. Low
response rates, however, does not necessarily lead to selection bias (5).
The cases in our registry -based and pilot case -control studies were
recruited from population -based cancer registries. In the U.S., population -based
194

cancer registries are a valuable source for recruiting population -based cases;
however, we do not have complete and up -to-date population registries (other
than for documenting vital statistics such as birth, death, or marriage) that tract
current contact information for the entire population or at least for the people in
the same geographic areas covered by the cancer registries. We must look to
other possible sources for recruiting population -based controls, and each source
has its own advantages and disadvantages. For example, population -based
controls may be recruited through telephone calls, mailings, or home visits, or by
utilizing other community networks such as community centers or clinics.
Recruitment through telephone calls using random digit dialing (RDD) or
telephone directories used to be favored methods for epidemiologic studies, but
the increasing use of cellular telephones (6) has meant that fewer landline phones
are the predominant phone for typical households and the portability of cell
phone numbers means that the numbers are not tied to specific geographic areas.
Studies have found that recruitment using RDD tends to underrepresent younger
people (7,8), as well as women and people of lower SES (9,10), and that cell
phone only users tend to differ on certain demographic and lifestyle
characteristics from landline users (11-13). Since excluding either landline or cell
phone users may lead to selection bias in research studies (14,15), surveys and
epidemiologic studies have begun incorporating dual landline- cell sampli ng
frames to recruit both landline and cell phone users (16,17).
In the studies that are interested in exposures that have a strong
geographic component such as those included in this dissertation assessing
breast cancer risk from exposure to pesticides, recruitment methods to obtain
195

controls via te lephone may not be very efficient or cost effective since our base
population included a specific geographic region. Address -based sampling may
be a useful method for such studies because it utilizes address listings from the
U.S. Postal Service and links the geocoded addresses to the geographic area of
interest (18). Studies can then use these geographic -linked addresses to recruit
participants through mailings or home visits. One disadvantage is t hat the
address listings may not be as complete for rural as for urban areas, which is a
particular concern for our studies where the exposure of interest occurs mainly in
rural areas such as the case for our studies of pesticide exposure.
Another method that has been used to identify population -based controls
from a specific geographic region that gave rise to the cases consists of following
pre-determined routes from a given case’s location to identify suitable
neighborhood controls (19,20). This method was used to recruit controls for the
study of Parkinson’s disease that was also the source of our pilot study controls.
These methods recruit neighborhood controls that may be similar to the cases on
factors that are related to the exposure, the disease, and other covariates such as
SES that may be related to the exposure and disease. The cost to recruit
neighborhood controls can be very high however, since it requires study st aff
knocking on doors and multiple visits to obtain a control.
Another method to identify population-based controls located in a specific
geographic region is to utilize community centers for recruitment efforts. Senior
citizen centers, churches, clinics, or other healthcare facilities may assist in
identifying members of the community as well as serve to increase participation
rates. A multi -site breast cancer study targeting minority women worked with
196

community representatives to specifically target thes e populations, and found
that the demographic distributions of key covariates were more similar for those
recruited from community outreach networks, particularly churches, than for
those recruited through RDD and that including both sources of controls le d to
their controls being more representative of their source population (21). Another
case-control study recruited controls from lists of fa mily practitioners in the
target region, noting that participation increased and refusals decreased when
people were recruited on behalf of their family practitioner compared to cold
calling (22).
The nature of recruiting for case -control studies may lead to different
sampling frames for cases and controls, which can lead to selectio n bias in
population-based studies. In order to examine the potential for selection bias
resulting from any or several methods of recruitment used to obtain population -
based participants for research studies, efforts should be made to collect
demographic data such as age and gender on both those that responded to the
recruitment efforts as well as the non -responders in order to evaluate non -
response. This information, however, is likely to be very difficult to obtain from
people who are not interested in pa rticipating or concerned about privacy from
disclosing such information.
The case -control study design will continue to be an important tool for
epidemiologic studies examining the etiology of diseases such as cancer, that
typically have long latency periods, and exposure to certain rare exposures such
as pesticide exposure. When case -control studies are population -based and
controls consist of a random sample of non- diseased persons from the same
197

source population, then we can make inferences on the absolute risk from a
specific cause by combining the estimates of relative risk and information on
overall disease risk (23). Selection bias in population-based studies however, can
affect the validity of these findings that may be the basis for public health
practice. Improving upon existing study design methods and finding new ways to
recruit participants, while minimizing the potential for selection bias, is therefore
critical as we continue to use case-control studies in public health research.

5.2 Public Health Impact
We explored whether or not breast cancer risk was associated with specific
pesticides that had strong biological plausibility or have been measured in the
ambient air at levels deemed of concern to public health. There are no current air
quality standards for the concentrations of pesticides measured in the ambient
air, although in highly agricultural areas local residents are exposed to pesticides
in the air from neighboring applications (24).
The majority of persistent organochlorine pesticides have been banned or
phased out of use in agriculture, and their decline in use ushered in an increase in
the application of organophosphate pesticides. Over the past decade the use of
organophosphate pesticides has also decreased and has mainly been replaced
with pyrethroids (25,26). A study conducted in 2012 found that agricultural land
and coastal watersheds in California that were historically contaminated by
organochlorine pesticides were contaminated more recently by
organophosphates, particularly chlorpyrifos, and pyrethroid pesticides (27). As
trends in pesticide usage change, we need to be mindful of the chemicals that are
198

being introduced to replace them in terms of their toxicological properties and
their potential to impact environmental systems as well as the public’s health.
Chemicals that are effective pesticides and are introduced to replace restricted or
banned pesticides may well share similarities to the chemicals they are replacing,
and yet may not be evaluated or in use long enough to have observed their
potential adverse effects.
Although the U.S. EPA placed new restrictions on agricultural uses of
chlorpyrifos starting in late 2012, which included reducing the application rates
and mandating certain buffer distances to decrease the potential for exposure
through pesticide drift (28), people were exposed in the past and so we may have
the public health impact of that exposure on the risk of breast cancer for decades
to come. The findings from the studies conducted in this dissertation support
state efforts to conduct more comprehensive evaluation of pesticides measured in
the ambient air and highlight the need for health risk assessments and complete
toxicological profiles on the myriad of chemical compounds used in pest control.
It is important to understanding the long -term health effects of exposure
to commercial pesticide application because it may detect preventable causes of
disease or it may identify specific chemicals requiring application policies
designed to protect the public’s health. Studies that evaluate disease etiology
from environmental exposures such as large -scale pesticide application must
develop new methods and improve upon existing ones in order to address the
inherent challenges for assessing relevant exposures that accumulate over
decades.
199

Chapter 5 References

1. Rothman KJ, Greenland S, Lash TL. Modern Epidemiology. 3rd ed.
Philadelphia, PA: Lippincott, Williams & Wilkins; 2008.
2. Wacholder S, McLaughlin JK, Silverman DT, et al. Selection of controls in
case-control studies. I. Principles. Am J Epidemiol. 1992;135(9):1019 –
1028.
3. Thompson CA, Arah OA. Selection bias modeling using observed data
augmented with imputed record -level probabilities. Annu Rev Public
Health. 2014;24(10):747–753.
4. Morton LM, Cahill J, Hartge P. Reporting participation in epidemiologic
studies: a survey of practice. Am J Epidemiol. 2006;163(3):197–203.
5. Stang A, Jockel KH. Studies with low response proportions may be less
biased than studies with high response proportions. Am J Epidemiol.
2004;159(2):204–210.
6. Blumberg SJ, Luke JV. Wireless substitution: Early release of estimates
from the National Health Interview Survey, January–June 2014.
Hyattsville, MD: U.S. Department of Health and Human Services, Centers
for Disease Control and Prevention, National Center for Health Statistics;
2014. (http://www.cdc.gov/nchs/data/nhis/earlyrelease/wireless201412.
pdf). (Accessed January 5, 2015).
7. Lee S, Brick JM, Brown ER, et al. Growing cell -phone population and
noncoverage bias in traditional random digit dial telephone health surveys.
Health Serv Res. 2010;45(4):1121–1139.
8. Clagett B, Nathanson KL, Ciosek SL, et al. Comparison of address -based
sampling and random -digit dialing methods for recruiting young men as
controls in a case -control study of testicular cancer susceptibility. Am J
Epidemiol. 2013;178(11):1638–1647.
9. Glasser SL, Stearns CB. Reliabil ity of random digit dialing calls to
enumerate an adult female population. Am J Epidemiol.
2002;155(10):972–975.
10. Brogan DJ, Denniston MM, Liff JM, et al. Comparison of telephone
sampling and area sampling: Response rates and within -household
coverage. Am J Epidemiol. 2001;153(11):1119–1127.
200

11. Livingston M, Dietze P, Ferris J, et al. Surveying alcohol and other drug use
through telephone sampling: a comparison of landline and mobile phone
samples. BMC Med Res Methodol. 2013;13:41.
12. Delnevo CD, G undersen DA, Hagman BT. Declining estimated prevalence
of alcohol drinking and smoking among young adults nationally: artifacts
of sample undercoverage? Am J Epidemiol. 2007;167(1):15–19.
13. Blumberg SJ, Luke JV, Cynamon ML. Telephone coverage and health
survey estimates: evaluating the need for concern about wireless
substitution. Am J Public Health. 2006;96(5):926–931.
14. Walsh MC, Trentham -Dietz A, Gangnon RE, et al. Selection bias in
population-based cancer case -control studies due to incomplete sam pling
frame coverage. Cancer Epidemiol Biomarkers Prev. 2012;21(6):881–886.
15. Blumberg SJ, Luke JV. Reevaluating the need for concern regarding
noncoverage bias in landline surveys. Am J Public Health.
2009;99(10):1806–1810.
16. Greenberg MR, Weiner MD. Keeping surveys valid, reliable, and useful: a
tutorial. Risk Anal. 2014;34(8):1362–1375.
17. Hu SS, Balluz L, Battaglia MP, et al. Improving public health surveillance
using a dual -frame survey of landline and cell phone numbers. Am J
Epidemiol. 2011;173(6):703–711.
18. Johnson TP, ed. Handbook of Health Survey Methods. Hoboken, NJ: John
Wiley & Sons, Inc; 2015.
19. Wu AH, Wan P, Bernstein L. A multiethnic population- based study of
smoking, alcohol and body size and risk of adenocarcinomas of the
stomach and esophagus (United States). Cancer Causes Control.
2001;12(8):721–732.
20. Bernstein L, Henderson BE, Hanisch R, et al. Physical exercise and reduced
risk of breast cancer in young women. J Natl Cancer Inst. 1994;86:1403–
1408.
21. Bandera EV, Chandran U, Zirpoli G, et al. Rethinking sources of
representative controls for the conduct of case –control studies in minority
populations. BMC Med Res Methodol. 2013;13(1):1–1.
22. Castaño-Vinyals G, Nieuwenhuijsen MJ, Moreno V, et al. Participation
rates in the selection of population controls in a case -control study of
colorectal cancer using two recruitment methods. Gac Sanit.
2011;25(5):353–356.
201

23. Gail MH. Encyclopedia of Biostatistics. Chichester, UK: John Wiley & Sons,
Ltd; 2005.
24. Wofford P, Segaw a R, Schreider J, et al. Community air monitoring for
pesticides. Part 3: using health- based screening levels to evaluate results
collected for a year. Environ Monit Assess. 2013;186(3):1355–1370.
25. Metcalfe M, McWilliams B, Hueth B, et al. The economic importance of
organophosphates in California agriculture. Sacramento, CA: California
Department of Food and Agriculture; 2002. ( http://www.cdfa.ca.gov/files/
pdf/OrganophosphatesCAAgriculture.pdf). (Accessed January 6, 2015).
26. Epstein L, Bassein S, Zal om FG. Almond and stone fruit growers reduce
OP, increase pyrethroid use in dormant sprays. California Agriculture.
2000;54:14–19.
27. Phillips BM, Anderson BS, Hunt JW, et al. Pyrethroid and
organophosphate pesticide- associated toxicity in two coastal watersheds
(California, USA). Environ Toxicol Chem. 2012;31(7):1595–1603.
28. Office of Pesticide Programs. Pesticide news story: new use restrictions on
insecticide chlorpyrifos address bystander risk from spray drift; EPA’s
partial response to chlorpyrifos petition denies claims. Washington, DC:
U.S. Environmental Protection Agency; 2012.
(http://www.epa.gov/oppfead1/cb/csb_page/updates/2012/chlorpyrifos.h
tml). (Updated October 10, 2012. Accessed July 5, 2014).

202

Bibliography

Agency for Toxic Substances and Disease Registry. Draft Toxicological Profile for
Endosulfan. Atlanta, GA: U.S. Department of Health and Human Services ;
2013. (http://www.atsdr.cdc.gov/toxprofiles/tp.asp?id=609&tid=113).
(Accessed September 30, 2013).

Agency for Toxic Substances and Disease Registry. Draft Toxicological Profile for
Toxaphene. Atlanta, GA: U.S. Department of Health and Human Services ;
2010. (http://www.atsdr.cdc.gov/toxprofiles/tp.asp?id=548&tid=99).
(Accessed September 22, 2013).

Agency for Toxic Substances and Disease Registry. Priority Data Needs for
Dichloropropenes. Atlanta, GA: U.S. Departme nt of Health and Human
Services; 2009 . (http://www.atsdr.cdc.gov/pdns/pdfs/pdn_doc_40.pdf).
(Accessed June 25, 2014).

Agency for Toxic Substances and Disease Registry. 2011. ToxGuide for
Dichloropropenes. Atlanta, GA: U.S. Departme nt of Health and Human
Services; 2011 . (http://www.atsdr.cdc.gov/toxguides/toxguide-40.pdf).
(Accessed June 25, 2014).

Agency for Toxic Substanc es and Disease Registry. Toxicological Profile for
Aldrin/Dieldrin. Atlanta, GA: U.S. Departme nt of Health and Human
Services; 2002. (http://www.atsdr.cdc.gov/toxprofiles/tp.asp?id=317&tid=
56). (Accessed September 22, 2013).

Agency for Toxic Substan ces and Disease Registry. Toxicological Profile for
Chlordane. Atlanta, GA: U.S. Department of Health and Human Services;
1994. ( http://www.atsdr.cdc.gov/toxprofiles/tp.asp?id=355&tid=62).
(Accessed September 22, 2013).

Agency for Toxic Substances and Disease Registry. Toxicological Profile for DDT,
DDE, and DDD. Atlanta, GA: U.S. Department of Health and Human
Services; 2002. (http:// www.atsdr.cdc.gov/toxprofiles/tp35.pdf). (Accessed
October 20, 2013).

Agency for Toxic Substances and Disease Registry. Toxicological Profile for
Dichloropropenes. Atlanta, GA: U.S. Departme nt of Health and Human
Services; 2008 .( http://www.atsdr.cdc.gov/toxprofiles/tp40.pdf). (Accessed
June 25, 2014).

203

Agency for Toxic Substa nces and Disease Registry. Toxicological Profile for
Endrin. Atlanta, GA: U.S. Department of Health and Human Services ; 1996.
(http://www.atsdr.cdc.gov/toxprofiles/tp.asp?id=617&tid=114). (Accessed
September 22, 2013).

Agency for Toxic Substances and Disease Registry. Toxicological Profile for
Heptachlor and Heptachlor Epoxide. Atlanta, GA: U.S. Department of Health
and Human Services ; 2007. (http://www.atsdr.cdc.gov/toxprofiles/tp.asp?id
=746&tid=135). (Accessed September 22, 2013).

Agency for Toxic Substances and Disease Registry. Toxicological Profile for
Methoxychlor. Atlanta, GA: U.S. Department of Health and Human Services ;
2002. (http://www.atsdr.cdc.gov/toxprofiles/tp.asp?id=778&tid=151).
(Accessed September 22, 2013).

Akkina JE, Reif JS, Keefe TJ, et al. Age at natural menopause and exposure to
organochlorine pesticides in Hispanic women. J Toxicol Environ Health A.
2004;67(18):1407–1422.

Alavanja MCR, Ross MK, Bonner MR. Increased cancer burden among pesticide
applicators and others due to pesticide exposure. CA: Cancer J Clin.
2013;63(2):120–142.

Althuis MD, Fergenbaum JH, Garcia -Closas M, et al. Etiology of hormone
receptor-defined breast cancer: a systematic review of the literature. Cancer
Epidemiol Biomarkers Prev. 2004;13(10):1558–1568.

Andersen HR, Vinggaard AM, Rasmussen TH, et al. Effects of currently used
pesticides in assays for estrogenicity, androgenicity, and aromatase activity
in vitro. Toxicol Appl Pharmacol. 2002;179(1):1–12.
Arcaro KF, Yang Y, Vakharia DD, et al. Toxaphene is antiestrogenic in a human
breast-cancer cell assay. J Toxicol Environ Health A. 2000;59(3):197–210.

Armes MN, Liew Z, Wang A, et al. Residential pesticide usage in older adults
residing in Central California. Int J Environ Res Public Health.
2011;8(8):3114–3133.

Arnold SF, Klotz DM, Collins BM, et al. Synergistic activation of estrogen
receptor with combinations of environmental chemicals. Science.
1996;272(5267):1489–1492.

Ashley-Martin J, VanLeeuwen J, Cribb A, et al. Br east cancer risk, fungicide
exposure and CYP1A1*2A gene -environment interactions in a province -wide
case control study in Prince Edward Island, Canada. Int J Environ Health Res
Public Health. 2012;9(12):1846–1858.
204

Bagga D, Anders KH, Wang H-J, et al. Organochlorine pesticide content of breast
adipose tissue from women with breast cancer and control subjects. Journal
Natl Cancer Inst. 2000;92(9):750–753.

Band PR, Le ND, Fang R, et al. Identification of occupational cancer risks in
British Columbia. A popul ation-based case -control study of 995 incident
breast cancer cases by menopausal status, controlling for confounding
factors. J. Occup. Environ. Med. 2000;42(3):284–310.

Bandera EV, Chandran U, Zirpoli G, et al. Rethinking sources of representative
controls for the conduct of case –control studies in minority populations.
BMC Med Res Methodol. 2013;13(1):1–1.

Bernstein L, Henderson BE, Hanisch R, et al. Physical exercise and reduced risk
of breast cancer in young women. J Natl Cancer Inst. 1994;86:1403–1408.

Blair A, Dosemeci M, Heineman EF. Cancer and other causes of death among
male and female farmers from twenty -three states. Am J Ind Med.
1993;23(5):729–742.

Blumberg SJ, Luke JV. Reevaluating the need for concern regarding noncoverage
bias in landline surveys. Am J Public Health. 2009;99(10):1806–1810.

Blumberg SJ, Luke JV. Wireless substitution: Early release of estimates from the
National Health Interview Survey, January–June 2014. Hyattsville, MD:
U.S. Department of Health and Human Services, Centers for Disease Control
and Prevention, National Center for Health Statistics; 2014.
(http://www.cdc.gov/nchs/data/nhis/earlyrelease/wireless201412. pdf).
(Accessed January 5, 2015).

Blumberg SJ, Luke JV, Cynamon ML. Telephone coverage and health survey
estimates: evaluating the need for concern about wireless substitution. Am J
Public Health. 2006;96(5):926–931.

Bonner MR, Han D, Nie J, et al. Positional accuracy of geocoded addresses in
epidemiologic research. Epidemiology. 2003;14(4):408–412.

Bow CJD, Waters NM, Faris PD, et al. Accuracy of city postal code coordinates as
a proxy for location of residence. Int J Health Geogr. 2004;3(1):5.

Bradman A, Schwartz JM, Fenster L, et al. Factors predicting organochlorine
pesticide levels in pregnant Latina women living in a United States
agricultural area. J Expo Sci Environ Epidemiol. 2007;17:388–399.

Bradman A, Whitaker D, Quirós L, et al. Pesticides and their met abolites in the
homes and urine of farmworker children living in the Salinas Valley, CA. J
205

Expo Sci Environ Epidemiol. 2007;17(4):331–349.
Briz V, Molina- Molina J -M, Sánchez-Redondo S, et al. Differential estrogenic
effects of the persistent organochlorine pesticides dieldrin, endosulfan, and
lindane in primary neuronal cultures. Toxicol Sci. 2011;120(2):413–427.

Brody JG, Aschengrau A, McKelvey W, et al. Breast cancer risk and historical
exposure to pesticides from wide- area applications assessed with GI S.
Environ Res. 2004;112(8):889–897.

Brody JG, Vorhees DJ, Melly SJ, et al. Using GIS and historical records to
reconstruct residential exposure to large -scale pesticide application. J Expo
Anal Environ Epidemiol. 2002;12(1):64–80.

Brogan DJ, Denniston MM, Liff JM, et al. Comparison of telephone sampling and
area sampling: Response rates and within -household coverage. Am J
Epidemiol. 2001;153(11):1119–1127.

Brophy JT, Keith MM, Watterson A, et al. Breast cancer risk in relation to
occupations with exposure to carcinogens and endocrine disruptors: a
Canadian case–control study. Environ Health. 2012;11:87.

Brouwer DH, Brouwer EJ, van Hemmen JJ. Estimation of long-term exposure to
pesticides. Am J Ind Med. 1994;25(4):573–588.

California Cancer Registry. Cancer Reporting in California: Standards for
Automated Reporting. California Cancer Reporting System Standards for
2012 Volume II. Sacramento, CA: California Department of Public Health;
2012. (http://www.ccrcal.org/DSQC_Pubs/V2- 2012/Vol_II_2012.pdf).
(Accessed July 20, 2013).

California Department of Pesticide Regulation. A Guide to Pesticide Regulation
in California. Sacramento, CA: California Environmental Protection Agency;
2011. (http://www.cdpr.ca.gov/docs/pressrls/dpr guide.htm). (Accessed May
31, 2013).

California Department of Pesticide Regulation. Summary of Pesticide Use Report
Data–2010. Sacramento, CA: California Environmental Protection Agency;
2011.
(http://www.cdpr.ca.gov/docs/pur/pur10rep/10sum.htm#Development).
(Accessed June 11, 2013).

California Department of Water Resources. California Land & Water Use. Land
Use Data. (http://www.water.ca.gov/landwateruse/lusrvymain.cfm).
(Accessed July 12, 2013).

Calle EE, Frumkin H, Henley SJ, et al. Organochlorines and breast cancer risk.
CA: Cancer J Clin. 2002;52:301–309.
206

Cassidy RAR, Natarajan SS, Vaughan GMG. The link between the insecticide
heptachlor epoxide, estradiol, and breast cancer. Breast Cancer Res Treat.
2005;90(1):55–64.

Castaño-Vinyals G, Nieuwenhuijsen MJ, Moreno V, et al. Participation rates in
the selection of population controls in a case- control study of colorectal
cancer using two recruitment methods. Gac Sanit. 2011;25(5):353–356

Cayo MR, Talbot TO. Positional error in automated geocoding of residential
addresses. Int J Health Geogr. 2003;2(1):10.

Charlier C, Albert A, Herman P, et al. Breast cancer and serum organochlorine
residues. Occup Environ Med. 2003;60(5):348–351.
Chen S, Zhou D, Yang C, et al. Modulation of aromatase expression in human
breast tissue. J. Steroid Biochem Mol Biol. 2001;79(1-5):35–40.

Chlebowski RT, Kuller LH, Prentice RL, et al. Breast cancer after use of estrogen
plus progestin in postmenopausal women. N. Engl. J. Med.
2009;360(6):573–587.

Clagett B, Nathanson KL, Ciosek SL, et al. Comparison of address -based
sampling and random -digit dialing methods for recruiting young men as
controls in a case -control study of testicular cancer susceptibility. Am J
Epidemiol. 2013;178(11):1638–1647.

Clarke M, Collins R, Davies C, et al. Tamoxifen for early breast cancer: an
overview of the randomised trials. Lancet. 1998;351(9114):1451–1467.
Cocco P, Kazerouni N, Zahm SH. Cancer mortality and environmental exposure
to DDE in the United States. Environ Res. 2000;108(1):1–4.

Cockburn M, Mills P, Zhang X, et al. Prostate cancer and ambient pesticide
exposure in agriculturally intensive areas in California. Am J Ind Med.
2011;173(11):1280–1288.

Cohn BA, Wolff MS, Cirillo PM, et al. DDT and breast cancer in young women:
new data on the significance of age at exposure. Environ Res.
2007;115(10):1406–1414.

Colditz GA. Epidemiology of breast -cancer. Findings from the Nurses' Health
Study. Cancer. 1993;71(4):1480–1489.

Colditz GA. Relationship between estrogen levels, use of hormone replacement
therapy, and breast cancer. J Natl Cancer Inst. 1998;90(11):814–823.

207

Coronado GD, Thompson B, Strong L, et al. Agricultural task and exposure to
organophosphate pesticides among farmworkers. Environ Res.
2003;112(2):142–147.

Cossette LJ, Gaumond I, Martinoli MG. Combined effect of xenoestrogens and
growth factors in two estrogen -responsive cell lines. Endocrine.
2002;18(3):303–308.

Costello S, Cockburn M, Bronstein J, et al. Parkinson's disease and residential
exposure to maneb and paraquat from agricultural applications in the central
valley of California. Am J Ind Med. 2009;169(8):919–926.

Coumoul X, Diry M, Robillot C, et al. Differential regulation of cytochrome P450
1A1 and 1B1 by a combination of dioxin and pesticides in the breast tumor
cell line MCF-7. Cancer. 2001;61(10):3942–3948.
Cox C. Pesticide drift: indiscriminately from the sk ies. Journal of Pesticide
Reform [electronic article]. 1995;15:1 –7. (http://sunridge.net/
assets/pdf/pesticide_drift.pdf).

Delnevo CD, Gundersen DA, Hagman BT. Declining estimated prevalence of
alcohol drinking and smoking among young adults nationally: artifacts of
sample undercoverage? Am J Epidemiol. 2007;167(1):15–19.

Demers A, Ayotte P, Brisson J, et al. Risk and aggressiveness of breast cancer in
relation to plasma organochlorine concentrations. Cancer Epidemiol
Biomarkers Prev. 2000;9(2):161–166.

Dolapsakis G, Vlachonikolis IG, Varveris C, et al. Mammographic findings and
occupational exposure to pesticides currently in use on Crete. Eur. J. Cancer.
2001;37(12):1531–1536.

Dorgan JF, Brock JW, Rothman N, et al. Serum organochlorine pesticides and
PCBs and breast cancer risk: results from a prospective analysis (USA).
Cancer Causes Control. 1999;10(1):1–11.

Duell EJ, Millikan RC, Savitz DA, et al. A population -based case-control study of
farming and breast cancer in North Carolina. Epidemiology. 2000;11(5):523–
531.

El-Zaemey S, Heyworth J, Fritschi L. Noticing pesticide spray drift from
agricultural pesticide application areas and breast cancer: a case -control
study. Aust N Z J Public Health. 2013;37(6):547–555.

El-Zaemey S, Heyworth J, Glass DC, et al. Household and occupational exposure
to pesticides and risk of breast cancer. Int J Environ Health Res.
208

2014;24(2):91–102.
Endocrine Disruptor Screening Program, Office of Chemical Safety and Pollution
Protection. Weight-of-Evidence: Evaluating Results of EDSP Tier 1
Screening to Identify the Need for Tier 2 Testing. Washington, DC: U.S.
Environmental Protection Agency; 2011. Federal Register (EPA -HQ-OPPT-
2010-0877-0021).

Engel LS, Hill DA, Hoppin JA, et al. Pesticide use and breast cancer risk amo ng
farmers' wives in the Agricultural Health Study. Am J Ind Med.
2005;161(2):121–135.

Epstein L, Bassein S, Zalom FG. Almond and stone fruit growers reduce OP,
increase pyrethroid use in dormant sprays. California Agriculture.
2000;54:14–19.

Falck FJ, Ricci AJ, Wolff MS, et al. Pesticides and polychlorinated biphenyl
residues in human breast lipids and their relation to breast cancer. Arch
Environ Health. 1992;47(2):143–146.

Farooq U, Joshi M, Nookala V, et al. Self -reported exposure to pesticides in
residential settings and risk of breast cancer: a case -control study. Environ
Health. 2010;9(30):1–9.

Fenske RA, Chensheng L, Barr D, et al. Children's exposure to chlorpyrifos a nd
parathion in an agricultural community in central Washington State. Environ
Res. 2002;110(5):549–553.

Fleming LE, Bean JA, Rudolph M, et al. Mortality in a cohort of licensed pesticide
applicators in Florida. Occup Environ Med. 1999;56(1):14–21.

Fuhrman BJ, Schairer C, Gail MH, et al. Estrogen metabolism and risk of breast
cancer in postmenopausal women. J Natl Cancer Inst. 2012;104(4):326–339.

Gail MH. Encyclopedia of Biostatistics. Chichester, UK: John Wiley & Sons, Ltd;
2005.

Gammon MD, Wolff MS, Neugut AI, et al. Environmental toxins and breast
cancer on Long Island. II. organochlorine compound levels in blood. Cancer
Epidemiol Biomarkers Prev. 2002;11:686–697.

García MA, Peña D, Álvarez L, et al. Hexachlorobenzene induces cell proliferation
and IGF-I signaling pathway in an estrogen receptor alpha -dependent
manner in MCF-7 breast cancer cell line. Toxicol Lett. 2010;192(2):195–205.

Gardner KM, Ou Shu X, Jin F, et al. Occupations and breast cancer risk among
Chinese women in urban Shanghai. Am J Ind Med. 2002;42(4):296–308.

209

Gilboa SM, Mendola P, Olshan AF, et al. Comparison of residential geocoding
methods in population -based study of air quality and birth defects. Environ
Res. 2006;101(2):256–262.

Glasser SL, Stearns CB. Reliability of random digit dialing calls to enumerate an
adult female population. Am J Epidemiol. 2002;155(10):972–975.
Goldberg D. The USC WebGIS Open Source Geocoding Platform. USC GIS
Research Laboratory Technical Report No. 11 ; 2009.
(http://spatial.usc.edu/wp-content/uploads/2014/03/gislabtr111.pdf).
(Accessed June 17, 2014).

Goldberg DW, Wilson JP, Knoblock CA. From Text to Geographic Coordinates:
The Current State of Geocoding. URISA Journal. 2007;19:33–47.

Goldberg DW. A Geocoding Best Practices Guide. North American Association of
Central Cancer Registries, Inc; 2008.
(http://www.naaccr.org/LinkClick.aspx?fileticket=ZKekM8k_IQ0%3d&tabid
=239&mid=699). (Accessed September 22, 2013).

Goldberg DW, Wilson JP, Knoblock CA, et al. An effective and efficient approach
for manually improving geocoded data. Int J Health Geogr. 2008;7:60.

Greenberg MR, Weiner MD. Keeping surveys valid, reliable, and useful: a
tutorial. Risk Anal. 2014;34(8):1362–1375.
Gunier RB, Harnly ME, Reynolds P, et al. Agricultural pesticide use in California:
Pesticide prioritization, use densities, and population distributions for a
childhood cancer study. Environ Res. 2006;109(10):1071–1078.

Gunier RB, Ward MH, Airola M, et al. Determinants of agricultur al pesticide
concentrations in carpet dust. Environ Res. 2011;119(7):970–976.

Harnly ME, Bradman A, Nishioka M, et al. Pesticides in dust from homes in an
agricultural area. Environ Sci Technol. 2009;43(23):8767–8774.

Healy MA, Gilliland JA. Quantifyin g the magnitude of environmental exposure
misclassification when using imprecise address proxies in public health
research. Spat Spatiotemporal Epidemiol. 2012;3(1):55–67.

Henderson BE, Feigelson HS. Hormonal carcinogenesis. Carcinogenesis.
2000;21(3):427–433.

Hu SS, Balluz L, Battaglia MP, et al. Improving public health surveillance using a
dual-frame survey of landline and cell phone numbers. Am J Epidemiol.
2011;173(6):703–711.
210

Hunter DJ, Hankinson SE, Laden F, et al. Plasma organochlorine levels and the
risk of breast cancer. N. Engl. J. Med. 1997;337(18):1253–1258.

Høyer AP, Grandjean P, Jørgensen T, et al. Organochlorine exposure and risk of
breast cancer. Lancet. 1998;352(9143):1816–1820.

Høyer AP, Jørgensen T, Grandjean P, et al. Repeated measurements of
organochlorine exposure and breast cancer risk (Denmark). Cancer Causes
Control. 2000;11(2):177–184.

International Agency for Research on Cancer. IARC Monographs on the
Evaluation of Carcinogenic Risk of Chemicals to Humans. Agents classified
by the IARC monographs. 2010;1-100. (http://monographs.iarc.
fr/ENG/Classification/ClassificationsAlphaOrder.pdf). (Accessed May 29,
2013).

International Agency for Research on Cancer. IARC Monographs on the
Evaluation of Carcinogenic Risk of Chemicals to Humans. Chlordane and
Heptachlor. 1991;53:115 –177. ( http://monographs.iarc.fr/ENG/Monograp
hs/vol53/mono53-8.pdf). (Accessed May 29, 2013).
International Age ncy for Research on Cancer. IARC Monographs on the
Evaluation of Carcinogenic Risk of Chemicals to Humans. Occupational
exposures in insecticide application, and some pesticides. 1999;53.
(http://monographs.iarc.fr/ENG/Monographs/vol53/volume53.pdf).
(Accessed May 31, 2013).

Iscan MM, Coban TT, Cok II, et al. The organochlorine pesticide residues and
antioxidant enzyme activities in human breast tumors: is there any
association? Breast Cancer Res Treat. 2002;72(2):173–182.

Jacquez GM, Meliker JR, Rommel RR. Exposure reconstruction using space -time
information technology. In: Geocoding Health Data - The Use of Geographic
Codes in Cancer Prevention and Control, Research, and Practice. Boca
Raton, FL: Elsevier; 2008.

Jacquez GM. Spatial and Spatio -temporal Epidemiology. Spatial and Spatio-
temporal Epidemiology. 2012;3(1):7–16.

Johnson TP, ed. Handbook of Health Survey Methods. Hoboken, NJ: John Wiley
& Sons, Inc; 2015.

Jones RR, DellaValle CT, Flory AR, et al. Accuracy of residential geocoding in the
Agricultural Health Study. Int J Health Geogr. 2014;13:37.

211

Kaaks R, Rinaldi S, Key T, et al. Postmenopausal serum androgens, oestrogens
and breast cancer risk: the European prospective investigation into cancer
and nutrition. Endocr Relat Cancer. 2005;12(4):1071–1082.

Kang HG, Jeong SH, Cho JH, et al. Chlropyrifos -methyl shows anti -androgenic
activity without estrogenic activity in rats. Toxicology. 2004;199(2 -3):219–
230.

Kasner EJ, Keralis JM, Mehler L, et al. Gender differences in acute pesticide -
related illnesses and injuries among farmworkers in the United States, 1998 -
2007. Am J Ind Med. 2012;55(7):571–583.

Kaushik PP, Kaushik GG. An assessment of structure and toxicity correlation in
organochlorine pesticides. J Hazard Mater. 2007;143(1-2):102–111.

Key T, Appleby P, Barnes I, et al. Endogenous sex hormones and breast cancer in
postmenopausal women: reanalysis of nine prospective studies. J Natl
Cancer Inst. 2002;94(8):606–616.

Key T, Pike MC. The role of oestrogens and progestagens in the epidemiology and
prevention of breast cancer. Eur J Cancer Prev. 1988;24(1):29–43.

Kojima H, Katsura E, Takeuchi S, et al. Screening for estrogen and androgen
receptor activities in 200 pesticides by in vitro reporter gene assays using
Chinese hamster ovary cells. Environ Res. 2003;112(5):524–531.

Kortenkamp A. Breast cancer, oestrogens and environmental pollutants: a re -
evaluation from a mixture perspective. Int J Androl. 2006;29(1):193–198.

Krieger N, Wolff MS, Hiatt RA, et al. Breast cancer and serum organochlorines: a
prospective study among white, black, and Asian women. J Natl Cancer Inst.
1994;86(8):589–599.

Krieger N, Waterman P, Lemieux K, et al. On the wrong side of the tracts?
Evaluating the accuracy of geocoding in public health research. Am J Public
Health. 2001;91(7):1114–1116.

Laden F, Hankinson SE, Wolff MS, et al. Plasma organochlorine levels and the
risk of breast cancer: an extended follow -up in the Nurses' Health Study. Int.
J. Cancer. 2001;91(4):568–574.

Lee S, Brick JM, Brown ER, et al. Growing cell -phone population and
noncoverage bias in traditional random digit dial telephone health surveys.
Health Serv Res. 2010;45(4):1121–1139.

212

Lemaire G, Mnif W, Mauvais P, et al. Activation of alpha - and beta- estrogen
receptors by persistent pesticides in reporter cell lines. Life Sci.
2006;79(12):1160–1169.

Liehr JG. Dual role of oestrogens as hormones and pro- carcinogens: tumour
initiation by metabolic activation of oestrogens. Eur J Cancer Prev.
1997;6(1):3–10.

Linko PP, Yeowell HNH, Gasiewicz TAT, et al. Induction of cytochrome P -450
isozymes by hexachlorobenzene in rats and aromatic hydrocarbon (Ah) -
responsive mice. J Biochem Toxicol. 1986;1(2):95–107.

Livingston M, Dietze P, Ferris J, et al. Surveying alcohol and other drug use
through telephone sampling: a comparison of landline and mobile phone
samples. BMC Med Res Methodol. 2013;13:41.

London L, Myers JE. Use of a crop and job specifi c exposure matrix for
retrospective assessment of long -term exposure in studies of chronic
neurotoxic effects of agrichemicals. Occup Environ Med. 1998;55(3):194 –
201.

Lopez-Carrillo L, Torres -Sanchez L, Moline J, et al. Breast -feeding and serum
p,p'DDT levels among Mexican women of childbearing age: a pilot study.
Environ Res. 2001;87:131–135.

Lovasi GS, Weiss JC, Hoskins R, et al. Comparing a single -stage geocoding
method to a multi -stage geocoding method: how much and where do they
disagree? Int J Health Geogr. 2007;6:12.
Luo Y, Zhang M. Spatially distributed pesticide exposure assessment in the
Central Valley, California, USA. Envir Pollut. 2010;158(5):1629–1637.

Ma H, Henderson KD, Sullivan -Halley J, et al. Pregnancy -related factors and the
risk of b reast carcinoma in situ and invasive breast cancer among
postmenopausal women in the California Teachers Study cohort. Breast
Cancer Res. 2010;12(3):R35.

Madigan MP, Ziegler RG, Benichou J, et al. Proportion of breast cancer cases in
the United States explained by well -established risk factors. J Natl Cancer
Inst. 1995;87(22):1681–1685.

Mandal TK, Das NS. Testicular gametogenic and steroidogenic activities in
chlorpyrifos insecticide- treated rats: a correlation study with testicular
oxidative stress and role of antioxidant enzyme defence systems in Sprague-
Dawley rats. Andrologia. 2011;44(2):102–115.

213

Marusek JC, Cockburn MG, Mills PK, et al. Control sele ction and pesticide
exposure assessment via GIS in prostate cancer studies. Am J Prev Med.
2006;30(2):S109–S116.

Maxwell SK. Generating land cover boundaries from remotely sensed data using
object-based image analysis: overview and epidemiological applic ation. Spat
Spatiotemporal Epidemiol. 2010;1(4):231–237.

Maxwell SK, Airola M, Nuckols JR. Using Landsat satellite data to support
pesticide exposure assessment in California. Int J Health Geogr.
2010;9(46):1–14.

Maxwell SK. Downscaling pesticide use data to the crop field level in California
using Landsat satellite imagery: paraquat case study. Remote Sensing.
2011;3(12):1805–1816.

Mazumdar S, Rushton G, Smith BJ, et al. Geocoding accuracy and the recovery of
relationships between environmental exposur es and health. Int J Health
Geogr. 2008;7:13

Medjakovic S, Zoechling A, Gerster P, et al. Effect of nonpersistent pesticides on
estrogen receptor, androgen receptor, and aryl hydrocarbon receptor.
Environ Toxicol. 2014;29(10):1201–1216.
Metcalfe M, McWilliams B, Hueth B, et al. The economic importance of
organophosphates in California agriculture. Sacramento, CA: California
Department of Food and Agriculture; 2002. ( http://www.cdfa.ca.gov/files/
pdf/OrganophosphatesCAAgriculture.pdf). (Accessed January 6, 2015).

Millikan R, DeVoto E, Duell EJ, et al. Dichlorodiphenyldichloroethene,
polychlorinated biphenyls, and breast cancer among African -American and
white women in North Carolina. Cancer Epidemiol Biomarkers Prev.
2000;9(11):1233–1240.

Mills PKP, Yang RR. Breast cancer risk in Hispanic agricultural workers in
California. Int J Occup Environ Health. 2005;11(2):123–131.

Mills PK, Yang R. Regression analysis of pesticide use and breast cancer
incidence in California Latinas. J Environ Health. 2006;68(6):15–14.
Morton LM, Cahill J, Hartge P. Reporting participation in epidemiologic studies:
a survey of practice. Am J Epidemiol. 2006;163(3):197–203.

Morton WE. Major differences in breast cancer risks among occupations. J.
Occup. Environ. Med. 1995;37:328–335.

214

Moysich KB, Ambrosone CB, Vena JE, et al. Environmental organochlorine
exposure and postmenopausal breast cancer risk. Cancer Epidemiol
Biomarkers Prev. 1998;7:181–188.

Moysich KB, Ambrosone CB, Mendola P, et al. Exposures associ ated with serum
organochlorine levels among postmenopausal women from Western New
York state. Am J Ind Med. 2002;41(2):102–110.

Muir K, Rattanamongkolgul S, Smallman -Raynor M, et al. Breast cancer
incidence and its possible spatial association with pesti cide application in two
counties of England. Annu Rev Public Health. 2004;118(7):513–520.

Mukherjee S, Koner BC, Ray S, et al. Environmental contaminants in
pathogenesis of breast cancer. Indian J Exp Biol. 2006;44(8):597–617.

Muñoz-de-Toro M, Durando M, Beldoménico PM, et al. Estrogenic
microenvironment generated by organochlorine residues in adipose
mammary tissue modulates biomarker expression in ERalpha -positive
breast carcinomas. Breast Cancer Res. 2006;8(4):R47.
Narayan S, Liew Z, Paul K, et al. Household organophosphorus pesticide use and
Parkinson's disease. Int J Epidemiol. 2013;42(5):1476–1485.

Nuckols JR, Ward MH, Jarup L. Using geographic information systems for
exposure assessment in environmental epidemiology studies. Environ Res.
2004;112(9):1007–1015.

Nuckols JR, Gunier RB, Riggs P, et al. Linkage of the California Pesticide Use
Reporting Database with spatial land use data for exposure assessment.
Environ Res. 2007;115(5):684–689.

O'Leary ESE, Vena JEJ, Freudenheim JLJ, et al. Pesticide exposure and risk of
breast cancer: a nested case-control study of residentially stable women living
on Long Island. Environ Res. 2004;94(2):134–144.

Office of Pesticide Programs. Pesticide news story: new use restrictions on
insecticide chlorpyrifos address bystander risk from spray drift; EPA’s
partial response to chlorpyrifos petition denies claims. Washington, DC: U.S.
Environmental Protection Agency; 2012. (http://www.epa.gov/
oppfead1/cb/csb_page/updates/2012/chlorpyrifos.html). (Updated October
10, 2012. Accessed July 5, 2014).

Okasha M, McCarron P, Gunnell D, et al. Exposures in childhood, adolescence
and early adulthood and breast cancer risk: a systematic review of the
literature. Breast Cancer Res Treat. 2003;78(2):223–276.

215

Onland-Moret NC, Kaaks R, van Noord PAH, et al. Urinary endogenous sex
hormone levels and the risk of postmenopausal breast cancer. Br J Cancer.
2003;88(9):1394–1399

Oostingh GJ, Wichmann G, Schmittner M, et al. The cytotoxic effects of the
organophosphates chlorpyrifos and diazinon differ from their
immunomodulating effects. J Immunotoxicol. 2009;6(2):136–145.

Pelletier CC, Doucet EE, Imbeault PP, et al. Associations between weight loss -
induced changes in plasma organochlorine concent rations, serum T(3)
concentration, and resting metabolic rate. Toxicol. Sci. 2002;67(1):46–51.

Pfleeger TGT, Olszyk DD, Burdick CAC, et al. Using a geographic information
system to identify areas with potential for off- target pesticide exposure.
Environ Toxicol Chem. 2006;25(8):2250–2259.

Phillips BM, Anderson BS, Hunt JW, et al. Pyrethroid and organophosphate
pesticide-associated toxicity in two coastal watersheds (California, USA).
Environ Toxicol Chem. 2012;31(7):1595–1603.
Phillips KP, Foster WG, Leiss W, et al. Assessing and managing risks arising from
exposure to endocrine-active chemicals. J Toxicol Environ Health B Crit Rev.
2008;11(3-4):351–372

Pike MC, Spicer DV, Dahmoush L, et al. Estrogens, progestogens, normal breast
cell proliferation, and breast cancer risk. Epidemiol Rev. 1993;15(1):17–35.

Quirós-Alcalá L, Bradman A, Nishioka M, et al. Pesticides in house dust from
urban and farmworker ho useholds in California: an observational
measurement study. Environ Health. 2011;10(1):19.

Rattenborg T, Gjermandsen I, Bonefeld -Jørgensen E. Inhibition of E2 -induced
expression of BRCA1 by persistent organochlorines. Breast Cancer Res.
2002;4(6):R12.

Reynolds P, Hurley SE, Goldberg DE, et al. Residential proximity to agricultural
pesticide use and incidence of breast cancer in the California Teachers Study
cohort. Environ Res. 2004;96(2):13–13.

Reynolds P, Hurley SE, Gunier RB, et al. Residential proximity to agricultural
pesticide use and incidence of breast cancer in California, 1988 –1997.
Environ Res. 2005;113(8):993–1000.

Reynolds P, Hurley S, Goldberg DE, et al. Regional variations in breast cancer
among California teachers. Epidemiology. 2004;15(6):746–754.
216

Ritz B, Costello S. Geographic model and biomarker -derived measures of
pesticide exposure and Parkinson's disease. Ann N Y Acad Sci.
2006;1076(1):378–387.

Rogan WJ, Gladen BC, McKinney JD, et al. Polychlorinated biphenyls (PCBs) and
dichlorodiphenyl dichloroethene (DDE) in human milk: effects on growth,
morbidity, and duration of lactation. Am J Public Health. 1987;77(10):1294–
1297.

Romieu I, Hernandez -Avila M, Lazcano -Ponce E, et al. Breast cancer, lactation
history, and serum organochlorines. Am J Ind Med. 2000;152(4):363 –
370.
Rossouw JE, Anderson GL, Prentice RL, et al. Risks and benefits of estrogen plus
progestin in healthy postmenopausal women - Principal results from the
Women's Health Initiative randomized controlled trial. JAMA.
2002;288(3):321–333.

Rothman KJ, Greenland S, Lash TL. Modern Epidemiology. 3rd ed. Philadelphia,
PA: Lippincott, Williams & Wilkins; 2008.

Rubin CH, Burnett CA, Halperin WE, et al. Occupation as a risk identifier for
breast cancer. Am J Public Health. 1993;83(9):1311–1315.

Rudel RA, Attfield KR, Schifano JN, et al. Chemicals causing mammary gland
tumors in animals signal new directions for epidemiology, chemicals testing ,
and risk assessment for breast cancer prevention. Cancer.
2007;109(S12):2635–2666.

Ruder EH, Dorgan JF, Kranz S, et al. Examining breast cancer growth and
lifestyle risk factors: early life, childhood, and adolescence. Clin Breast
Cancer. 2008;8(4):334–342.

Rull RP, Ritz B. Historical pesticide exposure in California using pesticide use
reports and land -use surveys: an assessment of misclassification error and
bias. Environ Res. 2003;111(13):1582–1589.

Rull RP, Gunier R, Behren Von J, et al. Residentia l proximity to agricultural
pesticide applications and childhood acute lymphoblastic leukemia. Environ
Res. 2009;109(7):891–899.

Rushton G, Armstrong MP, Gittler J, et al. Geocoding in cancer research. Am J
Prev Med. 2006;30(2):S16–S24.

Russell HH, Jackson RJ, Spath DP, et al. Chemical Contamination of California
Drinking Water. West J Med. 1987;147(5):615–622.
217

Shakeel MK, George PS, Jose J, et al. Pesticides and breast cancer risk: a
comparison between developed and developing countries. Asian Pac J
Cancer Prev. 2010;11(10):173–180.

Siegel R, Naishadham D, Jemal A. Cancer statistics, 2012. CA: Cancer J Clin.
2012;62(1):10–29.

Simcox NJ, Fenske RA, Wolz SA, et al. Pesticides in household dust and soil:
exposure pathways for children of agricultural f amilies. Environ Res.
1995;103(12):1126–1134.

Snedeker SM. Pesticides and breast cancer risk: a review of DDT, DDE, and
dieldrin. Environ Res. 2001;109(Suppl 1):35–47.

Sohoni P, Sumpter JP. Several environmental oestrogens are also anti -
androgens. J Endocrinol. 1998;158(3):327–339.

Soto AM, Chung KL, Sonnenschein C. The pesticides endosulfan, toxaphene, and
dieldrin have estrogenic effects on human estrogen -sensitive cells. Environ
Res. 1994;102:380–383.

St-Hilaire S, Mandal R, Commendador A, et al. Est rogen receptor positive breast
cancers and their association with environmental factors. Int J Health Geogr.
2011;10:32.

Stang A, Jockel KH. Studies with low response proportions may be less biased
than studies with high response proportions. Am J Epidemiol.
2004;159(2):204–210.

Stellman SD, Djordjevic MV, Britton JA, et al. Breast cancer risk in relation to
adipose concentrations of organochlorine pesticides and polychlorinated
biphenyls in Long Island, New York. Cancer Epidemiol Biomarkers Prev.
2000;9(11):1241–1249.

Strickland MJ, Siffel C, Gardner BR, et al. Quantifying geocode location error
using GIS methods. Environ Health. 2007;6(1):10.

Tamimi RM, Hankinson SE, Chen WY, et al. Combined estrogen and testosterone
use and risk of breast cancer in postmenopausal women. Arch Intern Med.
2006;166(14):1483–1489.

Teitelbaum SL, Gammon MD, Britton JA, et al. Reported residential pesticide use
and breast cancer risk on Long Island, New York. Am J Ind Med.
2007;165(6):643–651.

218

Thomas P, Dong J. Binding and activation of the seven -transmembrane estrogen
receptor GPR30 by environmental estrogens: a potential novel mechanism of
endocrine disruption. J. Steroid Biochem Mol Biol. 2006;102(1-5):175–179.

Thompson CA, Arah OA. Selection bias modeling using observed data augmented
with imputed record -level probabilities. Annu Rev Public Health.
2014;24(10):747–753.

United States Department of Agriculture, Economic Research Service.
Agricultural Productivity in the U.S. National Tables 1948–2011.
(http://www.ers.usda.gov/data-products/agricultural-productivity-in-the-
us.aspx#.Up6dwr--WXo). (Accessed December 3, 2013).

Usmani KA. Inhibition of the human liver microsomal and human cytochrome
P450 1A2 and 3A4 metabolism of estradiol by dep loyment-related and other
chemicals. Drug Metab Dispos. 2006;34(9):1606–1614.

Usmani KA, Rose RL, Hodgson E. Inhibition and activation of the human liver
microsomal and human cytochrome P450 3A4 metabolism of testosterone by
deployment-related chemicals. Drug Metab Dispos. 2003;31(4):384–391.

Valerón PF, Pestano JJ, Luzardo OP, et al. Differential effects exerted on human
mammary epithelial cells by environmentally relevant organochlorine
pesticides either individually or in combination. Chem Biol Interact.
2009;180(3):485–491.

van't Veer P, Lobbezoo IE, Martín -Moreno JM, et al. DDT (dicophane) and
postmenopausal breast cancer in Europe: case -control study. BMJ.
1997;315(7100):81–85.

Ventura C, Núñez M, Miret N, et al. Differential mechanisms of action are
involved in chlorpyrifos effects in estrogen -dependent or -independent breast
cancer cells exposed to low or high concentrations of the pesticide. Toxicol
Lett. 2012;213(2):184–193.

Verner M -A, Bachelet D, McDougall R, et al. A case study addressing the
reliability of polychlorinated biphenyl levels measured at the time of breast
cancer diagnosis in representing early -life exposure. Cancer Epidemiol
Biomarkers Prev. 2011;20(2):281–286.

Vieira VM, Howard GJ, Gallagher LG, et al. Geocoding rural addre sses in a
community contaminated by PFOA: a comparison of methods. Environ
Health. 2010;9:18.

Viswanath G, Chatterjee S, Dabral S, et al. Journal of Steroid Biochemistry and
Molecular Biology. J. Steroid Biochem Mol Biol. 2010;120(1):22–29.
219

Vo TT, Gladen BC, Cooper GS, et al. Dichlorodiphenyldichloroethane and
polychlorinated biphenyls: intraindividual changes, correlations, and
predictors in healthy women from the southeastern United States. Cancer
Epidemiol Biomarkers Prev. 2008;17(10):2729–2736.

Wacholder S, McLaughlin JK, Silverman DT, et al. Selection of controls in case -
control studies. I. Principles. Am J Epidemiol. 1992;135(9):1019–1028.

Walsh MC, Trentham -Dietz A, Gangnon RE, et al. Selection bias in population -
based cancer case -control studies due to incomplete sampling frame
coverage. Cancer Epidemiol Biomarkers Prev. 2012;21(6):881–886.

Wang A, Cockburn M, Ly TT, et al. The association between ambient exposure to
organophosphates and Parkinson's disease risk. Occup Environ Med.
2014;71(4):275–281.
Wang AA, Costello SS, Cockburn MM, et al. Parkinson's disease risk from
ambient exposure to pesticides. Eur J Epidemiol. 2011;26(7):547–555.

Ward MH, Lubin J, Giglierano J, et al. Proximity to crops and residential
exposure to agricu ltural herbicides in Iowa. Environ Res. 2006;114(6):893–
897.

Ward MH, Nuckols JR, Giglierano J, et al. Positional accuracy of two methods of
geocoding. Epidemiology. 2005;16(4):542–547.

Weichenthal S, Moase C, Chan P. A review of pesticide exposure and cancer
incidence in the Agricultural Health Study cohort. Environ Res.
2010;118(8):1117–1125.

Weinberg CR, Moledor ES, Umbach DM, et al. Imputation for exposure histories
with gaps, under an excess relative risk model. Epidemiology. 1996;7(5):490.

Whitsel EA, Quibrera PM, Smith RL, et al. Accuracy of commercial geocoding:
assessment and implications. Epidemiol Perspect Innov. 2006;3:8.

Wofford P, Segawa R, Schreider J, et al. Community air monitoring for pesticides.
Part 3: using health- based screening levels to evaluate results collected for a
year. Environ Monit Assess. 2013;186(3):1355–1370.

Wolff MS, Britton JA, Teitlebaum SL, et al. Improving organochlorine biomarker
models for cancer research. Cancer Epidemiol Biomarkers Prev.
2005;14(9):2224–2236.

Wolff MS, Toniolo PG, Lee EW, et al. Blood levels of organochlorine residues and
risk of breast cancer. J Natl Cancer Inst. 1993;85(8):648–652.
220

Wolff MS, Zeleniuch -Jacquotte A, Dubin N, et al. Risk of breast cancer and
organochlorine exposure. Cancer Epidemiol Biomarkers Prev.
2000;9(3):271–277.

Wong PS, Matsumura F. Promotion of breast cance r by beta -
hexachlorocyclohexane in MCF10AT1 cells and MMTV -neu mice. BMC
Cancer. 2007;7:130.

Woods N, Craig IP, Dorr G, et al. Spray drift of pesticides arising from aerial
application in cotton. J Environ Qual. 2001;30(3):697–701.

Wu AH, Wan P, Bernstein L. A multiethnic population- based study of smoking,
alcohol and body size and risk of adenocarcinomas of the stomach and
esophagus (United States). Cancer Causes Control. 2001;12(8):721–732.

Xu X, Dailey AB, Talbott EO, et al. Associations of serum conc entrations of
organochlorine pesticides with breast cancer and prostate cancer in U.S.
adults. Environ Res. 2010;118:60–66.

Yost K, Perkins C, Cohen R, et al. Socioeconomic status and breast cancer
incidence in California for different race/ethnic groups. Cancer Causes
Control. 2001;12(8):703–711.

Young HA. Use of a crop and job specific exposure matrix for estimating
cumulative exposure to triazine herbicides among females in a case- control
study in the Central Valley of California. Occup Environ Med.
2004;61(11):945–951.

Yu M, Tatalovich Z, Gibson JT, et al. Using a composite index of socioeconomic
status to investigate health disparities while protecting the confidentiality of
cancer registry data. Cancer Causes Control. 2013;25(1):81–92.

Zandbergen PA. A comparison of address point, parcel and street geocoding
techniques. Comput Environ Urban Syst. 2008;32(3):214–232.

Zandbergen PA. Influence of geocoding quality on environmental exposure
assessment of children living near high traffic roads. BMC Public Health.
2007;7:37.

Zandbergen PA, Green JW. Error and bias in determining exposure potential of
children at school locations using proximity -based GIS techniques.
Environ Res. 2007;115(9):1363–1370.
Zhan FB, Brender JD, DE Lima I, et al. Match rate and positional accuracy of two
geocoding methods for epidemiologic research. Annu Rev Public Health.
2006;16(11):842–849.
221

Zheng T, Holford TR, Mayne ST, et al. DDE and DDT in breast adipose tissue and
risk of female breast cancer. Am J Ind Med. 1999;150(5):453–458.

222

Appendix

Pesticide Exposure and Breast Cancer Pilot Study Questionnaire
PART A: YOUR CANCER HISTORY

A1. Have you ever been told by a doctor or other health care
professional that you had breast cancer? (A1BRSTDX)

00 No

01 Yes

A2. IF yes  When was the first time that a doctor or other health
care professional told you that you had it?
______________ ______________
MONTH (A2BRSTMO) YEAR (A2BRSTYR)

PART B: RESIDENTIAL HISTORY AND PESTICIDE EXPOSURE

The next set of questions is about where you lived, any pesticide
exposure you may have had and some of the activities you do at home.

B1. Were you born in the U.S.? (B1USYEAR)
01

Yes, I was born in the U.S. (Skip to B4)
02

No, I moved to the U.S. after I was born
B2. What year did you come to the U.S.?____________ (B2USMOVE)
(Probe: approximate year)

B3. In what country were you born? __________________ (B3BRTHPL)

B4. Have you ever lived on or near a farm? (B4FARMLIVE)

00 No 77 Can’t remember
01 Yes, on a farm 88 Refused
02 Yes, near a farm 99 Don’t know

223

B5. Have you ever lived within ½ mile of a facility such as a… (read
list)? B5CHEM, B5POWER, B5PULP, B5OILREF, B5LANDFL

Facility No Yes
Can’t
Rem.
Refu
sed
Don’t
know
A. Chemical plant 00 01 77 88 99
B. Power plant 00 01 77 88 99
C. Paper or Pulp mill 00 01 77 88 99
D. Oil refinery 00 01 77 88 99
E. Landfill site 00 01 77 88 99

B6. Since what age have you lived in California? ________ (B6CALIVE)

B7 Since what age have you lived in the counties of Fresno, Kern, or
Tulare? ________ (B7CNTYLV)

B8 Did you ever live in Mexico? (B8MEXLIVE)

00 No

01 Yes

B9. If yes, how many years did you live in Mexico? ________
(B9MEXYRS)

77 Can’t remember
224

B10. Now I would like to record the places where you have lived for 3 months or longer at any time of your
life. Let’s start with the place where you were born…(If subject answers “year when moved” do not ask
“How long”)
Address
(or cross streets if
complete address
unavailable)
77= Can’t remember.
88 = Refused
99 = Don’t know
(There are 25 sets in B10 now (A-P, R-Z)
B10ADDRQ does not exist
Address may have Street Name or just Cross
Streets
Year when moved How long Water
Supply
Farm
To
77 = Can’t rem.
88 = Refused
99 = Don’t know
This is approx.
Away
77 = Can’t rem.
88 = Refused
99 = Don’t know
This is approx.
Yrs.
77 = Can’t rem.
88 = Refused
99 = Don’t know
This is approx.
Mos.
77 = Can’t rem.
88 = Refused
99 = Don’t know
This is approx.

01 = Public supply
(tap water)
02 = Private well
03 = Filtered water
04 = Bottled water
05 = Other
77 = Can’t
remember.
88 = Refused
99 = Don’t know

1 = Yes
2 = No
77 = Can’t
remember.
88 = Refused
99 = Don’t know
A.
B10ADDRA

B10TOA B10AWAYA B10YEARSA B10MONTHA B10H20A B10FARMA
B.
B10ADDRB

B10TOB B10AWAYB B10YEARSB B10MONTHB B10H20B B10FARMB
C.
B10ADDRC

B10TOC B10AWAYC B10YEARSC B10MONTHC B10H20C B10FARMC
D.
B10ADDRD

B10TOD B10AWAYD B10YEARSD B10MONTHD B10H20D B10FARMD
E.
B10ADDRE

B10TOE B10AWAYE B10YEARSE B10MONTHE B10H20E B10FARME
F.
B10ADDRF

B10TOF B10AWAYF B10YEARSF B10MONTHF B10H20F B10FARMF
225

B11. How many glasses of water per day, and from what kind of source
did you drink on an average day when you were aged...?

(<25 years) (Now)
Tap (w/out filter) B11TAP25 B11TAPNOW
Tap (w/ filter) B11FTAP25 B11FTAPNOW
Bottled water B11BTL25 B11BTLNOW
Well water B11WELL25 B11WELLNOW
Other _________ B11OTHER25 B11OTHENOW

B12. Have you ever gardened for leisure or as a hobby? (B12GARDEN)

00 No
01 Yes
77 Can’t remember
88 Refused
99 I don’t know

B13. Have you ever used chemical products to kill insects or stop plant
diseases or weeds in your yard or garden? (B13KILLBUG)

00 No
01 Yes
77 Can’t remember
88 Refused
99 I don’t know

B14. Has anyone in your household ever used chemical products to kill
insects or stop plant diseases or weeds in your yard or garden?
(B14ANYKILL)

00 No
01 Yes
77 Can’t remember
88 Refused
99 I don’t know
226

B15. What kind of chemicals did you (or someone in your household) use on your lawn, in your yard or
garden?

What did you
use?

Chemical name
or purpose
77= Can’t
remember
88=Refused
99=Don’t know

How often applied?
00 = None,
01 = Rarely (once a year or less),
02 = Sometimes (2-11 times a year),
03 = Regularly (once a week to once a month)
Was it…?
Where did
you store
the
chemical?

Did you
protect
yourself
with?
00=No;
01=Yes
How often
did you
protect
yourself?
<25 YEARS

25-44
YEARS

45-54
YEARS
55-64
YEARS
65+
YEARS
Who did
it?
Spray
Granules
Bait
Powder

Garage
In the house
Storage Facility
Other: Specify

None
Gloves
Mask
Coveralls
Other: Specify
Never
Sometimes
Always
A B15CHEMA
00□
01□
02□
03□
B15APPL25A
00□
01□
02□
03□
B15APPL44A
00□
01□
02□
03□
B15APPL54A
00□
01□
02□
03□
B15APPL64A
00□
01□
02□
03□
B15APPL65A
Yourself?
□
Someone
else? □
B15DIDSFA
1 2 3 4
B15FORMA
1 2 3 4
B15STOREA
0 1 2 3 4
B15PROTA

0 1 2
B15UPROA
B B15CHEMB
00□
01□
02□
03□
B15APPL25B
00□
01□
02□
03□
B15APPL44B
00□
01□
02□
03□
B15APPL54B
00□
01□
02□
03□
B15APPL64B
00□
01□
02□
03□
B15APPL65B
Yourself?
□
Someone
else? □
B15DIDSFB
1 2 3 4
B15FORMB
1 2 3 4
B15STOREB
0 1 2 3 4
B15PROTB

0 1 2
B15UPROB
C B15CHEMC
00□
01□
02□
03□
B15APPL25C
00□
01□
02□
03□
B15APPL44C
00□
01□
02□
03□
B15APPL54C
00□
01□
02□
03□
B15APPL64C
00□
01□
02□
03□
B15APPL65C
Yourself?
□
Someone
else? □
B15DIDSFC
1 2 3 4
B15FORMC
1 2 3 4
B15STOREC
0 1 2 3 4
B15PROTC

0 1 2
B15UPROC
For what purpose? 01 □Weed control 02 □ Plant disease 03□ Other Specify________________ □ Not Answered

227

B16. Have you ever lived in a household with pets? (B16URPET)
00 No
01 Yes
77 Can’t remember
88 Refused
99 I don’t know

B17. Have you ever used any flea/tick chemicals on any of your pets?
00 No
01 Yes
77 Can’t remember
88 Refused
99 I don’t know

228

B18. For your pets did you personally or anyone in your household ever use chemicals (soaps, shampoos,
dips, powders) to kill fleas or ticks or other insects?

What did you
use?

Chemical name
or purpose
77= Can’t
remember
88=Refused
99=Don’t know

How often applied?
00 = None,
01 = Rarely (once a year or less),
02 = Sometimes (2-11 times a year),
03 = Regularly (once a week to once a month)
Was it…?
Where did
you store
the
chemical?

Did you
protect
yourself
with?
00=No;
01=Yes
How often
did you
protect
yourself?
<25 YEARS

25-44
YEARS

45-54
YEARS
55-64
YEARS
65+
YEARS
Who did
it?
Spray
Granules
Bait
Powder

Garage
In the house
Storage Facility
Other: Specify

None
Gloves
Mask
Coveralls
Other: Specify
Never
Sometimes
Always
A B18CHEMA
00□
01□
02□
03□
B18APPL25A
00□
01□
02□
03□
B18APPL44A
00□
01□
02□
03□
B18APPL54A
00□
01□
02□
03□
B18APPL64A
00□
01□
02□
03□
B18APPL65A
Yourself?
□
Someone
else? □
B18DIDSFA
1 2 3 4
B18FORMA
1 2 3 4
B18STOREA
0 1 2 3 4
B18PROTA

0 1 2
B18UPROA
B B18CHEMB
00□
01□
02□
03□
B18APPL25B
00□
01□
02□
03□
B18APPL44B
00□
01□
02□
03□
B18APPL54B
00□
01□
02□
03□
B18APPL64B
00□
01□
02□
03□
B18APPL65B
Yourself?
□
Someone
else? □
B18DIDSFB
1 2 3 4
B18FORMB
1 2 3 4
B18STOREB
0 1 2 3 4
B18PROTB

0 1 2
B18UPROB
C B18CHEMC
00□
01□
02□
03□
B18APPL25C
00□
01□
02□
03□
B18APPL44C
00□
01□
02□
03□
B18APPL54C
00□
01□
02□
03□
B18APPL64C
00□
01□
02□
03□
B18APPL65C
Yourself?
□
Someone
else? □
B18DIDSFC
1 2 3 4
B18FORMC
1 2 3 4
B18STOREC
0 1 2 3 4
B18PROTC

0 1 2
B18UPROC
229

B19. Have you ever used chemicals to kill insects in your home?
(B19BUGKILL)

00 No
01 Yes
77 Can’t remember
88 Refused
99 I don’t know

B20. Has anyone in your household ever used chemicals to kill insects
in your home? (B20HOMEBUG)

00 No
01 Yes
77 Can’t remember
88 Refused
99 I don’t know

230

B21. For what kind of insects did you (or someone in your household) use insecticides?

What did you
use?

Chemical name
or purpose
77= Can’t
remember
88=Refused
99=Don’t know

How often applied?
00 = None,
01 = Rarely (once a year or less),
02 = Sometimes (2-11 times a year),
03 = Regularly (once a week to once a month)
Was it…?
Where did
you store
the
chemical?

Did you
protect
yourself
with?
00=No;
01=Yes
How often
did you
protect
yourself?
<25 YEARS

25-44
YEARS

45-54
YEARS
55-64
YEARS
65+
YEARS
Who did
it?
Spray
Granules
Bait
Powder

Garage
In the house
Storage Facility
Other: Specify

None
Gloves
Mask
Coveralls
Other: Specify
Never
Sometimes
Always
A B21CHEMA
00□
01□
02□
03□
B21APPL25A
00□
01□
02□
03□
B21APPL44A
00□
01□
02□
03□
B21APPL54A
00□
01□
02□
03□
B21APPL64A
00□
01□
02□
03□
B21APPL65A
Yourself?
□
Someone
else? □
B21DIDSFA
1 2 3 4
B21FORMA
1 2 3 4
B21STOREA
0 1 2 3 4
B21PROTA

0 1 2
B21UPROA
B B21CHEMB
00□
01□
02□
03□
B21APPL25B
00□
01□
02□
03□
B21APPL44B
00□
01□
02□
03□
B21APPL54B
00□
01□
02□
03□
B21APPL64B
00□
01□
02□
03□
B21APPL65B
Yourself?
□
Someone
else? □
B21DIDSFB
1 2 3 4
B21FORMB
1 2 3 4
B21STOREB
0 1 2 3 4
B21PROTB

0 1 2
B21UPROB
C B21CHEMC
00□
01□
02□
03□
B21APPL25C
00□
01□
02□
03□
B21APPL44C
00□
01□
02□
03□
B21APPL54C
00□
01□
02□
03□
B21APPL64C
00□
01□
02□
03□
B21APPL65C
Yourself?
□
Someone
else? □
B21DIDSFC
1 2 3 4
B21FORMC
1 2 3 4
B21STOREC
0 1 2 3 4
B21PROTC

0 1 2
B21UPROC

01□ Ants 03□Spiders 05□Bees, hornets, wasps 07□Other crawling insects__________
02□Cockroaches 04□Termites 06□Flies 08□Other flying insects____________

231

B22. Have you ever hired a professional to spray or fumigate insects or kill weeds or plant diseases in your
home or garden? (B22HIREPRO)
00 No (If No, Skip to Section C)
01 Yes
B23. If yes, did
you hire the
professional to…
…….
How often did the professional apply chemicals per year?
00 = None,
01 = Rarely (once a year or less),
02 = Sometimes (2-11 times a year),
03 = Regularly (once a week to once a month)
Was it…?
<25 YEARS

25-44
YEARS
45-54
YEARS
55-64
YEARS
65+ YEARS

A Kill Weeds
B23WEEDS
00□
01□
02□
03□
B23APPL25A
00□
01□
02□
03□
B23APPL44A
00□
01□
02□
03□
B23APPL54A
00□
01□
02□
03□
B23APPL64A
00□
01□
02□
03□
B23APPL65A
01 □ Spray
02 □ Other
99 □ Don’t Know
B23FORMA
B Kill Insects
B23INSECTS
00□
01□
02□
03□
B23APPL25B
00□
01□
02□
03□
B23APPL44B
00□
01□
02□
03□
B23APPL54B
00□
01□
02□
03□
B23APPL64B
00□
01□
02□
03□
B23APPL65B
01 □ Spray
02 □ Other
99 □ Don’t Know
B23FORMB
C Kill Termites
B23TERMITE
00□
01□
02□
03□
B23APPL25C
00□
01□
02□
03□
B23APPL44C
00□
01□
02□
03□
B23APPL54C
00□
01□
02□
03□
B23APPL64C
00□
01□
02□
03□
B23APPL65C
01 □ Spray
02 □ Other
99 □ Don’t Know
B23FORMC
D OTHER: Specify
B23OTHER
00□
01□
02□
03□
B23APPL25D
00□
01□
02□
03□
B23APPL44D
00□
01□
02□
03□
B23APPL54D
00□
01□
02□
03□
B23APPL64D
00□
01□
02□
03□
B23APPL65D
01 □ Spray
02 □ Other
99 □ Don’t Know
B23FORMD
232

PART C: OCCUPATIONAL HISTORY AND PESTICIDE
EXPOSURE

The next questions will ask for information about your work
experience.

C1. In one of your jobs have you ever worked on a regular basis (once
a week or more) with…

Exposure
Worked
with
regularly
No Yes
A.
 Lead (C1LEAD)
 Copper (C1COPPER)
 Cadmium (C1CAD)
 Iron (C1IRON)
 Other (C1OTHMET)
specify___________________

00
00
00
00
00
00

01
01
01
01
01
01

B. Wood (C1WOOD) 00 01
C. Chemical solvents (C1SOLVENT) 00 01
D. Paint strippers (C1PAINT) 00 01
E. Fertilizers or pesticides (C1FERT) 00 01

C2. Have you ever worked on a farm, nursery, or orchard or ever
worked as a professional pesticide applicator? (C2APPLWRK)
If NO, Need to ASK C3-C7 (SKIP C8-C12) then continue questions beginning with ASKING C13
00 No
01 Yes
77 Can’t remember
88 Refused
99 I don’t know

233

C3: Please tell me all the jobs you have had that involved pesticides,
such as farm work, professional pesticide application, pesticide
manufacturing, gardener/ orchard work or any other job that has
exposed you to pesticides, starting with the earliest one. If your job title
changed while working at the same company or farm, please treat each
change as a new job.

For example, if you were a worker on the farm and then were promoted
to manager at the same farm, please treat those as two separate jobs. In
addition, we would like to know what your job title was, the name and
address of the farm or company you worked for, the company type
(such as nursery, chemical manufacturing or farm), a brief description
of your job responsibilities and the dates you worked there.

MAKE SURE TO ASK: are those ALL the jobs you had involving
pesticides?

Job Title

Company
Name and
Address
Company
Type
Job
Responsibilities
Period
(dates
A. C8JOBA

C8ADDRA C8TYPECOA C8RESPA C8PERIODA
B. C8JOBB

C8ADDRB C8TYPECOB C8RESPB C8PERIODB
C. C8JOBC

C8ADDRC C8TYPECOC C8RESPC C8PERIODC
D. C8JOBD

C8ADDRD C8TYPECOD C8RESPD C8PERIODD

234

C4: I am going to ask you more specific questions about your work as
a (say job title AND time period)

For EACH job title in previous table, fill out

Job
Title
(from
table
C8)
Where you
involved in
the mixing
and loading
of pesticides?

C9FLOAD

Did you
apply the
pesticides
yourself?

C9GAPPLY
Did you wash
your hands
with water at
work?

C9HWASH
Did you take
a shower
before going
home?

C9JSTORE
Did you
wash your
OWN work
clothes?

C9KWASH
A. C9FLOADA

□ Yes
□ No
99□Don’t Know
C9TYPEA

□ Yes
□ No
99□Don’t Know
C9WASHA

□ Yes
□ No
99□Don’t Know
C9PERIODAJ

□ Yes
□ No
99□Don’t Know
C9WACHA

□ Yes
□ No
99□Don’t Know
B.

C9FLOADB

□ Yes
□ No
99□Don’t Know
C9TYPEB

□ Yes
□ No
99□Don’t Know
C9WASHB

□ Yes
□ No
99□Don’t Know
C9PERIODBJ

□ Yes
□ No
99□Don’t Know
C9WACHB

□ Yes
□ No
99□Don’t Know
C.

C9FLOADC

□ Yes
□ No
99□Don’t Know
C9TYPEC

□ Yes
□ No
99□Don’t Know
C9WASHC

□ Yes
□ No
99□Don’t Know
C9PERIODCJ

□ Yes
□ No
99□Don’t Know
C9WACHC

□ Yes
□ No
99□Don’t Know
D. C9FLOAD

□ Yes
□ No
99□Don’t Know
C9TYPED

□ Yes
□ No
99□Don’t Know
C9WASHD

□ Yes
□ No
99□Don’t Know
C9PERIODD

□ Yes
□ No
99□Don’t Know
C9WACHD

□ Yes
□ No
99□Don’t Know

235

C5: CROP INFORMATION
What field crops were grown on the farm you worked on?
Field Crops Were pesticides used? Did you mix or apply
the pesticides?
A. (C10CROPA)

___________________
 Yes
 No
 Don’t know
C10PESTA
 Yes
 No
 Don’t know
C10MIXA
B. (C10CROPB)

___________________
 Yes
 No
 Don’t know
C10PESTB
 Yes
 No
 Don’t know
C10MIXB
C. (C10CROPC)

___________________
 Yes
 No
 Don’t know
C10PESTC
 Yes
 No
 Don’t know
C10MIXC

C6: What tree crops were grown on the farm you worked on?
TREE CROPS

Were pesticides used? Did you mix or apply
the pesticides?
A. (C11CROPA)

___________________
 Yes
 No
 Don’t know
C11PESTA
 Yes
 No
 Don’t know
C11MIXA
B. (C11CROPB)

___________________
 Yes
 No
 Don’t know
C11PESTB
 Yes
 No
 Don’t know
C11MIXB
C. (C11CROPC)

___________________
 Yes
 No
 Don’t know
C11PESTC
 Yes
 No
 Don’t know
C11MIXC

C7: What animals were raised on the farms you worked on?
ANIMALS
(# of animals)
Year
began
Year
ended
Were pesticides
used?
Did you mix or apply
the pesticides?
A. (C12ANMLA)

(C12NUMANA)
(C12BEGA)

(C12ENDA)

 Yes
 No
 Don’t know
C12PESTA
 Yes
 No
 Don’t know
C12MIXA
B. (C12ANMLB)

(C12NUMANB
(C12BEGB)

(C12ENDB)

 Yes
 No
 Don’t know
C12PESTB
 Yes
 No
 Don’t know
C12MIXB
C. (C12ANMLC)

(C12NUMANC)
(C12BEGC)

(C12ENDC)

 Yes
 No
 Don’t know
C12PESTC
 Yes
 No
 Don’t know
C12MIXC
236

Now I would like to find out more about any job you may have held
(paid or unpaid) during your lifetime that you haven’t already told me
about. Do you have your timeline ready?

C8. What was your first job that lasted a total of at least half a year?
Please start with your LAST job.

Job

(C13JOBTITLA-C13JOBTITLP)
Code
Year when Hours
How
long
Started Ended
Per
week
Years
Months
A.
Title: (C13JOBTITLA)
C13JOBCODEA
C13JOBBEGA
C13JOBENDA
C13JOBHRSA
C13JOBYRSA
C13JOBMOSA
Tasks: (C13JOBTASKA)
Industry: (C13JOBINDA)
Company: (C13COMPA)
Address: (C13COADDA)

B.
Title: (C13JOBTITLB)
C13JOBCODEB
C13JOBBEGB
C13JOBENDB
C13JOBHRSB
C13JOBYRSB
C13JOBMOSB
Tasks: (C13JOBTASKB)
Industry: (C13JOBINDB)
Company: (C13COMPB)
Address: (C13COADDB)

C.
Title: (C13JOBTITLC)
C13JOBCODEC
C13JOBBEGC
C13JOBENDC
C13JOBHRSC
C13JOBYRSC
C13JOBMOSC
Tasks: (C13JOBTASKC)
Industry: (C13JOBINDC)
Company: (C13COMPC)
Address: (C13COADDC)

237

PART D: DIET

Now I would like to ask some questions about your diet.

D1. Are you now following a special diet? (D1SPECDIET)

00 No (If No, Go to question D3)
01 Yes, Dr. prescribed
02 Yes, Self-prescribed
77 Can’t remember
88 Refused
99 I don’t know

D2. What kind of diet? (D2DIETTYPE)
Don’t give the alternatives; select all that apply. Probe if necessary.

Low cholesterol (D2DIETCHL)
Diabetic (D2DIETDIA)
Low fat (D2DIETFAT)
Triglyceride (D2DIETTRI)
Ulcer (D2DIETULC)
Potassium (D2DIETVIK)
Other (specify): ______________________________________(D2OTHDIET)
77 Can’t remember
88 Refused
99 Don’t know

238

D3. I will now ask you some general questions about your usual eating
habits during the past year…

General questions
How Often/ Much
Don’t eat Per day Per week Per mo.
A.
How often did you eat meat during the past year? D3NOMEAT D3MEATDAY D3MEATWEEK D3MEATMO
B.
How often did you eat poultry during the past year? D3NOFOUL D3FOULDAY D3FOULWEEK D3FOULMO
C.
How often did you eat seafood during the past year? D3NOFISH D3FISHDAY D3FISHWEEK D3FISHMO
D.
About how many servings of vegetables did you eat,
not counting salad or potatoes, during the past
year?
D3NOVEG D3VEGDAY D3VEGWEEK D3VEGMO
E.
About how many servings of fruit did you eat, not
counting juices, during the past year?
D3NOFRT D3FRTDAY D3FRTWEEK D3FRTMO

239

Next we would like to know about changes to your eating habits over time.
D4. Compared to the past year, as a (age period) did you usually eat (type of food) (more, less, same) . . . ?
Ask only about the age periods that are applicable.

Age period
<25 years 25-44 years 45-54 years 55-64 years 65+ years
A.
Compared to last year, how
often did you eat meat?
D4MEAT25
0 Didn’t eat
1 Less
2 Same
3 More
D4MEAT44
0 Didn’t eat
1 Less
2 Same
3 More
D4MEAT54
0 Didn’t eat
1 Less
2 Same
3 More
D4MEAT64
0 Didn’t eat
1 Less
2 Same
3 More
D4MEAT65
0 Didn’t eat
1 Less
2 Same
3 More
B.
Compared to last year, how
often did you eat poultry?
D4FOUL25
0 Didn’t eat
1 Less
2 Same
3 More
D4FOUL44
0 Didn’t eat
1 Less
2 Same
3 More
D4FOUL54
0 Didn’t eat
1 Less
2 Same
3 More
D4FOUL64
0 Didn’t eat
1 Less
2 Same
3 More
D4FOUL65
0 Didn’t eat
1 Less
2 Same
3 More
C.
Compared to last year, how
often did you eat seafood?
D4FISH25
0 Didn’t eat
1 Less
2 Same
3 More
D4FISH44
0 Didn’t eat
1 Less
2 Same
3 More
D4FISH54
0 Didn’t eat
1 Less
2 Same
3 More
D4FISH64
0 Didn’t eat
1 Less
2 Same
3 More
D4FISH65
0 Didn’t eat
1 Less
2 Same
3 More
D.
Compared to last year, how
much vegetables did you eat,
not counting salad or potatoes?
D4VEG25
0 Didn’t eat
1 Less
2 Same
3 More
D4VEG44
0 Didn’t eat
1 Less
2 Same
3 More
D4VEG54
0 Didn’t eat
1 Less
2 Same
3 More
D4VEG64
0 Didn’t eat
1 Less
2 Same
3 More
D4VEG65
0 Didn’t eat
1 Less
2 Same
3 More
E.
Compared to last year, how
much fruit did you eat, not
counting juices?
D4FRT25
0 Didn’t eat
1 Less
2 Same
3 More
D4FRT44
0 Didn’t eat
1 Less
2 Same
3 More
D4FRT54
0 Didn’t eat
1 Less
2 Same
3 More
D4FRT64
0 Didn’t eat
1 Less
2 Same
3 More
D4FRT65
0 Didn’t eat
1 Less
2 Same
3 More
240

PART E: PHYSICAL ACTIVITY

These questions are about physical activity and rest. Please think about
the average amount of time that you spent being physically active or
resting and include both time at work and leisure.

Strenuous exercise: we mean an activity that makes a person
breathe hard, such as active firefighting or
jogging (aerobics, bicycling on hills, tennis,
weight lifting, basketball)

Moderate exercise: are activities that are physically demanding but
do not usually make a person breathe hard:
such as harvesting crops manually, brisk
walking, golf, cycling on level streets,
swimming.

Light exercise: teaching, casual walking, light housework, light
gardening, standing, etc.

Let’s start with strenuous exercise, during (age period) how many hours
per day (or per week or per month) did you usually engage in strenuous
(moderate, light) exercise?

Type of activity
How many hours
Not at all Rarely Per day Per week Per mo.
At age… < 25 years
E1NOEX-
E15NOEX

E1DAYEX-
E15DAYEX
E1WKEX-
E15WKEX
E1MOEX-
E15MOEX
E1. Strenuous 0 1
E2. Moderate 0 1
E3. Light 0 1
At age… 25-44 years
E1. Strenuous 0 1
E2. Moderate 0 1
E3. Light 0 1
At age… 45-54 years
E1. Strenuous 0 1
E2. Moderate 0 1
E3. Light 0 1

241

Type of activity
How many hours
Not at all Rarely Per day Per week Per mo.
At age… 55-64 years
E1. Strenuous 0 1
E2. Moderate 0 1
E3. Light 0 1
At age… 65+ years
E1. Strenuous 0 1
E2. Moderate 0 1
E3. Light 0 1

E21. Have you ever competed in sports? (E21SPORTS)
00 No
01 Yes
77 Can’t remember
88 Refused
99 I don’t know

E22. IF YES: What type of sports have you competed in?

pe of Sports Age when
Code Started Ended
A.
B.
C.

242

PART F: PHYSICAL DEVELOPMENT

F2. What is your current height without shoes?
____ft(F2HEIGHTFT) ___in(F2HEIGHTIN) or ______cm (F2HEIGHTCM)

F3. What is your current weight without shoes? _________ (F3WEIGHT)

F4. At what age did you weigh the most? (F4HIWEIGHT)
01 Under 20 02 20-24 03 25-29 04 30-34
05 35-39 06 40-44 07 45-49 08 50-54
09 55-59 10 60-64 11 65-69 12 70+

F5. After the age of 20, what age did you weigh least? (F5LOWEIGHT)
02 20-24 03 25-29 04 30-34 05 35-39
06 40-44 07 45-49 08 50-54 09 55-59
10 60-64 11 65-69 12 70+

F6. What has been your pattern of weight gain or loss over your adult
life?
(F6WTPATT)
01 Weight has remained stable
02 Varied up or down by 10 pounds or less
03 Varied up or down by more than 10 pounds
04 Steadily increased
05 Steadily decreased

F7. What is the largest amount of weight you have lost when dieting?
(F7WTLOST)
01 Never dieted 04 11-20 pounds 07 41-50 pounds
02 0-5 pounds 05 21-30 pounds 08 > 50 pounds
03 6-10 pounds 06 31-40 pounds

F8. What is your current waist size? F9. Current hip size?
______inches (F8WAISTIN) ______inches (F9HIPINCH)

F11. How does your hip measurement compare to your waist
measurement? (F11HWRATIO)
01 Hips much bigger 04 Waist bigger 03 About the same
02 Hips bigger 05 Waist much bigger

243

PART G: REPRODUCTIVE HISTORY

G1. How old were you when you had your first menstrual period
(started bleeding regularly)? __________ (G1MENSES)

G2. Have you ever been pregnant? (G2PREGNT)
00 No  If NO, go to G7
01 Yes
77 Can’t remember
88 Refused
99 I don’t know

G3. How many pregnancies have you had (including any miscarriages
you may have had)? __________ (G3PREGNO)

G4. How many deliveries have you had (including any stillborn)?
__________ (G4DELIVR)

G5. Did you ever take the drug diethylstilbestrol (DES) to prevent
miscarriage?
(G5DES)
00 No
01 Yes
77 Can’t remember
88 Refused
99 I don’t know

G6. How old were you at the time of your first full-term pregnancy?
__________ (G6AGEPRG)

G7a. Did you ever breast feed? (G7BRSFED)
00 No  If NO, go to G7
01 Yes
77 Can’t remember
88 Refused
99 I don’t know

G7b. If YES, for how long did you breast feed? __________ (G7BRFED2)

244

G8a. Have you ever used birth control methods such as oral
contraceptives, birth control shots or injections, or implants (hormonal
methods) for one month or longer? (note: this does not include non-
hormonal methods such as IUD’s or condom use) (G8BRCTRL)
00 No  If NO, go to G9
01 Yes, and I am currently using hormonal birth control
02 Yes, but I no longer use any hormonal birth control method
77 Can’t remember
88 Refused
99 I don’t know

G8b. At what age did you use (method) first? Last? How long?
Type Used?
Age when How long
Started Ended Years Months
A. Birth control
pills
□ Yes
□ No
□ Don’t Know
G8PILLBEG G8PILLEND G8PILLYR G8PILLMO
B. Birth control
shots/injections
□ Yes
□ No
□ Don’t Know
G8SHOTBEG G8SHOTEND G8SHOTYR G8SHOTMO
C. Implants
□ Yes
□ No
□ Don’t Know
G8IMPBEG G8IMPEND G8IMPYR G8IMPMO

G9a. Have you ever had one or both ovaries removed? (G9OVARY)
00 No  If NO, go to G9
01 Yes, but only one or part of one
02 Yes, both at the same time
03 Yes, both but at different times

G9b. At what age did you have your ovary(ies) removed? __________
(G9OVRYAG)

G10a. Has your menstrual period stopped permanently? (G10MENOP)
00 No  If NO, go to G12
01 Yes
77 Can’t remember
88 Refused
99 I don’t know

G10b. At what age did you have your last period? __________ (G10MENOEND)
245

G11a. Have you ever taken or been injected with estrogen or
progesterone (or “progestin”) “female hormones” for symptoms of
menopause (after your period stopped permanently) or for other
reasons? (G11HRT)
00 No  If NO, go to G9
01 Yes, and I am currently taking estrogen
02 Yes, and I am currently taking progesterone or progestin
03 Yes, and I am currently taking combined estrogen/progestin
04 Yes, but I am no longer taking
77 Can’t remember
88 Refused
99 I don’t know

G11b. At what age did you use (method) first? Last? How long?

Type
Which one?
Age when How long
Started Ended Years Months
A. Estrogen only
□ Oral Premarin pills
□ Ogen pills
□ Estrace pills
□ Patch estrogen
□ Injection estrogen
□ Other
G11EST
BEG
G11EST
END
G11ES
TYR
G11EST
MO
B.
Progesterone or
progestin only
□ Oral Provera pills
□ Other

G11PRO
BEG
G11PRO
END
G11PR
OYR
G11PRO
MO
C.
Combined estrogen
and progestin
□ Prempro
□ Premphase
□ Other
G11CO
MBEG
G11CO
MEND
G11CO
MYR
G11COM
MO

246

PART H: SMOKING AND ALOCHOL USE

These next questions are about smoking and alcohol use.

H1. Have you ever smoked a total of 100 or more cigarettes in your
lifetime?
(H1EVERSMK)
00 No -----  If no, go to section H5
01 Yes, I currently smoke.
02 Yes, but I no longer smoke.

H2. How old were you when you first smoked regularly? _______yrs
(H2AGESMK)

H3. If you no longer smoke, how old were you when you last smoked?
________yrs (H3LASTSMK)

H4. On average, about how many cigarettes per day do you/did you
smoke during these periods?
Age/Time How many per day?
<-25yrs (H4CIGS25)
26-44yrs (H4CIGS44)

45-54yrs (H4CIGS54)

55-64yrs (H4CIGS64)
65+ (or past year) (H4CIGSPYR)

H5a. Did you ever drink alcohol at least once a week? (H5DRINK)
00 No  If NO, go to section I
01 Yes

H5b. At what age did you start drinking at least once a week?
_________ (H5DRKAGE)
247

H5c. Please tell me how much and how often you usually drank/drink
alcoholic beverages, where 1 drink is equal to…

1 can/glass/bottle of beer,
1 glass of wine/champagne/wine cooler or
1 cocktail, shot or mixed drink of liquor

Age/Time Drinks per week
1 0 2 1-3 3 4-10 4 11-17 5 18-24 6 24+
<-25yrs
(H5DRK25)

25-44yrs
(H5DRK44)

45-54yrs
(H5DRK54)

55-64yrs
(H5DRK64)

65+ (or past year)
(H5DRKPSTYR)

248

PART I: USE OF MEDICAL CARE SERVICES

I1. Have you ever had a mammogram? (I1MAMMO)
01 Yes
00 No  Skip to I5.

I2. Did you have the mammogram…? (I2WHYMAM)
01 As part of a routine physical exam or screening test
02 Because of a specific problem
03 As a follow-up to an earlier test or screening exam
99 Don’t know

I3. At what age did you have your first mammogram?_____ (I3MAMAGE)

I4. How many mammograms have you had in your lifetime?_____
(I4MANYMAM)

I5a. Have you ever had a benign breast condition (such as fibrocystic
breast disease, or cysts, mastitis, or a benign breast tumor)? (I5BRSTDZ)
00 No  skip to I15
01 Yes
99 Don’t know

I5b. What age were you when your benign breast condition
diagnosed?_____
(I5BRDZAGE)

I6a. Have you ever had another type of cancer? (I6OTHRCA)
01 Yes
00 No
99 Don’t know

I6b. If YES, have you ever had radiation therapy to the chest area for
treatment for another cancer (such as Hodgkin disease or non-Hodgkin
lymphoma? (I6RADTX)
00 No
99 Don’t know
01 Yes I16c.  If yes, at what age? ________ (I6RADTXAGE)

249

PART J: DEMOGRAPHIC AND BACKGROUND INFORMATION

We have arrived at the last section; the following are just a few more
questions about you.

J1. What is your date of birth? ____/____/______ (J1DATEBRTH)

J2. Are you Hispanic or Latino? (J2HISPLAT)
01 Yes
00 No

J3. What is your ethnic or racial background? (J3RACE)
01 White/European
02 White/Middle Eastern
03 Mexican, Central, South American, or other Hispanic
04 Black or African-American
05 American Indian or Alaskan Native
06 Asian Indian
07 Chinese
08 Filipino
09 Japanese (including Okinawa)
10 Korean
11 Vietnamese or other Southeast Asian
12 Other Asian
13 Native Hawaiian
14 Guamanian or Chamorro
15 Samoan
16 Other Pacific Islander
17 Some other race/ethnicity

J4. What is your current marital status? (J4MARRSTAT)
01 Married
02 Unmarried, living with a partner
03 Never married
04 Separated or divorced
05 Widowed

J5. How many times have you been married? ______ (J5NUMMARR)

250

J6. What is the highest degree or level of education you have
completed? (J6EDULEV)
01

Grade 4 or less
02

Grade 5-8
03

Grade 9-11
04

Grade 12, High School Graduate, or G.E.D.
05

Some College
06

Associate Degree (e.g AA, AS)
07

Bachelor’s Degree (e.g. BA, AB, BS)
08

Master’s Degree (e.g. MA, MS, MEd, MSW, MBA)
09

Professional Degree (e.g. MD, DDS, DVM, LLB, JD)
10

Doctorate Degree (e.g. PhD, EdD)

J7. What type of medical insurance did you have at the time of your
cancer diagnosis (cases)/ 6 months ago (controls)? (Check all that apply)
(J7MEDINS)
01

No insurance 04

Medicare
02

Private Insurance 05

Medicaid
03

HMO/Kaiser 06

Other:_____(J7OTHRINS)

J8. What about now? (Check all that apply) (J8CURMEDINS)
01

No insurance 04

Medicare
02

Private Insurance 05

Medicaid
03

HMO/Kaiser 06

Other:______(J8CUROTHR)
07 Same

J9. Which of the following choices best describes your work situation at
the time of your cancer diagnosis (cases)/ 6 months ago (controls)?
(Check all that apply) (J9EMPSTAT)
01

I worked full-time. 05

I was a student.
02

I worked part-time. 06

I was unemployed.
03

I worked occasionally. 07

I am disabled (unable to work).
04

I was a homemaker. 08

I was retired.

J10. Which of the following choices best describes your work situation
now)? (Check all that apply) (J10EMPCURR)
01

I work full-time. 05

I am a student.
02

I work part-time. 06

I am unemployed.
03

I work occasionally. 07

I am disabled (unable to work).
04

I am a homemaker. 08

I am retired.
251

J11. Have any of your relatives have/had breast cancer? Did your
mother, grandmother, sister, half-sister, daughter, aunt or niece ever
have breast cancer?

Mother

J11FAMHXM
Grand-
mother
J11FAMHXG
Sister

J11FAMHXS
Half-
sister
J11FAMHXHS
Daughter

J11FAMHXD
Niece

J11FAMHXN
Aunt

J11FAMHXA
01 Yes
00 No
99 DK
01 Yes
00 No
99 DK
01 Yes
00 No
99 DK
01 Yes
00 No
99 DK
01 Yes
00 No
99 DK
01 Yes
00 No
99 DK
01 Yes
00 No
99 DK

We have now finished the questionnaire and we greatly appreciate you
taking the time to participate in this study!

252

Abstract (if available)

Abstract There is strong evidence that a variety of pesticides affect hormone metabolism. Exposure to these chemicals may be involved in the development of breast cancer, but previous studies have had conflicting results. Current methodology is severely lacking for assessing cumulative exposures that amass over a lifetime. The objectives of this research were to assess breast cancer risk from ambient pesticide exposure using a Geographical Information Systems (GIS)-based approach to examine historical effects in the most agriculturally productive region in the U.S., where pesticide drift from neighboring application presents a major source of exposure. ❧ We conducted registry-based and pilot case-control studies employing a GIS-based exposure assessment to see if a comprehensive exposure model revealed true risk for breast cancer in postmenopausal women from specific, hormone-related pesticides. We evaluated risk from a selected group of organochlorine pesticides with biologically-plausible links to breast cancer and three other pesticides detected in ambient air monitoring at levels of concern to human health (chlorpyrifos, diazinon, and 1,3-dichloropropene). We observed no association between breast cancer and exposure to any of the pesticides in the registry-based study

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

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

PDF

The effects of hormonal exposures on ovarian and breast cancer risk

PDF

The role of heritability and genetic variation in cancer and cancer survival

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

The effects of tobacco exposure on hormone levels and breast cancer risk among young women

PDF

Carcinogenic exposures in racial/ethnic groups

PDF

Application of geospatial methods to cancer surveillance data to improve cancer prevention and control

PDF

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

PDF

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

PDF

Genes and hormonal factors involved in the development or recurrence of breast cancer

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

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

PDF

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

PDF

Identifying genetic, environmental, and lifestyle determinants of ethnic variation in risk of pancreatic cancer

PDF

Personal exposure to particulate matter PM2.5 sources during pregnancy and birthweight

PDF

Lifetime physical activity and its effects on breast cancer survival

PDF

Genetic and environmental risk factors for childhood cancer

PDF

Disparities in colorectal cancer survival among Latinos in California

PDF

Evaluation of new methods for estimating exposure to traffic-related pollution and early health effects for large population epidemiological studies

PDF

Factors that influence mammographic density: role of estrogen metabolism genes, biomarkers of inflammation, and lifestyle

Asset Metadata

Creator Tayour, Carrie (author)

Core Title The role of pesticide exposure in breast cancer

School Keck School of Medicine

Degree Doctor of Philosophy

Degree Program Epidemiology

Publication Date 02/10/2015

Defense Date 12/16/2014

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

Tag breast cancer,case control,Epidemiology,geocode,geographical information systems,GIS,OAI-PMH Harvest,pesticide

Format application/pdf (imt)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Cockburn, Myles (committee chair), Franklin, Meredith (committee member), Langholz, Bryan (committee member), Wilson, John P. (committee member), Wu, Anna (committee member)

Creator Email cnagy@usc.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c3-530650

Unique identifier UC11297302

Identifier etd-TayourCarr-3170.pdf (filename),usctheses-c3-530650 (legacy record id)

Legacy Identifier etd-TayourCarr-3170.pdf

Dmrecord 530650

Document Type Dissertation

Format application/pdf (imt)

Rights Tayour, Carrie

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