 

PLACE IN RETURN BOX to remove this checkout from your record.
TO AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

5I08 K:IProj/Aoc&Pres/CIRCIDateDue.indd

 

 

A POPULATION-BASED CASE-CONTROL STUDY OF PREGNANCY-
RELATED FACTORS AND MATERNAL BREAST CANCER RISK AMONG
YOUNGER WOMEN
by

Sarah Jean Nechuta

A DISSERTATION
Submitted to
Michigan State University
in partial fulﬁllment of the requirements
for the degree of
DOCTOR OF PHILOSOPHY
Epidemiology

2009

ABSTRACT
A POPULATION-BASED CASE-CONTROL STUDY OF PREGNANCY-
RELATED FACTORS AND MATERNAL BREAST CANCER RISK AMONG
YOUNGER WOMEN
By
Sarah Jean Nechuta
Though the etiology of breast cancer among younger women (s 50 years

of age) remains largely unknown, pregnancy is known to be a critically important
time in relation to a women’s subsequent risk of breast cancer. The biological
mechanisms underlying the observed short-term increase and long-term
decrease in breast cancer risk following childbirth, which depend on age at
pregnancy, are not clear. Variation in fetal growth (PG) or gestational age (GA) in
a woman’s own pregnancies may serve as indirect markers of the hormonal
environment during pregnancy. The overall goal of this dissertation research is
to investigate the associations between two under-investigated perinatal
exposures—PG and GA, as well as age at ﬁrst and last delivery, number of live
births, and maternal breast cancer risk (overall and for ductal and Iobular
histologic types) among parous Black and White Michigan (MI) women s 50
years of age. We conducted a population-based case-control study using linked
Ml Cancer Registry (1985-2004) and MI Live Birth records (1978-2004). Cases
were matched to controls 1:4 on maternal birth year and race (original sample:
8,251 cases and 33,004 controls). Using conditional logistic regression, we

examined the associations for breast cancer and age at first and last birth,

number of live births, GA and FG (deﬁned using BW percentiles both as a

continuous variable and categorized ((SGA) < 10th, (AGA) 10-90th (referent),
(LGA) > 90th)). Later age at first and last birth and multiparity were independently
associated with increased risks for both ductal and Iobular breast cancer, with
odds ratios (ORs) of similar magnitude. Some differences were found by race
(White, Black), including an increased risk of Iobular tumors for age at last birth 2
30 years (vs. < 30 years) among White women only (OR=1.70, 95% Cl: 1.21-
2.40). Delivery of an SGA or an LGA infant in a ﬁrst or last birth was not
signiﬁcantly associated with breast cancer risk, but among women with a last
birth at age 2 30 years, delivery of an SGA infant in a last birth was associated
with a reduced risk of breast cancer (OR=0.82, 95% CI: 0.68-0.98). A ﬁrst
delivery at < 32 weeks or at > 41 weeks (reference: 37—41 week) was associated
with reduced risks (ORs: 0.80, 95% CI: 0.62-1.04 and 0.92, 95% CI: 0.85-0.99,
respectively). In this large, population-based case-control study of parous women
s 50 years of age, we found limited evidence for an association between low or
high F6 and overall breast cancer risk. Delivery of an infant < 32 weeks in a first
birth may reduce breast cancer risk, but this ﬁnding was in contrast to our
hypothesis and the underlying biological mechanisms are not clear. Our results
for GA and breast cancer risk, as well as the inconsistent ﬁndings to date in this
area warrant future research to characterize these associations, including studies
with information on both biologic measures during pregnancy and other potential

confounding factors (e.g., maternal body size).

Copyright by
SARAH JEAN NECHUTA
2009

This dissertation is dedicated to Nate, thank you for your unfailing kindness, love,

and support.

I18.

 

VE

c0

ACKNOWLEDGEMENTS

I would like to begin by thanking my advisor and dissertation committee
chair, Dr. Ellen Velie, for her invaluable suggestions and guidance from the
beginning to the end of this process. I am also very grateful to my dissertation
committee members, Dr. Paneth, Dr. Gardiner, and Dr. Pathak, for their
encouragement and guidance throughout this process.

I would like to express my thanks and appreciation to Glenn Copeland
Director-Division for Vital Records and Health Statistics of the Michigan
Department of Community Health for his assistance in creating datasets for this
study and for offering numerous helpful suggestions and comments during
various stages of this project.

I also want to thank all my friends and family, for their encouragement and
support. In particular, I want to thank my mom, who instilled in me a love of
learning and has always believed in me and encouraged me to pursue my goals.

This study was made possible due to research support from the Michigan
State University T32 training program in perinatal epidemiology (NIH-T32
HDO46377), a grant from the National Cancer Institute (NlH-RO3CA128010), and
a student award program grant from the Blue Cross Blue Shield of Michigan

Foundation (1271 .sap).

vi

 

CH.
All:

CH»:
CAN

TABLE OF CONTENTS

LIST OF TABLES ................................................................................................. ix
LIST OF FIGURES .............................................................................................. xii
ABBREVIATIONS .............................................................................................. xiii
CHAPTER 1: INTRODUCTION AND AIMS .......................................................... 1
1. Introduction ............................................................................................ 1
2. Aims and Hypotheses ............................................................................ 5
3. Organization of Dissertation .................................................................. 7

CHAPTER 2: PERINATAL FACTORS AND MATERNAL BREAST CANCER

RISK: A REVIEW OF THE EPIDEMIOLOGIC LITERATURE ............................... 8
1. Abstract ................................................................................................. 8
2. Introduction ............................................................................................ 9
3. Methods ............................................................................................... 11
4. Preeclampsia, Pregnancy-Induced Hypertension, and Placental
Characteristics ......................................................................................... 15
5. Multiple Births ...................................................................................... 20
6. Gestational Diabetes ........................................................................... 25
7. Pregnancy Weight Gain ....................................................................... 28
8. Fetal Growth ........................................................................................ 31
9. Infant Gestational Age at Delivery ....................................................... 35
10. Infant Sex .......................................................................................... 39
11. Conclusion ......................................................................................... 40
12. Search Strategy ................................................................................. 42
13. Tables ................................................................................................ 44

CHAPTER 3: STUDY DESIGN, DATA PREPARATION, ................................... 64

AND METHODOLOGIC ISSUES ....................................................................... 64
1. Study Design and Study Population .................................................... 64
2. Data Sources and Measurement of Study Variables ........................... 67
3. Dataset Creation Procedures ............................................................... 74
4. Methodologic Issues ............................................................................ 95

CHAPTER 4: PREGNANCY-RELATED FACTORS AND RISK OF BREAST
CANCER BY HISTOLOGIC TYPE, A REGISTRY-BASED STUDY OF PAROUS

BLACK AND WHITE YOUNGER WOMEN ...................................................... 103

- 1. Abstract ............................................................................................. 103
2. Introduction ........................................................................................ 104

3. Methods ............................................................................................. 106

4. Results ............................................................................................... 112

5. Discussion ......................................................................................... 1 16

vii

6. Tables ................................................................................................ 122

CHAPTER 5: A POPULATION-BASED CASE-CONTROL STUDY OF FETAL
GROWTH, INFANT GESTATIONAL AGE AT DELIVERY, AND MATERNAL

BREAST CANCER AMONG YOUNGER WOMEN ................ . ......................... 133
1. Abstract ............................................................................................. 133
2. Introduction ........................................................................................ 134
3. Methods ............................................................................................. 136
4. Results ............................................................................................... 144
5. Discussion ......................................................................................... 150
6. Tables ................................................................................................ 159
EPI-LOGUE ...................................................................................................... 172
1. Summary of ﬁndings .......................................................................... 172
2. Limitations and Strengths .................................................................. 178
3. Conclusions ....................................................................................... 179
4. Recommendations ............................................................................. 181
REFERENCES ................................................................................................. 182

viii

 

LIST OF TABLES

Table 2.1. Preeclampsia, pregnancy-induced-hypertension, and maternal

breast cancer risk ............................................................................................... 44

Table 2.2. Multiple births and maternal breast cancer risk ................................. 48

Table 2.3. Gestational diabetes and maternal breast cancer risk ....................... 54
Table 2.4. Pregnancy weight gain and maternal breast cancer risk ................... 56
Table 2.5. Fetal growth and maternal breast cancer risk .................................... 58
Table 2.6. Infant gestational age at delivery and maternal breast cancer ........... 60
Table 2.7. Infant sex and maternal breast cancer risk ........................................ 62

Table 3.1. Summary of validation results for cancer to birth registry Iinkages....84

Table 3.2. Distributions of select characteristics by missing SSN status for
women diagnosed with cancer at S 50 years in the Michigan Statewide
Cancer Registry ................................................................................................ 101

Table 3.3. Distributions of select characteristics by missing SSN status for
women with a ﬁrst birth in the Michigan during 1978-2004 ............................... 102

Table 4.1. Distribution of characteristics for ductal and Iobular breast tumors,
Pregnancy Factors and Breast Cancer Registry-Based Study in
Michigan (1985-2004) ....................................................................................... 122

Table 4.2. Odds ratios for pregnancy-related factors and risk of breast cancer,
Pregnancy Factors and Breast Cancer Registry-Based Study in Michigan
(1985-2004) ...................................................................................................... 125

Table 4.3. Odds ratios for pregnancy-related factors and risk of breast cancer by
maternal race and maternal education at ﬁrst birth, Pregnancy Factors and
Breast Cancer Registry-Based Study in Michigan (1985-

2004) ................................................................................................................. 127

Table 4.4. Odds ratios for pregnancy-related factors and risk of breast cancer
By histologic type, Pregnancy Factors and Breast Cancer Registry-Based
Study in Michigan (1985-2004) ......................................................................... 128

Table 4.5. Odds ratios for pregnancy-related factors and risk of ductal and
Iobular breast cancer by maternal race, Pregnancy Factors and Breast Cancer
Registry-Based Study in Michigan (1985-2004) ................................................ 130

Table 4.6. Odds ratios for pregnancy-related factors and risk of ductal and
Iobular breast cancer by maternal education at first birth, Pregnancy Factors
and Breast Cancer Registry-Based Study in Michigan (1985-2004) ................. 131

Table 4.7. Distribution of age at first birth by maternal birth cohort among
controls, Pregnancy Factors and Breast Cancer Registry-Based Study in
Michigan (1985-2004) ........................................................................................ 132

Table 5.1. Descriptive characteristics for cases and controls, Pregnancy
Factors and Breast Cancer Registry-Based Study in Michigan (1985—2004)....159

Table 5.2. Odds ratios for infant birthweight, gestational age, and fetal growth
in ﬁrst birth and maternal breast cancer at age 5 50 years, Pregnancy Factors
and Breast Cancer Registry-Based Study in Michigan (1985-2004) ................. 162

Table 5.3. Adjusted odds ratios for infant birthweight, gestational age, fetal
growth and maternal breast cancer at age 5 50 years by age at ﬁrst birth and
years since ﬁrst birth, Pregnancy Factors and Breast Cancer Registry-Based
Study in Michigan (1985-2004) .......................................................................... 164

Table 5.4. Odds ratios for infant birthweight, gestational age, and fetal growth in
last birth and maternal breast cancer at age S 50 years, Pregnancy Factors and
Breast Cancer Registry-Based Study in Michigan (1985-2004) ........................ 165

Table 5.5. Adjusted odds ratios for infant birthweight, gestational age, fetal
growth in last birth and maternal breast cancer at age 5 50 years by age at last
birth and years since last birth, Pregnancy Factors and Breast Cancer Registry-
Based Study in Michigan (1985-2004) ............................................................... 167

Table 5.6. Adjusted odds ratios for gestational age and fetal growth in ﬁrst birth
and maternal breast cancer at age 5 50 years by histologic type, Pregnancy
Factors and Breast Cancer Registry-Based Study in Michigan (1985-2004)....168

Table 5.7. Adjusted odds ratios for gestational age and fetal growth in last births
and maternal breast cancer at age 5 50 years by histologic type, Pregnancy
Factors and Breast Cancer Registry-Based Study in Michigan (1985-2004)....169

Table 5.8. Adjusted odds ratios for gestational age and fetal growth in first and
births and maternal breast cancer by attained age, Pregnancy Factors and
Breast Cancer Registry-Based Study in Michigan (1985-2004) ........................ 170

Table 5.9. Adjusted odds ratios for gestational age and fetal growth in last births
and maternal breast cancer by attained age, Pregnancy Factors and Breast
Cancer Registry-Based Study in Michigan (1985-2004) .................................... 171

xi

 

 

 

 

 

 

 

 

 

 

 

 

 

Fm;

Al.»
UI-I

Foun
data.

igun

FGUW

(133:):

HOME
V

Cance:

LIST OF FIGURES

Figure 3.1. Overview of study design, study population, and data sources ......... 65
Figure 3.2. Dataset creation procedures conducted at MDCH ............................ 76
Figure 3.3. Summary of selection of cases .......................................................... 79

Figure 3.4. Summary of validation procedures at MDCH for the maternally linked

birth database ...................................................................................................... 90
Figure 3.5. Exclusions based on validation results for the maternally linked birth

data ...................................................................................................................... 93
Figure 3.6. Exclusions based on data cleaning work at MSU .............................. 94

Figure 3.7. Age at ﬁrst birth range for a 50 year old case by year of
diagnosis .............................................................................................................. 97

Figure 6.1. The maternal hormonal environment during pregnancy and breast
cancer ................................................................................................................ 177

xii

ABBREVIATIONS

Alpha-feto protein (AF P)

Appropriate for gestational age (AGA)
Benign breast disease (BDD)

Body mass index (BMI)

Conﬁdence intervals (Cls)

Grams (g)

Gestational age (GA)

Gestational diabetes mellitus (GDM)
Hazard ratios (HR)

Human chorionic gonadotropin (HCG)
International Classiﬁcation of Diseases for Oncology (lCD-O-3)
Incidence rate ratio (IRR)
lnsulin-like-growth-factor-l (lGF-l)
Large for gestational age (LGA)
Medical subject heading (MeSH)
Michigan Department of Community Health (MDCH)
Not reported (NR)

Odds ratio (OR

Pregnancy-induced hypertension (PIH)
Preterm delivery (PTD)

Relative risk (RR)

Small for gestational age (SGA)

xiii

Surveillance, Epidemiology, & End Results (SEER)
Sex-hormone binding globulin (SHBG)
Socioeconomic status (SES)

Standard deviation (SD)

Very preterm delivery (VTPD)

xiv

 

it
Ir“.

rE

ht
W

As

CHAPTER 1: INTRODUCTION AND AIMS

1. Introduction

Premenopausal breast cancer (< 50 years of age at diagnosis), is less
common than postmenopausal breast cancer (1, 2), but breast tumors in younger
women (particularly < 40 years of age) are associated with characteristics related
to poorer prognosis including later stage at diagnosis, estrogen receptor
negativity, higher tumor grade, and positive lymph node status(2-6). Five-year
survival rates are lower in women with breast cancer before age 40 (2, 4, 7, 8)
and studies have reported that younger women diagnosed with breast cancer
have lower overall quality of life than older women, as well as higher
psychological morbidity (9, 10). Despite this, the etiology of premenopausal
breast cancer remains largely unknown.

In the United States, the age-speciﬁc incidence of breast cancer differs by
race/ethnicity and socioeconomic status (SES) (1, 2, 11-14). Overall, White
women have higher age-adjusted rates of breast cancer compared to other
racial/ethnic groups (1, 13, 14) and women of higher SES have elevated rates of
breast cancer compared to women of lower SES (11, 15). In recent decades,
however, differences in incidence by SES have decreased among Black and
White women, but increased in other racial/ethnic groups e.g. Hispanics and
Asian-Americans (12). Among women < 40 years of age, Black women have
higher age-speciﬁc incidence rates of breast cancer compared to White women,
while after approximately age 45, incidence rates are higher among White

women (2, 16). Though breast cancer disproportionately affects young Black

women and is more aggressive (2), the literature on breast cancer etiology
among young Black women is sparse (17-32). Further, despite the fact that
differences in breast cancer incidence by race are reflected in differences by
SES (2, 12, 14, 19, 33, 34) , the contribution of SES to the increased incidence of
breast tumors in young Black as compared to young White women has not been
well-studied (1, 2, 12, 14, 19, 33).

Though the etiology of premenopausal breast cancer remains largely
unknown, pregnancy is known to be a critically important time in relation to a
women’s subsequent risk of breast cancer (35, 36). Women diagnosed with
breast cancer shortly following pregnancy are more likely to have advanced
breast tumors (37), poorer tumor prognostic factors (38), and increased mortality
risk compared to nulliparous premenopausal women (38-41). Earlier age at ﬁrst
birth and increased parity are established protective factors for long-term breast
cancer risk (35, 42, 43). Though not as well-established, evidence suggests that
pregnancy at any age is followed by a transient increase in breast cancer risk,
with possibly a stronger effect in women with a later age at ﬁrst birth (> 30 years)
(44-50), and that later age at any birth (besides ﬁrst) may also increase breast
cancer risk (50-60).

The biological mechanisms underlying the role of pregnancy in
premenopausal breast cancer are not clear (61, 62). The maternal hormonal
environment during pregnancy (63, 64) or the mammary cell microenvironment
following pregnancy have been hypothesized to promote tumor growth (61).

Perinatal factors such as twinning and fetal growth, which are associated with

 

du
me

79.

an altered maternal hormonal environment during pregnancy (63, 65, 66), could
provide insight into the biological mechanisms for the relation between
pregnancy and premenopausal breast cancer in mothers following pregnancy.

Research ﬁndings of the associations between perinatal factors and
maternal premenopausal breast cancer risk (maternal breast cancer refers to
breast cancer in mothers following one or more completed pregnancies) remain
inconclusive. Perinatal factors that have inconsistently been shown to influence
maternal breast cancer risk include high fetal growth (birthweight alone or
birthweight adjusted for gestational age as proxies for fetal growth) (63, 65),
delivery of an infant at earlier gestational ages, including preterm (< 37 weeks
gestation) and very preterm (< 32 weeks gestation) delivery (65-70), multiple
births (twining and higher order deliveries) (65-67, 71-81),
preeclampsia/pregnancy-induced hypertension (PIH) (65-67, 82-85), pregnancy
weight gain (86, 87), offspring gender (88, 89), and diabetes during pregnancy
(63, 66, 90, 91). Studies have found that the influence of perinatal factors may
be modiﬁed by time since index pregnancy (e.g., < 5 years compared to 2 5
years», age at index pregnancy (< 30 years; 2 30 years), infant gender for index
pregnancy, and/or birth order (e.g., exposure in ﬁrst birth or most recent birth)
(65, 73, 74, 77, 79). Endogenous hormones and growth factors that are elevated
during pregnancy and have been implicated in breast cancer etiology, may
mediate the associations between perinatal factors and breast cancer risk (65,
79, 92-94) and include estrogens (95-97), progesterone (98, 99), androgens

(100-102), human chorionic gonadotropin (HCG) (103, 104), insulin-like-growth-

factor-l (lGF-l) (105—108), prolactin (109), and alpha-feto protein (AFP) (92, 103,
110). For example, high fetal growth is associated with higher levels of maternal
serum lGF-I (105-107, 111) and lower levels of prolactin (109), which may
explain a positive association between fetal growth and maternal premenopausal
breast cancer risk.

The purpose of this study is two-fold. First, we will conduct a systematic
literature search to identify and synthesize all published studies of perinatal
factors and maternal breast cancer. Second, we will investigate the associations
between megnancy-related factors (later age at ﬁrst and last delivery,
number of live births, and two perinatal factors that have been very under-
investigated —fetal growth and gestational age at delivery (GA)), and
maternal breast cancer risk among parous Black and White women 50 years of
age or less (i.e., predominantly premenopausal women) in Michigan (MI), 1985-
2004. We will examine associations for risk of maternal breast cancer overall and
by two histologic tumor types (ductal and Iobular). We will conduct a registry-
linked population-based case-control study utilizing Ml Resident Birth files (1978-
2004), Ml Statewide Cancer Registry data (1985-2004), and Detroit Metropolitan
Surveillance Epidemiology End Results (Detroit SEER) Registry data (1978-
2004). This investigation can provide insight into possible biological mechanisms
for the role of pregnancy in premenopausal maternal breast cancer risk. The
study results may lead to the identiﬁcation of women who are at higher risk for

premenopausal breast cancer following childbirth and can be targeted for early

prevention efforts, possibly contributing to a reduction of the breast cancer
burden in premenopausal parous women.
2. Aims and Hypotheses
Aim 1. To systematically identify and descriptively summarize all published
studies on perinatal factors and maternal breast cancer risk, including maternal
conditions of pregnancy (preeclampsia, PlH, GDM, pregnancy weight gain) and
infant birth characteristics (fetal growth, GA, multiple births, sex). These perinatal
factors have been shown to reflect an altered hormonal environment during
pregnancy, and may provide insight into the biological mechanisms underlying
the role of pregnancy in breast cancer etiology. Given that that inﬂuence of many
breast cancer risk factors depend on menopausal status (or by proxy age at
diagnosis), we present ﬁndings stratiﬁed by menopausal status/attained age
whenever available.
Aim 2. To investigate the associations between later age at ﬁrst and last deliverv.
number of live births and breast cancer overall and by the two most common
histologic types (ductal and Iobular) among parous White and Black women 20-
50 years of age in MI, 1985-2004. We will examine potential modiﬁcation of the
above associations by race (White, Black), and maternal education at ﬁrst
delivery (a measure of SES).

We hypothesize that women with a later age at ﬁrst or last delivery (> 30
years) will have increased breast cancer risk compared to women with a younger
age at ﬁrst or last delivery ( s 30 years). We further hypothesize that later age at

ﬁrst and last delivery and multiparity will be more strongly associated with

increased risk for Iobular breast tumors as compared to ductal breast tumors,
given the development of lobules during pregnancy and that Iobular tumors have
been hypothesized to be more strongly associated with hormonal-related factors
in some studies (112).
Aim 3. To investigate the associations between fetal @wth, gestational age at
delivem (2 41 weeks (posterm), 37-41 weeks (term), 36-32 (preterm), < 32 weeks
gestation (very preterm», and breast cancer overall and by histologic type among
parous White and Black women 20- 50 years of age in MI, 1985-2004. We will
estimate fetal growth using published birthweight percentiles for gestational
weeks 24-44, described in the methods section in Chapter 5. We will examine
potential modiﬁcation of the above associations by race (White, Black), maternal
education at ﬁrst delivery, later age at delivery (ﬁrst and last), and shorter time
since delivery last delivery. We will examine the perinatal exposures in first births
and last births (among women with two or more births).
We hypothesize that women who deliver infants with higher fetal growth or have
a pretenn/very preterm delivery will have increased breast cancer risk compared
to women who deliver infants with lower fetal growth or do not have a
preterm/very preterm delivery.

3.1. We will examine the associations between fetal growth and

gestational age at delivery and breast cancer histologic type (ductal and

Iobular). The sample will be limited to cases with these histologic types

(n=5841) and their associated controls (n=21,325).

 

n)

We hypothesize that mothers who delivery infants with high fetal growth or
at earlier gestational ages may have stronger elevated risks associated
with Iobular tumors compared to ductal tumors. The overall rationale for
this hypothesis is given by the ﬁnding that Iobular tumors may be more
strongly associated with hormonal factors related to pregnancy. However,
this will be the ﬁrst study of these two exposures and histologic type of
breast tumors. Further, ﬁndings may vary for ﬁrst births compared to last
births, given development of the mammary gland during ﬁrst pregnancy.
3. Organization of Dissertation
This dissertation includes 5 chapters and an epilogue. Chapter 1 is the
introduction, aims, and hypotheses. Chapter 2 is a systematic review of all
epidemiologic studies of perinatal factors and maternal breast cancer
(manuscript 1 of the dissertation: ‘Perinatal factors and maternal breast cancer
risk: a review of the epidemiologic literature’). Chapter 3 describes the study
design, data preparation, and related methodologic issues. Chapters 4 and 5
are in the analytic manuscripts (abstract, introduction, methods, results, and
discussion) that correspond to aim 2 (manuscript 2: ‘Pregnancy-related factors
and risk of breast cancer by histologic type, a registry-based study of parous
black and white younger women’) and aim 3 (manuscript 3: ‘A population-based
case-control study of fetal growth, infant gestational age at delivery, and maternal
breast cancer among younger women’). The epilogue includes a brief summary
of ﬁndings, main study strengths and limitations, conclusions, and

recommendations for future research.

CHAPTER 2: Perinatal Factors and maternal breast
cancer risk: a review of the epidemiologic literature

1. Abstract

Objective: Many reviews have been published on perinatal factors and offspring
breast cancer risk; however, a review synthesizing the published literature on
perinatal factors and breast cancer in mothers using systematic search methods
is lacking. Methods: We conducted a systematic search to identify all published
studies of perinatal factors and maternal breast cancer. We used PUBMED (to
December 31, 2008) and identiﬁed 39 relevant articles. Results are summarized
for each factor followed by a discussion of the ﬁndings. Results: Though
inconsistent across 16 studies, evidence suggests multiple births may protect
against breast cancer. Preeclampsia was found to decrease risk by up to 20% in
all but two studies; results may be modified by infant sex. Breast cancer risk may
be increased by delivery at earlier gestational ages or elevated fetal growth in a
ﬁrst or last birth, but data are sparse. Infant sex does not appear to be associated
with breast cancer. Data on associations between gestational diabetes,
pregnancy weight gain and breast cancer risk are limited and conﬂicting.
Conclusions. Future research is needed to elucidate the associations between
perinatal factors and maternal breast cancer, including studies of potential

mechanisms for the role of perinatal factors in breast cancer etiology.

2. Introduction

Many review papers have been published on perinatal factors (reﬂecting
the in—utero environment) and later female offspring breast cancer risk (113-116).
Few reviews, however, have summarized epidemiologic ﬁndings for one or more
perinatal factor and subsequent breast cancer in mothers (i.e., maternal breast
cancer) (115, 117-119), and no review exists which synthesizes all the published
literature on perinatal factors and maternal breast cancer using systematic
search methods.

Epidemiologic studies of the association between perinatal factors and
maternal breast cancer remain inconclusive. Perinatal factors that have
inconsistently been shown to inﬂuence maternal breast cancer risk include fetal
growth (birthweight alone or birthweight adjusted for gestational age) (63, 65),
infant gestational age (GA) at delivery (65-67), multiple births (i.e., twining and
higher order deliveries) (71, 80, 81), preeclampsia and/or pregnancy-induced
hypertension (PIH) (66, 67, 82), placental characteristics (63, 120), pregnancy
weight gain (66, 87), gestational diabetes mellitus (GDM) (66, 90, 91), and infant
sex (65, 66). Perinatal factors are associated with altered levels of maternal
hormonal factors during pregnancy, including estrogens (95—97), progesterone
(98, 99), androgens (100—102), human chorionic gonadotropin (HCG) (103, 104),
lGF-I, lGF-l binding proteins (105-108), prolactin (109), and AFP (92, 103, 110).
These hormonal factors play a role in breast cancer etiology and have been
proposed to mediate the associations between perinatal factors and maternal

breast cancer risk (65, 79, 92, 93).

Pregnancy is known to be a critically important time in relation to a
women’s subsequent risk of breast cancer (35, 36). Earlier age at ﬁrst birth and
increased parity are established protective factors for long-term breast cancer
risk (35, 42, 43). Less-established is the ﬁnding that pregnancy at any age is
followed by a transient increase in breast cancer risk, with evidence for a
stronger effect in women with a later age at ﬁrst birth (i.e., > 30 years) (44-49).
The biological mechanisms underlying the role of pregnancy in breast cancer
etiology are not clear (61, 62), but several hypotheses, all of which involve
alterations in the maternal hormonal environment during pregnancy, have been
proposed to explain both the long-term decrease and short-term increase in
breast cancer risk. For example, the long-term reduction in risk associated with
early age at ﬁrst full-term pregnancy and parity may be due to horrnonally-
induced terminal differentiation of the mammary gland, resulting in permanent
changes in the breast tissue and protection against future carcinogens (36, 121).
Alternatively, molecular changes in mammary epithelial stem cells during
pregnancy, in response to the maternal hormonal environment, may inﬂuence
later cellular proliferation and DNA repair in the mammary gland (62). The short-
terrn increase in risk may be due to the promotion of tumor growth in response to
elevated pregnancy hormones (63, 64). It is difficult to directly study the maternal
hormonal environment during pregnancy and subsequent maternal breast cancer
(122); however, perinatal factors (e.g., multiple births, high fetal growth), which

may be proxies of this environment (63, 65, 66), can be used to provide insight

10

 

 

 

into the hormonal mechanisms underlying the inﬂuence of pregnancy on
mammary carcinogenesis.

The purpose of this review is to summarize the epidemiologic evidence for
the associations between perinatal factors and maternal breast cancer and to
identify areas for future research. We focused on perinatal factors that have been
shown to reﬂect an altered hormonal environment during pregnancy, and hence
may provide insight into the biological mechanisms underlying the role of
pregnancy in breast cancer, including maternal conditions of pregnancy
(preeclampsia, PlH, GDM, pregnancy weight gain) and infant birth characteristics
(fetal growth, GA at delivery, multiple births, sex). For each factor, we ﬁrst
describe and summarize the epidemiologic studies. This is followed by a
commentary on the possible explanations for the ﬁndings, with a brief discussion
on the postulated mechanisms that may help explain the associations between

perinatal factors and maternal breast cancer.

3. Methods

3.1 Search strategy and study selection

The electronic database PUBMED was searched systematically by one
reviewer (SN) for all articles published in peer-reviewed journals up to December
31, 2008. Searches included the medical subject heading (MeSH) “breast
neoplasms” and the keyword “breast cancer” and terms for the exposures of
interest (see appendix for terms and additional search strategy details). We
included articles that were peer-reviewed epidemiologic studies using population-

based case-control or cohort study designs that reported measures of

11

association (e.g., odds ratios (ORs), relative risks (RRs)) for one or more of the
exposures of interest and maternal breast cancer. We excluded hospital-based
case-control and cross-sectional studies (listed in appendix). The search strategy
identiﬁed 1,501 possible articles. If the article title appeared relevant or if
relevance was not clear from the title, the abstract was reviewed to conﬁrm the
study examined the outcome and exposure(s) of interest (n=211 abstracts). We
excluded 1,435 articles based on title/abstract review. Full text of articles were
obtained for review for both relevant studies and for studies where relevance was
not clear from abstract review alone (n=66). Thirty four of these articles met the
inclusion criteria and were included as original research studies, two articles
were reviews of one or more relevant exposures of interest, and 30 articles were
excluded for not meeting the inclusion criteria. The reference lists for each
included study as well as review papers were hand searched for additional
articles. Citations of all relevant studies were also searched in the citation index
Web of Science- part of ISI Web of Knowledge (123). An additional ﬁve relevant
reports not found in the PUBMED search were found through these methods. We
did not attempt to identify unpublished articles or abstracts from scientiﬁc
conferences.

We identiﬁed 39 relevant studies using the above search strategy. We
descriptively summarized the studies for each perinatal exposure and breast
cancer in tables, with the exception of placental characteristics, for which there
were only two studies (Tables 1-7). Many breast cancer risk factors (including

reproductive factors) differ in their effects by menopausal status (or attained age

12

 

 

(H

OH
fro
“It:

dfa

which is a proxy for menopausal status) (124-126) and hence we include results
stratiﬁed by menopausal status/age, whenever available. Data from each
identiﬁed study were abstracted directly from the published manuscript of each
individual study and tabulated by one author (SN). We did not conduct an overall
quality assessment for the studies; all identiﬁed reports that met inclusion criteria
are included in this review. Tables describe all studies that reported covariate-
adjusted measures of association. Studies that did not provide covariate-adjusted
measures of association with confidence intervals (Cls) were not tabulated, but
are referenced in the text. Four redundant articles were not included as separate
studies in the summary tables, but are discussed as appropriate in the text. One
of these articles had been recently updated in a new published report (82) and
only the new data is included; three articles were additional analyses on data
from previously published reports (89, 127, 128). Studies nested within a cohort
with follow-up were considered cohort studies. Age in the tables indicates age at
diagnosis or at follow-up (attained age) unless otherwise noted.

3.2 Methodologic Issues

Three main issues to consider when interpreting the studies include: 1) the

data source for exposure measurement (maternal self-report, birth registry,
medical records), 2) potential confounding factors, and 3) potential effect
modiﬁers. In addition, given that case-control studies are subject to systematic
errors that could inﬂuence measures of associations (e.g., recall bias and
participation bias), we include a separate discussion by study design, if possible.

Most case-control studies were registry-based, however, and are not subject to

13

 

 

is

be i
add
Imp:
factc
0” bl
as a

and t

some well-known systematic errors that plague case-control studies, because
exposure histories were collected from records prior to cancer diagnosis and
contact with participants is not required via the use of linked records. However,
registry-based studies may have limited information on potential confounding
factors and hence be more subject to residual confounding.

Several characteristics of the index pregnancy (e.g., age at index birth,
birth order) should be considered as potential effect modiﬁers of the associations
between perinatal factors and breast cancer. Previous studies have examined
perinatal exposures in ﬁrst births, any birth, and last births (among women with 2
or more deliveries). It is important to examine exposures in ﬁrst births given the
established critical role of a ﬁrst live birth on both the short- and long term risk of
breast cancer. Exposure in the last birth, among multiparous women, may also
be important because recency of exposure may reflect the growth-promoting
inﬂuence of an altered maternal hormonal environment on the development of
breast cancer. Alternatively, exposure in a birth prior to the most recent birth may
be inﬂuenced by subsequent births and the characteristics of those births. In
addition to birth order, age at the index birth and time since the index birth are
important potential modiﬁers. Later ages at ﬁrst and last birth are well-known risk
factors for breast cancer which may modify the time-related effects of pregnancy
on breast cancer risk (50). Finally, infant sex at index birth has been suggested
as a potential effect modiﬁer of the associations between other perinatal factors

and breast cancer (e.g., preeclampsia).

14

4. Preeclampsia, Pregnancy-Induced Hypertension, and
Placental Characteristics

4.1 Introduction

Preeclampsia is complication of pregnancy involving the placenta and
multiple organ systems which occurs in about 4-6% pregnancies in the US and is
a cause of both maternal and neonatal morbidity and mortality (129-133).
Preeclampsia is deﬁned by the presence of new onset hypertension, i.e., PIH
(systolic blood pressure 2 140 mm Hg or diastolic blood pressure 2 90 mm Hg)
and proteinuria (2 300 mg per 24 hours) after 20 weeks of gestation (134). The
etiology of preeclampsia is complex, and may originate with poor placentation
that may involve both genetic and immunological mechanisms (129, 135).
Abnormal placentation may lead to widespread endothelial dysfunction, which in
turn results in the clinical manifestation seen in preeclamptic pregnancies (129,
135). Maternal risk factors for preeclampsia include nulliparity, family history of
preeclampsia, pre-existing diabetes, insulin resistance, chronic hypertension,
multiple gestations, older maternal age, and high pre-pregnancy body mass
index (BMI) (134, 136).
4.2 Epidemiologic Studies

We found ﬁve case-control studies (65-67, 85, 137) and seven cohort
studies (63, 81, 83, 84, 120, 138, 139) of preeclampsia and/or PIH and breast
cancer risk (Table 2.1). Three of the twelve studies did not report measures of
association (63, 81, 120). Among the nine other studies, exposure measurement
was from birth registries for four studies, medical records for two studies, and

maternal self-report for three studies.

15

Cohort studies. Two large cohort studies using registry-based data in Norway
with overlapping populations (5,474 cases (2002) and 9,160 cases (2007))
reported that preeclampsia and/or PIH in a ﬁrst birth is associated with a
signiﬁcantly reduced risk of breast cancer (84, 138). Odds ratios were not
affected by age at breast cancer diagnosis (70). In contrast, a large cohort study
(the Jerusalem Perinatal Cohort Study) used medical records and reported a
signiﬁcantly increased risk of breast cancer in two reports, one in 2004 (82) and
also in an updated reported with further follow-up in 2008 (139). They also
reported an increased risk of cancer at any site (hazard ratio (HR) = 1.23 95% CI:
1.05-1.45), which is inconsistent with a recent cohort study in Utah of all cancer
sites and preeclampsia (140).
Case-control studies. Five case-control studies have reported evidence for an
inverse association between preeclampsia and/or PIH and breast cancer, though
only two studies reported signiﬁcant ﬁndings (85, 137). One US interview-based
case-control study with 1,310 cases examined the history of preeclampsia, PIH,
or both conditions in any pregnancy and breast cancer risk, and was able to
adjust for several known breast cancer risk factors (e.g., BMI, parity, age at
menarche, lactation, family history). This study found that preeclampsia and PIH
were inversely associated with breast cancer; results stratiﬁed by menopausal
status revealed a stronger association among postmenopausal women (137).
Two studies (both case-control) considered age at index birth and/or time
since index birth as effect modifiers of the association between preeclampsia and

breast cancer. One registry-based case-control study of ﬁrst births reported a

16

 

 

(I.

re
ar

ex

am

stronger inverse association between preeclampsia and breast cancer for women
> 30 years at ﬁrst birth and also in the ﬁrst three years after the first birth (65). An
interview-based case-control study found no evidence for effect modiﬁcation by
time since last birth (66).

In addition to age and time since birth, some studies (both case-control
and cohort) have examined effect modiﬁcation by length of gestation (84, 137),
fetal growth (84), and offspring sex (65, 128, 137-139). In the largest cohort
study, from Norway, Vatten and colleagues found the protective effect of
preeclampsia and/or PIH was limited to women who delivered a male infant and
in particular for preterm delivery of male infant (138). Troisi and colleagues
conducted additional analyses using data from the registry-based case-control
study of ﬁrst births initially reported on by lnnes and Byers (65), and found a
stronger protective effect of preeclampsia for women who delivered a male than
who delivered a female, but only among women > 30 years of age at first birth
(1 28).

Two studies have examined placental characteristics and subsequent
breast cancer risk (63, 120). Lower placental weight and other placental
characteristics that may represent reduced placental functionality, could also
reﬂect altered exposure to hormones and growth factors during pregnancy (141),
and may have implications for breast cancer risk (63). Cohn and colleagues
examined placental characteristics (e.g., placental weight, placental diameter) in
a small (146 cases) cohort study that followed-up participants of the Child Health

and Development Studies (120) and found some evidence for reduced breast

17

 

IEH

We

on:

cancer risk associated with lower placenta weight or reduced placental diameter.
Results from the Swedish registry-based cohort study conducted by Cnattingius
and colleagues that also looked at several other perinatal factors including fetal
growth, reported an increased risk of breast cancer per 100 gram increase in
placenta weight, (HR = 1.07 95% Cl : 1.02-1.13), adjusted for several other
perinatal factors, including offspring birthweight (63). Further, women with a ﬁrst
birth and second birth placenta weighing > 700 grams, were at twice the adjusted
risk of breast cancer compared to women for which both placentas weighed <
500 grams (63).
4.3 Summag

Most studies, based on both self-report and birth registry data have
reported that preeclampsia and/or PIH is associated with a decrease in breast
cancer risk (65-67, 84, 85, 137, 138). However, one large well-designed cohort
study (Jerusalem Perinatal Cohort study) found an increased breast cancer risk
among women with a history of preeclampsia. Authors of this cohort study
speculated reasons for discrepancies between studies including residual
confounding, varying exposure deﬁnitions (edema was a diagnostic requirement
for preeclampsia in their study), or genetic and environmental differences in the
populations studied (139). Overall, there is some evidence that infant sex and
length of gestation may be effect modiﬁers of the association between
preeclampsia and breast cancer, but results are not consistent across studies.
Only two studies of placental characteristics, have been conducted, one of which

was very small. However, this area of research is of interest because placental

18

characteristics are related to several other perinatal factors (including
hypertensive disorders of pregnancy), as well as steroid hormone and HCG
production during pregnancy, and hence may provide additional clues on the role
of pregnancy hormones in breast cancer.
4.4 Commerm

The protective effect of preeclampsia and/or PIH on breast cancer risk
may be mediated through lower levels of estrogen during pregnancy (65, 100,
117, 137), though studies of estriol and/or estradiol maternal blood levels during
preeclamptic pregnancies have been inconsistent (98, 100, 101, 142-146). Other
hormonal factors that have been implicated in breast cancer mechanisms and
found to be altered in preeclamptic pregnancies include higher levels of
progesterone (98, 117, 143), androgens (100, 101, 117, 142, 147), and HCG
(117, 148). However, a recent US study found limited evidence for associations
between androgens, estriol, and estradiol with blood pressure during the second
and third trimester, as well as change in blood pressure between trimesters
(149). Lower levels of maternal lGF-l during pregnancy may also play a role
(117, 137), but studies of maternal serum lGF-I and preeclampsia/PIH are
inconsistent and associations may depend on severity and/or length of gestation
for preeclamptic pregnancies (108, 150-155). Alpha-fetoprotein (AFP), which
may have “anti-estrogenic” effects in the breast tissue (117, 131), has been
shown to be elevated in women with preeclampsia/PIH (156-158). This protein
has been postulated to mediate the protective effect of preeclampsia/PIH on

breast cancer risk (65, 85 , 117). However, not all studies have reported

19

increased AF P in pregnancies with hypertension (148) and a small cohort study
did not ﬁnd that AFP explained a decrease in breast cancer risk associated with
elevated mean arterial pressure during pregnancy (83).

Alternative mechanisms to explain the association between
preeclampsia/PIH and breast cancer have been postulated (117, 140, 159).
Angiogenesis, (i.e., new blood vessel growth), is involved in tumor growth and
metastasis (160). Preeclamptic pregnancies have been shown to be
characterized by high levels of anti-angiogenic factors (e.g., soluble fms-like
tyrosine kinase-1) and low levels of pro-angiogenic factors (e.g., vascular
endothelial growth factor, placental growth factor) (161 ), which could lead to a
reduction in cancer risk (140). Gago-Dominquez and colleagues postulated that
lipid peroxidation may protect against breast cancer, possibly through inhibition
of cell proliferation as well as increased cell differentiation and apoptosis. They
hypothesized this may be a mechanism by which preeclampsia protects against
subsequent breast cancer, given studies have found increased lipid peroxidation
in preeclamptic pregnancies.

5. Multiple Births

5.1 Introduction

Multiple births may reﬂect an altered maternal hormonal environment,
including elevated levels of estrogens, progesterone, HCG, and AFP, which may
inﬂuence subsequent breast cancer risk (103, 162, 163). We use the term
“multiple births” to indicate both twins only or twins and higher order births, with

exact exposures deﬁnitions shown in Table 2. About 5% of multiple births in the

20

US in 2006 were triplets or higher order multiples (164). Multiple births may be
spontaneous or due to the use of assisted reproduction. A study of 13,206
pregnant women in the US during 1986-1991 reported that about 35% of twins
and 77% of higher order births were attributable to the use of assisted
reproduction (165). Women with multiple births are at increased risk for PIH and
GDM, and their infants are more likely to be born preterm and have lower
birthweights (164, 166-169). Factors associated with spontaneous multiples
include race/ethnicity, geographic location, maternal body size, maternal age,
maternal nutrition, parity, being born a twin, and family history of twinning (164,
166, 170-172).
ﬁgidemiologic Sturdies

We identiﬁed nine cohort (71, 76-81, 173, 174) and seven case-control
(65-67, 72, 73, 75, 175) studies of the association between multiple births and
breast cancer (Table 2.2). Several studies were of large sample size, and hence
could examine exposure in any birth, a last birth, or prior to the last birth as well
as effect modiﬁcation by age at index birth, time since index birth or infant sex.
The source of exposure data (e.g., birth records, maternal interview) is available
in the Table 2, but not discussed below given that the exposure is unlikely to be
misclassiﬁed regardless of data source.
Cohort studies. Three large cohort studies have reported a decreased risk of
breast cancer associated with a delivery of multiples for any birth, a last birth, or
a birth prior to the last birth (77, 78) The largest of these studies was conducted

in Sweden with 19, 368 cases. This study examined results by age (< 55 years, a

21

55 years) and found the protective effect was limited to women < 55 years and
was only signiﬁcant for exposure in any birth, OR = 0.85, 95% CI: 0.74-0.98. The
two other cohort studies were among women s 56 years of age. A fourth cohort
study of women < 50 years of age with 6,309 cases reported a small non-
signiﬁcant decrease in risk for exposure in any birth (76, 173). In contrast, one
large Danish cohort (9,495 cases) of latest births among women <58 years of
age, reported a non-signiﬁcant increase in breast cancer risk associated with a
last multiple birth.
Case-control studies. One US case-control study of women aged 20-54 years
with 3,918 cases reported a significant decrease in breast cancer risk associated
with delivery of multiples in a last birth only (72). Other case-control studies of
both younger and older women have found limited evidence for an association
between multiple births and breast cancer (66, 73, 75). In contrast, a US case-
control study in New York State with data on ﬁrst births only reported a non-
signiﬁcant elevation in risk of breast cancer for having a multiple birth (65).
Several studies have examined potential effect modiﬁers of the
association between multiple births and breast cancer. The US case-control
study of ﬁrst births in New York State reported a signiﬁcant increase in breast
cancer risk associated with a multiple birth among women with a later age at ﬁrst
birth ( > 30 years) and also for shorter time since first birth (5 5 years; > 6 years)
(65). Findings for the large Danish cohort study of latest births reported an
elevated risk for mothers diagnosed with breast cancer <5 years after the birth

(RR = 1.8; 95% CI: 1.1-2.8) (79). However, other studies that have investigated

22

 

\1

dl.
rei

W C

ler
$8!
res
COI‘.

hor

time since a multiple birth (66, 72, 73, 75-77), or age at index birth (66, 72, 73,
75-77) have not found evidence for effect modiﬁcation.

In addition to age at birth, time since birth, and birth order, other important
confounding and/or modifying factors that have been examined include use of
infertility treatments, length of gestation, and offspring sex. Studies could not
distinguish between spontaneous multiple births or multiples due to assisted-
reproduction, but three studies reported similar results when they either excluded
women who reported infertility treatments or a history of infertility problems, or
adjusted for these factors (66, 72, 73). Only one study considered gestational
length; results were similar after adjustment (72). Studies that have examined
same sex compared to different sex (a marker of zygosity) have found similar
results (72, 73, 78). Some studies have examined multiple births with all females
compared to all males, based on the hypothesis that infant sex is associated with
hormonal differences during pregnancy (66, 76). Two studies reported that the
non-signiﬁcant protective effect of a multiple birth was limited to women who
delivered all females compared to all males (66, 76).

5.3 Summagy

Based on three large European cohort studies, multiple births appear to
decrease risk of breast cancer among younger women, with possible
modiﬁcation by birth order (last, any), but results are inconsistent. The signiﬁcant
increase in risk for the ﬁrst ﬁve years following a multiple birth, reported in a US
case-control study of first births and in a large prospective cohort study in

Denmark of last births, however, suggests the protective effect is not universal

23

 

and
our

eslr

pres

in w

risk

COnr

Exp

388:.

and/or may depend on characteristics of the birth. The determinants of natural
multiples, spontaneous multiples, and zygosity may differ (discussed below);
hence, future studies that differentiate between the type of multiple births may
help clarify the association between multiple births and breast cancer.
5.4 Commenta_r_y

Postulated mechanisms to explain the association between multiple births
and breast cancer involve the role of altered levels of maternal hormonal factors
found in multiple pregnancies (65, 66, 72, 76-78, 80, 174). Elevated levels of
estrogens (104, 162, 176), progesterone (163), HCG (103, 104, 177) , and AFP
(103, 104, 110), have been found in multiple pregnancies compared to singleton
pregnancies. Investigators have hypothesized that the elevated estrogens found
in women with multiple pregnancies may help explain the increased short-term
risk found in some studies of parous women with multiple births (65, 74, 79).
Alternatively, elevated maternal HCG, progesterone, or AFP may mediate the
protective effect of multiple births on breast cancer risk in parous women (65, 66,
72, 76, 77, 80, 174). Other researchers have proposed that protective effect may
also be due to differences in hormone levels during the menstrual cycles and
post-pregnancy in women who deliver multiples (78). For example, SHBG has
been found to be higher in premenopausal women with a history of twins as
compared to women with only singleton births (162, 178).

The association between multiple births and breast cancer may also be
explained by characteristics of women who have multiple births that are also

associated with breast cancer (77, 78). As noted above, factors associated with

24

spontaneous multiple births include race/ethnicity, maternal body size, maternal
age, maternal nutrition, parity, being born a twin, and family history of twinning.
Further, women with a multiple pregnancy are at increased risk for PIH, GDM,
preterm delivery, and lower fetal growth. Though controversial, some studies
have shown that infertility drugs may be associated with breast cancer risk (179),
and it is not known if multiple births due to assisted-reproduction (compared to
spontaneous multiple births) have different inﬂuences on breast cancer risk.
Breastfeeding is a protective factor for breast cancer (180), and a few studies
were able to adjust for this factor (72, 73). The role of breastfeeding, however,
may be more complex given evidence that breastfeeding rates are lower in
women with multiple births who deliver preterm compared to women who deliver
full-tenn multiples or singletons (181).
6. Gestational Diabetes
6.1 Introduction

Gestational diabetes is deﬁned as any degree of glucose intolerance with
onset or ﬁrst recognition during pregnancy and in the US is usually screened for
at 24-28 weeks of pregnancy (182). Prevalence estimates for GDM vary (2-8%)
depending on the population, and consistent with the high correlation between
obesity and GDM as well as the increase in obesity in the US, rates of GDM
appear to be also increasing (183, 184). Risk factors for GDM include maternal
obesity, family history of diabetes, personal history of glucose intolerance, prior
delivery of a high birthweight infant, and non-White race/ethnicity (185, 186)

6.2 Epidemiologic Studies

25

Few epidemiologic studies have examined the association between GDM
and breast cancer (63, 66, 90, 91, 187) (Table 2.3). We found one registry-based
cohort study, one cohort study that used medical records for exposures, and two
interview-based case-control studies. One very small record-based study of
maternal fasting plasma glucose and a hospital admission for breast cancer is
included in Table 3, but not discussed below.

Cohort studies. Two record-based cohort studies (63, 90) have examined the
association between GDM and breast cancer. One of these studies did not ﬁnd
an association, though had only 10 exposed cases and included all types of
diabetes during pregnancy (gestational or pregestational (type I, type II» (63). In
contrast, in the Jerusalem perinatal cohort study, which used medical records for
exposure data (90), reported a signiﬁcant increase in risk of breast cancer for
history of GDM among women 2 50 years, but not among women < 50 years
was.

Case-control studies. Two US case-control studies of self-reported GDM (66,
91) reported conﬂicting ﬁndings. One study with 1,235 cases did not ﬁnd an
association between GDM and breast cancer risk (66), though when the study
examined modification by years since last birth (< 5 years, and 2 5 years) it found
a non-signiﬁcant protective effect of GDM in the ﬁrst ﬁve years and a non-
signiﬁcant elevated long-term risk (66). The other study, with 2,319 cases
conducted among non-Hispanic White and Hispanic women, reported an inverse
association between self-reported GDM and breast cancer among both pre- and

postmenopausal women; results were only signiﬁcant among postmenopausal

26

women (91 ). The authors further stratiﬁed the ﬁndings by age at onset of GDM
(i.e., age at index delivery), and found an inverse association for women < 35
years and a positive association for women 2 35 years (91).
6.3 Summagy

To date, studies of the association between GDM and breast cancer are
few. Results include a signiﬁcant protective effect in a large US case-control
study conducted in the Southwestern states which appears to be limited to
women < 35 years at index birth, a significant increased risk for women in the
large Jerusalem Perinatal Cohort study, and no association in a US case-control
study and Swedish cohort study. Given these discrepant ﬁndings, further studies
are needed on this association. In particular, studies that also have information
on post-pregnancy diabetes and biomarkers associated with diabetes
development may provide more data to determine if GDM is an independent risk
factor for breast cancer.
6.4 Commenta_ry

The association between type 2 diabetes outside of pregnancy and breast
cancer risk has been investigated in many studies, with evidence for a positive
association that is more consistent and stronger among postmenopausal women
(188). Several biological mechanisms have been proposed to explain an
association between type 2 diabetes and breast cancer risk. For example, type 2
diabetes is characterized by elevated levels of insulin and increased sex
hormones as well as decreased sex hormone binding globulin (SHBG ) (188,

189). These increased hormonal factors could activate cell pathways that lead to

27

increased cellular proliferation and decreased apoptosis (189). Hyperglycemia,
another feature of diabetes, may also be related to increased risk of breast
cancer, and glucose may promote cell growth (188). Several of these
mechanisms may provide reasoning for an increased risk of breast cancer
following GDM. In addition, women with GDM are more likely to develop type II
diabetes later in life (190), which could help explain an association between GDM
and breast cancer. Finally, given that obesity is important in both the
development of GDM (185), type 2 diabetes (189), and pre- and postmenopausal
breast cancer, future studies with information on pre-pregnancy BMI, gestational
weight gain, and pre and post-menopausal BMI could shed light on inconsistent
ﬁndings.
7. Pregnancy Weight Gain
7.1 Introduction

Researchers have hypothesized that weight gain during periods of
hormonal change over the life course (e.g., menarche, pregnancy, lactation,
menopause) may be of particular importance in relation to subsequent breast
cancer risk (191-193). Higher pregnancy weight gain may reﬂect increased
exposure to maternal hormonal factors during pregnancy (e.g., estrogens), which
could increase breast cancer risk(86). However, the study of pregnancy weight
gain is complicated by the interrelationships between pre-pregnancy body mass,

postpartum weight change, and adult adiposity.

7.2 Epidemiologic Studies

28

 

(I)

Few studies of the association of pregnancy weight gain and breast
cancer have been published (66, 86, 87, 120, 194, 195) (Table 2.4). Two of the
six studies did not report measures of association (120, 194). Measurement of
pregnancy weight gain was via self-report in three studies and from hospital
records in one study.

Cohort studies Two cohort studies of Finnish women with limited sample size
(< 150 cases), (86, 87) have examined the association between pregnancy
weight gain and breast cancer. One of these, with about 50% follow-up of the
original cohort, reported a positive association between estimated weight gain
during pregnancy > 15 kg (reference: 11-15 kg) and breast cancer risk. In
stratiﬁed analyses, the association was limited to postmenopausal women,
though authors noted that sample size for premenopausal cases was small (< 25
cases). To account for current BMI, the authors conducted a nested case-control
study among the cohort for women with available hospital records on weight and
height close to the time of diagnosis for cases (about 50% had data available),
and found a non-signiﬁcant positive association between pregnancy weight gain
and breast cancer, adjusted for later BMI (87). The other study found no
association between pregnancy weight gain and breast cancer risk in among
predominantly premenopausal women (86). Neither study reported information
on birth order with exposure (e.g., ﬁrst, last)

Case-control studies. Three US case-control studies, of women aged 35-79
years (194), 20-44 years (66), and < 50 years (195), have reported limited

evidence for an association between breast cancer risk and pregnancy weight

29

gain in the ﬁrst pregnancy (66, 194), or most recent pregnancy (195) or for
maximum weight gain in all pregnancies (66).
7.3 Summagy

At present, there is limited evidence for an association between pregnancy
weight gain and subsequent breast cancer risk, with the exception of perhaps an
association among postmenopausal women, which was reported in 1 study with
missing data on a high proportion of the cohort at follow-up. It is not clear,
however, if this association is due to the known relation between high BMI,
weight gain, and central adiposity and postmenopausal breast cancer risk (193),
given that women who gain more weight during pregnancy retain more weight
postpartum (196) and may also retain more weight into menopausal years (197,
198). Further, if the association is hypothesized to be due to the increase in
adipose tissue during pregnancy, increased body fat may be a better exposure
measure than pregnancy weight gain, which reﬂects several components (fetus,
placenta, amniotic fluid, increased blood volume, and adipose tissue) (199).
7.4 Comment_a_ry

Researchers have hypothesized that pregnancy weight gain may increase
risk due to an altered hormonal environment during pregnancy, in particular, due
to elevated estrogens (87). Investigations of maternal serum estrogens and
pregnancy weight gain, however, have been inconsistent. One early study
reported a positive association for weight gain up to the 31St week of gestation
and maternal estriol and total estrogens (200), though subsequent studies have

reported null results (96, 201-203). Alternatively, the underlying hormonal

30

mechanism may be due to lower levels of SHBG (96, 203) or progesterone (96)
that have been reported in women with high weight gain during pregnancy.
Recently, an animal study has reported that excessive weight gain during
pregnancy increased carcinogen-induced breast tumors and the ﬁndings could
not be accounted for via altered hormone levels of leptin, estradiol, or lGF-l
(204). Another possible mediator may be hyperinsulimia, which may be
associated with increased breast cancer risk (189), and also with higher weight
gain during pregnancy (191), though the relationship is complex and the role of
pre-pregnancy BMI, weight change following childbirth, as well as current body
size should be considered.

8. Fetal Growth

8.1 Introduction

Fetal growth may reﬂect circulating levels of maternal hormones or growth
factors important in breast cancer etiology (93, 95, 96, 105, 106, 109). Infant
birthweight is inﬂuenced by both duration of gestation and rate of fetal growth
(205). Fetal weight increases during pregnancy, with the highest weight gain in
the third trimester, due to increased fetal fat mass, and a peak at about 34-35
weeks (206, 207). Weight at birth (birthweight) alone or birthweight adjusted for
gestational age is used to estimate fetal growth because of the difficulties in
directly measuring the rate of fetal growth in-utero using prenatal ultrasounds
(208, 209). One method of adjustment for gestational age is to simply include
gestational age as a covariate in the model. Another approach is to adjust for

gestational age using reference birthweight percentiles for each gestational week

31

or to use the fetal growth ratio which is calculated by dividing the birthweight by a
reference median birthweight for the given gestational age (210). A thorough
discussion of the etiology of fetal growth is beyond the scope of this paper.
Brieﬂy, contributors to fetal growth that also may play a role in breast cancer
etiology include maternal anthropometry (height, pre-pregnancy body mass index
(BMI), gestational weight gain), maternal birthweight, parity, infant sex, race,
physical activity, medical complications during pregnancy (e.g., preeclampsia,
gestational diabetes), smoking, multiple births, maternal age, and socioeconomic
status (210-21 3).
8.2 Epidemiologic Studies

Six studies have examined fetal growth and breast cancer (63, 65, 79, 81,
120, 175) (Table 2.5). Two of these studies did not report measures of
association, and are not discussed further (81, 120). The remaining four studies
were all registry-based with three studies adjusting for gestational age as a
covariate and one study using birthweight alone as a proxy for fetal growth.
Cohort studies. Two cohort studies have examined the association between
fetal growth and breast cancer. A cohort study in Sweden with about 2,200 cases
reported a signiﬁcant increase in breast cancer risk per 500 gram increase in
birthweight for a ﬁrst birth (63). This study adjusted for gestational age as a
covariate (weeks: <37, 37, 38, 39, 40, 41, 2 42). A second cohort study in
Denmark with 3,874 cases reported a non-significant elevation in breast cancer

risk associated with delivery of an infant weighing >3500 grams compared to s

32

301

ex:
Si'J

ass

ha

wh

Il‘rl

3000 grams (79). This study adjusted for extremely preterm delivery (< 32 weeks,
2 32 weeks, unknown).

Case-control studies. Two population-based case-control studies have
examined fetal growth and breast cancer risk. A US registry based case-control
study of ﬁrst births reported a non-signiﬁcant reduction in breast cancer
associated with very low (< 1500 grams) and very high (2 4500 grams)
birthweight (65). This study adjusted for gestational age as a covariate (weeks: <
32, 32-36, 2 37). A second case-control study of birthweight alone and breast
cancer reported a non-signiﬁcant increased risk for lower and higher birthweight
in any birth (175).

Overall, consideration of effect modification by time or age at index birth
has not been reported. The exception is the cohort study conducted in Denmark,
where authors reported some evidence for a stronger increase in risk for the first
ﬁve years following delivery. No studies reported results by menopausal status--
though study populations were primarily among younger women (i.e., < 58 years
of age).

8.3 Summayy

In summary, evidence from two registry-based cohort studies that
adjusted for gestational age suggest high fetal growth in a ﬁrst or last birth may
be associated with as small (~10%) increased risk of breast cancer. However,
two registry-based case-control studies, one of ﬁrst births and one of any births

reported conflicting findings. The two case-control studies included women with

33

Ie ir

wt

ES

multiple births and one study did not adjust for gestational age, which could be
one factor to account for the differences.
8.4 Commengagy

The association between high fetal growth and increased breast cancer is
biologically plausible, given that high rates of fetal growth may be associated with
elevated maternal estrogens and IGF-l levels during pregnancy (79). Most
studies have found that fetal growth is positively associated with maternal estriol
levels (primarily of fetal origin), mainly in the third trimester (93, 95, 96, 109, 214),
while the evidence has been inconsistent for an association with maternal
estradiol (93, 97, 214), which has been consistently implicated in breast cancer
risk (215). Other hormones that may mediate a positive association between fetal
growth and breast cancer include higher maternal serum levels of lGF-1 or low
levels of IGF-binding protein-1, inconsistently associated with fetal
growth/birthweight (particularly during late gestation) (105-107, 111) or lower
levels of prolactin (109).

The observed association between fetal growth and breast cancer may be
due to non-measured environmental, social, or genetic factors. Some key factors
associated with both fetal growth and possibly breast cancer include maternal
height, pre-pregnancy weight, maternal birthweight, parity, gestational weight
gain, maternal age, smoking, hypertensive disease of pregnancy, and infant sex
(211, 216). Cnattingius and colleagues did adjust for several maternal factors
(maternal height, maternal BMI, pregnancy-induced hypertensive diseases),

which were attenuated compared to the age-adjusted associations. In addition,

34

I.“D

(T)

Vi

they also adjusted for placental weight (fetal growth depends on placental size
(217)), which further attenuated results. Finally, no study to date has estimated
fetal growth using approaches that are more appropriate than simple adjustment
for gestational age as a covariate (e.g., reference birthweight percentiles for each
gestational week).
9. Infant Gestational Age at Delivery
9.1 Introduction

Induced and spontaneous abortion, which reﬂect pregnancy interruption in
early gestation (primarily first trimester), and breast cancer have been well-
studied (218, 219). Few studies, however, have examined breast cancer and
variation in gestational length in the third trimester of live births. Pregnancies that
are shorter in duration during the third trimester have been hypothesized to
increase breast cancer risk, due to a possible lack of full terminal differentiation
of the mammary gland after a time of increasing hormone levels (65, 220).
Variation in infant gestational age at delivery has been examined using
established clinical cut points for preterm delivery (i.e., very preterm delivery
(VT PD), deﬁned as < 32 weeks gestation or preterm delivery (PTD), deﬁned as <
37 weeks of gestation), or other arbitrary categorical cut points (see Table 6).
From 1996-2006, the percentage of live singleton births in the US that were
preterm increased from 9.7% to 11.1%, while the percentage of very preterm
births was stable at 1.6% (221). PTD may broadly be divided into two types,
spontaneous or due to medical intervention. Briefly, key risk factors for PTD

include Black race, low and high maternal ages, lower socioeconomic status,

35

 

SI

In

In

multiple births, history of prior preterm birth, pregnancy complications (e.g.,
preeclampsia, diabetes), cigarette smoking, and low or high pre-pregnancy BMI
(222, 223).
9.2 Epidemiologic Studies

We identiﬁed seven studies of infant gestational age at delivery and
breast cancer (63, 65—70) (Table 2.6). Six studies were registry-based and one
study used an in-person interview for exposure assessment. Most birth registry
studies did not report how they estimated gestational age. Two studies reported
deﬁning gestational age using the date of the last menstrual period (LMP)
supplemented with the clinical estimate only (67) and both the clinical estimate
and ultrasound data (68), while one study reported using the clinical estimate
alone (65).
Cohort studies. Three of four registry-based cohort studies conducted in
Sweden, Norway, or Denmark reported evidence for an increased risk of breast
cancer associated with delivery of an infant at earlier gestational ages. Two
cohort studies reported a signiﬁcant trend for increasing breast cancer risk with
decreasing gestational age in a last birth (68) and a first birth (70). Both of these
studies reported increased risk for VPTD; the estimate for the smaller cohort
(1,363 cases), but not the larger (5,474 cases) was signiﬁcant. A third large
cohort study reported a signiﬁcant increase in breast cancer risk for PTD in a ﬁrst
birth among women 2 40 years of age, but not for women < 40 years of age (69).
Case-control studies. One small (275 cases) US registry-based case-control

study conducted in 1983 reported a non-signiﬁcant reduction in risk associated

36

with PTD and delivery at <30 weeks in a ﬁrst birth (67), while a recent larger US
registry-based case-control study for ﬁrst births reported a non-signiﬁcant
increased risk for VPTD, but not delivery at 32-36 weeks of gestation. Finally, a
population-based case-control study using maternal self-report for length of
gestation did not ﬁnd any association with breast cancer risk using varying
deﬁnitions and examining exposure in ﬁrst birth or ever (66).

Few studies have examined breast cancer risk associated with gestational
age by effect modiﬁers. Hsieh and colleagues reported that the increased risk
associated with PTD in a first birth may be limited to long-term risk (2 10 years
since delivery) (69). Melbye and colleagues examined effect modiﬁcation by
attained age, age at index birth, or parity, but analyses were based on small
sample sizes which limit interpretation (68). Other studies did not examine effect
modiﬁcation by age at index birth, years since index birth, or parity. Most studies
were among younger women (< 57 years) (63, 65-68) and only one study
reported results stratiﬁed by age (Hsieh et al., 1999) (69) with a meaningful
sample size.

9.3 Summar_'y

Based on ﬁndings from three large cohort and one large case-control
study, earlier gestational age, may increase breast cancer risk, with a VPTD
resulting in about a 20-70% increase in risk. However, one early small registry-
based case-control study found a non-signiﬁcant decreased risk for earlier
gestational ages, and two studies which did not report effect estimates (one US

case-control study using maternal self report and one Swedish registry-based

37

cohort study) reported null ﬁndings. Limited data exist, however, on whether the
association is modiﬁed by birth order (e.g., ﬁrst, last), age at birth, or time since
birth. Further, it is not clear if results are modiﬁed by age/menopausal status with
only one study reporting ﬁndings stratiﬁed by age.

9.4 Commentary

Researchers have proposed that PTDNPTD may increase breast cancer
risk due to the lack of complete terminal differentiation of the breast tissue during
pregnancy (65, 68). Hormonal factors increase fairly progressively until the end
of pregnancy, with the exception of HCG (224). The elevated levels of gestational
hormones, coupled with lack of complete terminal differentiation of the mammary
gland, could increase the susceptibility of the breast to the proliferating effects of
the hormones, and explain the increased breast cancer risk following a PTD and
VPTD (64, 65). Further, in prospective studies, shorter length of gestation at
delivery (in continuous weeks) (96) and a PTD (225) have been found to be
positively associated with maternal estrogen levels measured earlier in the index
pregnancy.

The etiology of PTD is complex. Risk factors for either induced or
spontaneous PTD that may also inﬂuence breast cancer risk include low or high
maternal BMI, cigarette smoking, maternal age extremes, low SES,
race/ethnicity, preeclampsia, multiple gestations, and a previous PTD (211, 216,
223). Most studies that have reported on PTD and breast cancer risk, have only
adjusted for age, parity, and/or age at ﬁrst birth. The fourth cohort study, which

reported null findings, did have information on several maternal factors (body

38

size, conditions of pregnancy, smoking), but did not report the measure of
association. The larger registry-based case-control study was able to additionally
adjust for education, race and other pregnancy characteristics (e.g.,
preeclampsia), however, they adjusted for birthweight in their ﬁnal fully adjusted

model as well, which limits interpretation of the ﬁnal model.

10. Infant Sex

10.1 Egidemiologic Studies

Maternal hormonal proﬁles during pregnancy may vary by infant gender,
which could inﬂuence subsequent breast cancer risk (96, 226). We identified 7
original studies of infant sex and breast cancer (63, 65, 66, 76, 79, 88, 175) and
two reports on additional analyses of the same cohorts (89, 127) (Table 2.7).
Almost all studies (both cohort and case-control), reported no association for
infant sex and breast cancer (63, 65, 66, 76, 175), including studies that
examined the role of infant sex in the short- and long- term effects of a last birth
(79, 89, 127). The one exception reported a reduced risk for women who had
male offspring, in particular among women with two or more births who reported
all males (compared to all females) and were < 40 years of age at diagnosis (88).
However, they did not consider birth order in their study (e.g., sex in last birth) or
modiﬁcation of results by the time since birth on breast cancer risk (127).
10.2 Summayy

Epidemiologic evidence to date provides little support for an overall

association between infant sex and breast cancer, though this factor may be an

39

effect modiﬁer for the associations between other perinatal factors (e.g., multiple
births, preeclampsia) and breast cancer.
10.3 Commentﬂy

Researchers have hypothesized that differences in gestational hormonal
proﬁles of mothers of female compared to male offspring may inﬂuence breast
cancer risk (128). However, as shown above, the epidemiologic literature
provides limited evidence to suggest an association between infant sex and
breast cancer. Studies, however, have shown some evidence for a role of infant
sex as a modiﬁer of the effects of other perinatal factors and breast cancer risk,
namely, preeclampsia (128, 138) and multiple births (66, 76). Brieﬂy, HCG levels
in the third trimester have been shown to be higher for women carrying female
fetuses (226), while levels of progesterone have been found to be lower (96).
Higher levels of AFP have been reported for women carrying males fetuses,
though results have been inconsistent (227-229).
11. Conclusion

Over the past thirty years, epidemiologic evidence has accumulated for
associations between perinatal factors and maternal breast cancer.
Preeclampsia and/or PIH is associated with a reduction in breast cancer risk in
most studies, with some evidence for a stronger reduction in risk for women with
male as compared to female preeclamptic pregnancies, which may depend on
additional pregnancy characteristics. One exception is the Jerusalem Perinatal
Cohort study, which reported a signiﬁcant increase in breast cancer risk for

women with a history of preeclampsia. Large cohort studies provide evidence

40

that having a multiple birth may protect against breast cancer, though two
studies, including a large Danish cohort study reported a significantly increased
risk in the ﬁrst ﬁve years following delivery of a multiple birth. Few studies of
GDM and breast cancer have been conducted, with a possible increased risk of
breast cancer found in one large cohort in Israel and a reduced risk of breast
cancer for GDM at < 35 years of age in a large case-control study in the
Southwestern US. Pregnancy weight has also been little studied and a summary
of ﬁndings suggests pregnancy weight gain itself may not be independently
associated with breast cancer. Two large registry-based cohort studies suggest
high fetal growth may slightly increase risk of breast cancer. Three large registry-
based cohort studies and one registry-based case-control study suggest that
delivery at earlier gestational ages, in particular a VPTD, may increase breast
cancer risk. Infant sex does not appear to be associated with breast cancer.

One rationale for the study of perinatal factors and maternal breast cancer
is that these exposures are proxies of the maternal hormonal milieu during
pregnancy, given that it is difﬁcult to directly study modiﬁcation of the maternal
hormonal environment and later breast cancer (though studies of pregnancy
hormones and later breast cancer risk have begun to accumulate (230-234)). It
has been shown, however, that some perinatal exposures may not accurately
reﬂect hormonal exposures during pregnancy (83, 214). Researchers have also
noted that comparison of hormonal factors during pregnancy between

populations may be biased due to the variability in plasma volume expansion

41

during pregnancy (235). Further, more data on hormonal factors during
pregnancy and the perinatal exposures is needed.

In conclusion, additional research in the area of perinatal factors and
maternal breast cancer is needed, including studies of perinatal exposures and
breast cancer risk with ample sample size to consider jointly several key effect
modiﬁers (e.g., time, age, birth order, menopausal status) with data on all
potential confounders, as well as studies of the maternal hormonal and metabolic
profiles of exposed women during and following pregnancy.

12. Search Strategy

Search terms and limits used for PUBMED searches
12.1 Search terms used to conduct searches in pubmed
Used keywords and medical subject headings in PUBMED:

a. Breast Cancer: Breast Neoplasms [Mesh] OR (“breast cancer”)

i. Every search for each exposure of interest included the medical
subject headings under “breast neoplasms” and the keyword “breast
canceri ,

b. Fetal-growth and birthweight: birthweight OR birthweights OR “birth
weight” OR “fetal growth”

0. Preterm delivery/length of gestation: premature OR preterm OR “preterm
birth” OR “preterm delivery” OR “length of gestation” OR “gestational
length” OR “pregnancy weeks” OR “pregnancy length” OR “weeks
gestation” OR “gestation” OR “gestation length” OR “gestational age”

d. Multiple births: twinning OR “multiple births” OR “multiple birth” OR “twins”
OR “twin” OR “multiple pregnancy” OR “multiple pregnancies” OR
“multiple fetuses” OR “twin pregnancy” OR “twin pregnancies” or “twin
birth” OR “multifetal gestation”

e. Preeclampsia: preeclampsia OR pre-eclampsia OR eclampsia OR
toxemia OR preeclamptic OR pre-eclamptic

f. Pregnancy-induced hypertension: “pregnancy-induced hypertension” OR
(pregnancy and hypertension) OR (pregnancy and “high blood pressure”)
OR “pregnancy-related hypertension”

9. Placental Characteristics: placenta OR placental OR “placental
characteristics”

42

h. Gestational Diabetes: “gestational diabetes” OR (gestational and diabetes)
or (diabetes and pregnancy) OR (“insulin resistance” and pregnancy) OR
(“glucose intolerance” and pregnancy) OR “diabetes during pregnancy”

i. Pregnancy weight gain: (weight and pregnancy) OR (“weight gain” and

pregnancy) OR “pregnancy weight gain”

Offspring sex: ((offspring and sex) or (offspring and sex))

Overall terms: “perinatal” OR “pregnancy factors” or “pregnancy

characteristics” OR “prenatal” OR (“birth characteristics” and offspring) OR

“pregnancy conditions” OR “pregnancy-related factors”

7r?"

12.2 Search Limits:

Articles published in English, studies of humans, the fields title/abstract.

43

 

 

 

 

 

 

268385 so $2-82
9:3 .89 .55 “9.: cm mom 8: .m> mob
5:8 .552 oEqmcmomm Awmdémdv 2.0 n m0 coacmtoqcf 39225 855 note:
Jam 8593 22 09mins. 2:5 523-5
69m oEqumomm ucm E29 E82 cmoE $06. $8 on? _._m
com 8535 :0 3565. on“ so ocm 9: 986m Eém 886 6:80.6me so comanE.
099-92.?
@2me .855
.525 «me E comma
mam ﬁEoﬁE Le 69mins. . . . 8c .m> mob 256$. C8
262.3 “we: Co 0E: quo Two 9 mm o H mm m_ono:m_waEm_ooan_ wmw mmmﬁcococmw
new 5:82 co 8:222 its. 6:“. as v BE .oscoobmmo ucm xmcugom
$593
one 292:8
$593 3030
29295 859.280
same 85882.
9:92:80 Co 35862 comtmano cozmSQoa 33m 9:986 33w >35

 

 

 

 

 

«.me 28ch “moms REES: new _coacotoafﬁmo:“3:738:35 .EmaEmeQd ._..N 638.

 

 

raw—am... .nr

44

 

.6 2% :52 .2228.
6:222: “m can .waﬂm
msonm Erma .28
8:222 m£ “m ucm om
8m 5 _s_m .55 as a mmm
.mmm 853m :2 n2wz€<
.mmm 8523 :o 8:822

:mmsmmocoEoi

$3-39 8.0 u :0
233322.58
89738 8.0 u :0
”6330:0255
”322m .mmsmao:mE >m

u222-xocw:m2m

 

85:
.w> 2qu282&
2:5 2:.

wmém

amen;
as;

nmm Toma?

252322
c0282.

6:80-080

wwumuw “5ch

5 5 88
.._m cm Eek

 

.2me 6:: :_ ommmsomﬁ
2m 2:62 SN: :8ch
:82: new 2382035

comic: 5:22

o5 E xom actamto Co 22
6:2 mEmem 9 62039.8
225 3368 5:362
.co_§mwm 3235::
62:82.2 ozaanm

.xom ESE 59:23 “m

can Ecozﬂmmm 2:925:22:
Eat: 525038

.82 :25 :2: cm

can 2:22: .mmm 3226
a: 8322 29:2

20 2% can 8:362

so 3:38 :o 8:822

ANT Tmmdv mwd u m0

3: .m> mob
2qu282¢
”:2: 2E

mm-NN

Nmodw
\Nmmd

mmmw 3:9.

comm:
-Emamm

thOO-meo

82m 8%:

38
38 .203
new $5:

 

.mBEm

638056 95 .__2m .55
:2: cm mmﬂbtma .82 .27.
due 8235 :2 3632
.m2m 2823mm new

mom uoEmzm co 8:022

 

2.2-2.8 3.8 n ma
2.78.8 5.0 H mm

 

 

Lo>oc .m> IE

$29. .m> 2me0 .2
2:5 22

 

vv row

 

Neé
5mm;

 

Nmmrémmw

32225
523.2

.0bcoo mme

 

82m 825

as 82
__m 8 Each

 

 

88:80 .3 2:3

45

 

 

 

 

 

 

 

 

a: a m m .2 Rap? 5:5 2:. -8v :88 :28 -8528
63-8.8 3.0 u m:
555:8 22:2 88-82
=83 .2 5:0 x2: 8.0-8.8 8.0 u m: s
x8 actuate 0:0 2:08 ”222 , 008: 25:02
_2:mE 5:8 £05285 28:8 occamto >m 28:0: .w> IE .2853:
5 52.2 5:5 a: a 8m 8.0-8.9 8.0 um: 558 928.89: 5%. 8 58: B8
.08 8:58 5: Baas: ”.555 5:5 a2: :2 52.0, :28 .5 a 88>
82-82
3.8.8 5.: H mm
3:8 58 “28> om m 088 82:02
5:5 a: a. 8m 88:85 3.0-8.8 33 u m: 522 .9 1.: -228:
20 00:8 28:28 28» om v 220:0 28220020 mmﬁmwm cum: Noom
.08 5258 5: Baa? «.8 8:58 5 5:5 at 8-8 v E}. :28 .5 a 58>
.28 852: 82 ::5
x85 9: 2 59855 a: 32-82
mm>> : gm>wc mEzmwm END
.80 :8 0:05 8:222 8m. 78.8 .5... u mo 50...”. 8:002
5505:: .82 0:0 6062.:
8:285 5:925: a: E. 78.9 8.0 u :0 22m 1.: .2222; 8220 :25:
8 8m. $08820 x85 828:.
a m8 .8 2a=5< .92 28.5.8.8 23 u :0 205 225289: :8 38 88 :5
5:5 5:52: :0 5222,. 5:5 85 :31: 58 :28 a 88.22:
22:00:80
8 IE 5 228.89: a: 35.8 8.0 u :0
0:0 £05880 ”6880:8282; :05:
2:2:50 28:8 8020 En. 75.8 mod n m0 .m> 86:88?

 

 

 

 

 

 

 

 

 

.8228 .8 2%»

46

.38 5:: 826:2 .8338: 502520 5 0585500 .5 090..
.38 .5 555800 2 82 2% 59: 03-325: 0000055 5.; .800 38 .5 5 .250 5 902: :0 00; >030 05» .

r.
.039 05 :_ 2000:0000 08:08:50 0:0 00630 020 05 00:0: .00:000:000 0:__000E

0:0 000::0 80288 E90000 000: 0:0 00000: :0 00: 0_ >030 05 50 .980 meow .._0 00 :0:0> £05 0000 05000003 000205 >030 050
000V 5:: 00320. 0007000? 00000 E00595 0:0E00_0>0o 0:0 5.00: 05:0 05 0:000 00:00_0_t00 00 00:0: 005

.0_0:0:00 05.: 0:0 00000 mum; 003 :0_0:0t00>: 0000.0:->0:0:00:0 :0:— 0N_0 0_0E0m0

.0295 00:00:80 $00 0 0_ 0_:._.0

.0_n0__0>0 0:03 00:000: 5:5 :0Es 00380 00: 00 -0000:00_0 002, :00:00 :0:>>0

. .0_00__0>0
: 30-3260 000. 00 :00> 00 0:000: >:u0_00: 5:5 00:: 00 :00> .000 00-2,0__0:0_00:00_0 :00:00 :000:£ 8:00:00 0000 00 0:00> 0000_0:_ 0:00>

SN: 0_00E0_000:m

00: 00000 N 28 0:0 80050000 00 05000:: 000030005080 05 :000: 8: 00 :5 >50 050805090 00 00x00. .800 008 .._0 00
:000:0:0_m 20 00 00:00 05500.059 0:0 5.001 0.20 05 00 9:092:00 000: 005 >020 0:0 .00_< .2>0:0:00:0 00 90286 0>_0:0t00>£
A000 .38 0_00E0_000:0v :8 0.00... 05 :_ 00.00858 00: 0:0 80060000 00 005000;: 0000100009550 0000: 00: 20 005 020300

 

38-52 50:0.
00:000:
0m>mc I 50522 cam—9 mOON
>500 20 5:5 65.00.; :3 "0: .0> 000505090 _OVA :2 .5 5 5090.2

 

 

 

 

 

 

 

 

 

00228 .8 050»

47

 

8929.6

 

 

 

 

 

 

 

 

 

 

E. 72.9 :e u mo __m .9 22.22 $2-82
its am. 9 BE
.3ch Em 5% 39295 929m 825
“me am mom 5,. 8622 29296 c8595
.85 28388 Em 62-2.9 8.0 u mo __m .9 22.22 28 at mm? .._m
0mm 8586 :0 8:222 its “mm... 3.8 $55 63:00-38 am 5882.
003.30an
$.me BE:
ummmn
.moumcgoo .2 29:62? . . . c2295 -bumamm $8
02 ém>=mo “we .o 9:: k3 Tow 8 mm _‘ n am 32? met: mmm mm? cozmcmn
ucm 5:80. :0 vacuums. ”FEE “BE mv v 85 65:00.88 ucm xmccﬂoa
359$
mom £03.50
859?. )8me
$225
858:8 oxommv
958800 55632 “.0 93mm: comthEoo 5:938 355 acmﬁmu >36 32w

 

 

mxm: .mocmo Emmi .mEQmE new 2:5 9 £22 ~.~ 2.3...

48

 

 

.mmum_gm>oo LOF “coEumzbm

ummmn

 

 

 

 

 

 

 

 

02 63:3 ho Eamon mcoumacﬁ -EEmmm 85v
éozau B 2% .mmm am. 738 Be u mo __m .9, 2E 30.8 82 €on
umEmtm.btmacovmcem2 ”:En>c< omz \mrﬂm 65:09émmo acmcoﬂo
220.96
3 t 78.8 8.0 u «.0 __m .9 03:32
5:5 32 2 BE
.owsmaosE 22295
a mam 95 .2229: am. 78.8 v: u «.0 __m .9, 2&3: 52-82
E mam __s_m was,“ ”5% 53
59.3ng Socmcmma 2,9235 85m. 8ch
E5158 6.: am mam :3th 2.229% 39323
§ 8382 9% 2m mam E. 72.8 3.0 u «6 __m 23.2: r N :3 at 32
853m 8 88228 SE >5. 3 v 586 6:80-88 _._m .m 505
33-58 «3 u mo
”mkmm> mn-mm
5.838 F? u «.0 82985 __m 2332
”98> 3-8 ”5.5 an. 2 BE
8973.8 8.? n mo
”9mm; aﬁmm 220.93
63.88 3.0 u «.0 __m .9, 29%:
”98> mméw 5:5 63
89.8”?
an. 788 Be u «.0
2:5 “98> 2-3 32225 85w BE:
9,: B 525: 95 .85 we 5. 75.8 o: n m0 2.229% __m 89323
a mam .8528. 6 358 98> 3-8 .2, E5 22:35 F N Sod at $2.6
.mmm 853m .2 87,22 “mom umsmﬁm >m its >c< aﬁom :omd .obcooémmo am 832

 

 

 

82:80 .~.~ 2%»

49

 

awfomov 3.0 u «.0

0mm _. -wmmr

 

 

 

 

.ctﬁ. E20“ 6.59» mm A bmwmn cmcmgw
22.22: a mam Em 2:8 33.58 8.0 u 20 2.2295 _.m 222%”.
.2 U222? .52.. 2o 28:. 28> mm v .2, 2.22 2222: r N $282 :8 82
can 52. 8 8206.2 ”mom 852% 2m. 225 >c< mm m 28.2 29.8 _..m a 853
.mmEmoma
ozaEnm .mﬁaEmBmmi
.xmm EEC. 5.62.8 6
wow .mcozﬂwmm £99555
E2... 25.2838 82 -E:
.82 .222 222 a 8.2m U225
0mm ﬁEQmE .mmm umEmum ummmn
.8 828.22 2.22.3 -6663. 38
2o 2% Eu 882$. 5.328 m: n 20 2229.5 .2, 2223.2 $3: 38 .225
“6 3:28 co 8:832 .529 62.... 3-8 \Nmmd. .obcoo mwmo ucm wocE
9.22me
62.88 3 H mm ._m .9 2.5
2:5 mm. 2 5...”.
2.06.9.7,
6.348 8.0 H mm ._m .2, 225 82-82
.225 222 a ”2.... 23..
mom .529 6.: 2m 393:8 26.29:. 365 82:5
.39 .96 .mmm .2 852.232 2.92926 c0982..
.35 222.28% 25 23.88 .23 H mm ._m .2, 2.5 2 N No: 88 82
mom 8522a :0 8622 2:5 22 3 .8 an? .228 88 __m .m .22»
88-22
{mEcwo

 

 

 

 

 

 

 

 

82.280 .~.~ 222.

50

 

£5.7me 8.0 H mm.

9.92me
Ehv .9 223.2

”5.5 mm. 2 5E

 

 

 

9.90.9.6
83-8.8 one u mm. __m .9, 98.22 52.8?
SE Em...
comma >m2coz
.mchmcmma E8“ 2229.6 4:883”.
=8 6 .852. 828 £5 30. 72.8 3.0 u mm. __m .9, 22:22 $§2 at 88 ._m
.mmm 8596 5.. 8389a ”£5 >94 8.8 $5. :28 6 c3292
29296
5. 7.0.08 5.0 u mo __m .9 was;
its “mm. 9 Bra
29996
23.38 Ed u «0 __m .9 2;: $9-52
5:5 63
.55 umwmn :3mb
8.2 .3755 um mom new 2929.6 -Emamm
3:8 5.. 8582 .55 .6 33.3.8 go u «0 __m .9, 55 F N in? at 52
56> 5522. :0 02292 its >c< on v Ban...» tocoo :_m 5 39:2
in. 78.8 3.0 u mo
Emmi mm A mcoumasm
so. 738 86 u «.0 __m .9, 29%.).
”2mg mm v ”£5 «mm. 9 Sta

 

an. 78.8 3.0 u «0
”9mm; mm A
A8. 73.8 8.0 u mo
”9mm; mm v

 

2.20.me
__m .9, mazes.
H£5 “mm;

 

 

 

 

 

 

82:80 .~.~ «38.

51

 

KEN-ome-

 

 

 

.ww_ocmcmma 8me :3mb
.6 59:3: 96 .ctEEEQ 2299.6 .6563.
525 a «mm .858 sad-:8 35 n mm __m .w> 55 m.2 E:
.mmm 8596 52 8582 55 22 mz 55 F :28 538 .._m 8 5
82-52
umwmn :8mb
.5505. ho £52 50 2mm mcoumasm -Emamm
can. .55 «we a mam 555 3972.8 3.0 H mm __m .9 05552 «.2 85 ”Bow
.6 .852 .5 8583. 55 22 on v 58.5 :28 .._m a .282
587$?
.55 5m. 2m 55 a was 83-5.8 8.5 H mm
ncm 36:2me ho .onEzc “28> om A 083 mow-Em 8:5
.2 8583. .5265 :m. 78.8 Noe H mm 222me -52QO
65:. ho .85 can 58> £5 ”29.; 8 w ._m .9, 553.2 58.2 88 38
co ummoaxmc: 2 858.2 ”man 8585 >m ”FEB 22 555-2 63 5:00 _._m 8 .282
35789
59:50
.555 8565 29:9me 8me
.25 .m55 ,6 .853: .55 .5295”. at
we a mam .858 558.8 8.70: E H mm 552?; 9 055.22 30.58 82 55.02
.mmm 855m 52 853.2 55 58.. 8 v 58.5 :88 Ba £5525

 

 

 

 

 

 

 

 

 

 

8258 .3 25¢

52

.535 :32 m 53> 5503 95:5 mmmmo 50:8 $35 50 525:2.

Emu 958 m5 0.53 2 >25 .65 «m5 5ng 6c on new mmmzmcm 0cm,

.mco_m:_oxm .3068 E 855:6 8m: 35 £965 .83: moon 282 new CD mm? .._m am 383 53> 3255 “m5 Emu 0mm: >536 $5.;
.536 9983 m mm 08205 2 = 8:9. .mco_m:_oxm new :9me

333 E 805st 9m 995 £965 .Ct mm? .._m “m 853 >5 nmw: cozﬂaaoa 33m m5 53> 8955 3th “m5 Emu 0mm: 32m wEhm
do? E 805 3.32.2 .EmE__oEm am @935

520 8c 2 £892th 5 mmcm. mm<o

.5295 853:8 o\comv

.m_nm__m>m 295 8.89 55 55> 3305 Ho: on 63836 83 .8ch cmc>>u

.m_nm__m>m

5 E:->>O__£ «mm. yo 56> 2 Boom. €569 55. “9: ho am; 5.3 a:->>o__otm_m0cmm_u .850 new 558:8 Emu B 28> $865 233

.A F 5 3850 $35

ucm £659me .m> 2.33 ozomEn .2 7: mm 58556-5: m5 .8 02.6%. 6: $3 3295 8:35:00 m 3383 tecoo 8.0.3-982
m: 95 $3.96. 83 ﬂow... A EV mBm... o5 E 855552... Ho: 9m 55083 8 993mm,: umumaﬁméﬁtgoo roam. .0: 2o 55 moﬁamm

8258 .3 29...»

53

 

Sofa-.8

8;" mo ”25> mm m

5.0-5.8

3.0 u «.0 émm> mm v
two-5.8 8.0 u 5

 

 

 

 

 

5855558,;
83 -58
:l u 5 “98> mm M
53-8.8 38-82
59055.05 5.0 n «O me> mm v
Emu-=3 5 555:: :N. 598 2.0 u 5 5355 85m 85:
95 55> 3 5m =2m 53555555 5 2n. 5556 >5 5 595-5
.25 5555 .2 .5532 5.50 5 wow 95 >55: 0: .m> $50 mrmd :3 Son
.25 5595 5 5555. 955 525555 E 550. >c< 2-3 v 55d 5550 mme .._m 5 55:3.
.395 5505 ucm
.>59moEEmE 65555 «mar-om?
5 mam .=2m 5950 .555
85 5 5355 .39 .35 39225 85me 555
.wmm 5595 .8 5532 595-5
.85 25985 Em 3. 72.8 E H mm 55: .m> .25 8: 88 82
mom 5595 5 558.2 55 >c< vv -ow Emu;- 6550 ammo ._m 5 655
A253
mum 20550
5592 33.3
5295
85350 exommv >
355500 555082 5 05522 555500 55.55 >caw 5:995 53w 53$

 

 

 

 

 

 

 

RE. 5050 555 .9555 v5 555% 555330 6." Sam...

54

.2: 53 ._ 538 55.55555 .0 555959 55555 5:5 5555 5 mob >:<0

 

 

 

 

.2555
t Ana-26:2 52 5 55> 2 Boom. >555. 55 5.5 5 55> 5.5V 5525550555 5050 55 550250 95 5 95> 55055 953
.2an m5 5 555555 5: 55 55.085 5 5.555 55355-55550 55. 5: 55 55 553mm
35.52.
.5% 859955. 5.3. s C u m:
55 .5555 .595 ”on m 5.005. 55.9
2:55 .550 55cm .55 3.508 0.5 u m: 50522
3238 .9. 5m .550 55 .8 v 8.. .5 68
.55 5595 .o. 5532 550555 5 5mm 5 .55: .w> 550 5>m 5.9 55.5 550 .5 5 5.5...
Comm-5.: For n mm 5.556
5.5 55 Sig-5.8 w:- H mm 5.55.». 59-29
.55 .95 5.3 >825... 5598.8 85 n 5 8.1.?
x25 5 5.8 5%. .9. 55 03 www.85- 88..
55 >55 55:55 555 5.555 253 >555”. ucmzoow
5 555 5 m5onm 55 .55 mm 505 5 58:5 mmu
5:555 .55 .o. 55:35 @559 5 m3 m2 .9 550 £va 5w>>mo
.5995
9555 599555 >55
55:55 92 5 @5555
55mg 5255 5555
.15. .55 5 >550 room-N55
.5535 559 65on5
.55 5:555 .592. .55 595 5535
5.5 5 5mm .55 595 .55 . >555.
585me .55 .9. .m 55.58 .3 n 5 09. 9 $.85 om now. P 5. .8. meow .5
won 555:. .0. 85:5 55 5.... V 5a 95 550 5 35:59.0

 

 

 

 

 

 

 

 

82580 .5 033

55

 

 

 

 

 

 

 

 

 

 

 

 

5.5 555 w m 5:05:55 Cmmvoe 9v 59-59 0:25”.
5 55 60.5025 >me 55
.0. 553.5,... .5555. 50.55 5:95; >05:5.n. >535 555. 68
55.9555 05 5 mm: 55 5555500 mam meow .5 5
55 5:55 :0 5:055. 55v 5.5 >:< 5-5 .5 5:00 52.205255:
.05 5.5
0:008 5.5 5.... m :5: 58-59
m.5: 5:00 5.5 555 :55 c.0550: one u mm 3.: .m> m7
55:55-05 55 >550 55555550.". By: 55 50.005. 525“.
.>05:5.a x25. 5 05 $553.8 0; u xx 5553 >05:5.n. 5550...
5.5 5.5 5 55 5:05:55 55555855 “5555500 mwod :8 voom :5
5 55 .0. 5535.... 5:55 555555 >m 53 5.5 >:< Eém 59 5:00 5 55:5.
.3: 25:89.80 a. 72.0. o: H mm 0% A
5.0 05 .955 3.75.8 mud H mm 0?.on
5:005 5:505. 5:05:5E 6.73.8 o: u mm 05-0.5 59-52
.5 omm ._s.m 5.9.8 5.... 8.5-8.0. 8.0 u :: 5.5-0.5
7.... a 355.5: .89 .27. Cum. 3 5.5-5.5 5.22... 8.5 85.5
.08 855 .0. 5.8.5 Av. 53. 3 H mm 0.8 m 58.8-...
.85 0.598% 9.5 .3; 5m .5925 .8. 82
55 5:55 :0 5:055. .55 55... 3. .8 253mm 55:00 550 .5 5 5.0...
5.53
55 20.500
555?. 5550
529:. 5055500 $58
555500 55.082 5 55.555. 525.500 55.500 >55 5:255 >55 >55

 

 

N5.. 5050 555 5:55;. v5 5.5 5:952, >055

 

9n. .V.N 533.

56

: 32-326», uwm_ “—o .69» 2 Boom. 36?: 5.5 “2: ho 59» rag 3-326358%? 8260 cam c2638 Emu ho Emma 2865 98>

.=>_m E82 “moE m5 2 m>=m_m._ 5:335 as: 5 5mm :66; mm 8550

0
.m_nm__m>m

n

.36: 50:8 685 _mm:ma0cm§woa go 05 ncm >ocmcm9a ~ch 5 Ema £90; 5. co=m_oowmm o: 8:80: Ava:
.Ahmcta mSmoaxm £5 mEmem Ho: Eu Sn .xmmz, 5a 355 E90? ”6 $3.63: $98 588 685 *0 85202 $533.03 959205
ho 95: m 3:3me 89: was 9: E 858888 Ho: 9m 539038 6 $588.. uQmivmémtgoo tone 6c Eu “9: wmﬁamm

 

 

 

 

 

 

 

 

 

dam
ucm _=>_m_ .50ch $85
.6 rem: 2E5 98> no. 738 Ed u «.0 mm N 88.8?
m mg 05 5 EmaoEEmE : v 72.8 No.0 u «.0 mm om
588:8 .55 Emmy 3 om v 2,9225 855 8%:
E: E mom 5:8 .mwmum ommcmco 9.9322.
.mmm 85sz be 8332 59m; 52$ 8: 3m: 88 .._m
.mmm 8586 :0 856.2 5:5 “mm; on v Eon; _o:coo-mmmo “0 89¢qu
Sm 5.5 9.88 .55
“we m 509 m>mc 2:8
55 x85 .3: :58 95% 5.32: 8d u mo +8
_s_m_ 2 $56 .588 2.93on 3 n mo 8-9
385 6 been :53 5.73.3 8.0 u «.0 2-9

 

 

8:528 .vd 2%»

57

 

oommémmm

 

 

 

 

 

 

5.5 steed 552:6 2.3.8 3 H mm - 8me
2m 2:5 .0 595: .55 c.7ddv dd n mm 8% 88 -Ewaom at
as an mdm dozmd 5228 cum: de 88 w 82 .9562
.0? 852m .2 Baa? 05:5 33 8 v mzidd :28 new £3.82,
.co_§mmm
_m§_=:E 69583
932% .mﬁaEﬂomma
.me 2&5 .td dd. 73.2 Ed n «0 83 N
N 97% .mm vv 52% 5 A3. 7R9 «dd u mo 83788
mdm .mcodmmmd 5:838 at 73.9 add u «.0 88.83 82 -Rd;
.89 5.5 we s cm”: dd ddvmddmm 35w 8%:
add 55%.: .mdm 8595 63$ 8 odd u mo ddvmdddm Ewan
5.. 8322 59:8 so. 73 9 odd n mo $2-88 -Emamm 33
do 2% Em 8528. E. 73.8 «dd u mo 82 v N39 38 .996
do 3,58 do 8562 5:5 E: 3% Ram 6.28-88 Em 85..
.mdm acodgmmd 5. 68¢ 8d 722
$3.6m Ho: 20 .mmumtgoo . .0 {9:ch
.9 Edema? 02 av. 78.9 2: n mo -83 9, 68¢ Av 889
62:3 do 558; dads. -Emamm E :
.bmzau do 9% 6% Eddie ﬁe u mo dog .9 88m vv mmddm 82 .556
8596 5:8 :0 822.22 its >c< omz \m an _9Eoo-mwmo new 520
A233
mam 205:8
umcﬁaxx \wmmmo
9325 88228 $de
mucoEEoo co=m_00mmm do 9:822 A8 EgmscEm cozmSaoq 33w 9:9on >035 33m

 

 

 

 

 

 

 

mxw: Bocmo Enos 5599: van 5265 .30“. .m.~ 235...

58

2:5 :22ch 2 8:5: N

.ooo.v-oom.~ w_ 965 8.8999 9: 65 can 55 Em be 9m 3.3me m5 “m5 mszmmm o; €99» 6: 5% x85 am 3:8 95 9.65 8553!.
«60.0 doc m_ £392th 80 mass. mm<o
.m_nm__m>m

t E:->>o__2 ﬁg ”6 ﬁg 2 Boom. >559 5.5 SE 80 ﬁg :93 a:->>o__otm_mo:mm_u 50ch new cozom=ou Emu ho 28> .2865 2mm?“

.89 .38 2an 9: c_ um~tmEEzw 6: 9m cozﬂoommm ho

mmSmmmE 8536999550 tone 6c Eu 55 @235 A82 .6 mum v.o 9mm) mm .08 6380555 v6 cmEo; rm: 5an 53305an
>_c_mE 3205 863m 29.65 @935 05 “—o Em 88 8:89 8: 90; madman $330:de “6 99:38 mmﬂmamuw _mm:mao:oE 3 £383

 

.btma $8233

.89 E 9685 _mc_mm>
53:68 wmuonmﬁ .wmmmmwﬁ
965:3? 8805
$8835 .ctE 80 2:58
52626 >=Eﬂ .mEonm Som-~wmw
._s_m .2555 .292

 

 

 

 

 

 

 

 

 

RENEE .xmm ENE .88. A2. 78.8 N? n m: . ddwm 889 5826
N .3 dv dd dd .5 .58 3773.: :e n m: .83 w> do? A -82de
man acoﬁumd .55 “we a 8855 d odd .8 dm 8d. F 5 88 88 __N
mom RENEE .8 8532 mats 55 v 38 58d :28 6 39.820
8265.5 6.3. 8 E H mm 8?, A 32-82
.38; mm N .933 ~de 3788 3 H mm 85.88 {9:50

 

82:80 .3 23¢

59

 

 

 

 

 

 

 

$778.88.? H mm 3N
Cum: odd 8
E. 78.8 «3 u mm 8.8.
8078.8 x: H mm 5-8 «8782
A8 7K8 dd F n mm 8-5
25.8-8.8 N: H mm 8-8 8me 85:3
5.5 E: 8 m8 sod-8. 8 8.8 H mm 5-8 488%
Em 5:8 60:8 5228 $378.: :8 n ma 8 v 88:. 88 82
.08 85sz § 8582 8:5 53 $2 88; :28 .._m a 83.22
mddr--dp
.55 E: a Nam 885 BE:
.23de 25 N8 8598 $8 8e 5 N 889
.2 8382 82:8 63.88 5d n «.0 8.8 -82de $8
6 2% 95 8829.: 8.8-8.8 m: u «.0 mm v 88.8 38 .986
8o 858 :0 8:222 “55 E”. 3.8 883 688-38 Em $5;
082-82
386 8%:
. an. 738 5d H mm 252
38:98 .2 69:39.3 u . . . nm N .m> mm v -bﬁmmm ﬁg
02 252.8 EEO 2:: dad 78 8 8 d n mm 5 N .9 on v 9.3 82 5258..
new 5582 co 8:822 its “mi 3 v ER 6850.030 95 $5.28
3.59:
can 208:8
30592 \mmwmo
omen-as 85350 $38 @385
mEmEEoo 533082 “o 9335. 0mm Ecozﬁmmo cozﬂaaoa 355 9:9wa >36 35%

 

 

 

 

 

 

 

th058 «805 ESQ-NE new bm>=mv “m mam Ecozﬂmmm ESE .wN win...

60

._m>._mE_ mocwuccoo excomu

.m_nm__m>m 2m; $802 5.5 :32, 3305 Be on -umwocmm_u mm; 50:8 5550

.o_nm__m>m

t Ans-26:2 89 80 80> 9 E89 3362 5.5 “we “o 52” rag nus-32636836 5250 cam cozom=oo Emu ho 28> 9855 98?.
.588 $me 5599: new bm>__wu 8 0mm accumﬁmm E35 .8 £392

:3: 8:89 Eon ucm 80 .88 2an m5 5 umwcmEEzm “0: 9m mcozgoommm ho 3539: 8338-9398 team. 6: Eu “9: 363mm

 

 

 

 

 

 

 

 

 

 

 

Emmy 03 8. N 82$?
2:5 be $25: .92 Em 33-8.8 Q: H mm 8-5 5202
55 E. a mom «magmas 82-5.8 :4 H mm mm-Nm 8mg
.6 8:8 5228 83.58 N? H mm mm v 8 $38 328m at moon
.0? 8295 § 8582 55 mi -8 v :56 .._m 6 gem-S
:28
89-29
A8. 78. 8 one n «0
6.59» 0v N vwwmn cwvmgm
5.5 E: an. :28 8e u «.0 9.83 imam”.
a mom Le 8382 .59» ”28> ow v S N .w> 9.8; 5v 8 2: a8 82
£5 55ng 8 8655. mam 8535 3 £5 mi «.2 BEN :28 .._m um 52-.
63-3.: «5 u xx 5%:
.w> DPQ> ..m>m

 

82:80 .3 29¢

61

 

 

 

 

 

 

 

Ear-$9
x.mEcmo
5.5 E.9m.a >_mEm.uxm ummmn
Em 2:5 .0 .852. 5.5 .3553. at
.2: a 08 3:8. 525.8 a. to. 5 3 H mm 00.3 .2, 29%. 30.08 82 .2355.
.03 8596 .o. 8623‘ H5.5 “mm. mm v \mmvd tozoo ucm Emu—22S
.co=m~mmm REESE
.oﬂcmoma ozazbm
.QmaEmBmma 59:2. Hm
mom Egan-ammo 59955.3
ESE 60.588 30.. .R?
.mom. .ctB 6.: am $.me umgcs
wmm .mEEmE 6mm umcﬁrm uwmmn
5. 8.3.2 59.6.. 32%”. 68
.o 2% EN 8528. $3.88 8.. u mo .22. .2, was». $0.8 $8 .996
.o 3:38 :0 8582,. “5.5 “mi mm-NN \Nmmd 6.28-38 van 3...:
89.2.9
{9..ch
.mm.m_.m>oo .o. EmEHmaﬁm 88..
oz 602.3 .o .958. 8.95. 452%”. at.
$52.2. .0 Emu .09 So. 73.8 Se n «.0 8 $.95 25. xmw mmodm mm? .E.2m
852m Erma co 8:822 HEB >c< 12 Ram _o.Eoo-owmo ucm 53.0
3.33
mom m_o.Eoo
8592 .830
29.9... 005350 8&9
wacmEEoo 8.6682 80 8.335. 5889.50 5:938 .635 ncgwmu >35 >36

 

 

 

 

 

 

 

Nxm: .mocmo “mm-65 8:6me new xmw Emt: Hm 23m...

62

._ 83-326. .3 .o .3) 2 Boom. >562 5.5 .2: .0 :8.» rod. q:->>o__otm_wocmm_u .mocmo ucm cozom=oo Emu .o 28> 9865 98>

.28 2:3 8.29.50

.m_nm__m>m

9

1.850635 ﬁESmE ucm xwm Emu—E c0053 5.5688
o: 09.0%. 508 80 :8. V 9an 9.. c. um~_.mEE:m .0: 9m co_.m_oommm .o 3.3me 3638-23.98 team. .9. 06 55 8.9.5..

 

29.6. .m> was.

 

 

.8. 78.9 8.. u my: 5.5 an... 52-89
0 $3.02
m .m 88.
5:8 new .tocooéts .8. 28.9 8.0 n my: 3:8. > was. SE2 535$. 65 32 _._m
6% 852928612 05.3 2. 8-8 .85. :28 .0 c3225?
6.17%.... 8.. u mo
”noxzz
c0759 2.. u mo
”moﬁE __<
”98> 9. N
3.3-8.8 3.0 u mo 39.22
onm 852
.5933 8.0 u «0 0835. __m “.88 585
.ctB 5.: “G mum Dcm mmm ”mm—NE =< .m> £05.03 m:o.wq_n {$5601
8596 5. 8.8.2 .8. “mam. 9. v 9.9% 8.52.56 08.2 88 82
LED ESQ—NE :0 09.332 w_mocmm_v um wmm >m xmm “GEE Vmuor \wNQN tOLOO _._m aw £0_wI

 

 

 

 

 

 

 

 

8:528 ....~ 2%;

63

CHAPTER 3: STUDY DESIGN, DATA PREPARATION,
AND METHODOLOGIC ISSUES

1. Study Design and Study Population
1.1 Overview of Study Design

We conducted a population-based, case-control study among parous MI
women 20-50 years of age to investigate the associations between pregnancy-
related factors and maternal premenopausal breast cancer risk. This study
utilized Ml Resident Birth ﬁles (MI birth ﬁles), Ml Statewide Cancer Registry data
(MSCR), and Detroit Metropolitan Area Surveillance, Epidemiology, and End
Results (SEER) Registry data. Figure 3.1 summarizes the study design, study

population, and data sources.

64

 

 

38w. w. m e. ”83 285:3
£5 2,: use: 235322-5955

 

 

 

V

 

82-82 ”8.3
BSEtoo gEm 2,: 59:25.

 

 

$56 $0ch :3
£2.22 “E29”.

 

5mm 82 5:68:22 :33

 

 

 

3onme ”355mm .8ch

\

 

z

 

 

coon

$53.me 5322.2

mwmr

 

 

/

 

.85 £5 :2 9: c_ or m mom E Amﬁocmﬂu
588 2 5:3 £5 9: we 2 EVE: .38
.39 mch :2 E om-om mom 5 5250 “much
basin H9: 0282: ._o 2_w-c_ 5:5 nomOCmmﬁ
mmﬁEﬂ xoﬂm Lo 2E>> ”330 .8ch «modem

 

 

whmr

 

 

.mmm v6 28> E w «m €me

.33 «Ha =9qu 5 5 $838: 2569
50:8 :2 E 5250 So BBQ: 0: Emoxo
£38 8 29:0 Qﬁaﬁ seam .3033:
m2: ctﬁ EmEmm: :2 E0: 862mm ”£05.50

 

 

 

.w_o=coo .8 mﬁocmmﬁ So

23 938 8:0ng ”ammo .2 £8530 So 063 Emu 8:939 >36 “m 32 ._o 98> om 0mm 052, can voom
-wkmw @550 EmeP. :2 mm om-®_‘ mom um 5:5 02. “mu: m 53> cmEoz, xoﬂm Lo 2E>> 9m; mEmEoEmQ m_n_m__m_

 

cozﬂaom >335 IL

$058 San cam .cosmiqoa >036 699% >an So 2,9290 ._,.n 2:9”.

 

 

65

1.2 Study Population
1.2.1. Cases

Eligible breast cancer cases were identified from the MSCR (1985-2004) and
linked to their ﬁrst live birth in the MI birth ﬁles (1978—2004). Eligibility criteria
included: (1) diagnosed with in situ or invasive ﬁrst primary breast cancer
between 1985 and 2004 in the MSCR; (2) age 20-50 years at breast cancer
diagnosis; (3) no previous diagnosis of any cancer with the exception of basal
and squamous cell carcinoma; (4) White or Black race based on Ml birth ﬁle; (5)
ﬁrst live birth in MI at age 16 years or older during 1978—2004; and (6) residing in
MI at time of diagnosis. The study reference date for cases was the case’s date
of diagnosis.

1.2.2. Source Population
The source population (the population that gave rise to the cases in this
study) could not be deﬁned until after linkage of eligible cases to the MI birth ﬁles.
The source population included parous White and Black women aged 20-50
years during 1985-2004, who had a first live birth at age 2 16 years while residing
in MI during 1978-2004, and who are assumed to not have been lost to follow-up
(i.e., died before study reference date or moved out of MI prior to the study
reference date). In addition, the source population has restrictions on the
Possible range of age at ﬁrst births that depend on both the age at study
reference date and year of study reference date. This is due to the use of pre-

existing data ﬁles to identify our study cases with restrictions on both the year of

66

ﬁrst birth (1978-2004) and year of diagnosis (1985-2004). This issue is discussed
further in the methodologic issues section below (see section 4.1).
1.2.3. Controls
Eligible controls were selected from the source population described

. above, identiﬁed in the MI birth ﬁles (after linkage of the birth ﬁles to MSCR). The
eligibility criteria for controls included: (1) no history of cancer in MI between
1985-2004 or in the Detroit SEER Registry for the years 1978-1984 (area
covered by Detroit SEER accounted for 43.6% of Ml’s population in 1980; 42.1%
in 1990) (240).; (2) age 20-50 years at study reference date (individually-matched
case’s diagnosis date); (3) White or Black race based on Ml birth ﬁle; and (4) ﬁrst
live birth in MI at age 16 years or older during 1978-2004. Control Sampling
M. We individually matched four controls from the source population to
each eligible case on calculated maternal year of birth (+/- 1 year) and maternal
race (White; Black). Controls were required to have their first live birth in MI prior
to the study reference date. The ﬁrst step in control sampling was to identify the
sample of eligible controls that matched a single case on maternal race and
calculated maternal year of birth. This sample was further limited by requiring the
ﬁrst live delivery date of controls to be prior to the date of diagnosis for the case.
Four controls were then randomly selected from this sample and matched to the

case with a unique number.

2. Data Sources and Measurement of Study Variables

2.1 Michigan Statewide Cancer Regjsﬂ

67

Cancer data in MI has been collected since 1947, but the MSCR was not
fully established until January 1,1985. Ml Hospitals, physician ofﬁces, and
laboratories (~180 facilities) are required to report in situ and invasive
malignancies (with the exception of non-genital basal and squamous cell
carcinomas) to the MSCR. In addition, nursing homes, hospice care facilities,
and 15 other state registries provide case information to the MSCR. The registry
includes National Cancer Institute (NCI) SEER registry data for metropolitan
Detroit area residents. The MSCR is a member of the North American
Association of Central Cancer Registries (NAACCR) and is certiﬁed as meeting
all NAACCR standards for quality, completeness, timeliness, and unresolved
duplicate records. Data quality criteria for the MSCR (2004) are as follows: case
ascertainment (2 95%), passing edits (99.8%), cases identiﬁed from death
certiﬁcates only (1.7%), missing sex (0.04%), missing age, (0.0%), and missing
race (2.56%)(241).

2.2 Metropolitan Detroit Cancer Surveillance System

The Metropolitan Detroit Cancer Surveillance System (MDCSS) or the
Detroit SEER registry has conducted comprehensive cancer surveillance since
1969 for the Detroit metropolitan area which includes Wayne, Oakland and
Macomb counties (242). The registry became an ofﬁcial part of NCl’s SEER
Program on January 1, 1973. NCl’s SEER program includes 18 population-based
cancer registries across the US that comprise about 26 percent of the US
population (243). The population covered by MDCSS included approximately 3.9

million people in 1990, encompassing 42% of the state population (240). Since

68

1973, data on approximately 692,300 SEER reportable cases have been
collected by MDCSS (242). MDCSS data have been used extensively for cancer
research (244-246). MDCSS collects data on patient demographics, primary
tumor site, tumor morphology and stage at diagnosis, first course of treatment,
and follow-up for vital status (243). The registry also collects social security
number (SSN) for cancer patients and has a unique registry number which
enables MDCSS patients to be identified in the MDCH statewide cancer registry.
Data quality criteria for the MDCSS (2004) are as follows: case ascertainment
(QC-94%), passing edits (100%), cases identiﬁed from death certiﬁcates only
(1.0%), missing sex (0.0%), missing age, (0.0%), and missing race (2.24%)
(241).
2.3 Michigan Birth Certiﬁcate Data

The Michigan Department of Community Health (MDCH), Vital Records
and Data Development Section has maintained computerized records of all Ml
births since 1970 (MI birth ﬁles) (247). We elected to initiate this study in 1978
because maternal SSN, the main record linkage variable used in this study, is
missing for > 11% of mothers prior to 1978 (248). Ml birth data is used at the
county, state, and national level to monitor maternal and infant health trends and
to create vital statistics reports (249). Ml birth certificate forms are completed by
mothers after delivery and by the hospital staff, ﬁled and reviewed at the local
registrar ofﬁce, and forwarded to the registration unit at MDCH where the
certiﬁcates are checked for completeness, consistency, and appropriateness.

The certiﬁcates are then coded using standard instructions. Data items available

69

 

from the birth certiﬁcate have changed over the study years (1978-2004), with a
major revision in 1989. This revision included new methods for the collection of
pregnancy complications and the addition of data items including obstetrics
procedures, labor complications, maternal risk factor data (e.g., alcohol use,
weight gain during pregnancy, tobacco use), and the addition of maternal
identifying information (e.g., name, address, and date of birth) to the electronic
birth ﬁles.

MDCH Maternally-linked birth dataset. MDCH currently has available a
maternally-linked live birth dataset of MI birth ﬁles (1989-2004) created using a
multi-stage deterministic approach to link live delivery records for the same
women (250). This ﬁle has been used for research purposes (250, 251).

2.4 Measurement of Stgciy Variables

The data sources for the study variables were the MI Birth ﬁles and
MSCR. The statistical deﬁnitions of the variables used in the analyses for each
aim are defined in the statistical analysis section of the methods for Chapter 4
and 5.

2.4.1 Case-Control Matching Variables

Data from both the MI birth files and MSCR were used to match controls
to cases on maternal race at first live delivery and maternal year of birth. Prior to
1989, maternal DOB was not electronically available in the MI birth ﬁles, hence
age at ﬁrst delivery (years) and year of ﬁrst infant birth were used to estimate
maternal year of birth for all cases and controls for consistency. Controls were

required to have had their ﬁrst delivery prior to study reference date (i.e.,

70

matched case’s date of diagnosis). Maternal race was self-reported White or
Black race and information on ethnicity was not available prior to 1997 in the MI
birth ﬁles.
2.4.2 Outcome

The study outcome was primary breast cancer status (yes/no), obtained
from MSCR. Cases were women diagnosed with ﬁrst primary in-situ or invasive
breast (210) cancer (see eligibility criteria, section 1.2. 1). Controls were
women without a history of cancer diagnosis. Breast cancer histologic type was
available for all cases. Histologic types with ample sample sizes for analyses
included ductal (International Classiﬁcation of Diseases for Oncology (ICD-O-3)
code 8500, 8521, 8541, 8543) and Iobular (ICD-O-3 code 8520).
2.4.3 Exposures

The main exposures of interest (i.e., age at ﬁrst and last delivery, years
since last delivery, birthweight, fetal growth (birthweight-for-gestational age), and
infant gestational age), were assessed for each live birth delivery using the MI
birth ﬁles and were available for all study years 1978-2004. Again, please note
that the statistical deﬁnitions of the variables used for the analyses are deﬁned in
the analytic sections for manuscript 2 (chapter 4) and manuscript 3 (chapter 5).
Maternal age at each live delivery was continuous and in years. Years since
index delivery was defined using both data from the MI birth ﬁles and MSCR (i.e.,
reference mo/year — delivery mo/year). Gestational age (GA) is calculated by
MDCH registry staff as the interval between first day of the women’s last normal

menstrual period (LMP) and date of delivery (252). Gestational age in weeks

71

 

based on the clinical estimate was also used. Infant birthweight in grams was
available for each week. It is important to note that we obtained raw data on
gestational age and birthweight from the MI birth ﬁles and we conducted our own
cleaning of these variables using published methods (please see section 4.3
and chapter 5).
2.4.4 Covariates

Covariates measured using the MI birth ﬁles and not described above
include multiple births for each live delivery (deﬁned as the birth of two or more
children from a single term of pregnancy), maternal education at each live
delivery (to be deﬁned further in the statistical analysis sections of Chapter 4 &
5), number of prior children now living and number of prior children now dead at
each live delivery (used to determine number of live births or parity), prior
stillbirths/miscarriages/abortions (note: the quality of this variable is uncertain),
infant year and month of birth, and infant gender at each live delivery (male;
female). Variables will be considered as potential confounders, effect modiﬁers,
and mediators where appropriate.
2.4.5 Accuracy of Michigan Birth Certiﬁcate Data: Exposure and Covariates.

Information from MI birth ﬁles used in this study (described above in
Measurement of Study Variables) have been validated in several states in the
US, though we were unable to identify validation studies speciﬁc to Ml birth ﬁles
(253-266). Overall, most studies have found that demographics (e.g., maternal
race, ethnicity, and age) are accurately recorded (sensitivity and speciﬁcity

>93%) compared to both in-person interviews (254) and medical records (255,

72

256, 258, 260). Most investigators have also found that number of previous live
births (257, 260) and birthweight categorized (258-260, 263) are accurately
reported on the birth certiﬁcate compared to medical records. For example, a
large (n = 33,000) study in Ohio, reported high sensitivity and speciﬁcity for
birthweight 2 3000 9 vs. < 3000 g (99.4% and 98.8%, respectively) and prior
pregnancy (95.3% and 97.5%, respectively) comparing birth certiﬁcates to
medical records (260).

Though LMP is known be ﬂawed in its assessment, even from antenatal
medical records, it is widely used to estimate gestational age (267). The
agreement percentage for LMP (within one week or less) between birth certiﬁcate
data and medical records has been shown to be above 90% in two previous
validation studies (257, 258). Some studies, however, have reported agreement
< 80% (264, 265) and have found overestimation of gestational age by LMP to be
higher among infants with low birthweight (265). A recent study compared birth
certiﬁcate LMP (using California live birth records from 2002) to a population-
based database of prenatal records (considered the gold standard). This study
reported a high proportion of misclassiﬁcation for both preterm and post-term
births and also found that about 15% of preterm births were missed using LMP
from the birth record, although overall agreement in LMP (within 1 week)
between the two data sources was 89% (268). We used published approaches to
clean gestational age and birthweight which are discussed below (please see

section 4.3 and chapter 5 methods).

73

3. Dataset Creation Procedures
3.1 Summary of Dataset Creation Procedures Conducted at MDCH

This study involved the linkage of pre-existing registry data from the Ml
birth ﬁles, the MSCR, and MDCSS during the years 1978-2004. The dataset
creation procedures are shown in ﬁgure 3.2. In Step 1, we identiﬁed all female
breast cancer cases (in-situ and invasive) diagnosed .<. 50 years of age from the
MSCR (1985-2004). In Step 2, we linked eligible cases to their ﬁrst live birth at
age 16 or older in MI during 1978-2004 using simple deterministic linkage with
maternal SSN as the only linkage variable. In Step 3, we conducted several
linkage validation/quality control procedures for the birth-cancer linkages, which
are described in detail below. In Step 4, we identiﬁed eligible controls from the
birth ﬁles and individually matched four controls to each case (see Control
Sampling Strategy in section 1.2.3). In Step 5, we linked all controls ( n =
33,004) to the MDCSS using SSN alone to determine if controls were diagnosed
with cancer at 5 50 years of age during the study years, 1978-2004. We then
excluded controls with cancer (n = 95) and re-selected new controls.

The next goal was to identify all live births for study cases and controls in
the MI birth ﬁles. In step 6, we linked study participants to the existing MDCH Ml
maternally-linked birth dataset (1989-2004) to identify any additional births during
these study years. In step 7, we expanded this maternally-linked birth dataset
speciﬁc for cases and controls to locate births prior to 1989 in the MI Birth ﬁles
(1978-1988). We again used simple deterministic linkage with maternal SSN

alone for the initial linkage. Finally, in step 8, we conducted several linkage

74

validation and quality control procedures for the maternal birth-birth linkages,
described in detail below. The ﬁnal analytic ﬁle includes breast cancer data from
the MSCR for cases (1985-2004) and maternally-linked birth data for all study

participants with a ﬁrst birth after 1978 (1978-2004).

75

Figure 3.2. Dataset creation procedures conducted at MDCH

 

 

 

Step 1. Identified all female breast cancer cases (in-situ and invasive)
diagnosed <_ 50 years of age from the MSCR (1985-2004).

 

l

 

 

Step 2. Linked eligible cases to their first live birth at age 16 or older in MI
during 1978-2004 using simple deterministic linkage with maternal SSN.

 

l

 

 

Step 3. Linkage validation and quality control procedures for step 2.

 

l

 

 

Step 4. Identiﬁed eligible controls and individually matched to cases (1 :4).

 

l

 

 

Step 5. Sent matched controls to MDCSS to determine if controls were
diagnosed with cancer during 1978-2004, using SSN as the only linkage
variable. Excluded controls with cancer and re-selected new controls.

 

l

 

 

Step 6. Linked the study participants (cases and controls) to the existing MI
maternally-linked birth dataset (1989-2004).

 

l

 

 

Step 7. Expanded maternally-linked birth dataset for cases and controls with
births prior 1989 using maternal SSN.

 

l

 

 

Step 8. Linkage validation and quality control procedures for step 6 and 7.

 

 

iv

Final study ﬁle of MSCR data for cases (1985-
2004) and maternally-linked birth data for all
study participants 1978-2004)

 

 

 

 

 

76

 

 

 

 

 

 

 

 

 

3.2 Selection and Linkage Procedures: Breast Cancer Cases

We identiﬁed all ﬁrst primary female breast cancer cases (in-situ and
invasive) diagnosed S 50 years of age from the MSCR for the study years, 1985-
2004 (n = 33, 941) (see Figure 3.3). After excluding cases missing SSN in the
cancer ﬁle (n = 477, 1.4%), we linked cases to the MI birth ﬁles (1978-2004)
using simple deterministic linkage with maternal SSN as the only linkage variable
(see Step 2 in Figure 3.1). SSN is the only linkage variable used because other
identifying variables such as maternal date of birth (DOB) and name are not
available in MI birth files prior to 1989 (248). However, SSN is a unique identiﬁer
and has been shown to have high validity as a linkage variable both in previous
work and in the current study (our validation results are described below). For
example, Simon and colleagues conducted a study. using MDCSS data linked to
Ml birth data (1989-1994) and compared two linkage approaches, SSN alone
and deterministic linkage with seven linkage variables (maternal first name,
maternal last name, maternal maiden name, child’s last name, maternal alias,
maternal DOB, maternal SSN) and 14 steps (245). All linkages were manually
reviewed by hand and were considered valid if all linkage variables from the birth
data matched all linkage variables from the cancer data. They found that use of
SSN alone identiﬁed 98% of valid linkages that would have been found using the
seven identifiers (245).

After linkage with SSN alone, a total of 19,480 cases did not link to the MI
birth ﬁles, 1978-2004 (Figure 3.3). Reasons for non-linkage and ineligibility

included: 1) had a ﬁrst birth or all births outside of MI, 2) had all births prior to

77

1978, 3) were nulliparous, or 4) were missing SSN in the MI birth ﬁles (note: is
impossible to know how many cases were missing SSN in the birth ﬁle and not
linked because the only way to know this would be to link them to the birth ﬁle).
Eligible women may have not linked due to inaccuracies in SSN alone linkage.
We conducted several validation procedures to demonstrate the validity of SSN
alone linkage, which are described below.

Based on SSN alone, 13,984 (41.2% of cases diagnosed between 1985
and 2004), were linked to a birth in MI resident ﬁles (Figure 3.3). We then
excluded non-ﬁrst births between 1978-2004 (n = 5,322), women with month/year
of diagnosis 5 month/year of ﬁrst delivery (n = 173), women who were not Black
or White race based on the birth record (we based race on the birth record
because this was all that was available for controls) (n = 53), women not residing
in MI at ﬁrst birth (n = 7) and women < 16 years atﬁrst birth (n = 16, 0.1% of

8,429 otherwise eligible women). The case sample size prior to validation work

was 8,413 eligible cases.

78

Figure 3.3. Summary of selection of cases

 

 

 

33,941 Breast Cgcer Cg§§§

Female in-situ or invasive ﬁrst primary breast cancer cases at age 20-50

years at diagnosis among Black or White women residing in MI at time of
diagnosis, 1985—2004 (no duplicate cancer records)

 

1

 

 

Exclude missing SSN in Cancer Registry: n = 477

 

 

l

 

 

Exclude 19,480 assumed nulliparous, deliveries outside of MI, or
deliveries prior to 1978, or missing SSN in birth ﬁle

 

 

1

 

 

41.8% of 33,464

13,984 Eligible Cases linked to Ml Birth Files (1978-2004)

 

 

I

 

 

Exclude non-ﬁrst births: n = 5,322

 

 

V

 

 

8,662 cases linked to ﬁrst birth 1978-2004

 

 

 

 

Not residing in MI

at 1st birth (n =7) 4"

<age 16 at 1St
birth (n =16)

 

 

 

Additional
exclusions
n = 249

K>

 

 

 

 

 

l

 

Date at diagnosis
5 date of delivery
(n = 173)

Not BlackNVhite
race on birth
record (n= 53)

 

 

 

8,413 eligible cases

 

 

l

 

 

Validation of First Birth-Cancer Linkage:

Exclude n = 162

 

 

I

 

 

Validation/Data Cleaning of Birth-Birth Linkage

Exclude n = 64

 

 

l

Breast Cancer Cases n = 8,187

 

79

 

 

 

3.3 Case Linkage Validation Procedures and Results

We conducted several linkage validation and quality control procedures for
the ﬁrst birth-cancer linkages. These procedures included a) identifying invalid
linkages by checking women with inconsistencies in maternal age based on birth
compared to cancer data, b) manual veriﬁcation of a random sample of linkages,
c) checking multiple ﬁrst birth records linked to the same women, and d)
comparing simple deterministic linkage (SSN alone) to multi-stage deterministic
linkage (SSN, maternal name, infant’s last name, maternal DOB). We had two
primary goals in conducting the validation work. F irst, to demonstrate that SSN
alone is a valid linkage approach we asked, ‘Were cases truly matched with their
ﬁrst birth or were cases falsely matched to a wrong child?’ Second, we were
interested in completeness. How certain were we that all eligible women who had
a ﬁrst live birth in MI and were diagnosed with breast cancer at s 50 years in MI
are included in the study? Below is a description of each validation procedure
(summarized in Table 3.1).

a) Maternal age check. For this procedure we calculated the case age at
first birth using the age on the cancer record and the year of first delivery from
the birth record. This age was compared with maternal age on the birth record. If
the age did not match within 1 year, the linkage was manually veriﬁed. Manual
verification of the linkages involved using additional identifying variables to verify
the cancer to birth linkages by hand. The variables needed to validate the
linkages were available in electronic ﬁles only for births after 1988. For earlier

births, paper birth certiﬁcates were checked to verify linkages. A birth-cancer

80

linkage was considered valid if maternal SSN, ﬁrst name, and last name on the
cancer record matched maternal SSN, ﬁrst name, and last name (maiden,
child’s, or father’s) on the birth record. We identiﬁed 227 linkages where age at
ﬁrst birth based on cancer record did not match within 1 year to age at first birth
based on the birth record, 1978-2004. Of these, 154 were found to be invalid
after manual review, with a total percent invalid linkages among cases = 1.83%
(154/8413). Among, the 227 linkages ﬂagged by the maternal age check as
possible invalid linkages, 73 were valid and 154 were invalid for a percent invalid
among these ﬂagged linkages = 67%. This procedure demonstrated that about
70% of cases with unmatched age are truly invalid linkages. Further, if we
exclude all cases with unmatched age, we will exclude about 0.8% of truly valid
ﬁrst birth to cancer linkages (73/8,413).

b) Manual verification for random sample of cases. We selected a
random sample of cases linked to their ﬁrst birth (n = 299 for 1978-1988 births; n
= 292 for 1989-2004 births). Linkages were considered valid if SSN, age (+/- 1
year), ﬁrst name, and last name on the cancer record matched the SSN, age,
ﬁrst name, and last name (including maiden name, child’s last name, or father’s
last name) on the birth record. Linkages where last name only did not match out
of the 4 record linkage variables were also considered valid (women could have
married or remarried in-between birth and cancer diagnosis). For cases linked to
their ﬁrst birth in 1978-1988, the percent valid = 97.7%, for cases linked to their
ﬁrst birth in 1989-2004, the percent valid = 98.3%, and for all years combined,

the percent valid = 98.0% (Table 3.1). In addition, we determined that the number

81

ofmve
excluc
vahdlr
hghw
image

anag

hnked
InaCCr
bﬁhr
deten
above
0f 1hr
inacc

agec

birth
hnka
ﬁlm]

We {E

”to?

\

Vari

of invalid linkages identiﬁed by a maternal age check was 11 out of 12. When we
excluded the invalid linkages identified by the maternal age check the percent
valid increased to 99.8%. This demonstrates that linkage via SSN alone is a
highly valid linkage approach for this dataset and that checking for discrepancies
in age between linked records can identify close to all of the false positive
linkages that may occur with SSN alone linkage.

c) Multiple linked records check. We conducted a check for multiple
linked ﬁrst birth records for the same case to identify invalid linkages and/or
inaccurate data. We found 110 breast cancer cases linked to more than one ﬁrst
birth record (n = 224 records). The 224 linkages were manually veriﬁed to
determine if they were valid linkages (using the same approach described in the
above paragraph). We found the following: 182 linkages were valid first, second
or third births for the same case (the birth registry parity variables were
inaccurate), 39 linkages were invalid (34 would have been caught by maternal
age check), and 3 were duplicate birth certiﬁcate records.

d) Multi-stage deterministic and simple deterministic linkage for ﬁrst
births 1989-2004. We compared two linkage approaches for birth-cancer record
linkages (multi-stage deterministic and simple deterministic linkage with SSN
alone). In the multi-stage deterministic approach, additional linkage variables
were used to ensure no linkages were missed. We used seven linkage variables
(mother’s ﬁrst name, mother’s last name, maiden name, infant’s last name, alias,
mother’s date of birth, SSN) and 14 successive matching steps, the same

variables and steps used by Simon et al., 2004 (245). The comparison of the two

82

We y
aSSL

sam

link;
C89

Hex

linkage approaches was made only for births in 1989-2004, because prior to
1989 maternal personal identiﬁers such as name and DOB are not available in
the electronic Ml birth ﬁles and hence cannot be used for computerized linkage.
Using simple deterministic linkage (SSN alone) 2,414 cases were linked to their
ﬁrst births; using multi-stage deterministic linkage, 2,444 cases were linked. The
percentage of possible cases missed using the SSN alone approach compared
to the deterministic approach was 1.2%. We also checked the reverse and
conﬁrmed that all cases found by simple deterministic linkage were also found by
multi-stage deterministic linkage. This procedure shows that linkage using only
SSN resulted in fairly complete linkage. Because we do not have computerized
information on maternal name and DOB prior to 1989, we have to assume that
we would ﬁnd a similar low percentage of missed cases during 1978-1988. This
assumption is reasonable given the similar results obtained in our random
sample manual veriﬁcations of SSN linkage for cases with ﬁrst births prior to
1989 and after 1989. See Table 3.1 for the results.

In summary, we found a total of 162 invalid/duplicate birth to cancer
linkages which were excluded from the final case ﬁle (see Table 3.1). The total
case sample size at this stage (prior to additional validation procedures after the

next phase of the study described below) was 8,251 eligible cases.

83

 

Percent invalid among cases with Percent invalid cases women with
unmatched age: unmatched age:
9311 51 = 61.6% 61/76 = 80.3%

 

Percent invalid among cases with unmatched age for all birth years:
‘ 154/227 = 67.8%

 

 

Percent valid: 292/299 = 97.7 Percent valid: 287/292 = 98.3

Number of invalids that would have been Number of invalids that would have
b aoe check (n=6/7 ) been caught by age check (n=5/5 )

Percent valid for all birth years: 579/591 = 98%

Results from manual veriﬁcation:
182 linkages were valid ﬁrst, second or third births
39 linkages were invalid (5 not caught by age check)
3 duplicate birth certiﬁcate records were found

30 ‘new’ cases found using multi-stage deterministic
2444 total number of cases found using multi-stage deterministic

Percentage of cases missed using SSN alone = 1.2
(Have to assume same quality for 1989 prior)

 

Total Invalid Linkages/Duplicate Records : 162*

 

 

 

*These invalid linkages were excluded from the ﬁnal case ﬁle.

84

four contrr
1981) and
1.2.3). We
dagnosec
variable. I
2 controls
inaccurac
duplicate
left 95 cor
2004 (77
Cancers.

these 95

Controls f

and Com

3.4 Control Selection and Matching Procedures

We identiﬁed eligible controls from the birth ﬁles and individually matched
four controls to each case (n = 33,004) on maternal year of birth +/- 1 (1935-
1981) and race (White; Black) (see Control Sampling Strategy, in section
1.2.3). We then had all controls matched to MDCSS to determine if controls were
diagnosed with cancer at s 50 years of age. Using SSN alone as the linkage
variable, MDCSS registry staff identiﬁed 179 linkages. Out of these 179 linkages,
2 controls were linked to males cancer cases (we are assuming this is due to
inaccuracies in SSN reported in the birth file and/or the MDCSS ﬁle), there was 1
duplicate record, and 81 were linked to cancers occurring in 2005 or later. This
left 95 controls diagnosed with cancer in MDCSS during the study years, 1978-
2004 (77 diagnosed prior to 1985; 18 diagnosed in 1985 or later). Of these
cancers, 12 were breast cancer (4 were diagnosed after 1984). We excluded
these 95 controls (95l33,004 = 0.29% of study controls) and re-selected new
controls from the eligible control file by hand.
3.5 Creation of Maternally-linked Birth Dataset for Cases and Controls

The next step was to create the maternally-linked birth dataset for cases
and controls. The goal was to identify all children born in MI to study cases and
controls. We had 33,004 cont_rgl§ and 8,251 cases prior to linkage of additional
births (beyond ﬁrst births) for study women.
3.5.1 Birth-Birth Linkage Procedures

The goal of this step was to identify and link all live births for study cases

and controls to create a ﬁnal study data file of maternally-linked birth records (for

85

live b
and i
for n
to lir
data
for S
vain

3.5

link

‘6

women with more than one live birth). The procedures used for linkage of births
were similar to those used for linkage of cases to their ﬁrst live birth. We had an
additional resource, however, because MDCH currently has a maternally-linked
live birth dataset (1989-2004). This dataset has been used for research purposes
and was created using a multi-stage deterministic approach to link birth records
for multiple deliveries for the same women (250). The ﬁrst step for our study was
to link our cases and controls to this existing database. We then linked the
database to the MI birth ﬁles for 1978-1988 using SSN alone to identify all births
for study participants during these years. The following sections describe the
validation and quality control work for birth-birth linkages for cases and controls.
3.5.2 Birth-Birth Validation Procedures and Results

Prior to any validation or quality control work, we had 84,747 records
linked to 41,255 cases and controls. Because of the large number of linkages,
manual veriﬁcation of all birth-birth linkages or even a random sub-sample of
birth-birth linkages was not feasible given study time constraints. Instead, we
used procedures to identify possible invalid birth-birth linkages for study cases
and controls through several “checks”. The checks included looking for
discrepancies between infant birth year, maternal age, and parity at each birth for
women with 1 or more births (i.e., we compared birth 2 to birth 1, birth 3 to birth 2
and 1, and so on). For birth certiﬁcate data, parity is determined by combining
‘nowlive' and ‘nowdead’, which indicate the number of children now living and the

number of children born alive but now dead at the current live birth.

86

We identiﬁed 1,668 women (4% of all participants) with birth-birth linkages
where birth year, maternal age, or parity did not match or were implausible
across birth histories (ﬁgure 3.4). Of these 1,668, we generated computer reports
to validate all birth-birth linkages for women with ﬁrst births in 1989 or later (n =
461) because we could use computerized records for these pregnancies. We
used the same approach to validate birth-birth linkages as was used to validate
the birth-cancer linkages (see section 3.3). Similar to the criteria for a valid birth-
cancer linkage, a birth-birth linkage was considered valid if maternal SSN, ﬁrst
name, and last name on the first birth record matched maternal SSN, ﬁrst name,
and last name on the subsequent birth record.

For these 461 women with all their births after 1989, we determined why
there were inconsistencies in birth year, age, and parity. The reasons for
inconsistencies for women with validated birth-birth linkages included (Figure
3.4): .

1) Multiple births (the checking mechanism we used identiﬁed these

women as possible problems because of the way ‘nowlive’ and ‘nowdead’

and birth year are recorded for multiple births);

2) Inaccurate parity data;

3) Selected as a control twice (this was due to inaccurate ‘nowlive’ or

’nowdead’ variables (i.e., the same women had more than one “first” birth

due to inaccurate data and we happened to select two of her births as

controls»; and

87

As d!

parti.

parti

to es

ECO

4) Now ineligible (this was due to linkage to earlier births prior to what
we had thought was the participant’s ﬁrst birth). These women are now
ineligible because they were linked to earlier births that were prior to 1978

or births that occurred out of MI.

5) Unknown in parity variables

6) Maternal age inconsistency for subsequent births.

We also found a few women with invalid birth-birth linkages (Figure 3.4). f."
As described later, we excluded the invalid birth records, but not the study
participant.

While checking the 461 women with ﬁrst births after 1989, we found 155
participants who we had thought had a ﬁrst birth in 1989 or later, but were linked
to earlier births before 1989 when we created the maternally linked birth ﬁle
(Figure 3.4). This is the same issue that led to the women that were selected as
a control twice (number 3 above) and “now ineligible” women above (number 4)
above. What we had thought was the ﬁrst birth, was actually not the ﬁrst birth
because the parity data was inaccurate. For these 155 women, we validated their
birth-birth linkages with the birth certiﬁcate paper birth files using the same
manual validation procedure described above and also for birth to cancer
linkages (see section 3.3). Among the 155 women, we identiﬁed 48 women with
1 or more invalid linked births, and 35 of these women had an age that didn’t
match across birth-records (73%) (Figure 3.4). Among the 106 women with valid
birth histories, only 1 women had age that did not match across birth records

(0.94%) (Figure 3.4).

88

lOOwh
Of thes
inconsi
Whhbu
For all 1
Year ar
hand.\
among
numbe
births t
incons'

The l'lL

Given the birth data is not computerized prior to 1989, we next identiﬁed a
random sample (n = 100) of the 1207 women who had their ﬁrst birth prior to
1989 and where discrepancies were identiﬁed in maternal age, parity, and birth
year order across their pregnancies (ﬁgure 3.4). We validated the birth-birth
linkages for these 100 women using the same approach used above and also for
birth to cancer linkages (see section 3.3). We identiﬁed 22 women (22%) of the
100 who had one or more invalid births linked to their “ﬁrst” birth (or index birth).
Of these 22 women, 82% would have been identiﬁed by checking for
inconsistencies in maternal age across births. Of the 88 women with all valid
birth-birth linkages, 4 or about 4.5% had inaccurate maternal ages (Figure 3.4).
For all the women with validated birth histories, we ordered their births by birth
year and the parity data (‘nowlive’ and ‘nowdead’ variables) were corrected by
hand. We also hand ﬁxed parity data for women that we did not manually validate
among the original problematic sample prior to 1989 (1207- 100 = 1107). This
number was further reduced with exclusions due to new ineligibility (linked earlier
births that were prior to 1978 or out of state births) and women with
inconsistencies in age across births (described below and in shown ﬁgure 3.5).

The number left from the original 1207 that were hand ﬁxed was 762.

89

Figure 3.4. Summary of validation procedures at
MDCH for the maternally linked birth database

n = 84, 747 births to 41, 255 women

 

 

Identiﬁed 1668 women (4%) with one or more possible invalid birth-birth linkages by
comparing maternal age at each birth, parity at each birth, and birth year order. This
included 5,571 births (6.6%) of total births.

/ \

 

 

 

 

 

“First” Births before 1989* “First” Births after 1989*
(Have to check paper ﬁles to (Can validate on computer)
validate) n= 461 women (1.1%)

 

 

 

 

n = 1,207 women (2.9%)
I

 

 

1

What are the issues after manual
veriﬁcation?

 

Valid birth-birth linkages (n=331)
- Multiple births (n=19)
. Inaccurate parity data
(n=284)
- Selected as a control twice
(n=8)
. Linked to new birth --now
ineligible (n=7)
- age wrong at 2 1 births
(n=8)
- Unknown value in parity
(n=5)
Invalid linkages (n=5)

Unable to validate (n=155)
New earlier births prior to 1989 so
have to check paper ﬁles to validate

l

 

 

 

 

 

 

 

 

 

Random sample to manually Results for validation of births
verify with the paper ﬁles for 155 women with “new” births
because cannot check all 1200 by prior 1989.
hand (n=100).
Mild births: n=22 1gv_alid birthts): n=48

Wrong age: 18 (81.8%) Number with wrong age: 35 (73%)
All births valid: n=88 All births valid: n=106

Wrong age: 4 (4.5%) Number with wrong age: 1 (0.94%)

 

 

 

 

 

 

 

 

 

*Due to inaccuracies in parity data, after creation of the maternally-linked birth database,
original ﬁrst births for some study participants were found to be non-ﬁrst births, with linkage to
earlier “new” births. Some of these new earlier births occurred prior to 1978 or out of state,
which made the study participant ineligible.

90

created
The LE
eligible
these :
Of link;
we de
tudy
invali
CODCl
births
20-2
iSee
birth
hlS‘ir

DOS

Applying the results from the validation procedures described above, we
created the ﬁnal study dataset. These procedures are summarized in Figure 3.5.
The ﬁrst step was to delete birth records determined to be invalid and linked to
eligible study participants (n = 99 records). This ensured that we did not delete
these study participants in the next step (if they had an age discrepancy because
of linkage to a birth record that was not one of their births). In the second step,
we deleted women with discrepancies in reported age across births (0.6% of
study participants), with the goal of removing the large majority of women with
invalid births linked to them. By removing these study participants, several
concerns arise. These include: (1) we will not identify all women with invalid
births linked to them using this exclusion method (estimated to be missing about
20-25% among the 4% of the study population with possible problematic linkages
(See Figure 3.4)); (2) we are excluding the entire participant, not just the invalid
birth record; and (3) some women with inaccurate ages may have valid birth
histories (~ 1-4.5%, see ﬁgure 3.4) among the 4% of the study population with
possible problematic linkages. The third step was to delete women who were
identiﬁed as ineligible with creation of the maternally-linked birth database and
linkage of “new” earlier births to cases and controls. lneligibility was due to: 1)
ﬁrst birth at < age 16 years; 2) first live birth prior to 1978; or 3) earlier births
outside of MI. In gepﬁ we deleted the controls matched to any case that was
deleted in previous steps. In stpLS we identiﬁed women selected as a control
more than once and marked the matched case for control re-selection. In M,

we re-selected controls for 325 cases where 1 or more controls were deleted in

91

above steps or had duplicate control selection (327 controls (2 cases needed 2
controls». Finally, we linked additional births for the new controls. The ﬁnal

sample size was 8,187 cases and 32,739 controls.

92

Figure 3.5. Exclusions based on validation results
for the maternally linked birth data

 

41 , 255 women (84, 747 births)

 

Step 1. Delete invalid birth records linked to eligible study
participants (n = 99 records)

 

 

 

T

 

Step 2. Delete women with un-matched age across births
n = 237 (38 cases; 199 controls)

These women were all among the original 1668 women

identified with problematic records.

 

 

 

l

 

Step 3. Delete women who are “new” ineligibles*
n = 148 (26 cases; 122 controls)

 

 

 

I

 

Step 4. Delete controls matched to cases deleted due to
ineligibility (n = 251)

 

 

 

l

 

Step 5. Identiﬁed women selected as controls twice and
marked cases for control re-selection (n = 17)

 

 

 

l

 

Step 6. Selection of new controls for 325 cases with < 3
controls after above exclusions (n = 327 controls)

 

 

 

l

 

Step 7. Linkage of additional births for new controls

 

 

 

(n = 559 records)

 

Sample Size Prior to Data Cleaning at MSU:
8,187 cases; 32,739 controls

Note: for nine cases- only have three controls

 

 

 

 

 

 

*Reasons for new ineligibility include earlier births that were prior to age 16,
prior to 1978 or births not in MI.

93

3.6. Data Cleaning-Management at MSU

Figure 3.6 displays additional exclusions made for either women or births
during the data management phase at MSU with the final study ﬁles from MDCH.
It is important to note that data management work, including additional study
population exclusions made, is mostly described as relevant in chapter 4 and

chapter 5.

__Figure 3.6. Exclusions based on data cleaning work at MSU
8,187 cases; 32,379 controls (83,380 births)

 

 

Checked the “new” controls selected at MDCH (n = 327)
for linkage issues. Deleted 2 controls and their 4 birth
records due to ineligibility.

t

Found 11 births found to be ineligible while cleaning date
of last live delivery variable.

7

Excluded 3 cases and 12 controls due to missing month
of diagnosis.

1

Excluded 1783 birth records that were due to
pregnancies that occurred on or after study reference
date.

f

81,545 records (8184 cases; 32,725 controls)

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

94

4. Methodologic Issues

4.1 ldenti in the Stud base and Im lications for Bias

We used control selection procedures that ensured that controls were
selected from the same population that gave rise to the cases. Selection of cases
in this study was limited by the year of ﬁrst birth (1978-2004) and year of ﬁrst
cancer diagnosis (1985-2004). Because of these two data restrictions, we
selected a group of cases where the age at ﬁrst birth range is restricted, in
particular for older cases, because cases with older ages at diagnosis (e.g., 40
years) have a larger plausible age range at which a first birth could occur as
compared to cases diagnosed at younger ages (e.g., 30 years). Thus, we know
that our source population is not representative of all women in MI with a first live
birth at age 16-50 during the years 1978-2004 who could have been diagnosed
with breast cancer at age 50 years or less (assuming women did not move out of
state or die).

To understand how the age at ﬁrst birth restriction impacts the age
distribution of our study population, we created excel spreadsheets that list the
age at first birth ranges by case age and year of cancer diagnosis. These
spreadsheets demonstrate how as a case’s age increases, the likelihood of
capturing the entire true range of age at ﬁrst birth for a case population is
reduced (and hence it is reduced for controls as well). For example (see Figure
3.7), at the most extreme, for a 50 year old case diagnosed in 1985 in our study
population, her possible age at ﬁrst birth range is limited to 43-50 years because

she had to have a ﬁrst birth in 1978 or later. This implies that our sample will not

95

include any 50 year old cases diagnosed in 1985 who had a ﬁrst birth at age 5 42
years. To address this issue we will conduct analyses stratiﬁed by attained age
and examine age at first and last birth by maternal birth cohort and years of study

(please see Chapter 4 and Chapter 5 discussion sections for more details).

96

 

 

 

aa av av nv av av vv av Nv vv av aa aa ma aa aa va aa Na wa aa am am RN am am vN
aa av av 5v av av vv av Nv _.v av aa aa a aa aa va aa Na Fa aa am am RN am am
aa av av wv av av vv av Nv _.v av aa aa a aa aa va aa Na Fa aa am am RN am
aa av av hv av av vv av Nv Fv av aa aa a aa aa va aa Na _.a aa am am RN

aa av av nv av av vv av «v v av aa aa a aa aa va aa Na _.a aa am am

aa av av mv av av vv av Nv _.v av aa aa ma aa aa va aa Na Pa aa am

aa av av nv av av vv av Nv _.v av aa aa ma aa aa va aa Na Fa aa

aa av av mv av av vv av Nv _.v av aa aa a aa aa va aa Na Fa

aa av av nv av av vv av Nv _.v av aa aa a aa aa va aa Na

aa av av nv av av vv av Nv _.v av aa aa a aa aa va aa

aa av av mv av av vv av «v v av aa aa a aa aa va

aa av av nv av av vv av Nv vv av aa aa a aa aa

aa av av nv av av vv av Nv Pv av aa aa a aa

aa av av nv av av vv av Nv _.v av aa aa Ra

aa av av nv av av vv av Nv Fv av aa aa

aa av av 5v av av vv av Nv _.v av aa

aa av av nv av av vv av Nv Pv av

aa av av Nv av av vv av Nv _.v

aa av av nv av av vv av Nv

aa av av mv av av vv av

va aa No a aa aa aa a aa aa va aa Na Fa aa aa aa a aa aa va aa Na Fa aa an a\.

38-32 28» 2E5 >38 .2 5.5 sat a 8<

voom 5 ca
aoom E on
Noam E on
room 5 ca
ooom E om
mam? 5 am
am? 5 ca
33 5 ca
aaar 5 am
ma? 5 ca
vaaw 5 ca
aaaw 5 am
Name 5 ca
53 5 ca
came 5 ca
aaae 5 ca
aaav E on
name 5 ca
aaar 5 ca
aaar E ca
.6 an oa<

 

 

m_wocam_o .5 56> 3 88 20 66> oa m .8 69.9 5.5 “we: am ma< Ha 6.59".

97

4.2 Incomplete Follow-up: Cases and Controls
4.2.1 Movement Out of State

Cases may have had a ﬁrst live birth in MI, moved out of state and had
additional birth(s), moved back into the state, and then be diagnosed with breast
cancer in MI before age 50. Controls may have also had a ﬁrst live birth in MI and
then moved out of state and had additional births. The original proposed solution
was to obtain address histories for a sub-sample of randomly selected cases and
controls. We would have used this data to obtain an estimate of the number of
women who have a ﬁrst live birth in MI and then moved out of state. We would
have also used this data to estimate the likelihood of selected controls moving
out of state during the study years (1978-2004) and before the reference age by
key factors (i.e., SES, infant birthweight ) to determine the possible effect of bias
on the study effects estimates.

In lieu of data on study participants who have moved out of MI, we have
data from a US census report for characteristics of persons who moved out of
any state during the period 1990-1995. Interstate moving rates were higher for
White non-Hispanics (compared to Black non-Hispanics, Asian and Pacific
Islanders, and Hispanics) (269). Moving rates increased as education level
increased, the rates were the highest for those with a professional or graduate
degree. Moving rates were higher for married individuals and those with higher
incomes. The age groups with the highest moving rates were 25-29 years and
those older than age 65 years (269). This same report examined characteristics

of outmigration from the Midwest, but did not look at characteristics of persons

98

who moved out of speciﬁc states. Please see chapter 4 and 5 discussion
sections for the implications of this potential bias in the context of the study
results.

4.2.2 Time Lag between Exposure and MSCR creation (1978-1984)

Due to the restrictions of the data ﬁles, we have a time lag between
exposure and the start of statewide cancer registry surveillance (1978-1984). To
address this issue, we linked selected controls to MDCSS to identify any cancers
diagnosed before age 51 during 1978-1984. We found 77 controls diagnosed
with cancer (8 were breast cancers). We excluded these women and selected
new controls. This data is reassuring, with only 8 women from the study
population diagnosed with breast cancer during 1978-1984 in MDCSS, which
accounted for 44% of the population in MI in 1980 (240).

4.2.3 Controls that Died before Reference Date

It is possible that selected controls may have died before the study
reference date. One proposed solution to address this was to link selected
controls to the National Death Index and exclude those who died before their
matched case was diagnosed with breast cancer, but this would be very costly
($5.46 per control). It is reassuring however, that the probability of death in this
young population is very low (270).

4.3 Use of Birth Certiﬁcate Data for Exposure Measurement: Birthweight

and gestational age

Two published approaches for cleaning gestational age, Alexander et al.,1996

(271) and Zhang and Bowes, 1995 (209), are widely used in perinatal

99

epidemiology research. The method proposed by Zhang and Bowes involves
using birthweight-gestational age combinations for weeks 25-35 to identify births
for replacement with the clinical estimate of gestational age and if the clinical
estimate is missing or is the same poor quality as LMP-based gestational age
then the births will be excluded. The Alexander et al., method provides cut points
for implausible birthweights for gestational weeks 20-38. A recent study
conducted by Parker and colleagues compared the use of these two approaches
and concluded that there is no optimal approach and with both approaches,
higher birthweights are more likely to be excluded, higher risk women’s births are
more likely to be excluded, and the edits modify lower gestational ages more
than higher gestational ages (272). We decided to use the Zhang and Bowes
method given this method excludes fewer births and we have the clinical
estimate of gestational age available for all study years in MI, 1978-2004.
Additional details on the exclusions made are described in the methods section
of chapter 5.
4.4 Exclusions and Selection Bias: SSN, Age, and Race

We compared characteristics for women diagnosed with cancer in MSCR
at age 50 or less during 1978-2004, by missing SSN status (yes (n=200,380; no
(n=15,437)). Compared to cases with SSN, those missing SSN were younger,
diagnosed with cancer at earlier years, had a higher percentage of in-situ
cancers or unstaged/unknown cancers at diagnosis, and a greater percentage

were missing race (Table 3.2).

100

1978
3.3).
in Mr
ﬁrst |

mISs

Table 3.2. Distributions of select characteristics by
missing SSN status for women diagnosed with cancer at

 

 

 

 

 

 

 

 

 

 

s 50 years in the Michigar' Statewide Cancer Registry
No Yes
n (%) n (%)
Age at Diagnosis (years)
20—24 16,271 (8.0) 2,722 (17.6)
25-29 23,543 (11.8) 3,488 (22.6)
30-34 26,676 (13.3) 2,860 (18.5)
35-39 30,369 (15.2) 2,158 (14.0)
40-44 41,061 (20.5) 1,960 (12.7)
45—50 62,460 (31.2) 2,249 (14.6)
Year of Diagnosis
1985—1989 31,487 (15.7) 3,726 (24.1)
1990-1994 42,100 (21.0) 3,561 (23.1)
1995-1999 55,880 (27.9) 4,000 (25.9)
2000-2004 70,913 (35.4) 4,150 (26.9)
Stage at Diagnosis
In situ 72,392 (36.1) 9,819 (63.6)
Localized 57,136 (28.5) 1,993 (12.9)
Regional 26,572 (13.3) 516 (3.3)
Distant metastases 16,645 (8.3) 350 (2.3)
Unknown/unstaged 27,635 (13.8) 2,759 $9)
Race
Black 164,626 (82.2) 9,189 (59.5)
White 25,500 (12.7) 988 (6.4)
Other 2,448 (1 .2) 226 (1.5)
Missing 7,806 (3.9) 5,034 (32.6)

 

We compared select available characteristics of ﬁrst births in MI during
1978-2004 by missing SSN status (yes (n=1,460,217); no (n=38,119)) (Table
3.3). Compared to women with SSN, women missing SSN at their ﬁrst live birth
in MI were younger and had their ﬁrst birth in earlier years. Very few (~ 0.04%)
ﬁrst live birth records were missing maternal age and 0.4% of records were

missing race.

101

Table 3.3. Distributions of select characteristics by missing SSN
status for women with a ﬁrst birth in the Michigan during

 

 

 

 

 

 

 

1978-2004
No Yes
n (%) n (%)
Age at First Live Birth (years)
<16 24,430 (1.7) 4,813 (12.7)
1619 318,473 (21.8) 14,277 (37.8)
20-24 477,751 (32.7) 9,559 (25.3)
25-29 397,428 (27.2) 6,284 (16.6)
30-34 184,170 (12.6) 2,250 (6.0)
35-39 50,204 (3.4) 550 (1.5)
2 40 7,525 (0.52) 72 (0.2)
Missing Age at First Birth
Yes 236 (0.02) 314 (0.8)
No 1,459,981 (99.9) 37,805 (99.2)
Year of First Live Birth
1978-1982 279,512 (19.1) 17,363 (45.5)
1983-1987 266,889 (18.3) 11,293 (29.6)
1988-1992 290,420 (19.9) 2,054 (5.4)
1993-1997 269,358 (18.5) 405 (1.1)
1998-2004 354,038 (24.3) 7,004 (18.3)
Race
Black 1,183,612 (81.3) 27,121 (71.4)
White 239,111 (16.4) 8,702 (22.9)
Other 32,394 (2.2) 2,147 (5.7)
Missing Race
Yes 5,100 (0.4) 149 (0.4)
No 1,455,117 (999 37,970 (99.6)_

 

 

 

 

 

102

CHA
brea
of p:

1.Al:

weir-s
risk 0'
young
hetp r

etok:

asso
oanc
raﬁo

and

risk

CHAPTER 4: Pregnancy-related factors and risk of
breast cancer by histologic type, a registry-based study
of parous black and white younger women

1. Abstract

Pregnancy-related factors such as age at ﬁrst birth and parity have been
well-studied in relation to risk of breast cancer, yet few studies have examined
risk of breast cancer by histologic type for these factors, in particular among
younger women. The identiﬁcation of differences in risk by histologic type can
help elucidate biological mechanisms for the role of pregnancy in breast cancer
etiology. We conducted a population-based case-control study among parous
Michigan (MI) women aged 5 50 years with singleton births using linked Ml
Cancer Registry (1985-2004) with Ml Live Birth records (1978-2004). Cases
(n=7,837) were matched 4:1 on maternal birth year and race to controls
(n=30,159). We used conditional logistic regression models to examine
associations between age at ﬁrst and last birth, number of live births, and breast
cancer risk overall and by histologic type (ductal, Iobular). Later age at ﬁrst (odd
ratio (OR) per 5 year increase = 1.16, 95% Conﬁdence Interval (CI): 1.13-1.20)
and last birth (OR per 5 year increase = 1.11, 95% CI: 1.04-1.18), and multiparity
vs. uniparity (OR for 2 births = 1.36, 95% CI: 1.28-1.44, 3 births = 1.29, 95% CI:
1.19-1.40, 4 births = 1.25; 95% CI: 1.11-1.42), were independently associated
with increased breast cancer risk, with ORs of similar magnitude by histologic
type. Results were similar by maternal education (5 high school, > high school).
Some differences were observed by race (White, Black), including an increased

risk of Iobular tumors for age at last birth 2 30 years (vs. < 30 years) among

103

White women only (OR = 1.70, 95% CI: 1.21-2.40). Pregnancy-related factors
were not signiﬁcantly associated with risk of ductal or Iobular tumors among
Black women. Our results suggest that among parous women s 50 years of age,
later age at ﬁrst and last birth and multiparity are associated with increased risk
of both ductal and Iobular breast cancer with associations of similar magnitude by
histologic type.

2. Introduction

Later age at ﬁrst birth is a well-established breast cancer risk factor (35,
42, 43). Later age at last birth has also been shown to increase breast cancer
risk independently of age at ﬁrst birth, though there have been fewer studies and
ﬁndings have been less consistent (48, 54, 55, 57, 59, 60, 273). Multiparity has
also been well-studied, but ﬁndings have been inconsistent across populations
and by age at diagnosis and/or menopausal status (24, 25, 35, 43, 124, 274-
276). Few studies, however, have examined these factors and breast cancer risk
by histologic type among younger women (112, 277-281). Understanding how
pregnancy-related risk factor proﬁles differ by risk of histologic breast cancer
subtypes could generate biological hypotheses for the inﬂuence of pregnancy on
mammary carcinogenesis (280, 282).

Several studies have examined associations between pregnancy-related
factors and breast cancer risk for different histologic types among
older/postmenopausal women only or mixed age populations (112, 277-279, 281-
286). Research in young women is needed given that the distribution of tumor

characteristics and the inﬂuence of reproductive risk factors on breast cancer risk

104

have both been shown to vary by age/menopausal status (3-6, 124, 287), and
few studies have examined risk in younger women (288). We found only one
study that reported ﬁndings by tumor histologic type stratiﬁed by menopausal
status. Li et al., reported ﬁndings from a population-based case-control study of
risk for both ductal and Iobular breast cancer and age at ﬁrst birth among
premenopausal White and Black women (288). They reported some evidence
that later age at ﬁrst birth was more strongly related to increased risk of Iobular
as compared to ductal tumors among White premenopausal women only
(compared to Black women).

Using data from a large, population-based case-control study conducted
using state-wide birth and cancer registry data in Michigan (MI), we ﬁrst
examined the associations of age at ﬁrst and last birth, and number of live births
with breast cancer risk overall. We next examined these associations by
histologic breast cancer subtype, focusing on the two most common subtypes
(ductal and Iobular), which had adequate samples sizes. We further examined
the associations for pregnancy-related factors and breast cancer overall and for
ductal and Iobular tumors by race and maternal education. We hypothesized that
later age at ﬁrst and last birth and multiparity would be more strongly associated
with increased risk for Iobular breast tumors as compared to ductal breast
tumors, given the development of lobules during pregnancy and that Iobular
tumors have been hypothesized to be more strongly associated with hon'nonal-

related factors in some studies (112).

105

3. Methods

3.1 Study Design

We conducted a population-based, case-control study among parous Ml
women aged 5 50 years who had a live birth in MI during 1978-2004 at age 16-
50. This study was registry-based, using linked MI birth ﬁles (1978-2004) and the
MSCR (1985-2004). We created a complete live birth history for cases and
controls through linkage of all of a woman’s Ml births. The study protocol was
approved by the institutional review board at Michigan State University and the
Michigan Department of Community Health (MDCH).

3.2 Study Population

Cases. Eligible breast cancer cases were identiﬁed from the MSCR
(1985-2004) and linked to their ﬁrst live birth in the MI birth ﬁles (1978-2004).
Eligibility criteria included: (1) diagnosed with in situ or invasive ﬁrst primary
breast cancer between 1985 and 2004 in the MSCR; (2) age 20-50 years at
breast cancer diagnosis; (3) no previous diagnosis of any cancer with the
exception of basal and squamous cell carcinoma; (4) White or Black race based
on MI birth ﬁle; (5) ﬁrst live birth in MI at age 16 years or older during 1978-2004;
and (6) residing in MI at time of diagnosis. The study reference date for cases
was the date of diagnosis.

Controls. Eligible controls were selected from the MI birth ﬁles (after
linkage of the birth ﬁles to MSCR). Eligibility criteria included: (1) no history of
cancer in MI between 1985-2004 or in the Detroit Surveillance, Epidemiology, &

End Results Registry (SEER) for the years 1978-1984 (area covered by Detroit

106

SEER accounted for 43.6% of Ml’s population in 1980; 42.1% in 1990) (240); (2)
age 20-50 years at study reference date (individually-matched case’s diagnosis
date); (3) White or Black race based on MI birth ﬁle; and (4) ﬁrst live birth in MI at
age 16 years or older during 1978-2004. @ntrol Sampling Strat_egy. We
individually matched four controls to each eligible case on maternal year of birth
(+l- 1 year) and maternal race (White; Black). Controls were required to have
their ﬁrst live birth in MI prior to the study reference date.

3.3 Data Sources

MSCR. The MSCR, which was fully established in 1985, is a member of
the North American Association of Central Cancer Registries (NAACCR) and is
certiﬁed as meeting all NAACCR standards for quality, completeness, timeliness,
and unresolved duplicate records. Data quality criteria (2004) are as follows:
case ascertainment (2 95%), passing edits (99.8%), cases identiﬁed from death
certiﬁcates only (1.7%), missing sex (0.04%), missing age, (0.04%), and missing
race (2. 6%) (241). Available data items utilized for the present study included
stage, behavior, histologic type, age and date at diagnosis.

Ml birth ﬁles. MDCH-Vital Records and Data Development Section has
maintained computerized records of all Ml births since 1970. We elected to
initiate this study in 1978 because maternal SSN, the main record linkage
variable used in this study, is missing for 2 10% of mothers prior to 1978 and <
10% after 1978. Data items available from the MI birth certiﬁcate have changed
over the study years (1978-2004), with a major revision in 1989. Data items

utilized for the present study, available for each live delivery, included maternal

107

age,nr
(deﬁne
mmew

chtdre

a.)
h
n

honrt
sthy
tehral

the hi

bhrh
a Bar
Iinka

tier,

age, month/year of delivery, clinical estimate of gestational age, multiple births
(deﬁned as the birth of two or more children from a single term of pregnancy),
maternal education, number of prior children now living, and number of prior
children now dead.

3.4 Dataset Creation Procedures

We conducted several steps to create the ﬁnal analytic ﬁle of cancer data
from the MSCR for cases (1985-2004) and maternally-linked birth data for all
study participants (1978-2004). Cancer to birth linkages. First, we identiﬁed all
female breast cancer cases diagnosed s 50 years of age during 1985-2004 from
the MSCR (n=33,941). Second, we linked eligible cases to their ﬁrst live birth at
age 16 or older in MI during 1978-2004 using simple deterministic linkage with
maternal social security number (SSN) as the only linkage variable (n=8,662;
excludes 477 cases missing SSN in the MSCR, and 19,480 cases assumed
nulliparous, ﬁrst delivery outside of MI, ﬁrst delivery prior to 1978, or missing SSN
in birth ﬁle). Additional reasons for exclusions after linkage included: not residing
in MI at 1st birth (n=7), < age 16 years at 1st birth (n=16), date at diagnosis 5
date of delivery (n=173), and non-Black or White race (n=53), leaving 8,413
eligible cases.

Next, we conducted linkage validation/quality control procedures for the
birth-cancer linkages. Two main ﬁndings from this work include: (1) we found, in
a random sample of 591 linkages manually veriﬁed, that the proportion of correct
linkages was 98%, and (2) for the years 1989-2004, when additional maternal

identiﬁers were available, comparing SSN alone linkage with multi-stage

108

eemms
waenﬁs
t2%.Af
ihkages
marau
onmah
lmkedr
wened
2004.

re-set

Dank
hnke
Dﬁor
aga'
dete
hay
Dar
SLR
ag.

aC:

deterministic linkage (which used additional identiﬁers to ensure no linkages
were missed), the percentage of possible cases missed using SSN alone was
1.2%. After the validation work, we excluded 162 cases with invalid/duplicate
linkages. Then we identiﬁed eligible controls from the MI birth ﬁles and
individually matched four controls to each case (cases=8,251; Controls=33,004)
on maternal year of birth +/- 1 (1935-1981) and race (White, Black). Finally, we
linked all controls to the Detroit SEER registry using SSN to determine if controls
were diagnosed with cancer at s 50 years of age during the study years, 1978-
2004. We then excluded controls with cancer (n=95; 0.29% of study controls) and
re-selected new controls.

Birth-Birth Linkages. We next identiﬁed all live Ml births for study
participants. First, we linked participants to the existing MDCH Ml maternally-
linked birth dataset (1989-2004) (250) and then expanded this to include births
prior to 1989 in the MI Birth ﬁles. We used maternal SSN alone for linkages. We
again conducted validation procedures, including manual veriﬁcation, to
determine the accuracy of linkages among 4% of the participants identiﬁed as
having possible invalid birth-birth linkages (i.e., non-matching maternal age,
parity, birth year order) and found an estimated 20% had invalid linkages in this
subset. We excluded 68 cases and 267 controls with invalid linkages or where
age at each delivery (a good indicator of invalid birth-birth linkages) did not match

across live birth histories.

3.5 Data analyses

109

Outcome. Histology information was available from the MSCR for both in-
situ and invasive cases and is based on medical records and pathology reports
when available. Histology was deﬁned using International Classiﬁcation of
Diseases for Oncology (ICD-O-3) codes with ductal breast cancer deﬁned as
ICD-0-3 code 8500 and Iobular breast cancer deﬁned as lCD-0-3 code 8520. We
did not examine other rarer histologic sub-types due to limited sample sizes, in
particular for stratiﬁed analyses.

Exposures and covariates. Study exposures were categorized using cut
points selected a pn'on' based on previous literature (57), including age at ﬁrst
birth in years (< 20 (reference), 20-24, 25-29, 30-34, 2 35), age at last birth in
years (< 25 (reference), 25-29, 30-34, 35-39, 2 40) among women with 2 2 live
births, and number of live births (1 (reference), 2, 3, 2 4). We also examined age
at ﬁrst and last birth per ﬁve year increase. Matching factors included race
(White, Black) and maternal year of birth (continuous). Covariates included year
of ﬁrst live birth (1978-1983, 1984-1988, 1989-1993, 1994-1998, 1999-2004),
maternal education at ﬁrst birth (< high school (HS), 2 high school), and years
since last birth (< 5, 2 5). For cases only additional variables were year of
diagnosis (1985-1989, 1990-1994, 1995-1999, 2000-2004) and stage at
diagnosis (in-situ, invasive).

Demographic and pregnancy-related factors were compared between
cases and controls, and also between ductal and Iobular cases using frequencies
and proportions and means (where appropriate). We examined separately the

associations between age at ﬁrst birth, number of live births, and age at last birth

110

and breast cancer risk overall and separately by histologic type (ductal, Iobular).
Odds ratios and 95 percent conﬁdence intervals (Cls) were obtained by ﬁtting
conditional logistic regression models to the data, using maternal race and
maternal year of birth as conditioning variables. Covariates and study exposures
were considered as potential confounders when not the main effects of interest,
selected a priori based on previous literature. Each potential confounder was
tested individually in conditional logistic regression models for each exposure
and breast cancer risk. Though not all potential confounders resulted in a change
of 2 5 percent for the main effects parameter estimate, we adjusted all ﬁnal
models for the same covariate set (age at ﬁrst birth, maternal education, and
number of live births).

We also examined associations for pregnancy-related factors and both
overall breast cancer risk and breast cancer risk for ductal and Iobular tumors by
race and maternal education at ﬁrst birth. For these stratiﬁed analyses, we
dichotomized study exposures (age at ﬁrst birth and last birth (< 30 years, 2 30
years) and number of birth (uniparous, multiparous». To test for multiplicative
interactions, we used the likelihood ratio test to compare models with and without
the multiplicative term for the potential modiﬁers and exposure of interest. For
subgroup analyses and analyses by histologic type, we present only the results
from fully adjusted models. P-values for case-case comparisons were calculated
by ﬁtting unconditional logistic regression models to compare ductal and Iobular
cases, adjusted for age at ﬁrst birth, number of births, maternal education, race,

and maternal year of birth (as appropriate) (11). SAS version 9.2 was used for all

111

analyses. All tests were two-sided and p-values < 0.05 indicated statistical
signiﬁcance.
3.6 Original Sample and Analﬂic Sample for the Present Study

The original study sample of eligible women included 8,251 cases and
33,004 controls. During validation of linkages, as described above, we excluded
331 women (64 cases; 267 controls). We also excluded 15 women (3 cases, 12
controls) missing study reference month and 66 cases diagnosed with breast
cancer during pregnancy. Finally, we excluded 264 controls who were matched
to excluded cases.

The initial eligible sample for the present study included 8,118 cases and
32,461 controls. We then excluded 247 cases and 1,130 controls with
missing/implausible ( < 24 weeks) gestational length for any birth. We required
participants to have pregnancies lasting at least 24 weeks to compare to
previous studies. A total of 34 cases and 115 controls were missing information
on education. In sum, 3.5% of cases and 3.8% of contrOls were excluded due to
missing/rmplausible data. Finally 1,057 controls that were matched to excluded
cases were also excluded. The ﬁnal analytic sample size was 7,837 cases and
30,159 controls. Sample sizes for analyses by histologic subtypes of breast

cancer are shown in the Tables 4-6.
4. Results

Table 4.1 displays descriptive and pregnancy-related characteristics for
ductal and Iobular breast tumors. The mean age at diagnosis was 40.4 (standard

deviation (SD): 5.61) for ductal cases (67.7% of all cases) and 43.1 (SD: 4.63)

112

for Iobular cases (8.7% of all cases). Compared to cases with Iobular tumors,
cases with ductal tumors were more likely to be younger at diagnosis, be
invasive tumors , have 5 HS. education, and be younger at ﬁrst and last birth.
Number of births did not differ by tumor histologic type.

Table 4.2 shows the distributions by case-control status for age at ﬁrst
birth, number of live births, and age at last birth. Controls were more likely than
cases to have earlier age at ﬁrst and last birth. The mean and median ages at
ﬁrst birth were 27.3 (SD: 5.04) and 27.0 for cases, and 26.7 (SD: 5.23) and 26 for
controls. Data from the National Center for Health Statistics based on the US.
natality ﬁles reported that the mean age at ﬁrst birth increased from 21.4 in 1970
to 24.9 in 2000, while the median age of mother increased from 25.4 to 27.1
(289). The mean number of live births for cases was 2.2 (SD: 0.87) and for
controls 1.9 (SD: 0.92). Among women with 2 2 live births, the mean age at last
birth for cases and controls was 31 (SD: 4.49) and 30.5 (SD: 4.66), respectively.
Controls were more likely than cases to have earlier ages at ﬁrst and last birth.

In models conditioned on age and race, later age at ﬁrst birth and later
age at last birth were signiﬁcantly associated with increased risk of breast cancer
and adjustment for potential confounders did not substantially alter the results
(Table 4.2). ORs adjusted for maternal education at ﬁrst birth, number of live
births, and age at ﬁrst birth (for last birth models only), were elevated for each
category for age at ﬁrst birth (reference = < 20 years) and last birth (reference =
< 25). For example, women with a ﬁrst birth at 2 35 years had an 80% increase in

risk compared to women < 20 years at ﬁrst birth, while women with a last birth at

113

2 40 years had a 36% increase in risk compared to women < 25 years (Table
4.2). In models conditioned on age and race, women with 2 or 3 live births
(compared to uniparous women), had an increased risk of breast cancer, with
similar ORs for further adjusted analyses (Table 4.2).

We further examined the associations between pregnancy-related factors
and breast cancer risk by race (White, Black) and maternal education at ﬁrst birth
(5 HS, > H.S.) (Table 4.3). Age at ﬁrst birth was associated with increased risk
of breast cancer for both Black and White women and women of lower and
higher education at ﬁrst birth. Multiparous women had an increased risk of breast
cancer for 2, 3, or 2 4 births compared to uniparous women for all subgroups with
the exception of Black women, who had a decreased risk for 2 4 births (pintemson
for race = 0.04). Later age at last birth was also associated with increased risk of
increased breast cancer risk; ORs tended to be higher in magnitude for Black
women compared to White women (pinteraction = 0.01) and among women with >
H.S. education compared to H.S. education or less (Pinteractron=0.41).

Results for adjusted associations between pregnancy-related factors and
risk of ductal and Iobular breast cancer are displayed in Table 4.4. Later age at
ﬁrst birth for each age group (reference: < 20 years) was signiﬁcantly associated
with increased risk of ductal breast cancer. Results were similar for Iobular
cancer except the OR for age at ﬁrst birth at 20—24 years was close to 1.0 and
not signiﬁcant. In the case-case comparison, the p-value was signiﬁcant for a
difference in ductal compared to Iobular tumors for age at ﬁrst birth. ORs for

number of live births were elevated and fairly similar for both ductal and Iobular

114

breast cancer risk. Later age at last birth also increased risk of breast cancer for
both histologic types, for Iobular breast cancer only the OR for 30-34 years at last
birth (reference: < 25 years) was signiﬁcant (OR = 2.11, 95% Cl: 1.17, 3.83).

We further examined pregnancy-related factors and risk of ductal and
Iobular breast cancer by race (Table 4.5). Among White women, later age at ﬁrst
birth 2 30 years (reference: < 30 years) was associated with increased risk of
both ductal and Iobular cancer. Among Black women, age at ﬁrst birth was not
signiﬁcantly related to risk of ductal or Iobular breast cancer, and the non-
signiﬁcant OR for Iobular cancer was in the opposite direction compared to White
women (OR = 0.62, 95% Cl: 0.26, 1.48). Multiparity was associated with
increased risk of both ductal and Iobular breast cancer among White women
only. Later age at last birth (2 30 years; < 30 years) was associated with a 70%
increase in risk of Iobular breast cancer (OR = 1.70, 95% Cl: 1.21, 2.40), but not
related to risk of ductal breast cancer among White women, while ORs for Black
women were in the direction of reduced risk and nonsigniﬁcant. Tests for effect
modiﬁcation by race on a multiplicative scale were signiﬁcant only for age at last
birth and risk of ductal breast cancer (pimmaion = 0.01).

ORs for ductal and Iobular breast cancer by maternal education at ﬁrst
birth (5 HS; > H.S) are shown in Table 4.6. For women of both education level
groups, age at ﬁrst birth 2 30 years and multiparity was associated with
increased risk of ductal cancer, while multiparity and age at last birth 2 30 years

were associated with increased risk of Iobular breast cancer.

115

5. Discussion

In this large, population-based registry-linked case-control study of parous
Black and White MI women S 50 years of age, later age at ﬁrst and last birth, and
multiparity were associated with increased risk of breast cancer overall and both
ductal and Iobular breast cancer. Later age at ﬁrst (odds ratio per 5 year increase
= 1.16, 95% CI: 1.13-1.20) and last birth (OR per 5 year increase = 1.11, 95% CI:
1.04-1.18), and multiparity vs. uniparity (OR for 2 births = 1.36, 95% Cl: 1.28-
1.44, 3 births = 1.29, 95% CI: 1.19-1.40, 4 births = 1.25, 95% CI: 1.11-1.42),
were independently associated with increased breast cancer risk, with ORs of
similar magnitude by histologic type. Results were similar by maternal education
(5 high school, > high school). Some differences were observed by race (White,
Black), including an increased risk of Iobular tumors for age at last birth 2 30
years (vs. < 30 years) among White women only (OR = 1.70, 95% CI: 1.21-2.40).
Further, associations by histologic type among Black women tended to be
nonsigniﬁcant and risk did not appear to be elevated, but sample size was
limited.

Few studies have examined the effect of age at ﬁrst birth and risk of ductal
or Iobular breast cancer among younger women. We identiﬁed only two reports,
both using the Women’s Contraceptive and Reproductive Experiences Study
(CARE) case-control data, which have examined reproductive factors and
histologic type among younger/premenopausal women. Results from the ﬁrst
report suggested that later age at ﬁrst birth was associated with an increased risk

for Iobular tumors (e.g., OR for 25—29 years vs. S 19 years = 2.31, 95% Cl: 1.11-

116

2.47, OR for 2 30 years vs. 5 19 years = 1.61, 95% CI: 0.89-2.91), but not ductal
tumors ( (e.g., OR for 25-29 years vs. 5 19 years = 1.13, 95% CI: 0.96-1.33, OR
for 2 30 years vs. S 19 years = 1.05, 95% CI: 0.86-1.27) and ﬁndings were not
modiﬁed by age (280). A second study, using the same data, reported ﬁndings
for risk of both ductal and Iobular breast cancer and age at ﬁrst birth jointly by
menopausal status and race (for premenopausal women only) (288). They found
some evidence that later age at ﬁrst birth was more strongly related to increased
risk of Iobular as compared to ductal tumors among White women and
associations were null among Black women.

We found only one study of later age at last birth and histologic type of
breast cancer, and ﬁndings were similar for ductal and Iobular tumors, though the
study was among women 50 years or older (278). As with age at ﬁrst birth,
studies of number of live births and breast cancer by histologic type are among
older women or mixed age populations and/or use nulliparous women as the
reference group (112, 277-281), which hinders comparison to our study.

The maternal hormonal milieu during pregnancy is thought to underlie the
associations between pregnancy-related factors and breast cancer. Researchers
have hypothesized that later age at ﬁrst birth may be more strongly associated
with Iobular tumors as compared to ductal tumors, given that Iobular tumors have
been proposed to be more hormonally sensitive and are more likely to be
estrogen receptor positive (282, 290). However, in the present study associations

for pregnancy-related factors and risk of ductal and Iobular cancer were similar.

117

In our study we also conﬁrmed the well-known association for increased
breast cancer risk overall for later age at ﬁrst birth and the less consistently
reported increased risk for later age at last birth (26,46, 54, 55, 57, 59, 60, 273).
Number of births has also been well-studied in relation to breast cancer (24, 124,
274, 275), but has been less studied among younger women, and in particular
among parous younger women. Among parous younger women, some studies
have reported an increased risk of breast cancer for increased number of births
(35,43). However, two large US. population-based case control studies (i.e., the
Women’s CARE study and the Cancer and Steroid Hormone Study), have not
reported an increased risk for parity among women 35-49 years (25) and 20-55
years (276), or evidence for an effect of years since last birth on breast cancer
risk. For example, in contrast to our ﬁndings, the CASH study of women aged 20-
50 years of age reported a signiﬁcant protective effect for increasing number of
births and breast cancer risk, independent of age at ﬁrst birth (276).

Several limitations must be considered when interpreting the results of our
study. First, we only had pregnancy-related variables available in the birth
registry to consider as covariates. However, several other studies with this
information have not found that adjustment for other known breast cancer risk
factors substantially alters age at birth associations or have also only adjusted for
pregnancy-related factors (278, 281, 288). A second limitation is the potential for
underestimates of parity due to potential movement out of state for women who
had more than one live birth. This could lead to an underestimation of the

number of live births, incorrect classiﬁcation of a birth as a last birth (for age at

118

last birth), and lack of complete adjustment for number of live births as a
confounder. Further, it could be said that it is more likely that controls moved out
of state and had additional births because cases were required to have both their
births and cancer diagnosis in MI; hence it is possible that our controls were
more likely to have underestimated number of births and ORs for the association
between multiparity and breast cancer risk could be biased away from the null.
Third, the source of histology data for this study was from the statewide MI
cancer registry, which routinely abstracts tumor histology data from pathology
reports and medical records, and though information is complete, there is no
centralized pathology review to assure consistency and accuracy in histological
classiﬁcation.

Several limitations of our study are related to the use of registry-linked
data. First, SSN alone was used to link cancer and birth data as well as
additional births for study women. However, our validation work demonstrated
that linkages were correct and complete above 98% (see Chapter 3, Methods).
Simon and colleagues also demonstrated that SSN was a highly valid approach,
with only 2% of possible linkages missed using SSN as the only linkage variable
(245). Second, we had to exclude women missing SSN in either the cancer or
birth registry. Few women, however, were missing SSN in the cancer registry
(1.4% of cases) or birth ﬁles (1-5% among all births per year) and missing SSN is
unlikely to be related to case-control status. Third, outcome misclassiﬁcation is a
concern because some controls may have developed breast cancer either during

the time lag between exposure and the start of the MSCR (1978-1984) or in a

119

different state. However, breast cancer diagnosis, before the age of 50 is rare.
Further, we found by linking to the Detroit SEER registry that only 8 controls in
Detroit were diagnosed with breast cancer before age 51 during 1978-1984; if
this estimate is extended to the entire state, only about 0.05% of our controls
could have been misclassiﬁed.

Our study population has a much higher percentage of women with later
age at ﬁrst births and last births, as compared to other population based studies.
Table 4.7 diSplays distributions for age at ﬁrst birth overall and by birth cohort for
controls, with a comparison to a recent birth registry-based study among younger
women conducted in Sweden (63). We expected these higher distributions,
which are due to the data restrictions of our study (i.e., ﬁrst live birth in 1978-
2004 and year of ﬁrst diagnosis during 1985-2004). Further, we ensured that
controls were selected from the same population that gave rise to the cases by
using the same selection criteria for cases and controls, with the exception of
cancer diagnosis, and requiring controls to have their ﬁrst live birth before the
matched case’s date of diagnosis. Hence, the higher age at ﬁrst and last birth
distribution should only inﬂuence the generalizability of our study ﬁndings, and
not bias the odds ratios.

Main strengths of our study include the large sample size, population-
based state-wide design, and use of registry data. Registry-based studies are
less subject to key bias found in case-controls studies, because exposure data is
collected before breast cancer diagnosis, though recall for important reproductive

events such as age at births and number of births is unlikely to be a large

120

concern. Further, using existing registry data provided us the opportunity to
select controls from the same study base as the cases, without being subject to
potential bias due to low participation rates. In addition, we had high quality
cancer registry data that meets NAACCR standards.

In this large, population-based registry-linked study of parous women s 50
years of age, we found that later age at ﬁrst and last births, as well as multiparity
were associated with an increased risk of breast cancer overall and for both
Iobular and ductal histologic types. We did not ﬁnd that associations varied by
histologic type among all women. Some differences were found by race and
education subgroups, most notably a 70% signiﬁcant increase in risk of Iobular,
but not ductal breast cancer among White women only. Very few studies,
however, have examined histologic type among younger women and results
require conﬁrmation in future studies. Future studies that jointly examine
histology and other tumor characteristics (e.g., estrogen receptor status) can
further contribute to the understanding of pregnancy-related factors and risk of

breast cancer by histologic type.

121

£6 9,: see so 28>

 

 

283 8.8va 2.68.83 859:.
is; 8m 3m: NS 37..-...
w_mocam_u 6 5058 “much 6 maﬁa
:80.v 8.5 an 8.38ch 38-88
3va SN 3. 5 ~82 82-82
63: 8 6.3V OR 32.8?
3.8 mm 3.3 «mm 82-82
28:35 20 50>
:83 32:6 $.36; ovm
23$ :2 69¢ $3 9. v
6693 madcamﬁ am oa<
P8.0 :48 6 8:58 xogm
6.29 mmm 6mg 2: 922,
com. .6522).

g 2 ea c

@8ch $8.85

6265-6 6336390.. 3339030

 

5.838: 28.222 5 256 889258“.
50:60 685 new 9.906“. 56521 .2923 6665
6:50. new 50:6 .8 wozmtoaomamco 20 2259320 . _..v 633.

main... .a

122

 

F_.aaa.v

taaa

awaaav

666666

FFvaaa

Paaaa.a

6.666 666
26.6: 66

6. a 6
6.666 6:
66$ «66
6.666 E
666 66

6.6: 6
6.666 662
66$ 66
266$ e62
666 66

26$ 66

66$ 66?
66$ 66
6.666 666

6666 6:-
24.666 666

6. a 6
6.$ 66
£6: 62
$666 6:
6.666 266

:6: 666.4
6.666 666.2

6.6V v2
66: 666
266$ 663
666$ 66:
65 666

65 6:.
6.666 666.2
6.666 266.6
6.466 626.?
A666 F66

2666 666
66: 666
66$ 626.6
6. F66 6662

6.466 666.6
266$ 664.6

:66 6:
6.2 666
66$ 66
6.666 663
66$ 24.6

a N
a v
v5.5 “mm. @056 mamm>

64 N

66-66

46-66

66-66

66 v

46.86 5%. am. a 66<

66 N

666

66-66

66-66

66 v

6.82 Em 62E 6 66<

vm
a
N
6

6526 62.. a 28:52
.856 66.: A
669.856 66:6 6
66:6 .66 a 868666
6666-6662
6663662
66676662
666E662
66676662

865260 .5. 6.666

123

62:3 62. 668 .o 6 £3, .5an 629:6;V
6:6: 56 ~69 96:66-56 666666526626
.236 662.6 66 ar 5 N.- .a_. 6666 69:95 6662656
62660666 65:66 66 669 626666-120r

nos—5:00 ._..v 636...

124

.Eta Eon :_ 9m. _m>m_ mod w a “a 2.85ch >__mo_~m=mum $5 .3395; ascozﬁcoo
mm mom. ﬁEQmE new 55 be Em.» .mESmE 53> .2808 56859 2650. $829.8 B 32.8 mmor

 

8:73.: 5.. .33,: 2.. 35% m .53 5.5 mg a mm<
5.73:8; 5.3”.52... 6.8 $6 as it SN
:~..-§.:3.F £3-33: @3086 @238; 8-8
3.73:5; 63.3.52; 6.8303 :.§§~.~ 3:8
33-3.58; 54-3.53; 8.533 6&5? 8-3.
$0553.33 $0553.33 C.o:omo.m at can mmv
memob 5.5 $2 a 0?.

5.73.53; amgmemoe 3.3 83 a.& 8m X
83-2.52; 3.3-8.52; 6.0593 8.0383 n
3.73.38; 5.78.3»: ﬁgmomﬁ 3.32% N
8893933 $893933 :15 50.2 :Svomvd F
2% 9,: .6 .8532

8.4.4-2.: 2.. 5.73.: 3.. $5.: .0. be 5.5 E... a mm<
85-3.53; $178.58.. :9 Kim 3.8 N8 9%
a..~-$.:$.. aqvmmim: $8586 6.883 3-8
A351: 8.. 5.7352... 3.3 50.: €83.85 8:3
$172.:m3 5272.35. 3.52:5 A335? 58
$823233 $893933 :9 83. 6.9 3:. omv

Emma 55 we a 0?.

 

Ncoeome mo 2033 mo 33 c g c

 

antenna Cmgus
32:50 330

 

Avoomémm E amaze—2 E >25 ommmmégmamm .8ch $.35 95 920m“.
5:235 .5050 Emmi ho xw: EB 2203 nmﬁ_m.->ocmcmma .2 worm. mnuO .Né $33.

125

.mctﬁ m>__ 908 .o m 55, :mEQs mcoE<m
£5 E: a. 8:838

.mEQmE new .923 9: .0 .882: .55 “9: am mom 82:05 85 20008 .8399 mchEocoo 2 8:68 5N
Ema—5:00 .Né Baa...

126

85.5 m>__ EOE B N £5, cmEoz, 9.9:...N

.55 .8. .m cozmonum EEQmE ucm .mctB m2. .0 .383: its .9: 8 wow .0. 883.6... 0cm 8.3ng mchzﬁcoo
mm mom. 8599: new 5.5 .o 89. 8529: 55. 230E 53852 2652 8.55950 E 8.26.. mmo.

8.8-2.: 8.
888.: 8..
5.8.81: 8..
8.8-8.: .5
$055.0: 8..

8.8.: 8..
8.8.: 8.
8.8.: .1.
60:23.8: 8..

8.88.: 88
8.8.8.: 8.8
8.8-8.: 8.
8.8-8.9 01.
$055.0: 8..

8. .89 8..
35-2.: 81.
8.8.: 8..
8.-.... 8.
3055.9: co.

81.8.: 8..
8.8.: 8..
81.8.: 8..
60:83.9: oo.

8.8.: 8..
8.8-8.: 8.
a...8.: 8..
8.81:8.
80:989. oo.

8.8-8.8 «.8
88.8.8 8..
88-8.: 8.
8.8-81: 8.8
$9.999: co.

85-8.8 8.0
8.8.9 8..
31.8.: 8..
$8889: 00..

8.8.8 8..
3.8-8.: 8..
8.8.: 8..
61.8.9 8. .
30:90.9: 8..

8. .89 8.. 9. N
35-..: m1. 8-8
8.8.: 8.. 88
8.8.: 8. 8-8
$055.8: 8.. mm v
NAM-L83 5.3 as a. 8<

8.81:8. v N
$118.8... 8
£118.: 8.. N
.8555: co. .
8:5 9,. .o .8532

8.88.: 8.. 8 N
8.8-8.: 8. 88
8.8.: 8. 8-8
.8. .8. : 8. 88
3055.0: co. om v

3.83 5.5 a... a 8<

 

.2308: m0

.5088: m0

.588: m0

.588. m0

 

 

©33Vﬂm0mm0
Cozmoavm

.m.I A

FdeHmmmmO

8888 .w: w

nmwumwwmo
.85

88.8880
322,

 

88-88: 8952.2 5 83m

omwmméﬁamm .850 8QO new 20.0mm 38391 .55 .2... .m cozmoanm EEQmE ncm
mom. 859m... B Eocmo 8mm... .0 xm: new 923. BEE-58:93 8. 839 £60 .né min...

127

 

 

8.0 81.8.8 80 m 8. .-88 8.. 8. 0.. N
3.8-8.8 88.. 0.. A00. .-008 8.. 88 8-8
08.2.: ...N 88 8.8.: 8. 81. 88
88.8.8 8.. ... 8. .-.... : 8.. 8 ... 8-8
80289: 00.. 8 $08.89: 00.. 88 8 v
...933 :5: 8m. .0 09.
8.0 6.8.88 .1. 8 8.81:8. 88 v N
8.8.: 8.. 8. 8.13.: 8.. m8 8
8.8.: 8. 0.8 $1 .8: 8. 0.8 N
80989: 00.. 88 8090.2: 00.. .8. .
8:5 9,: .0 .8202
8.0 8.88.: 8.8 8 58-81.85 o... 8 N
88.0.: 8. 8. 5.8-8.: 8.. 08.. 8-8
88-8.: 8.. .8 8.81: 8.. .88 8-8
$588 8.. 8. ...-.31: 8.. 8.8.. 8-8
600989: 00.. 8 8088.2: 00.. .8 8 v
A988 58 .9... a m0<
mag-n. ~20 80mm: m0 @030 .20 8mm: mo 90me

m

 

8:53

.3000

 

Avoom-mmm: :mm_:0__2 5*035 gamméﬁamm 50:06 .mm9m 0cm 90.0mm. >0:m:m9n_ .85
060.28: E .0058 ...-.99: .0 xm: 0cm 9203 0939-5539: :2 8:9 £0.00 .16 03m...

128

.3350 EN; 25 938 .930. at. ”22:50 mmmd new $80 .903 9.0.8 mctﬁ m2. 20E 5 m 5;, 5an 9.085.
.239 “.0 59.56 new .55 BE 5 wow £388..» .009 .mmm ..8 87.38 .mmSwoaxm >33 9: .9 mboEB 6:50. .w> .903
9388 2 E was $89: c2399.. 2650. 6556505 5.8583 .Gwmv mcomtmano $8.38 ..8 9m mm:_m>-am
.6 5.muw_o:coo Nomnwmmmov 5.5 “we “m 8:838

.mEQmE new .935 9: we .moEsc its «m;— 5 mum go“— umuwaﬁm 02m 0cm. $39.9 mchEccoo mm

mom. 5539: new 5.5 .o 86> .mEQmE 53> 232: 53852 2660. 6:86:00 3 32.8 3.928 .m> 5.30; mmou
.vawdﬁgezoo ”83"?»me 5% we a 8:838

5529: van 655 02. 3 595: its “e: um mum .2 3333 86 new mmBmtm> mc_:o_u_ucoo mm mom. .mESmE

new 5.3 ho 50> .mEQmE 53> 3808 c2393. 2662 $5238 B 32.3 2928 .w> .9023 mmoF

8:528 .3 wank

129

05.5 w>__ 0.0:. .0 m 5.2, e080; meoe.<N

.8003. 5.5 um. .0 mom 0% .3008 2:5 .0 .852. Sodas 5.5 a... a 00m 0% $50; 0.220 ”$0.0"... 5.5

.09 .w wow new .cldu 3 05.5 .0 .0053: .3mdu3 5.3 .0... .w wow new e080; 00_e>> ”030:0. 0w 0.0.5 65.5 .0 .wnEae new 5.5
.9: .w wow .eozwosnw .wmw .0. nwﬁanw 0.0:.3 .ano. .m> _w.0:n 0.w0E00 0. m_wnoe. e0_mmw.mw. 0:062 _weoz_ne00e: E0... n0>ewnv
00w. .5 meowewaeso wmw0-www0 w .0. 829.0 .55 .0... .w eozwoanw _we.w0we. new .230 9: .0 .wnEse 5.5 .0... 0w wow .0. nwumanw
new w_nw_.w> meEoEneoo w 0w 5.5 .0 .90.. _we.w.we. 5.2, 2.80:. eo_mmw.mw. 0:052 _weo_._ne00 050000-009 .5 82.8 mmor

 

5.98830 2 30.75830 3: 834882.. 0% 3.078803 $0.0 0mm
3290.933 2 305.8233 09 8085.233 a. $8989.80 83 8v
«3.88 5.5

.mew 091

80.75.8000 mm 8078.803 N8 35.28%; 3. attain? 22., 928:3:
605.9088; mm 8890.088... omm 305.3088; mt 889080.800... 3m; 3059::
2:3 9:

.0 .8252

63-8830 3 83.880: 03 5.78.803 8N $3-583? 0:... 0mm
$8989.03 9. $8989.80 8% $8989.03 0% 3890.088; «mam 9V
3.88%.“.

00.5w 0m<

 

:0 £68 «.0 wmmwo :0 $8. m0 wwwwo :0 $8. 10 wmwwo :0 o\ommv m0 wmwwo

 

.w_:00._ 6.030 .w_:00._ .9030

e0E0>> x0w_m e0E0>> w._e>>

 

uvoomémm: eaﬁoi e. anw nwmwméaﬁwm .8er .mw0.m new 0.203 >0ewem0.n_ .00w.
_we.0.wE .5 .088 Ewen .w_:n_0_ new _w.0:n .0 x0: new 0.203 nw.w_w.->0eweow.0 .0. 003w. mnnO .mé 03w...

130

.0590 9: 0.05 .0 N 0.3, 5E0; me0E<

.80"... 80 .09 8 8m. 8m $.08. 8.8 .o .858 ....oduaw

80 .9. .m 08 80 .0.: A .83"... 80 .8. .m 80 8.. 08..."... 8.8 .o .858 08.0"... 80 .8. .0 08 Ba .0.: 0 86:9

0w 052. .0590 .0 50E:e new 0.5 .0.... .w wow .wow. .mmw .0. n0.0:..nw .0.0E:. .w_:00_ .0> 555 050.50 0. 0556 5000.00. 000.02
_we0_._ne00e: 80.. 5253 550:5 >0 05050.50 00w0-w0w0 w .0. 00:_w>-0 05:0 9: .0 50Eae new 5.5 .0._. 5 wow .0. 5.035
new 00_0w_.w> me_e0_._ne00 0w 50w. 5505:. new 0E0 .0 50> 5555:. 0.3., 05n0e. 5000.05. 2.062 55528 >0 nw>_5n 0m0.

 

63-3.8... .8. ..N.-.m.88. 0%.. Stu-8.88.. 8. 8.7.0.88. 8. omm
58999.8. .0 $8999. 8.. .8 $8999. 8.. 8 68999. 8.. 08 00v
~88... 5.0

.095 wm<

.0...-8...8.. 8. ...0..-.0...n.... 8. 5..-»...8. .0. 8178.88. 80.. 8988.2
50555.30... 00 5055.088. 000 5055.088. 3 50555.30. .0. 000595
890 9,.

.0 59.52

30.880... 8. 8.8.8.... 8... $0.88... 8 .00..-.N....n.. .... omN
$8999. 8.. 08 58999. 8.. .0.. 88999.8. ..N 58999. 8.. 08. 00v
.9808...

.905 wm<

 

:0 $mm. mo 050w0 :0 $8. 10 000w0 20 $00. m0 000w0 :0 $mm. m0 050w0

 

5300.. 5.0.5 5300.. 5.0.5

e0_.w0:nw .0.: A 0000 .0 e0_.w0:nw .0....

 

..voow-wwmt 595:2 5 >55 n00wm->..0_mmm 5050 .055 new 0.0.0w... >0ewemw.n_ .030 .0... .w 5.50.60
5:55.: >0 5050 .055 5302 new .805 .0 x0: new 0.0.0w. n0.w_0.->0ewemw.0 .0. 00.5. 0nnO .0... 0.0w...

131

.0090 08 80 .-.-8. 880 05.5 .9. £5, 8an £00950 89:0 09888 .85 :88 w s 05.5 __w .9 8.85% 5.5 .9... .0 09..

.w_0Ew0 we. .0 $00 .0. wow .0 0.ww> on w .w .088

 

 

 

N... 0 8.3 ~00 8.0. 80.. 30.8000 :0. .090 .92
a... 8.0. 8 6.0..000. 2.8.08... 3.0.08 3.8.08.0 3-8
0.8 .000. 0.0 8.0. 90.0 8.3.. 00.... o 6.8. .8. .. 8.8
0.8 6.00. 000 6.00. 80... a. .0. 0.0.0 o .30. 8.0 «0-8
0.. 6.8.08 3.2.2.0 o o 2.0. 8v...- 8v
.0.ww>.
5.0 .0... .w 00<
9 S... e A... c S... c n... ...... c
.008 .._w .0 0280.80 80.05. 80 .000. 08 .03. .8780. .90.

.09. 5.5 6590.2

 

$80.80.. 8022.2 5 >030 000008.009. .880
.0ww.m new 0.0.0wn. >0ewemw.n_ .0_0..e00 meoEw ..0000 0E0 _we.w.we. >0 02m .0._. .w wow .0 eon—.0505 He ﬂew...

132

CHAPTER 5: A population-based case-control study of
fetal growth, infant gestational age at delivery, and
maternal breast cancer among younger women

1. Abstract

Variation in fetal growth (FG) or gestational age (GA) in a woman’s own
pregnancies may serve as indirect markers of the hormonal environment during
pregnancy, which may play a role in both the short-term increase and long-term
decrease in risk of breast cancer following childbirth. We conducted a population-
based case-control study among parous Michigan (MI) women aged 5 50 years
with singleton births using linked Ml Cancer Registry (1985-2004) with Ml Live
Birth records (1978-2004). Cases (n=7,591) were matched 4:1 on maternal birth
year and race to controls (n=28,382). Using conditional logistic regression, we
examined the associations for breast cancer and GA (< 32 wks, 32-36 wks, 37-
41 wks, 2 42 wks) and FG (deﬁned using BW percentiles for GA ((SGA) < 10th,
(AGA) 10-90th (referent), (LGA) > 90th) in both ﬁrst and last births. Delivery of an
SGA or an LGA infant in a ﬁrst or last birth was not signiﬁcantly associated with
breast cancer risk. However, among women with a last birth at age 2 30 years,
delivery of an SGA infant in a last birth was associated with a reduced risk of
breast cancer (OR=0.82, 95% CI: 0.68-0.98). A ﬁrst delivery at < 32 weeks or at
> 41 weeks (reference: 37-41 week) was associated with a reduced risk (ORs:
0.80, 95% CI: 0.62-1.04 and 0.92, 95% CI: 0.85-0.99, respectively). Our study
provides limited evidence for an association between low or high FG and overall
breast cancer risk among women s 50 years of age. Delivery of an infant < 32

weeks in a ﬁrst birth may reduce breast cancer risk, though this ﬁnding was in

133

contrast to our hypothesis and the underlying biologic mechanism is not clear.
Little work has been done in the area of infant birth characteristics and maternal
breast cancer and future studies with information on biologic measures during
pregnancy and other potential confounding factors (e.g., maternal body size) are
needed to further characterize these associations.

2. Introduction

Breast cancer following childbirth has been shown to be associated with a
poorer prognosis and higher mortality risk (as compared to nulliparous women of
the same age) (38—41). Little is known, however, about the hormonal
mechanisms underlying the role of pregnancy in breast cancer development
among premenopausal parous women. The observed short-term increase and
long-term decrease in breast cancer risk following childbirth, which depends on
age at pregnancy, has been proposed to be due to maternal hormonal factors
during pregnancy (61, 62, 291). The direct study of maternal hormones during
pregnancy and subsequent breast cancer risk, however, is difﬁcult due to
methodologic and practical issues (e.g., long time span between exposure
measurement and breast cancer diagnosis) (122).

In lieu of biologic measures, infant birth characteristics such as fetal
growth or gestational age at live delivery, which may reﬂect altered exposure to
maternal hormones during pregnancy, could serve as proxy measures of the
maternal hormonal environment (65, 66, 79). Few studies, however, have
examined fetal growth (estimated using birthweight adjusted for gestational age)

(63, 65, 79, 81, 120, 175) or infant gestational age at delivery (including preterm

134

(> 37 weeks of gestation) and/or very preterm (< 32 weeks of gestation delivery»
(63, 65-70) in a women’s own pregnancies and subsequent maternal breast
cancer risk. Though limited, evidence from two large registry-based cohort
studies of birthweight adjusted for gestational age and breast cancer suggests
that delivery of a high birthweight baby in a ﬁrst birth (2 4500 9 vs. 2500-3499 9)
(63) or most recent birth (2 3.75 kg vs. 5 3.0 kg) increases breast cancer risk,
and the effect may be more pronounced in the ﬁrst ﬁve years following delivery
(79). One case-control study (65) and three cohort studies (68-70) have reported
evidence for an increased risk of breast cancer for earlier gestational ages at
delivery, though other studies have reported conﬂicting ﬁndings (63, 66, 67).
Finally, the etiologies of breast tumor histologic sub-types may vary (282), but no
study has examined variation in the associations between the infant birth
characteristics and risk of breast cancer for different histologic types.

In summary, studies of the association between infant gestational age and
fetal growth for a woman’s own pregnancies and breast cancer risk are sparse
and several questions remain. To estimate fetal growth, previous researchers
have used birthweight alone or adjusted for gestational age as a covariate
(birthweight is inﬂuenced by both duration of gestation and rate of fetal growth
(211)). No study has deﬁned fetal growth using percentiles for infant birthweights
for each gestational week based on published national reference values, which
allows for more complete adjustment for gestational age and hence improves
estimation of the independent effect of fetal growth on breast cancer risk (210).

Second, though studies have examined exposure in ﬁrst births, given the

135

importance of a ﬁrst full-term (full-term is defined as a pregnancy lasting 2 24
weeks in breast cancer research) pregnancy in relation to subsequent breast
cancer risk (35, 36), as well as exposure in last births, no study has examined
exposure to fetal growth and gestational age in both ﬁrst and last births.

In light of the few studies and remaining questions, we conducted a
registry-linked, population-based case-control study among parous Michigan (MI)
women s 50 years (predominantly premenopausal) using Ml Resident Live Birth
ﬁles 1978-2004 (Ml birth ﬁles) and MI Statewide Cancer Registry data 1985-2004
(MSCR). We investigated associations between fetal growth and gestational age
overall and the two most common histologic types (ductal and Iobular) in ﬁrst and
last births. We also examined associations of interest by the a priori effect

modiﬁers age at ﬁrst/last birth and years since ﬁrst/last birth.
3. Methods

3.1 Study Design

We conducted a population-based, case-contrOl study among parous Ml
women aged 5 50 years who had a live birth in MI during 1978-2004 at age 16-
50. This study was registry-based, using linked Ml birth ﬁles (1978-2004) and the
MSCR (1985-2004). We created a complete live birth history for cases and
controls through linkage of all of a woman’s Ml births. The study protocol was
approved by the institutional review board at Michigan State University and the
Michigan Department of Community Health (MDCH).

3.2 Study Population

136

Cases. Eligible breast cancer cases were identiﬁed from the MSCR
(1985-2004) and linked to their first live birth in the MI birth ﬁles (1978-2004).
Eligibility criteria included: (1) diagnosed with in situ or invasive ﬁrst primary
breast cancer between 1985 and 2004 in the MSCR; (2) age 20-50 years at
breast cancer diagnosis; (3) no previous diagnosis of any cancer with the
exception of basal and squamous cell carcinoma; (4) White or Black race based
on Ml birth file; (5) ﬁrst live birth in MI at age 16 years or older during 1978-2004;
and (6) residing in Ml at time of diagnosis. The study reference date for cases
was the date of diagnosis.

Controls. Eligible controls were selected from the MI birth ﬁles (after
linkage of the birth files to MSCR). Eligibility criteria included: (1) no history of
cancer in MI between 1985-2004 or in the Detroit Surveillance, Epidemiology, &
End Results Registry (SEER) for the years 1978-1984 (area covered by Detroit
SEER accounted for 43.6% of Ml’s population in 1980; 42.1% in 1990) (240); (2)
age 20-50 years at study reference date (individually-matched case’s diagnosis
date); (3) White or Black race based on Ml birth ﬁle; and (4) first live birth in MI at
age 16 years or older during 1978-2004. Control Sampling Strategy. We
individually matched four controls to each eligible case on maternal year of birth
(+/- 1 year) and maternal race (White; Black). Controls were required to have
their ﬁrst live birth in MI prior to the study reference date.

3.3 Data Sources
MSCR. The MSCR, which was fully established in 1985, is a member of

the North American Association of Central Cancer Registries (NAACCR) and is

137

certiﬁed as meeting all NAACCR standards for quality, completeness, timeliness,
and unresolved duplicate records. Data quality criteria (2004) are as follows:
case ascertainment (2 95%), passing edits (99.8%), cases identiﬁed from death
certiﬁcates only (1.7%), missing sex (0.04%), missing age, (0.04%), and missing
race (2. 6%) (241). Available data items utilized for the present study included
stage, behavior, histologic type, age and date at diagnosis.

Ml birth ﬁles. MDCH-Vital Records and Data Development Section has
maintained computerized records of all MI births since 1970. We elected to
initiate this study in 1978 because maternal SSN, the main record linkage
variable used in this study, is missing for 2 10% of mothers prior to 1978 and <
10% after 1978. Data items available from the MI birth certiﬁcate have changed
over the study years (1978-2004), with a major revision in 1989. Data items
utilized for the present study, available for each live delivery, included maternal
age, month/year of delivery, last menstrual period (LMP) estimate of gestational
age (calculated as interval between first day of the women’s last normal
menstrual period and date of delivery), clinical estimate of gestational age, infant
birthweight, multiple births (deﬁned as deﬁned as the birth of two or more
children from a single term of pregnancy), maternal education, infant gender,
number of prior children now living, and number of prior children now dead.

3.4 Dataset Creation Procedures (please see Chapter 3 for more details)

We conducted several steps to create the ﬁnal analytic ﬁle of cancer data
from the MSCR for cases (1985-2004) and maternally-linked birth data for all

study participants (1978-2004). Cancer to birth linkages. First, we identiﬁed all

138

female breast cancer cases diagnosed s 50 years of age during 1985-2004 from
the MSCR (n=33,941). Second, we linked eligible cases to their ﬁrst live birth at
age 16 or older in MI during 1978-2004 using simple deterministic linkage with
maternal social security number (SSN) as the only linkage variable (n=8,662;
excludes 477 cases missing SSN in the MSCR, and 19,480 cases assumed
nulliparous, ﬁrst delivery outside of MI, ﬁrst delivery prior to 1978, or missing SSN
in birth ﬁle). Additional reasons for exclusions after linkage included: not residing
in MI at 1st birth (n=7), < age 16 years at 1st birth (n=16), date at diagnosis 5
date of delivery (n=173), and non-Black or White race (n=53), leaving 8,413
eligible cases.

Next, we conducted linkage validation/quality control procedures for the
birth-cancer linkages. Two main ﬁndings from this work include: (1) we found, in
a random sample of 591 linkages manually veriﬁed, that the proportion of
linkages correct was 98%, and (2) for the years 1989-2004, when additional
maternal identiﬁers were available, comparing SSN alone linkage with multi-
stage deterministic linkage (which used additional identiﬁers to ensure no
linkages were missed), the percentage of possible cases missed using SSN
alone was 1.2%. After the validation work, we excluded 162 cases with
invalid/duplicate linkages. Then we identiﬁed eligible controls from the MI birth
ﬁles and individually matched four controls to each case (cases=8,251;
Controls=33,004) on maternal year of birth +/- 1 (1935-1981) and race (White;
Black). Finally, we linked all controls to the Detroit SEER registry using SSN to

determine if controls were diagnosed with cancer at S 50 years of age during the

139

study years, 1978-2004. We then excluded controls with cancer (n=95; 0.29% of
study controls) and re-selected new controls.

Birth-Birth Linkages. We next identiﬁed all live Ml births for study
participants. First, we linked participants to the existing MDCH Ml maternally-
linked birth dataset (1989-2004) (250) and then expanded this to include births
prior to 1989 in the MI Birth ﬁles. We used maternal SSN alone for linkages. We
again conducted validation procedures, including manual verification, to
determine the accuracy of linkages among 4% of the participants identiﬁed as
having possible invalid birth-birth linkages (i.e., non-matching maternal age,
parity, birth year order) and found an estimated 20% had invalid linkages in this
subset. We excluded 68 cases and 267 controls with invalid linkages or where
age at each delivery (a good indicator of invalid birth-birth linkages) did not match
across live birth histories.

3.5 Study Variables

Gestational age data cleaning. We used published methods to clean
gestational age (209, 210, 271). First, we used the clinical estimate to replace
missing/implausible values(i.e., < 20 weeks or > 44 weeks) for the LMP estimate
of gestational ages for 3,837 births (4.6% of all births) (210, 271). Next, we used
published cut points for gestational weeks 25-35 to identify births with implausibly
high birthweights (e.g., week 25, exclude birthweights > 1250 9); see Zhang and
Bowes (1995) for all cut points (209)) and substituted the clinical estimate for the
LMP estimate, when available (615 births; 0.7% of births) (209). After the above

described modiﬁcations, 57 cases and 303 controls were excluded due to

140

missing/implausible gestational age, gestational age 20-23 weeks (to compare to
previous breast cancer studies which exclude pregnancies lasting < 24 weeks
(291)), or implausible birthweight for gestational age where the clinical estimate
unavailable for substitution.

Exposures. Birthweight was deﬁned using a priori categories based on
previous studies (< 1500 9, 1500-1999 9, 2000-2499 9 , 2500-3499 9 (reference),
3500—3999 9, 4000-4499 9, 4500 g) (63, 65). We deﬁned gestational age both as
a continuous variable (24-44 weeks), as well as using clinically relevant
deﬁnitions of gestational age (< 32 (very preterm delivery (VPTD)), 32-36
(preterm delivery (PTD)), 37-41 (term) (reference), 42+ (posterm)) (292). Fetal
growth was deﬁned using published birthweight percentiles based on US Natality
data for each week of gestation (24-44 weeks) for percentiles 1-99 (210). We
examined this as a continuous variable. We also categorized this variable using a
priori categories (271, 272). Small for gestational age (SGA) included infants with
birthweights < 10th percentile by gestational week, appropriate for gestational age
(AGA) included infants with birthweights between the 10-90th percentile by
gestational week (reference), and large for gestational age (LGA) included
infants with birthweights > 90th percentile by gestational week. Birthweight
adjusted for gestational age as a covariate has been used by all previous studies
as an estimator of fetal growth; however, we use the term “fetal growth’ in the
Tables and Results to indicate when we are using the birthweight percentiles for
each gestational week definition. All exposure variables were created for first and

last births among women with 2 2 live births.

141

Covariates. All covariates were created for ﬁrst births and last births
(except parity, maternal education, and age at reference) and were categorized
based on previous studies (57, 66, 68, 69, 79). Categorical variables included
age at ﬁrst birth in years (< 20 (reference), 20-24, 25-29, 30-34, 2 35), age at last
birth in years (< 25 (reference), 25-29, 30-34, 35-39, 2 40), age at reference in
years (< 25 (reference), 25-29, 30-34, 35-39, 40-44, 2 45), years since ﬁrst birth
and last birth (< 5, 2 5 (reference)), infant gender (female, male (reference)),
maternal education at ﬁrst birth (< high school, 2 high school (reference)). Parity
(i.e., number of live births) was determined by the sum of number of prior
children now living/now dead at the last birth (1, 2, 3, 2 4). Binary categories for
the potential modiﬁers included years since ﬁrst/last birth (same as above), age
at ﬁrst/last birth (< 30 years, 2 30 years), and age at reference (< 40 years, 2 40
years) were also selected a pn'ori based on previous studies (66, 68, 69, 79).
3.6 Original Sample and Analytic Sample for the PresenM

The original study sample of eligible women included 8,251 cases and
33,004 controls. During validation of linkages, as described above, we excluded
331 women (64 cases; 267 controls). We also excluded 15 women (3 cases, 12
controls) missing study reference month and 63 cases diagnosed with breast
cancer during pregnancy. For the present analyses, we excluded 1,115 women
(231 cases; 884 controls) with 2 1 multiple birth. Finally, we excluded 1,149
controls that were matched to excluded cases.

Our initial eligible sample for the present study included 7,890 cases and

30,692 controls with singleton births. We excluded 57 cases and 303 controls

142

with missing/implausible gestational age, 16 cases and 42 controls with
missing/implausible birthweight, 33 cases and 116 controls missing education or
infant gender, and 193 cases and 728 controls with missing/inaccurate parity
data (3.8% of cases and 3.9% of controls excluded due to missing data). Finally
we excluded 1,121 controls matched to excluded cases. The ﬁnal analytic
sample size was 7,591 cases and 28,382 controls.

3.7 Statistical analyses

Demographic and pregnancy-related factors were compared between
cases and controls using frequencies and proportions. Statistical signiﬁcance
testing of these differences was conducted using the chi-square tests for
categorical variables (age at reference, year of ﬁrst live birth, education at ﬁrst
birth, infant sex in ﬁrst birth, infant sex in last birth), the Mantel-Hansel chi-square
for trend for variables with three or more ordered categories (parity, age at ﬁrst
birth, age at last birth, birthweight, gestational age, and fetal growth).

We examined separately the associations between breast cancer risk and
birthweight, gestational age, and fetal growth in first births for all women and last
births for women with 2 2 live births. Odds ratios (ORs) and 95 percent
conﬁdence intervals (Cls) were obtained by ﬁtting conditional logistic regression
models to the data, using maternal race and maternal year of birth as
conditioning variables. Several potential confounders were considered, selected
a pn'on' based on previous literature, including age at ﬁrst birth and last birth,
maternal education at ﬁrst birth, parity, gestational age in ﬁrst and last birth, and

fetal growth in ﬁrst and last birth. Each factor was tested individually in

143

conditional logistic regression models for each exposure and breast cancer risk.
Though not all potential confounders resulted in a change of 2 5 percent for the
main effects parameter estimate, we adjusted all ﬁnal models for ﬁrst birth
exposures for the same covariate set (shown in Table 2) and all ﬁnal models for
last birth exposures for the same covariate set (shown in Table 4).

To examine whether associations were modiﬁed by covariates
hypothesized a pn'on' to be potential effect modiﬁers (age at ﬁrst/last delivery,
years since ﬁrst/last delivery, age at reference, race, and maternal education),
we included a multiplicative term for the potential modiﬁer and exposure of
interest. We used the likelihood ratio test to compare models with and without the
interaction term. The Wald test was used to test for trend for ordered categorical
variables.

We also conducted conditional logistic regression analyses by histologic
type. Specifically, we modeled the associations for gestational age and fetal
growth in ﬁrst and last births separately with the following outcomes with ductal
cases compared to matched controls and Iobular cases compared to matched
controls (sample sizes for ﬁrst birth and last birth analyses are shown in Table 6).
For subgroup analyses and analyses by histologic type, we present only the
results from fully adjusted models. SAS version 9.2 was used for all analyses. All

tests were two-sided and p—values < 0.05 indicated statistical significance.

4. Results

4.1 Descrigtion of Samgle

144

Table 5.1 displays descriptive characteristics for cases and controls. By
design, cases and controls had similar attained age and race distributions. Cases
were more likely than controls to have a ﬁrst live delivery later in the study years,
> high school education at ﬁrst birth, 2 or 3 live births, and be older at their ﬁrst
birth. Infant gender in ﬁrst and last birth did not differ for cases and controls. Over
75% of ﬁrst deliveries for both cases and controls occurred prior to 1989. About
38% of cases were diagnosed before the age of 40 and 62% between 40-50,
with a mean age at diagnosis of 41 years (SD) =5.64).

4.2 BirthweightI gestational age, and fetal growth in ﬁrst birth

Table 5.2 reports the overall associations between birthweight (an
estimate of fetal growth), gestational age, and fetal growth (estimated via
established reference birthweight percentiles for gestational week) in ﬁrst births
and breast cancer risk. Birthweight. In unadjusted models, birthweights of <
1500 g and 1500-1999 (vs. 2500-3499 9) were associated with a reduced risk of
breast cancer (< 1500 9 OR = 0.80, 95% Cl: 0.61, 1.05; < 1500-1999 9 OR =
0.73, 95% CI: 0.56, 0.93), while birthweights above 4000 or 4500 g were
suggestively associated with a slight elevated risk of breast cancer (ORs,
respectively, 1.04 (95% Cl: 0.95, 1.14) and 1.08 (95% Cl: 0.89, 1.32)). Only the
ORs for delivery of an infant weighing 1500-1999 9 were signiﬁcant. Adjustment
for gestational age in ﬁrst birth, age at first birth, parity, and infant gender in ﬁrst
birth did not appreciably alter results (Table 5.2).

Gestational age. Women with a ﬁrst live delivery at < 32 weeks or 32-36

weeks had a nonsigniﬁcant decreased risk of breast cancer compared to women

145

who delivered a term infant, with similar unadjusted and adjusted ORs (adjusted
OR for < 32 weeks = 0.80, 95% Cl: 0.62, 1.04; for 32-36 weeks OR = 0.95, 95%
Cl: 0.85, 1.05). Women with a post-tenn ﬁrst delivery had a signiﬁcant decreased
risk of breast cancer (OR = 0.89, 95%Cl: 0.83, 0.96) and adjustment for potential
confounders (age at ﬁrst birth, parity, infant gender in ﬁrst birth, and fetal growth
in ﬁrst birth) did not appreciably alter the OR (Table 5.2).

Fetal Growth. ORs for delivery of an SGA infant or an LGA infant were
nonsigniﬁcant and close to 1.0, with similar results for unadjusted and adjusted
models (Table 5.2). Additional adjustment for education at ﬁrst birth did not
appreciably alter the results for any models for the three exposures of interest
(data not shown). We also examined fetal growth for all birthweight percentiles
(1-99) by gestational week in unadjusted and adjusted models (adjusted for age
at ﬁrst birth, parity, infant gender in ﬁrst birth, and gestational age in ﬁrst birth),
rather than using the SGA, AGA, and LGA categories. We found limited evidence
for an association per 1 percentile (adjusted OR = 1.00, 95% CI: 1.00-1.00; p for
trend = 0.18) or per 10 percentile change in birthweight-for-gestational age
(adjusted OR = 1.01, 95% CI: 100-102).

We next examined potential modiﬁcation of associations between
birthweight, gestational age, and fetal growth and breast cancer risk by age at
ﬁrst birth and years since ﬁrst birth (Table 5.3). For birthweight adjusted for
gestational age and fetal growth, results were similar by age at first birth (< 30, 2
30), with close to null results for low birthweight (< 2500 g) or delivery of an SGA

infant and a slight nonsignificant increase in risk for high birthweight (2 40009) or

146

delivery of an LGA infant. For gestational age, the nonsigniﬁcant decreased risks
observed for delivery at < 32 weeks or 32-36 weeks (vs. term) were similar by
age at ﬁrst birth; the decrease in risk for a posterm ﬁrst delivery was stronger and
significant in women 30 years or older at their ﬁrst birth (OR = 0.80, 95% Cl:
0.68, 0.94). We found limited evidence for modiﬁcation of results by years since
ﬁrst birth. Statistical tests for multiplicative interactions were not significant (Table
5.3).
4.3 BirthweightI gestational age, and fetal growth in last birth

Table 5.4 reports the overall associations between birthweight, gestational
age, and fetal growth in last births and breast cancer risk among women with 2 2
live births. Adjusted results for lower birthweights were all fairly close to 1.0 and
nonsigniﬁcant, as were results for higher birthweights with a slight increase
above 1.0 for the highest birthweight category 2 4500 9 (vs. 2500-3499) adjusted
OR = 1.08, 95% Cl: 0.88, 1.31. In contrast to ﬁrst births, adjusted ORs for a last
delivery at < 32 weeks of 32-36 weeks (vs. term) were above 1.0, though
nonsignificant. A posterm last delivery was unrelated to breast cancer (adjusted
OR = 0.97, 95% CI: 0.86-1.09). Delivery of an SGA infant in a last birth as
compared to delivery of an AGA infant, was associated with a non-signiﬁcant
reduction in risk of about 12%; the OR was slightly attenuated after adjustment
for potential confounders. Delivery of an LGA infant in a last birth was unrelated
to breast cancer risk (Table 5.4). Adjustment for education did not appreciably

alter any of the findings in Table 5.4 (data not shown).

147

We further examined whether age at last birth or years since last birth
modiﬁed the associations between birthweight, gestational age, and fetal growth
and breast cancer risk (Table 5.5). In general, results for birthweight and
gestational age were similar to overall ﬁndings, and all ORs were non-signiﬁcant
as were tests for multiplicative interaction. As for overall results, delivery of an
SGA infant (reference: AGA infant) was associated with a reduced risk of breast
cancer, but this ﬁnding was only found among women 2 30 years at last birth or
with 2 5 years since last birth (OR among women 2 30 years at last birth: 0.82,
95% Cl: 0.68, 0.98; OR for 2 5 years since last: 0.88 95% Cl: 0.75, 1.02).
Delivery of an LGA infant among women < 30 years at last birth was associated
with a non-signiﬁcant decrease in breast cancer risk of about 10%. The
multiplicative interaction between fetal growth and age at last birth was signiﬁcant
(Table 5.5).

4.4 Birthweight, gestational rile. and fetamowth by histologic gyge

We further examined the associations between gestational age and fetal
growth and breast cancer by the two most frequent histologic tumor sub-types
(Tables 5.6 and 5.7). The mean age at diagnosis for cases by histologic type was
as follows: 40 years (SD=5.64) for ductal cases and 43 years (SD=4.70) for
Iobular cases.

Overall, none of the effect estimates were signiﬁcant for analyses by
histologic types. Findings for risk of ductal tumors for both ﬁrst and last births for
the three exposures of interest, tended to parallel overall ﬁndings. For ﬁrst births

(Table 5.6), a non-signiﬁcant reduction in risk was found for ductal tumors (OR =

148

0.86, 95% CI: 0.63-1.17), but not Iobular tumors. For last births (Table 5.7), a
VPTD was associated with a non-significant increased risk for ductal tumors and
a reduction in risk of both ductal and Iobular tumors for delivery of an SGA infant.
Sample sizes became quite small for Iobular tumors and. We did not examine the
exposures of interest by other histologies because of too small sample sizes.
4.5 Gestational age and fetal growth by attained age, race, and education

Attained age. Results by age at study reference date (< 40 years, 40-50
years) for the associations between breast cancer risk and gestational age and
fetal growth in ﬁrst births and last births are presented in Tables 5.8 and 5.9,
respectively. For gestational age, results were fairly similar by age, though the
protective effect of a posterm delivery was signiﬁcant only among women < 40
years (OR = 0.87, 95% CI: 0.77-0.98). For fetal growth, ORs were below 1.0 for
both delivery of an SGA or LGA infant among women < 40 years, while for
women 40-50 years the OR for delivery of an LGA infant was above 1.0 and non-
signiﬁcant (OR =1.11, 95% CI: 0.99-1.25). Results for last births were fairly
similar by age, though delivery at earlier gestational ages was associated with
non-significant increased risks only for women < 40 years and a posterm delivery
in a last birth was associated with a non-signiﬁcant reduction in risk only for
women 40-50 years of age.

Race and maternal education. We also examined the associations
between breast cancer risk and gestational age and fetal growth in both ﬁrst and
last births by maternal race and education at ﬁrst birth (data not shown). Overall,

results were fairly similar by race (White; Black) and education at ﬁrst birth (5

149

H.S.; > H.S). However, for gestational age in ﬁrst births, protection for delivery at
32-36 weeks (vs. term) was only found for Black women OR = 0.68, 95% Cl:
0.53, 0.89). Further, the OR for decreased risk due to a posterm ﬁrst delivery (vs.
term) was stronger and only signiﬁcant among Black women (Black: OR = 0.65
95% Cl: 0.50, 0.80; White: OR = 0.95, 95% Cl: 0.88, 1.03, pinteraction= 0.005) and
women with > H.S. education (5 HS: OR = 0.99, 95% Cl: 0.89, 1.09; > H.S. OR
= 0.86, 95% Cl: 0.77, 0.95, pinteractlon = 0.07).
5. Discussion

Using data from a large, registry-linked, population-based case-control
study among Ml women s 50 years of age, we found limited evidence for an
association between low or high F6 and overall breast cancer risk among
women 5 50 years of age. Delivery of an infant at < 32 weeks in a ﬁrst birth may
reduce breast cancer risk, though this finding was in contrast to our hypothesis
and the underlying biological mechanism is not clear. We also found that a
posterm delivery was associated with an eight percent decreased risk of breast
cancer

We found that low or higher birthweights adjusted for gestational age in a
ﬁrst birth were associated with breast cancer risk. We observed a 23-30%
reduction in risk for lower birthweights. Results were statistically signiﬁcant only
for the association between breast cancer risk and delivery of an infant weighing
1500-1999 9 compared to 2500-3499 9. These findings for ﬁrst births are
consistent with two previous large registry-based cohort studies which estimated

fetal growth by birthweight adjusted for gestational age as a covariate. The ﬁrst of

150

these, conducted in Sweden with 2,216 cases for ﬁrst births, reported a non-
signiﬁcant increase in breast cancer risk for delivery of an infant weighing 2 4500
vs. 2500-3499 9 and a signiﬁcant increase in breast cancer risk per 500 9
increase in infant birthweight (63). This study also reported a non-significant
reduction in risk for birthweight < 2500 9 vs. 2500-3499. A second study in
Denmark of infant birthweight in the most recent birth, with 3,874 cases, reported
a non-signiﬁcant elevation in breast cancer risk associated with delivery of an
infant weighing >3500 grams vs. S 3000 grams, but did not look at lower
birthweights and breast cancer risk (79). A US registry-based case-control study
of 2,522 cases, similarly found a non-signiﬁcant reduction in risk for very low (<
1500 g) birthweight in ﬁrst births, but not for birthweights 1500-1999 9. This study
also reported a non-signiﬁcant reduction in risk for very high birthweights (2 4500
9) (65)-

When we examined the association between fetal growth, deﬁned using
birthweight percentiles for each gestational week, and breast cancer risk, we
found limited evidence for an overall association for fetal growth in a ﬁrst birth
and breast cancer risk, though ORs were in the hypothesized direction (i.e., slight
decreased risk for delivery of an SGA infant and slight increased risk for delivery
of an LGA infant). For last births, delivery of an SGA infant was associated with a
non-signiﬁcant reduced risk of breast cancer, which was limited to women 2 30
years at last birth and for women with 2 5 years since last birth in stratiﬁed
analyses. No previous study has examined fetal growth estimated using

percentiles for infant birthweights for each gestational week based on published

151

national reference values, which allows for more complete adjustment for
gestational age and hence improves estimation of the independent effect of fetal
growth on breast cancer risk (210).Elevation in risk of breast cancer for high fetal
growth and decreased risk for low fetal growth may reﬂect altered exposure to
maternal hormonal factors during pregnancy that are important to breast cancer
etiology. Most studies have found that fetal growth is positively associated with
maternal estriol levels (primarily of fetal origin), mainly in the third trimester (93,
95, 96, 109, 214) while studies did not ﬁnd an association with maternal estradiol
(93, 97, 214), which has been consistently implicated in breast cancer risk (215).
Other hormones that may mediate a positive association between fetal growth
and breast cancer risk include higher maternal serum levels of IGF-1 or low
levels of IGF binding protien-1, which have been inconsistently associated with
fetal growth (particularly during late gestation) (105-107, 111), or lower levels of
prolactin (109).

We found that delivery of an infant at < 32 weeks gestation in a ﬁrst birth
was associated with a non-signiﬁcant 20% reduction in breast cancer risk, which
was in the opposite direction of our hypothesis and most previous literature to
date. For last births, we found slight non-signiﬁcant elevations in breast cancer
risk of about 9-10% for delivery at < 32 weeks and 32-36 weeks. We also found a
signiﬁcant decreased risk of breast cancer associated with a posterm ﬁrst
delivery and breast cancer risk, which was more pronounced and signiﬁcant only
among women < 40 years in age and not among women 40-50 years in stratiﬁed

analyses. Our ﬁndings for a protective effect on breast cancer risk for delivery at

152

early gestational ages in a ﬁrst birth is not consistent with one case-control and
three cohort studies. Two European registry-based cohort studies, have reported
a signiﬁcant trend for increasing breast cancer risk with decreasing gestational
age in both last birth (68) and first births (70). Both studies reported an increased
risk for VPTD; the estimate for the smaller cohort (1,363 cases), but not the
larger cohort (5,474 cases), was signiﬁcant. A third cohort study with 2,318 cases
reported a significant increase in breast cancer risk for PTD in a ﬁrst birth among
women 2 40 years of age, but not for women < 40 years of age (69) and one US
case-control study reported a non-signiﬁcant increased risk of breast cancer for a
ﬁrst delivery at < 32 weeks (65). In contrast, and similar to our ﬁndings, a small
US registry-based study with 275 cases reported a non-signiﬁcant reduction in
risk associated with PTD and delivery at <30 weeks in a ﬁrst pregnancy (67).
Finally, a registry-based cohort study (63) and a population-based case-control
study using self-report data (66) reported no association between gestational age
and breast cancer.

Pregnancies that are shorter in duration during the third trimester have
been hypothesized to increase breast cancer risk, due to possible lack of full
terminal differentiation of the mammary gland followed by a time of increasing
hormone levels (65, 220). However, gestational age is inversely correlated with
AFP (229), which has been hypothesized to decrease breast cancer risk through
inhibiting estrogen-dependent tumor (94, 230, 231, 233, 293), and hence could
explain our ﬁndings of a protective effect of VPTD in a ﬁrst birth. Alternatively, our

ﬁndings could be attributed to confounding by factors that we could not adjust for,

153

such as low or high prepregnancy BMI or a more accurate measure of SES (222,
223)

We found a signiﬁcant decreased risk of breast cancer associated with a
posterm ﬁrst delivery and breast cancer risk, which was more pronounced and
signiﬁcant only among women < 40 years in age stratiﬁed analyses. To the best
of our knowledge, only one study has investigated posterm delivery and breast
cancer risk. In contrast to our ﬁndings, a cohort study in Sweden reported no
association for delivery at > 40 weeks (231), though they examined exposure in
last births. The mechanism for our ﬁnding of a protective effect of a posterm ﬁrst
delivery, which appears to be mostly among women < 40 years, is not clear. One
possible explanation could be maternal obesity, given that high maternal pre-
pregnancy BMl is associated with posterm delivery (294) and ovenNeight/obesity
during early adulthood may increase premenopausal breast cancer risk (295).
However, this explanation does not ﬁt with the null ﬁndings for last births.

In this ﬁrst investigation of gestational age and fetal growth by histologic
type of breast cancer, we did not ﬁnd much evidence that fetal growth or
gestational age were associated with risk of ductal or lobular tumors. Findings
were non-signiﬁcant, though the direction of the ORs for ductal tumors (the most
common histologic type (68% of breast tumors)) tended to parallel overall
ﬁndings. Sample sizes were small in several strata, in particular for exposures in
last births and risk of Iobular tumors.

Percentages of very low birthweight (1 .1), low birthweight (6.3), VPTD

(1 .1), and PTD (8.1) for ﬁrst births among controls in our study were fairly similar

154

to NCHS data for 1996-2004 (very low birthweight: 1.1% for all years, low
birthweight: 6.0-6.5%, VPTD: 1.6% for all years, PTD: 9.7-10.8%) (221).
However, our posterm birth rates for both ﬁrst births (15.2) and last births (9.9)
were high compared to data from the 1980/early 1990’s (296), which
encompasses most of the years of the current study.

Several limitations must be considered when interpreting the results of our
study. First, the use of registry data for study exposures and covariates meant
we had little control over how variables were measured. Overall, most studies
have found that demographics (e.g., maternal race, ethnicity, and age) are
accurately recorded (sensitivity and speciﬁcity >93%) compared to both in-person
interviews (254) and medical records (255, 256, 258, 260). Most investigators
have also found that number of previous live births (257, 260) and birthweight
(258-260, 263) are accurately reported on the birth certiﬁcate compared to
medical records. The accuracy of gestational age from birth certiﬁcates using
LMP, however, is a concern (272). Studies have shown, there is misclassiﬁcation
at lower gestational ages where birthweights appear to be higher than plausible
for a given week and also at late gestational ages (> 41 weeks), which may
incorrectly include term births (209, 272). Gestational age quality is dependent
on year in the US, however, and has improved over the years (297). To address
this concern, we used published cleaning methods, (209, 271) to reduce
misclassiﬁcation in our data.

We were also concerned that our ﬁndings may have differed from other

studies due to the way we deﬁned gestational age (e.g., LMP-based estimate)

155

and methods we used to ‘clean’ LMP. Though it is difficult to compare both our
deﬁnition and cleaning approach to other studies, because this information is not
reported in detail (63, 66, 67, 69, 79), we conducted two sensitivity analyses to
compare ﬁndings. First, we re-analyzed the association between breast cancer
risk and birthweight and gestational age in ﬁrst births, using the clinical estimate
of gestational age, with no modiﬁcations and excluding only births with
gestational age < 24 weeks or missing values. Results were similar to the
presented ﬁndings. We also re-ran the analyses using the LMP estimate of
gestational age with these same exclusions applied and again found similar
results.

Several limitations of our study are related to the use of registry-linked
data. First, SSN alone was used to link cancer and birth data as well as
additional births for study women. However, our validation work demonstrated
that linkages were correct and complete above 98% (see Chapter 3, Methods).
Simon and colleagues also demonstrated that SSN was a highly valid approach,
with only 2% of possible linkages missed using SSN as the only linkage variable
(245). Second, we had to exclude women missing SSN in either the cancer or
birth registry. Few women, however, were missing SSN in the cancer registry
(1.4% of cases) or birth ﬁles (1-5% among all births per year) and missing SSN is
unlikely to be related to case-control status. Third, we may have missing
exposure or inaccurate parity data if women moved out of state and had
additional births, which could be a possible explanation for the null ﬁndings for

last births. Fourth, outcome misclassiﬁcation is a concern because some controls

156

may have developed breast cancer either during the time lag between exposure
and the start of the MSCR (1978-1984) or in a different state. However, breast
cancer diagnosis, before the age of 50 is rare. Further, we found that only 8
controls were diagnosed with breast cancer before age 51 during 1978-1984 by
linking to the Detroit SEER registry, which if this estimate is extended to the
entire state, indicates only about 0.05% of our controls could have been
misclassiﬁed. 5

Finally, as have most previous studies, we only had variables available in !

 

the birth registry to consider as covariates. For example, we did not have
information on maternal pre-pregnancy BMI, which is an important determinant of
both birthweight and length of gestation and also breast cancer risk. Hence,
residual confounding cannot be ruled out as a possible explanation for our
ﬁnding. However, most previous studies have also only had information on
pregnancy-related factors such as age at deliveries, number of births, race,
maternal education, and infant gender (65, 67-70, 79).

Main strengths of our study include the large sample size, population-
based state-wide design, and use of registry data. Registry-based studies are
less subject to key bias found in case-control studies, because exposure data is
collected before breast cancer diagnosis. Further, using existing registry data
provided us the opportunity to select controls from the source population that
gave rise to the cases, without being subject to potential bias due to lower
participation rates among controls. In addition, we had high quality cancer

registry data that meets NAACCR standards. Finally, this is the ﬁrst study to

157

examine fetal growth defined using birthweight percentiles for gestational week,
to examine exposures in both ﬁrst and last births, and to investigate birth
characteristics and breast cancer risk for speciﬁc histologic sub-types.

In summary, we found limited evidence that fetal growth influences breast
cancer risk overall and some evidence that earlier or late gestation in a ﬁrst birth
may be associated with a reduction in breast cancer risk. We did ﬁnd a non-
signiﬁcant reduction in risk for delivery of an SGA infant in last birth, but not an I-

LGA infant, which was statistically signiﬁcant among women older than 30 years

 

at their last birth in stratiﬁed analyses. This ﬁnding could reﬂect the inﬂuence of
lower levels of maternal estrogens or IGF-l levels during pregnancy on
subsequent breast cancer risk. The suggestion of a reduced breast cancer risk
for a ﬁrst delivery at < 32 weeks could reﬂect elevated levels of maternal alpha-
fetoprotein, but this ﬁnding is in the opposite direction of our hypothesis and most
previous literature to date. Our ﬁnding of a protective effect of a posterm delivery
in a ﬁrst birth, which has not been previously reported, warrants further study.
Few studies have examined infant birth characteristics and maternal breast
cancer and further work is needed to characterize these associations among
different populations, including studies with information available on other

potential confounding factors (e.g., body size).

158

 

 

3.8V 42% 6.3 «Ba 313
GS 246 ads 55 8-8
ANN: 8% ANN: m8 3:8
8.3 3m 33 «mm 8-3
2.9 09 $9 mm mmv
A833 3% 855%: um om<
«Ed 33 mmw adv omm 52-22
8.3 48.2 55 Bad 82-82
2483 80.9 8.3 meme 82-9.2
$3 30 24.3 42 92.52
5.5 .6 23> .mEoﬁS.
«bro Ros 48...” 3.3 mom xogm
:38 $98 3.3 83 922,
5.5 “mg: E mom. 5505.).

g c g c

N38": 53":
o:_m>-a w_o:coo mommo

 

159

238385
39:25. E 53w omwmm->:m_mmm .6950 385 can 920m“.
>ocmcm9n. .m_o=coo new $28 8.. mozmtmsomcmco ozﬁzowmo ....m min...

833 .m

ﬂ500.0 v

m58.0 v

m58.th

m58.?

mmd

89 Bud
8.8V 83
6.08 $4.9
3.3 Sm.»
3.: 83

6.3 oz}
:8: mm:
3.3 95.9
$.08 32:

ewe £3:
63 39?

Ass m5
6.9 83
6.4: Sma-
ammv 9;
Gas 83:

Avdwv mmod

8.8 NB
3.8 85
£88 83
6.3 m 5;
8.8 Be

as Em

$8: 2.3
8.3 83.
3.5 Stu

€69 «mow;4
6.3L wmmd

a: E
8.9 4%
Re: 08;
$5 $3
2.5 £3

3va meta

+8

8-8

8-8

«New

om v

$.82 55 Em a 0?.
v

1—NC‘0AI

magma

.828 29... A
828:8 :9: w
05% E: a 8:833
38.82

82.3?

82-82

82-59

82-22

5.5 92. E: .6 28>

9. N
82:80 .3 «33

160

.wctﬁ m>__ N N 53> cmEoB mcoE<m
.mxmm; vm M 5:38....
.o 59.9 5:5 2:5 o>__ ..o 69E? 2 3:8 cm 2.5 m>__ .0.. 3.89 5.5 cm: 2:0 o>>.v
.525 .2: .m w. .o .t .0. 60mm coEoz, $62050
6:9. .6. .8... manna-Eu 6852.932

8
.co_.m_oommm 3.3% 5.. .mm. Swag-EOm

 

m8... 6...... 866 66... 8.8 6656”.
6. .6. 8.6 6.86. 86.8 222

656 so 2 :8 235

68.0 66... 68.6. 6.9.. N86 238
:6. 36.1 6.8. 866 222

. £6 .2: 5 .8 as...

n.58.: 88. was 68. um. ovm
6.2. 486 6.6.. 5. 8-8

6.8. $66 6. .... 86.8 3-8

6. .m. .666 6.8. .66.. 8-8

69.05.. .3. v8 8v

6.8... 5.6 ...-s .6 8.4
835.80 ...... 833

O

161

88 6.8
E 96 _m>m_ mod w c .6 88556 262.268 $5 6966:? 65:36:00 mm mom. 652.9:
new 5.5 .o .69. 6599: 56> m_mnoE 86859 2.662 6:036:00 E 826.. 895 6106

 

 

 

6...-66. 8.. 6...-86.8.. 66. 866 66. 68 <9
68999. 8.. 68999.8. 666.6...8 6.8. 866 <o<
66.-86. 66 66.-86. 86 6.... .86 6.... .8 <06
652.8386".
666-86. 86 666-86. 86 66.. 6.6... 66 .. 86.. .8 N
68999. 8.. 68999.8. 6.8.68.6 6.8.866 ...-.6
66.86. 86 66.86.86 6... .8. 66. 6.6 8-8
66.-86.86 66.-86.86 .... 86 66. K 8v
.3695.
mum Econ—6.600
66.-.66.... 66.-86.8.. 6... 86 6... 8. 88m
6...-86.8.. 3.86.8. 66. .86 66. .8 83.68..
66. .-.-66. 86 66. .86. 86 6.8. E6 68. 686 88-88
68999. 8.. 68999. 8.. 6.8.86. 6.6.9.6 8888
6...-86..6. 5.76.6.8. 6.4. 8... 66. ..6 8888
66686.86 66686.86 6... 86 66. 8 86.86.
6...-.66.R6 66.-66.86 .... 68 66. 8 88v
6. 3662.66

0.668. «0 6.668. «6 .9. c 3.. c

68.8"... :86": .

£9500 wwwmu

 

162

5668-86.. 86322 5 636 6686-2669. 688 .896
new 8966... 288666. .868. om w mum .6 50:8 .6665 ﬁEmﬁE new 5.5 .65
5 £265 36.. new .096 .mco_.m.mmm 290385 .56.... .o. 62.6.. 6260 .~.m 636...

..xmmz, 6:98.666 .3. sow A 6.2963539 53. 6.5.5 - (m... .w__.cmo.ma sow-o.
6... 503.8 6.299585 53> 9:95- <0< 9:583 so. v 6.5963239 55. 6.56.5- <0wu

.5263 :99 20 29955.5 26 89.6 56.: m... 8ch 986E
.0. 2:0. ctﬁ .8... 5 wow 6:28.63 .9065 59.: m... m_ 660E mum 6:98.666 8ch 26.69.:
.9 28. 5.5 .9... s 5296 6.9 .88 .56 .96 s .886 895 .56 .9: .6 86 6.. 8.8.3.

8_ 88
A. 252.680 68.8.. ...8 668.26.

69.5.5“... ..N m 636...

163

ii ‘0; ... L11”?

.2003 .mcozﬂmmm >5 58 A 2:995:30

53> 2:95 - <04 6.5523 58.? m5 009200 25995220 53> 2:85- <0< 6.50828 53 v 2:995:30 53> 2:85 - (Ow
.comtm 59: m5 2 539m
.98 20 29035.5 29.; 239: .2 2:8 5.5 .25 5 wow 500.88% .9850 58: m5 2 EnoE mom .mcSSmmm 9ch £200.: .2 2:3 5:0
52: c_ 5305 .98 .580 02c :_ 53% ESE .Amumtaoaam :93: 5.5 “2: 8 wow 5983903 :25 55 52: 8:6 2mm> >5th .8 382nm
9m 05 mmzmtms 05005608 8 89 8:58.: cam 55 “.o 0mm> .mEEmE 53> $209: 02898: 2860. .mcoEucoo >0 82.20 mmom

 

 

 

at 788 Be N8 8.788 00.0 8 8.788 N: 88 E788 8.? 8w <9
$085.68 00. F v in $290.68 00. F 88 3009890 00. F 8. F 689998 00. F >0 r .... <o<
80. 78.8 8.0 83 cm. 7:8 80 8 A8. 78.8 004 80 80. 78.8 8.0 m8 <8
052,20 Em“.
>00. 78.8 8.0 :0; 8?. 78.8 8.0 8 $00.88 8.0 08 8. 78.8 V0.0 v8 8 N
8009890 00+ 8% $8988 004 «8 $8928 00.. 3.0; 608588 00.? 80;. :18
60. 78.8 8.0 E. 8.788 8.0 8 A2. 78.8 8.0 82 :0. 78.8 «0.0 88 8-8
>00. 78.8 8.0 8 £08-88 8.> m 8. 78.8 8.0 8 80. 78.8 8.0 8 8 v
Amxmmsv mom
_m:ozmumm0
at 78.8 80> 08 8. 78.8 8.0 8 an. 78.8 0: >8 2:. 78.8 00.? 88 000v N
85.890 004 08.8 389998 00.? Ev $085.68 008 80; 809890 008 83 008-08
60. 78.8 8.0 88 8. 788 N08 8 am. 788 8.0 wt 60. 78.8 5.0 88 008 w
8 290385
0088 m0 50.88 2088 m0 888 0088 m0 38.88 0088 m0 80888
$80 $me $me 8..me
2mm> m N 28> m v 2mm> on m 2mm> on v
5.5 we 8% 28> 5.5 E: a m0<

 

«$08.88: 8022.2

5 >35 ummmm->5m_mmm 08.50 892m 8.6 220mm. >ocmcmmi .525 “2: 85m 2mm> new 5.5 *2: *m ¢Q A
w mam 5a 588 835 3589: new 5306 .98 6mm .mcozﬂmmm 5:995:53 Eat: 5“ 8:8 .3 m mam
. . 8 00.0
80¢. .8

.m 208

164

5 wow .mcozﬁmmm Erma .55 “mm. E 53% E92 .593 “mm. «m mom .53 .9: am mum .8 3633 om_<o

.mmﬁmtms mchzﬁcoo 3205 25 £822

a

$8.8ch 2:5 02. 908 .o 92 5:5 5:8; 83.05 mﬁmh .mmﬁmcg mchzﬁcoo
mm 89 .mEQmE ocm 5L5 M6 $9» 6529: 53> 2258 56359 2360. .mcozﬁcoo B 09ch mmom

 

 

 

63-3.0390 63-8986 3.33va 2.3 3: <9
$823933 3893933 8.353: 32:03 <o<
€972,956 23-50590 6.: $3 69 «3“ <8
£93566“.
33-8950 93-3930 89 85 :9 mm». N:
$899933 $899233 $838.2 8.983 :15
$3.398: 63-398; $9 9.3 :9 2m 8-8
63.399: 30.7398; :9 m: :9 mm mmv
Amxmmgv mom aces—330
23-898; 83-39:? 89 NE :9 N2 89;
5.7598; 5.7595: aﬁrmmw. 8.83% 83183
60.75.0590 30.78.0390 6.933 8.99me 88-88
8892933 3899933 3338.» :99 53 $3.88
$4.89 5: 83-8934 39 :3 :9 our 85:88
63-3986 3.7%.95: :9 v2 :9 mm 82-83
63-50:? :97399; 69 o: :9 mm 82v
9 £92.55

059°8va 0:09.8va 29 c 3.9 c

88.2% catmus

£2200 mmmmo

 

Emmi new 920mm >ocmcmmi 98> on w mam E .8ch “mambo .mEQmE ncm 5.5
“mm. E 5265 .93 ucm .wmm .mcozﬁmmm .Em_m>>ct_n 285 8.. moan. £60 in min...

«9838: $963 5 >85 umwmméaamm .8ch

165

.903 .mcgﬁmom :9 58 A £36393 5;, 35.2 - <9 .23528

58-8 2: 5922 259395 5:, 252- <0< 63528 52 v 98632:: 55, megs - <8
.98? SE: 9: 9 wow .mcozﬂmmm
was: 5.5 mm. 5 £265 .99 new .833 59: m5 2 2365 663590393 925 5:5 “mm.
nmsczcoo {m can...

166

.28; 58:88 >9 :8 A 959895 5;, was: - <9 65:85:

58.? 9: c8258 9355::5 53> 9:95- <0< 65888 59 v 9cm_m>>::_n 5:5 9:9:_ - «Gm:

.Eta Eon :_ 9m _m>m_ mod w c 8 E8536 >__8=m=9m mmo .Avomdm u 5 2:5 82. 99: .5 02: 53> :mEoz, 9

c9_E__ mBmh .comtm £9: 05 9 wow 553989 98:; .289: :2 >_:ov 5:5 “mm. :_ £265 :99 .388 EmE 9: m_ £39m .99 :o Egosctﬁ
88:2, 289: :2 >_:ov 5:5 9mm. 8 wow .mcoz98m .55 9mm. :_ .898 9:95 :3: “8. 85m 98> .55 6:: 9m mom >9er :8 09.0.38

:8 838m; mchzﬁcoo mm 89 $599: :8 5:5 :0 50> .899: 5:5 280:: :2898: 2860. 8:036:00 >9 82:8 98>, mmom

 

 

 

80.739 8.0 9% 83-59 8.0 09 39.739 .89 Em 5.799 8.0 «mm <9
@8989 03 8 F .m @8299 8.? So @8989 03 :3 $8999 8. F on? <2
:0. 729 8.0 gm an. 799 8.9 mm 88-89 8.0 me am. 739 5.9 :9 <8
95265 .98...
:o. 789 3.0 95 :m. 789 m3 8 :F. 739 03 mom a: 729 8.0 8m 9. N
3898: 8.9 88 $8999 03 83 A8555: 2: 83. $8989 8.9 83 3.2
83-89 o: :8 51-5989 5 am. 789 Be 09 @3499 3.. «9 8-9,
5.509 2: R 23-59 m: : 63-39 3.0 om avaéwov o? 9 mm v
3805 wow
3:039me
A9. 789 m3 mum am. 789 :3 :8 am. 7:99 2: 9m 5. 799 8.: EN 88 N
38289 8.? 3 F .m $8989 03 am 38989 03 23 68999 8.9 «:3 88-83
:N. 759 Se 2: EV. 389 8.0 ow 3N. 2:9 8.0 m : at 799 9: mm 28 w
9 £9895
599 m0 :8.mu5 59:9 :0 signs 59:9 :0 ammmus 2099 mo 65. HE
880 880 880 8me
98> m N 28> m v 98> om N 98> om v
5:3 :8. 8% 28> 5:5 “mm. a 0?

 

>388va 5952.2 5 >85 :88
.998m .880 585 ucm 990m“. 5:885 its .92 85m 98> ncm 5:5 9mm. 8 wow >2 98> on w mam 8 80:8
885 .88ng ncm 5:3 98. :_ £265 .99 6mm .mcoz98m 9:995:33 9:95 :9 8:9 960 8639: .m.m 23¢...

167

 

. ‘0 rr

.933 .9253: 9 58 A 999595 5:, 995 - <9 958.8.

58.? 9: 5928 9:995:59 5:5 9:95- <O< 55590: so? v 9:995:95 5:5 9:95- <0mn
.83 38 90-8. 9.7.:

85390 99:2 5.32 ucm comm mooo $0-00. m5m: 85890 952, 99:3 .990 .50.? 59:
m5 9 £265 .99 923 >_:ov 5.5 99: 5 mam .mcozmummm .3093 59: m5 9 0mm .mcozmummm

90:2, >_:ov :59 9: 5 £265 _99 5:5 9: 5 53% E95 .593 9: 9m mom 5:9 :9 “99:5
ucm $39.9 m5:05u:oo mm mom: ucm :59 90 50> 5599: 5:5 996.: :o_mm9m9 9550.
5:36:00 >9 09:3 69:8 .w> .932 :5 69:8 .m> .903: :9 2993mm :9mezmm 99> mmom

 

«5.729 8.0 mm €3.39 9: an... <9
68993 8.9 mmm $8999 03 S3 <0<
6N. 7:9 3.0 2 so. 739 3.0 9% <8

“2565 .99.
E. 78.9 ~90 «m :8. 99.9 3.0 mm: 9. N
689999 9: F 3 $829»: on: no? in
am. 789 3.0 t. 60. 78.9 8.: 9% 8-8
60.939 8.0 m E. 78.9 8.: mm 8 v

9093 wow $53930

 

209.09 m0 .330 :09va KO 330

 

Am_o._9:oon~mm€ Homoummmmov 8506 Fu9obcoo ”5 _. .mummmmov
5.30.. .925

 

3838: 59:22 5 >86 839323: 588 585
:5 990m“. 5:935 .95 060.29: >9 98> om w mum am .858 9985 .mE9mE
ncm 55 9c 5 £265 .99 ncm mom .9039me :9 8:9 998 v99€< 66 «En...

168

 

.9095 .229me 9 58 A ngoéts 5? 9:95 - <9 65:850.

58-9 9: 5923 33635.5 55, 9:85- <0< 65583 so? v 9:063:39 5;) 3:35 - <Owa
.83 88 90.00. mg»: 853% 99:3 5.30. 2m 83 88 90-8. 9%: 8:533

92> 99:3 .2030 .239 02. 92: .o 92 55, $80; 2 8:8: wmmzmcm 5.5 “m3 .comtm 59: m5 2
68E mom 65:980. 923 2:9 5.3 “mm. c_ 5266 .93 .comtm :68 05 m_ 5265 _m§ 0..ch 2:3
5.5 “mg 6 wow _mco_§mmm Erma .55 “mm. 5 52mm ESE .55 “mg 3 com .55 6.: am mom .8
3623 new 8399 mchEucoo mm mom. van 5% 9o .99” EEQmE £5, m_m_uoE 86859 2560.
6:86:00 3 329% .928 .m> 5:50. new .9360 .m> .903 .8 29933 8558 903 mmom

 

48.739 2: mm 33-8.9 So so <9
689039 8.? 3m 382929 03 89.0. <0<
am. 739 Ed mm Ga. 7.89 8.0 mum <8

95266 Emu.
5. 78.9 who 5 6?. 73.9 «3 mm «v N
389929 8.? am 38993 03 $3 :13
an. 739 :3 cm Cm. 75.9 m E mmm 8-8
an? F9 3.0 m 6N.~-~m9 93 mm mm v

@6029 wow .mcozﬁmmw

 

20$mmv m0 mmmmo 20$m0v m0 330

 

Ammo. Pum_obcooumvvummwmov 3 565.8280 ”www.mummmmov
5.30.. .3030

 

«9838: c.3225. 2 >25 nwmmméuanmi .850 gem
new 290m“. 5:235 .33 060.29; B Emmi om w mom “m .8ch Emmi .mEQmE
new 2:5 “mm. c_ £32m .209 new mum ﬁaoﬁﬁmm 5». .832 mace 3623‘ Ed win...

169

58-8 2.. 5928 9:93.95 5? 93.5- <0< £828 58 v 959955 5;, 9:95 - <8

.23; _wco_.w.mmm >9 sow A ﬂanges 55, new: - <9 6.9.023

n
Ea 28 2

9w _w>w_ mod w n aw «ewo:_em_m >__wo:m:wﬁ mmo .cowtw e_wE 05 w_ 5265 .98 922, £ng .8 zeov
etﬁ 6.: e_ mmw _weo:wummm 608:... 59.. 9: w_ wow _weo:wummm 9mg, m_wnoe. .8 2:8 5.5 3.: 5
5.58m _w~8 .55 3.: e_ .wnewm E85 .55 5.: «w wow éewq .8 nwuwzﬁw new mw_nw_.w> me_eo:_neoo
mw mow. new 5.5 :o .wg _we.8we. 5:5 930:. 563.3. 2562 _weo:_neoo .5 nw>_.mn 9m; mmow

 

 

 

m.wm> om-ov

2w?» on v

dm. 78.9 :8 5. A3. 73.9 8.0 .8 <9
.8583 8.: mid A8853 03 83 <o<
60. 789 8d 2m 8: 73.9 8.0 m3 <8

952,299“...

:5. 759 8d 08 893.9 Bo 8m Nv N
305.999 2: 89m 389999 8.: m :3 81m
3.78.936 own 63.898: «2 8-8
63.89 2.0 9 83909 8d 8 .8 v
Amxmwzc wow _weo:wuww0

209.89 «.0 38.?9 59°99 mo 68qu
mmmmo mmmmo

 

«98389 $955. ... 39m nmmmméaamm
.moewo :wwwﬁ new m.80wn_ >oewemm.n_ .mow nwe_wzw .3 .8er :wwwﬁ _we.8we. new
wees new 7...: 5 £265 38 new mow _weo:wummm .8 8:9 mnno n8m:..n< .w.m saw...

170

 

.58; 889$: 2: so: A 9:99:55: :5; 9:95 - <0: 6::85:
58-8 :5 8928 93625: :5; 9:95- <o< 65:85: :8 v 9:90;:5: :5; 9:95 - <09:

.95.: 50:

:_ 8m 55. mod w a 9m 58:5:5 2599998.: 35 .9085 :5... m£ 9 £26.: .88 05:3 985:: .8
2:8 5.5 “mm. :_ m:m 58:83: 508:0 :_m:. 9: 9 m:m 55:88: 80:2, 985:. .8 2:8 5.5 “mm.
:_ £26.: .88 ...:5 9mm. :_ .85: 9:85 5:5 59 um m:m 5:5 5.: 9m m:m >58 .8 8:83.88 ucm mom.
558.9: ucm 8.5 :o .mm> _m:.8m:. 55> 986:. 55338 o:9:o_ 5:995:00 >5 82.8: 885 8:0:

 

 

3.8-3:: 5.: at. 83-8.98: 3: <0:
3899:: 8. F :8: 30:99:: :3 23 <o<
6:. 8-8:: mm: mom 3. 8-8:: mm: ..8 <8
::§o.:_9mu_

6:. 78:: 3,: 8: 5. 8-8:: :3 N8 .9. N
58:99.9: :3 5 E: 30:99:: :3 «No. 8 :1:
83-3.98: :8 83-59:? m8 :3:
6:. 78.9 mm: 8 8998:: :3 :N N: v
Amxmmé mmm 55:89-90

59.8: :0 58:89 5959 :0 68.?5
$30 $80

 

83> 86:

88> ov v

 

93:98:: $955.2 5 >85 8892998:
.858 Emmi :5 280mm 5:99.29 .3: 3:58: >5 .858 5:85 5589:
ucm 98.5 “mm. :_ £26.: .88 ucm m:m 55:89:: .8 8:8 860 n8win< .:.m 253.

171

EPl-logue
1. Summary of findings

This dissertation began with a systematic review of the literature to identify
studies of the associations between perinatal factors in a woman’s own
pregnancies and maternal breast cancer risk. This work was initiated due to an
interest in understanding the biological mechanisms underlying the role of
pregnancy in relation to maternal breast cancer risk. Direct study of maternal
hormones during pregnancy and subsequent breast cancer is difﬁcult due to
methodologic and practical issues (e.g., long time span between exposure
measurement and breast cancer diagnosis) (122). In lieu of biologic measures,
perinatal factors, which may be proxy measures of this environment (65, 66, 79),
can be examined (118). As an example, ﬁgure 6.1 summarizes the proposed
relationships between fetal growth and later maternal breast cancer, as well as
the potential biological mechanisms underlying this association. After an
extensive literature review and considering the study design and data limitations
for conducting a registry-based birth and cancer linked study in MI, we decided to
focus on the two perinatal exposures that have been least studied, gestational
age and fetal growth, and which are measured fairly well and were available in
the MI birth ﬁles for all years of the planned study.
Aim 1. (Perinatal factors and maternal breast cancer risk: a review of the
epidemiologic literature)

To systematically identify and descriptively summarize all published

studies on perinatal factors and matemal breast cancer risk, including maternal

172

 

conditions of pregnancy (preeclampsia, PIH, GDM, pregnancy weight gain) and
infant birth characteristics (fetal growth, GA, multiple births, sex). These pen'natal
factors have been shown to reﬂect an altered hormonal environment during
pregnancy, and may provide insight into the biological mechanisms undertying
the role of pregnancy in breast cancer etiology. Given that that inﬂuence of many
breast cancer risk factors depend on menopausal status (or by proxy age at

diagnosis), we present ﬁndings stratiﬁed by menopausal status/attained age

 

whenever available.

In this ﬁrst paper, we summarized data on published studies for seven
perinatal exposures, including preeclampsia/PIH, multiple births, GDM,
pregnancy weight gain, fetal growth, gestational age, infant sex. Brieﬂy, we found
across 16 studies, that multiple births may protect against breast cancer.
Preeclampsia was found to decrease risk by up to 20% in all but two studies;
results may be modiﬁed by infant sex. Breast cancer risk may be increased by
delivery at earlier gestational ages or elevated fetal growth in a ﬁrst or last birth,
but data is sparse. Infant sex does not appear to be associated with breast
cancer. Data on associations between gestational diabetes, pregnancy weight
gain and breast cancer risk is limited and conﬂicting. In summary, we found that
additional research in the area of perinatal factors and maternal breast cancer is
needed, including studies of perinatal exposures and breast cancer risk with
ample sample size to consider jointly several key effect modiﬁers (e.g., time, age,
birth order, menopausal status) with data on potential confounders, including

non-pregnancy-related factors.

173

Aim 2. (Pregnancy-related factors and risk of breast cancer overall and by
histologic type, a registry-based study of parous black and white younger
women)

To investigate the associations between later age at ﬁrst and last delivery.
mber of live births and breast cancer overall and by the two most common
histologic types (ductal and lobular) among parous White and Black women 20-
50 years of age in MI, 1985-2004. We will examine potential modiﬁcation of the
above associations by race (White, Black), and maternal education at ﬁrst
delivery (a measure of 858).

As hypothesized, we found that women with a later age at ﬁrst or last birth
had increased breast cancer risk. We also found that multiparity was associated
with increased risk compared to women who had only one birth. Further, we
found that later age at ﬁrst and last birth increased risk of both ductal and lobular
breast cancer. Contrary to our hypothesis, we did not find evidence that risks for
age at ﬁrst and last birth were more strongly associated with lobular as compared
to ductal breast cancer. We also found that multiparity (versus primiparity)
increased breast cancer risk similarly for both ductal and lobular cancer. Some
differences by race were found, in particular, we found that among White women,
later age at last birth significantly increased lobular breast cancer risk, but was
not associated with ductal cancer. Among Black women, associations between
pregnancy-related factors and risk of ductal or lobular tumors were non-
signiﬁcant, though number of lobular cases were small. Notably, we found a non-

significant reduction in risk associated with both later age at ﬁrst birth and risk of

174

qmg'1(.z

lobular tumors and later age at last birth and risk of ductal tumors. These
pregnancy-related factors were associated with increased risk among White
women.

Aim 3. (A population-based case-control study of fetal growth, infant
gestational age at delivery, and maternal breast cancer among younger
women)

To investigate the associations between fetal growth, gestational ag§_at
delivery (2 41 weeks (posterm), 37—41 weeks (term), 36-32 (preterm), < 32 weeks
gestation (very pretemi)), and breast cancer overall and by histologic type among
parous White and Black women 20- 50 years of age in MI, 1985-2004. We will
estimate fetal growth using published birthweight percentiles for gestational
weeks 24-44, described in the methods section in Chapter 5. We will examine
potential modiﬁcation of the above associations by race (White, Black), maternal
education at ﬁrst delivery, later age at delivery (ﬁrst and last), and shorter time
since delivery last delivery. We will examine the perinatal exposures in ﬁrst births
and last births (among women with two or more births).

Using data from a large, registry-linked, population-based case-control
study among Ml women 5 50 years of age, we found that low or higher
birthweights adjusted for gestational age in a ﬁrst birth, an approach commonly
used by cancer epidemiologists to estimate fetal growth, were associated with
breast cancer risk, including a 23-30% reduction in risk for lower birthweights and
an 11% increase in risk for high birthweights. Results were statistically signiﬁcant

only for the association between breast cancer risk and delivery of an infant

175

weighing 1500-1999 9 compared to 2500-3499 9. When we examined the
association between fetal growth, deﬁned using birthweight percentiles for each
gestational week, and breast cancer risk, we found limited evidence for an overall
association for fetal growth in a ﬁrst birth and breast cancer risk, though ORs
were in the hypothesized direction (i.e., slight decreased risk for delivery of an
SGA infant and slight increased risk for delivery of an LGA infant). For last births,
delivery of an SGA infant was associated with a non-signiﬁcant reduced risk of
breast cancer, which was limited to women 2 30 years at last birth and for
women with 2 5 years since last birth in stratiﬁed analyses.

We found that delivery of an infant at < 32 weeks in a ﬁrst birth was
associated with a non-significant 20% reduction in breast cancer. risk, which was
in the opposite direction of our hypothesis and most previous literature to date.
For last births, we found slight non-signiﬁcant elevations in breast cancer risk of
about 940% for delivery at < 32 weeks and 32-36 weeks. We also found a
signiﬁcant decreased risk of breast cancer associated with a posterm ﬁrst
delivery and breast cancer risk, which was more pronounced and signiﬁcant only
among women < 40 years in age and not among women 40-50 years in stratiﬁed

analyses.

176

 

 

 

 

.3ch
$35

 

 

Eman_m>mo ucm
£390 9:20 EwEEms.
co 882E. _mcoE._oI
nogm_9->ocmcmmi

so cozm§_< 3:56.”.

 

 

 

 

 

 

 

dogma ooEw
Mali ucm 50230
a mam .3 SEE:

AS c_29a9£-mca_m .mEoﬁE E
56205 $529: a
ocozouwomoa .mEmﬁE H

_.-n_o_ .mEmﬁE e

E 662.8 a

653 ﬁEmﬁE a

”St: boumxoowmm

on :2: 5305 ‘88 :9:

Socmcmoi 9.36:0“.
vcm 9:50 EoEco._>:m
.mcoEBI no.3? 53:30.“. .__

 

 

 

 

 

 

I

55:35

mc_>>o__0n_ can @550
EmanE>oo new £390
25.0 meEms. 853:5

 

 

 

 

 

 

 

 

toumoEE co ucoaou 9.05223

5205 9.8-29.1 F

56205 e

532:

new 220m”. £280 33-5.35 e
5305833 25:05 cmEsl e
ocoaoymwmoi a

mammozmw e

«coﬁcohgcm
_wcoE.oz->o:~:moE ._

 

 

 

.8ch 585 EEQmE ucw .5266 RE Socmcmma
mESo EmEco.._>:m .mcoan: .mEQmE 9:. .rd 959".

177

2. Limitations and Strengths

Two main limitations should be considered when interpreting our results.
First, it is possible that parity data is inaccurate for some women due to
movement out of state for women who had more than one live birth, which could
lead to underestimation of number of live births, incorrect classiﬁcation of a birth
as a last birth (for age at last birth), and lack of complete adjustment for number ”-3
of live births as a confounder. Further, it could be said that it is more likely that

controls moved out of state and had additional births because cases were

 

required to have both their births and cancer diagnosis in MI; hence it is possible i
that our controls were more likely to have underestimated number of births and
ORs could be biased away the null.

The other key limitation is that our study population has a much higher
percentage of women with later age at ﬁrst births and last births, as compared to
other population-based studies. Table 4.7 displays distributions for age at ﬁrst
birth overall and by birth cohort for controls, with a comparison to a recent birth
registry-based study among younger women conducted in Sweden (63). We
expected these higher distributions, which are due to the data restrictions of our
study (i.e., ﬁrst live birth in 1978-2004 and year of ﬁrst diagnosis during 1985-
2004), and we carefully ensured that controls were selected from the same
population that gave rise to the cases, so this should only inﬂuence the
generalizability of our study ﬁndings, and not the bias the odds ratios.

Main strengths of our study include the large sample size, population-

based state-wide design, and use of registry data. Registry-based studies are

178

less subject to key bias found in case-control studies, because exposure data is
collected before breast cancer diagnosis. Further, using existing registry data
provided us the opportunity to select controls from the source population that
gave rise to the cases, without being subject to potential bias due to lower
participation rates among controls. In addition, we had high quality cancer
registry data that meets NAACCR standards. Finally, this is the ﬁrst study to
examine fetal growth defined using birthweight percentiles for gestational week,
to examine exposures in both ﬁrst and last births, and to investigate birth
characteristics and breast cancer risk for speciﬁc histologic sub-types.
3. Conclusions

To date, few studies have been conducted in the area of perinatal factors
in a women’s own pregnancy and subsequent maternal breast cancer risk. The
ﬁrst manuscript of this dissertation will be the ﬁrst published review focusing
speciﬁcally on perinatal factors and maternal breast cancer and using systematic
search methods. This paper can help guide researchers interested in this area of
research when they are planning and conducting studies. The main conclusion of
this work is that future research is needed to elucidate the associations between
perinatal factors and maternal breast cancer, including studies of potential
mechanisms for the role of perinatal factors in breast cancer etiology.

The second manuscript, which examined the a priori effect modiﬁers for
the third aim of this dissertation, also makes an important contribution. Though
ages at ﬁrst and last birth and low parity have been well-studied, the associations

of these factors by histologic type, in particular among younger women, have

179

 

been little described. We found that later age at ﬁrst and last births, as well as
multiparity, were associated with an increased risk of breast cancer overall and
for both lobular and ductal histologic types. We did not ﬁnd that associations
varied by histologic type among all women. Some differences were found by race
and education subgroups, but very few studies have examined histologic type
among younger women and results require conﬁrmation in future studies,
particularly given limitations in sample size in subgroups.

The third paper of this manuscript is the ﬁrst study to examine fetal growth
deﬁned using birthweight percentiles for gestational week, to examine infant birth
characteristics in both ﬁrst and last births, and to investigate at infant birth
characteristics and breast cancer risk for speciﬁc histologic sub-types. We found
limited evidence that fetal growth inﬂuences breast cancer risk overall and some
evidence that earlier or late gestation in a ﬁrst birth may be associated with a
reduction in breast cancer risk. We did ﬁnd a non-signiﬁcant reduction in risk for
delivery of an SGA infant in last birth, but not an LGA infant, which was
statistically signiﬁcant among women older than 30 years (vs. < 30 years) at their
last birth in stratiﬁed analyses (OR = 0.82, 95% CI: 0.68-0.98). This ﬁnding could
reﬂect the inﬂuence of lower levels of maternal estrogens or IGF-l levels during
pregnancy on subsequent breast cancer risk. The suggestion of a reduced breast
cancer risk for a ﬁrst delivery at < 32 weeks gestation (vs. 37-41 weeks) (OR =
80, 95% CI: 0.62-1.04) could reﬂect elevated levels of maternal alpha-fetoprotein,
but this ﬁnding is in the opposite direction of our hypothesis and most previous

literature to date. Our ﬁnding of a protective effect of a posterm delivery in a ﬁrst

180

 

birth (OR for > 42 weeks vs. < 37-41 weeks gestation = 0.92, 95% CI: 0.85-0.99),
which has not been previously reported, warrants further study.
4. Recommendations

In summary, few studies have been conducted in the area of perinatal
factors in a women’s own pregnancy and subsequent maternal breast cancer
risk. The study of breast cancer in premenopausal parous women is of particular
importance given the increasing trend in delayed childbearing (289) and because
women diagnosed with breast cancer shortly following pregnancy are more likely
to have advanced breast tumors (37) and poorer tumor prognostic factors (38),
as well as increased mortality risk compared to nulliparous premenopausal
women (38-41). Additional research in the area of perinatal factors and maternal
breast cancer, which incorporates epidemiologists trained in both perinatal and
cancer epidemiology (113), is needed. Future studies of perinatal factors and
maternal breast cancer should have ample sample size to consider jointly several
key effect modifiers (e.g., time, age, birth order, menopausal status) with
information on other potential confounding factors (e.g., body size). To improve
biological hypotheses for studies of perinatal factors and breast cancer, future
work is also needed to further characterize the maternal hormonal and metabolic

profiles of exposed women during and following pregnancy.

181

 

10.

11.

REFERENCES

Joslyn, S. A., Foote, M. L., Nasseri, K., Coughlin, S. S., and Howe, H. L.
Racial and ethnic disparities in breast cancer rates by age: NAACCR
Breast Cancer Project. Breast Cancer Res Treat, 92: 97-105, 2005.

Baquet, C. R., Mishra, S. l., Commiskey, P., Ellison, G. L., and DeShields,
M. Breast cancer epidemiology in blacks and whites: disparities in

incidence, mortality, survival rates and histology. J Natl Med Assoc, 100:
480-8, 2008.

Talley, L. |., Grizzle, W. E., Waterbor, J. W., Brown, D., Weiss, H., and
Frost, A. R. Hormone receptors and proliferation in breast carcinomas of

equivalent histologic grades in pre- and postmenopausal women. Int J
Cancer, 98: 1 18-27, 2002.

Yankaskas, B. C. Epidemiology of breast cancer in young women. Breast
Dis, 23: 3-8, 2005.

Klauber-DeMore, N. Tumor biology of breast cancer in young women.
Breast Dis, 23: 9-15, 2005.

Pratap, R., and Shousha, S. Breast carcinoma in women under the age of
50: Relationship between p53 immunostaining, tumour grade, and axillary
lymph node status. Breast Cancer Res Treat, 49: 35-9, 1998.

American Cancer Society. Breast Cancer Factors & Figures, 2005-2006.
Atlanta: American Cancer Society, Inc.

Kroman, N., Jensen, M. B., Wohlfahrt, J., Mouridsen, H. T., Andersen, P.
K., and Melbye, M. Factors influencing the effect of age on prognosis in
breast cancer: population based study. ij, 320: 474-8, 2000.

Avis, N. E., Crawford, 8., and Manuel, J. Quality of life among younger
women with breast cancer. J Clin Oncol, 23: 3322-30, 2005.

Wenzel, L. B., Fairclough, D. L., Brady, M. J., Cella, 0., Garrett, K. M.,
Kluhsman, B. 0., Crane, L. A., and Marcus, A. C. Age-related differences

in the quality of life of breast carcinoma patients after treatment. Cancer,
86: 1768-74, 1999.

Kelsey, J. L., and Bernstein, L. Epidemiology and prevention of breast
cancer. Annu Rev Public Health, 17: 47-67, 1996.

182

12.

13.

14.

15.

16.

17.

18.

19.

20.

21.

22.

Krieger, M, Chen, J. T., Waterman, P. D., Rehkopf, D. H., Yin, R., and
Coull, B. A. Race/ethnicity and changing US socioeconomic gradients in
breast cancer incidence: California and Massachusetts, 1978-2002
(United States). Cancer Causes Control, 17: 217-26, 2006.

Smigal, C., Jemal, A., Ward, E., Cokkinides, V., Smith, R., Howe, H. L.,
and Thun, M. Trends in breast cancer by race and ethnicity: update 2006.
CA Cancer J Clin, 56: 168-83, 2006.

Baquet, C. R., and Commiskey, P. Socioeconomic factors and breast
carcinoma in multicultural women. Cancer, 88: 1256-64, 2000.

Krieger, N. Exposure, susceptibility, and breast cancer risk: a hypothesis
regarding exogenous carcinogens, breast tissue development, and social
gradients, including black/white differences, in breast cancer incidence.
Breast Cancer Res Treat, 13: 205-23, 1989.

SEER. SEER Fast Stats.

http://seer.cancer.gov/faststats/selections. php?run&output=1 &data=1&stat
istic=3&cancer=553&year=200803&sex=3&age=1&series=race&race=2;3.
Last accessed August 26, 2008.

Jones, L. A., and Chilton, J. A. Impact of breast cancer on African
American women: priority areas for research in the next decade. Am J
Public Health, 92: 539-42, 2002.

Marie Swanson, G., Haslam, S. 2., and Azzouz, F. Breast cancer among
young African-American women: a summary of data and literature and of
issues discussed during the Summit Meeting on Breast Cancer Among
African American Women, Washington, DC, September 8-10, 2000.
Cancer, 97: 273-9, 2003.

Krieger, N. Is breast cancer a disease of afﬂuence, poverty, or both? The
case of African American women. Am J Public Health, 92: 611-3, 2002.

Austin, H., Cole, P., and Wynder, E. Breast cancer in black American
women. Int J Cancer, 24: 541-4, 1979.

Mayberry, R. M. Age-specific patterns of association between breast
cancer and risk factors in black women, ages 20 to 39 and 40 to 54. Ann
Epidemiol, 4: 205-13, 1994.

Mayberry, R. M., and Stoddard-Wright, C. Breast cancer risk factors
among black women and white women: similarities and differences. Am J
Epidemiol, 136: 1445-56, 1992.

183

 

23.

24.

25.

26.

27.

28.

29.

30.

31.

32.

Brinton, L. A., Benichou, J., Gammon, M. D., Brogan, D. R., Coates, R.,
and Schoenberg, J. B. Ethnicity and variation in breast cancer incidence.
Int J Cancer, 73: 349-55, 1997.

Hall, I. J., Moorman, P. G., Millikan, R. C., and Newman, B. Comparative
analysis of breast cancer risk factors among African-American women and
White women. Am J Epidemiol, 161: 40-51, 2005.

Ursin, G., Bernstein, L., Wang, Y., Lord, S. J., Deapen, D., Liff, J. M.,
Norman, 8. A., Weiss, L. K., Daling, J. R., Marchbanks, P. A., Malone, K.
E., Folger, S. G., McDonald, J. A., Burkman, R. T., Simon, M. S., Strom,
B. L., and Spirtas, R. Reproductive factors and risk of breast carcinoma in
a study of white and African-American women. Cancer, 101: 353-62,
2004.

Palmer, J. R., Wise, L. A., Horton, N. J., Adams-Campbell, L. L., and
Rosenberg, L. Dual effect of parity on breast cancer risk in African-
American women. J Natl Cancer lnst, 95: 478-83, 2003.

Schatzkin, A., Palmer, J. R., Rosenberg, L., Helmrich, S. P., Miller, D. R.,
Kaufman, D. W., Lesko, S. M., and Shapiro, S. Risk factors for breast
cancer in black women. J Natl Cancer Inst, 78: 213-7, 1987.

Zhu, K., Caulﬁeld, J., Hunter, 8., Roland, C. L., Payne-Wilks, K., and
Texter, L. Body mass index and breast cancer risk in African American
women. Ann Epidemiol, 15: 123-8, 2005.

Palmer, J. R., Rao, R. S., Adams-Campbell, L. L., and Rosenberg, L.
Height and breast cancer risk: results from the Black Women's Health
Study (United States). Cancer Causes Control, 12: 343-8, 2001.

Palmer, J. R., Rosenberg, L., Harlap, S., Strom, B. L., Warshauer, M. E.,
Zauber, A. G., and Shapiro, S. Adult height and risk of breast cancer
among US black women. Am J Epidemiol, 141: 845-9, 1995.

Beiler, J. S., Zhu, K., Hunter, S., Payne-Wilks, K., Roland, C. L., and
Chinchilli, V. M. A case-control study of menstrual factors in relation to
breast cancer risk in African-American women. J Natl Med Assoc, 95: 930-
8, 2003.

Laing, A. E., Demenais, F. M., Williams, R., Kissling, G., Chen, V. W., and
Bonney, G. E. Breast cancer risk factors in African-American women: the
Howard University Tumor Registry experience. J Natl Med Assoc, 85: 931-
9, 1993.

184

 

33.

35.

36.

37.

38.

39.

40.

41.

42.

43.

44.

Krieger, N. Social class and the black/white crossover in the age-speciﬁc
incidence of breast cancer: a study linking census-derived data to
population-based registry records. Am J Epidemiol, 131: 804-14, 1990.

Hall, S. A., and Rockhill, B. Race, poverty, afﬂuence, and breast cancer.
Am J Public Health, 92: 1559; author reply 1560, 2002.

Kelsey, J. L., Gammon, M. D., and John, E. M. Reproductive factors and
breast cancer. Epidemiol Rev, 15: 36-47, 1993.

Russo, J., Moral, R., Balogh, G. A., Mailo, D., and Russo, l. H. The
protective role of pregnancy in breast cancer. Breast Cancer Res, 7: 131-
42, 2005.

Wohlfahrt, J., Andersen, P. K., Mouridsen, H. T., and Melbye, M. Risk of
late-stage breast cancer after a childbirth. Am J Epidemiol, 153: 1079-84,
2001.

Phillips, K. A., Milne, R. L., Friedlander, M. L., Jenkins, M. A., McCredie,
M. R., Giles, G. G., and Hopper, J. L. Prognosis of premenopausal breast
cancer and childbirth prior to diagnosis. J Clin Oncol, 22: 699-705, 2004.

Kroman, N., Wohlfahrt, J., Andersen, K. W., Mouridsen, H. T.,
Westergaard, T., and Melbye, M. Time since childbirth and prognosis in
primary breast cancer: population based study. ij, 315: 851-5, 1997.

Daling, J. R., Malone, K. E., Doody, D. R., Anderson, B. 0., and Porter, P.
L. The relation of reproductive factors to mortality from breast cancer.
Cancer Epidemiol Biomarkers Prev, 11 : 235-41, 2002.

Whiteman, M. K., Hillis, S. 0., Curtis, K. M., McDonald, J. A., Wingo, P. A.,
and Marchbanks, P. A. Reproductive history and mortality after breast
cancer diagnosis. Obstet Gynecol, 104: 146-54, 2004.

MacMahon, 8., Cole, P., Lin, T. M., Lowe, C. R., Mirra, A. P., Ravnihar, B.,
Salber, E. J., Valaoras, V. G., and Yuasa, S. Age at ﬁrst birth and breast
cancer risk. Bull World Health Organ, 43: 209-21, 1970.

Bernstein, L. Epidemiology of endocrine-related risk factors for breast
cancer. J Mammary Gland Biol Neoplasia, 7: 3-15, 2002.

Hsieh, C., Pavia, M., Lambe, M., Lan, S. J., Colditz, G. A., Ekbom, A.,

Adami, H. 0., Trichopoulos, D., and Willett, W. C. Dual effect of parity on
breast cancer risk. Eur J Cancer, 30A: 969-73, 1994. .

185

 

45.

46.

47.

48.

49.

50.

51.

52.

53.

54.

55.

56.

Liu, 0., Wuu, J., Lambe, M., Hsieh, S. F., Ekbom, A., and Hsieh, C. C.
Transient increase in breast cancer risk after giving birth: postpartum
period with the highest risk (Sweden). Cancer Causes Control, 13: 299-
305, 2002.

Bruzzi, P., Negri, B, La Vecchia, C., Decarli, A., Palli, D., Parazzini, F.,
and Del Turco, M. R. Short term increase in risk of breast cancer after full
term pregnancy. ij, 297: 1096-8, 1988.

Lambe, M., Hsieh, C., Trichopoulos, D., Ekbom, A., Pavia, M., and Adami,
H. 0. Transient increase in the risk of breast cancer after giving birth. N
Engl J Med, 331: 5-9, 1994.

Albrektsen, G., Heuch, l., and Kvale, G. The short-term and long-term
effect of a pregnancy on breast cancer risk: a prospective study of
802,457 parous Norwegian women. Br J Cancer, 72: 480-4, 1995.

Janerich, D. T., and Hoff, M. B. Evidence for a crossover in breast cancer
risk factors. Am J Epidemiol, 116: 737-42, 1982.

Albrektsen, G., Heuch, I., Hansen, 8., and Kvale, G. Breast cancer risk by
age at birth, time since birth and time intervals between births: exploring
interaction effects. Br J Cancer, 92: 167-75, 2005.

Wohlfahrt, J., and Melbye, M. Age at any birth is associated with breast
cancer risk. Epidemiology. 12:68-73, 2001.

Decarli, A., La Vecchia, C., Negri, E., and Franceschi, S. Age at any birth
and breast cancer in Italy. Int J Cancer, 67: 187-9, 1996.

Trichopoulos, D., Hsieh, C. C., MacMahon, B., Lin, T. M., Lowe, C. R.,
Mirra, A. P., Ravnihar, B., Salber, E. J., Valaoras, V. G., and Yuasa, S.
Age at any birth and breast cancer risk. Int J Cancer, 31: 701-4, 1983.

Hsieh, C. C., Chan, H. W., Lambe, M., Ekbom, A., Adami, H. 0., and
Trichopoulos, D. Does age at the last birth affect breast cancer risk? Eur J
Cancer, 32A: 118-21, 1996.

Lambe, M., Hsieh, C. 0., Chan, H. W., Ekbom, A., Trichopoulos, D., and
Adami, H. 0. Parity, age at first and last birth, and risk of breast cancer: a
population-based study in Sweden. Breast Cancer Res Treat, 38: 305-11,
1996.

Kvale, G., and Heuch, l. A prospective study of reproductive factors and
breast cancer. ll. Age at ﬁrst and last birth. Am J Epidemiol, 126: 842-50,
1987.

186

 

 

57.

58.

59.

60.

61.

62.

63.

64.

65.

66.

67.

Chie, W. C., Hsieh, C., Newcomb, P. A., Longnecker, M. P., Mittendorf,
R., Greenberg, E. R., Clapp, R. W., Burke, K. P., Titus-Ernstoff, L.,
Trentham-Dietz, A., and MacMahon, B. Age at any full-term pregnancy
and breast cancer risk. Am J Epidemiol, 151: 715-22, 2000.

Robertson, C., and Boyle, P. Age at any birth and breast cancer risk. Stat
Med, 17: 431-45, 1998.

MacMahon, B., Purde, M., Cramer, D., and Hint, E. Association of breast
cancer risk with age at ﬁrst and subsequent births: a study in the
population of the Estonian Republic. J Natl Cancer Inst, 69: 1035-8, 1982.

Negri, B, La Vecchia, C., Duffy, S. W., Bruzzi, P., Parazzini, F., and Day,
N. E. Age at ﬁrst and second births and breast cancer risk in biparous
women. Int J Cancer, 45: 428-30, 1990.

Schedin, P. Pregnancy-associated breast cancer and metastasis. Nat Rev
Cancer, 6: 281-91, 2006.

Sivaraman, L., and Medina, D. Hormone-induced protection against breast
cancer. J Mammary Gland Biol Neoplasia, 7: 77-92, 2002.

Cnattingius, S., Torrang, A., Ekbom, A., Granath, F., Petersson, G., and
Lambe, M. Pregnancy characteristics and maternal risk of breast cancer.
Jama, 294: 2474-80, 2005.

Campagnoli, C., Abba, C., Ambroggio, S., and Peris, C. Pregnancy,
progesterone and progestins in relation to breast cancer risk. J Steroid
Biochem Mol Biol, 97: 441-50, 2005.

lnnes, K. E., and Byers, T. E. First pregnancy characteristics and
subsequent breast cancer risk among young women. Int J Cancer, 112:
306-1 1 , 2004.

Troisi, R., Weiss, H. A., Hoover, R. N., Potischman, N., Swanson, C. A.,
Brogan, D. R., Coates, R. J., Gammon, M. D., Malone, K. E., Daling, J. R.,
and Brinton, L. A. Pregnancy characteristics and maternal risk of breast
cancer. Epidemiology, 9: 641-7, 1998.

Polednak, A. P., and Janerich, D. T. Characteristics of first pregnancy in

relation to early breast cancer. A case-control study. J Reprod Med, 28:
314-8, 1983.

187

 

 

68.

69.

70.

71.

72.

73.

74.

75.

76.

77.

78.

79.

Melbye, M., Wohlfahrt, J., Andersen, A. M., Westergaard, T., and
Andersen, P. K. Preterm delivery and risk of breast cancer. Br J Cancer,
80: 609-13, 1999.

Hsieh, C. C., Wuu, J., Lambe, M., Trichopoulos, D., Adami, H. 0., and
Ekbom, A. Delivery of premature newborns and maternal breast-cancer
risk. Lancet, 353: 1239, 1999.

Vatten, L. J., Romundstad, P. R., Trichopoulos, D., and Skjaerven, R.
Pregnancy related protection against breast cancer depends on length of
gestation. Br J Cancer, 87: 289-90, 2002.

Wyshak, G., Honeyman, M. S., Flannery, J. T., and Beck, A. S. Cancer in
mothers of dizygotic twins. J Natl Cancer Inst, 70: 593-9, 1983.

Jacobson, H. l., Thompson, W. D., and Janerich, D. T. Multiple births and
maternal risk of breast cancer. Am J Epidemiol, 129: 865-73, 1989.

Nasca, P. C., Weinstein, A., Baptiste, M., and Mahoney, M. The relation
between multiple births and maternal risk of breast cancer. Am J
Epidemiol, 136: 1316-20, 1992.

Hsieh, C. C., Goldman, M., Pavia, M., Ekbom, A., Petridou, E., Adami, H.
0., and Trichopoulos, D. Breast cancer risk in mothers of multiple births.
Int J Cancer, 54: 81-4, 1993.

Dietz, A. T., Newcomb, P. A., Storer, B. E., Longnecker, M. P., and
Mittendorf, R. Multiple births and risk of breast cancer. Int J Cancer, 62:
162-4, 1995.

Albrektsen, G., Heuch, |., and Kvale, G. Multiple births, sex of children and
subsequent breast-cancer risk for the mothers: a prospective study in
Nonrvay. lnt J Cancer, 60: 341-4, 1995.

Lambe, M., Hsieh, C., Tsaih, S., Ekbom, A., Adami, H. 0., and
Trichopoulos, D. Maternal risk of breast cancer following multiple births: a
nationwide study in Sweden. Cancer Causes Control, 7: 533-8, 1996.

Murphy, M. F ., Broeders, M. J., Carpenter, L. M., Gunnarskog, J., and
Leon, D. A. Breast cancer risk in mothers of twins. Br J Cancer, 75: 1066-
.8, 1997.

Wohlfahrt, J., and Melbye, M. Maternal risk of breast cancer and birth

characteristics of offspring by time since birth. Epidemiology, 10: 441-4,
1999.

188

 

80.

81.

82.

83.

84.

85.

86.

87.

88.

89.

90.

Neale, R. E., Purdie, D. M., Murphy, M. F., Mineau, G. P., Bishop, T., and
Whiteman, D. C. Twinning and the incidence of breast and gynecological
cancers (United States). Cancer Causes Control, 15: 829-35, 2004.

Mogren, l., Stenlund, H., and Hogberg, U. Long-term impact of
reproductive factors on the risk of cervical, endometrial, ovarian and
breast cancer. Acta Oncol, 40: 849-54, 2001.

Paltiel, 0., Friedlander, Y., Tiram, E., Barchana, M., Xue, X., and Harlap,
S. Cancer after pre-eclampsia: follow up of the Jerusalem perinatal study
cohort. ij, 328: 919, 2004.

Richardson, B. E., Peck, J. D., and Wormuth, J. K. Mean arterial pressure,
pregnancy-induced hypertension, and preeclampsia: evaluation as
independent risk factors and as surrogates for high maternal serum alpha-
fetoprotein in estimating breast cancer risk. Cancer Epidemiol Biomarkers
Prev, 9: 1349-55, 2000.

Vatten, L. J., Romundstad, P. R., Trichopoulos, D., and Skjaerven, R. Pre-
eclampsia in pregnancy and subsequent risk for breast cancer. Br J
Cancer, 87: 971-3, 2002.

Thompson, W. D., Jacobson, H. I., Negrini, B., and Janerich, D. T.
Hypertension, pregnancy, and risk of breast cancer. J Natl Cancer Inst,
81: 1571-4, 1989.

Hilakivi-Clarke, L., Luoto, R., Huttunen, T., and Koskenvuo, M. Pregnancy
weight gain and premenopausal breast cancer risk. J Reprod Med, 50:
81 1-6, 2005. ,

Kinnunen, T. |., Luoto, R., Gissler, M., Hemminki, E., and Hilakivi-Clarke,
L. Pregnancy weight gain and breast cancer risk. BMC Womens Health, 4:
7, 2004.

Hsieh, C., Wuu, J., Trichopoulos, D., Adami, H. 0., and Ekbom, A. Gender
of offspring and maternal breast cancer risk. Int J Cancer, 81: 335-8,
1999.

Wohlfahrt, J., and Melbye, M. Gender of offspring and long-term maternal
breast cancer risk. Br J Cancer, 82: 1070-2, 2000.

Perrin, M. C., Terry, M. B., Kleinhaus, K., Deutsch, L., Yanetz, R., Tiram,
E., Calderon-Margalit, R., Friedlander, Y., Paltiel, 0., and Harlap, S.
Gestational diabetes and the risk of breast cancer among women in the
Jerusalem Perinatal Study. Breast Cancer Res Treat, 2007.

189

 

91.

92.

93.

94.

95.

96.

97.

98.

99.

100.

Rollison, D. E., Giuliano, A. R., Sellers, T. A., Laronga, C., Sweeney, C.,
Risendal, B., Baumgartner, K. B., Byers, T., and Slattery, M. L.
Population-based Case-Control Study of Diabetes and Breast Cancer Risk
in Hispanic and Non-Hispanic White Women Living in US Southwestern
States. Am J Epidemiol, 2007.

Vatten, L. J., Romundstad, P. R., Odegard, R. A., Nilsen, S. T.,
Trichopoulos, D., and Austgulen, R. Alpha-foetoprotein in umbilical cord in

relation to severe pre-eclampsia, birth weight and future breast cancer
risk. Br J Cancer, 86: 728-31, 2002.

Troisi, R., Potischman, M, Roberts, J., Siiteri, P., Daftary, A., Sims, C.,
and Hoover, R. N. Associations of maternal and umbilical cord hormone

concentrations with maternal, gestational and neonatal factors (United
States). Cancer Causes Control, 14: 347-55, 2003.

Bennett, J. A., Zhu, S., Pagano-Mirarchi, A., Kellom, T. A., and Jacobson,
H. I. Alpha-fetoprotein derived from a human hepatoma prevents growth of

estrogen-dependent human breast cancer xenografts. Clin Cancer Res, 4:
2877-84, 1998.

Kaijser, M., Granath, F., Jacobsen, G., Cnattingius, S., and Ekbom, A.
Maternal pregnancy estriol levels in relation to anamnestic and fetal
anthropometric data. Epidemiology, 11 : 315-9, 2000.

Wuu, J., Hellerstein, S., Lipworth, L., Wide, L., Xu, B., Yu, G. P., Kuper,
H., Lagiou, P., Hankinson, S. E., Ekbom, A., Carlstrom, K., Trichopoulos,
D., Adami, H. 0., and Hsieh, C. C. Correlates of pregnancy oestrogen,
progesterone and sex hormone-binding globulin in the USA and China.
Eur J Cancer Prev, 11 : 283-93, 2002.

Petridou, E., Panagiotopoulou, K., Katsouyanni, K., Spanos, E., and
Trichopoulos, D. Tobacco smoking, pregnancy estrogens, and birth
weight. Epidemiology, 1: 247-50, 1990.

Tamimi, R., Lagiou, P., Vatten, L. J., Mucci, L., Trichopoulos, D.,
Hellerstein, S., Ekbom, A., Adami, H. 0., and Hsieh, C. C. Pregnancy
hormones, pre-eclampsia, and implications for breast cancer risk in the
offspring. Cancer Epidemiol Biomarkers Prev, 12: 647-50, 2003.

Walsh, S. W. Progesterone and estradiol production by normal and
preeclamptic placentas. Obstet Gynecol, 71: 222-6, 1988.

Troisi, R., Potischman, N., Roberts, J. M., Ness, R., Crombleholme, W.,
Lykins, D., Siiteri, P., and Hoover, R. N. Maternal serum oestrogen and

190

7"".

 

 

 

101.

102.

103.

104.

105.

106.

107.

108.

109.

androgen concentrations in preeclamptic and uncomplicated pregnancies.
Int J Epidemiol, 32: 455-60, 2003.

Atamer, Y., Erden, A. C., Demir, B., Kocyigit, Y., and Atamer, A. The
relationship between plasma levels of leptin and androgen in healthy and
preeclamptic pregnant women. Acta Obstet Gynecol Scand, 83: 425-30,
2004.

Baksu, A., Gurarslan, H., and Goker, N. Androgen levels in pre-eclamptic
pregnant women. Int J Gynaecol Obstet, 84: 247-8, 2004.

Barnabei, V. M., Krantz, D. A., Macri, J. N., and Larsen, J. W., Jr.
Enhanced twin pregnancy detection within an open neural tube defect and

Down syndrome screening protocol using free-beta hCG and AFP. Prenat
Diagn, 15: 1131-4, 1995.

Wald, N., Cuckle, H., Wu, T. S., and George, L. Maternal serum
unconjugated oestriol and human chorionic gonadotrophin levels in twin
pregnancies: implications for screening for Down's syndrome. Br J Obstet
Gynaecol, 98: 905-8, 1991.

Boyne, M. S., Thame, M., Bennett, F. l., Osmond, C., Miell, J. P., and
Forrester, T. E. The relationship among circulating insulin-like growth
factor (lGF)-l, IGF-binding proteins-1 and -2, and birth anthropometry: a
prospective study. J Clin Endocrinol Metab, 88: 1687-91, 2003.

Chellakooty, M., Vangsgaard, K., Larsen, T., Scheike, T., Falck-Larsen, J.,
Legarth, J., Andersson, A. M., Main, K. M., Skakkebaek, N. E., and Juul,
A. A longitudinal study of intrauterine growth and the placental growth
hormone (GH)-insulin-like growth factor I axis in maternal circulation:
association between placental GH and fetal growth. J Clin Endocrinol
Metab, 89: 384-91, 2004.

Murphy, V. E., Smith, R., Giles, W. B., and Clifton, V. L. Endocrine
regulation of human fetal growth: the role of the mother, placenta, and
fetus. Endocr Rev, 27: 141-69, 2006.

Kocyigit, Y., Bayhan, G., Atamer, A., and Atamer, Y. Serum levels of
leptin, insulin-like growth factor-I and insulin-like growth factor binding
protein-3 in women with pre-eclampsia, and their relationship to insulin
resistance. Gynecol Endocrinol, 18: 341-8, 2004.

Mucci, L. A., Lagiou, P., Tamimi, R. M., Hsieh, C. C., Adami, H. O., and
Trichopoulos, D. Pregnancy estriol, estradiol, progesterone and prolactin

in relation to birth weight and other birth size variables (United States).
Cancer Causes Control, 14: 311-8, 2003.

191

ﬁri'r‘.“ .
3:

 

 

110.

111.

112.

113.

114.

115.

116.

117.

118.

119.

120.

121.

Johnson, J. M., Harman, C. R., Evans, J. A., MacDonald, K., and
Manning, F. A. Maternal serum alpha-fetoprotein in twin pregnancy. Am J
Obstet Gynecol, 162: 1020-5, 1 990.

Holmes, R., Montemagno, R., Jones, J., Preece, M., Rodeck, C., and
Soothill, P. Fetal and maternal plasma insulin-like growth factors and
binding proteins in pregnancies with appropriate or retarded fetal growth.
Early Hum Dev, 49: 7-17, 1997.

Reeves, G. K., Pirie, K., Green, J., Bull, 0., and Beral, V. Reproductive
factors and speciﬁc histological types of breast cancer: prospective study
and meta-analysis. Br J Cancer, 100: 538-44, 2009.

Potischman, N., and Troisi, R. ln-utero and early life exposures in relation
to risk of breast cancer. Cancer Causes Control, 10: 561-73, 1999.

Xue, F., and Michels, K. B. Intrauterine factors and risk of breast cancer: a
systematic review and meta-analysis of current evidence. Lancet Oncol, 8:
1088-100, 2007.

Forman, M. R., Cantwell, M. M., Ronckers, C., and Zhang, Y. Through the
looking glass at early-life exposures and breast cancer risk. Cancer Invest,
23: 609-24, 2005.

Ruder, E. H., Dorgan, J. F., Kranz, S., Kris-Etherton, P. M., and Hartman,
T. J. Examining breast cancer growth and lifestyle risk factors: early life,
childhood, and adolescence. Clin Breast Cancer, 8: 334-42, 2008.

lnnes, K. E., and Byers, T. E. Preeclampsia and breast cancer risk.
Epidemiology. 10: 722-32, 1999.

Troisi, R., Potischman, N., and Hoover, R. N. Exploring the underlying
hormonal mechanisms of prenatal risk factors for breast cancer: a review
and commentary. Cancer Epidemiol Biomarkers Prev, 16: 1700-12, 2007.

Velie EM. Nechuta S, Osuch JR. Lifetime Reproductive and
Anthropometric Risk Factors for Breast Cancer in Postmenopausal
Women. Breast Disease, 22: 1 -19, 2006.

Cohn, B. A., Cirillo, P. M., Christianson, R. E., van den Berg, B. J., and
Siiteri, P. K. Placental characteristics and reduced risk of maternal breast
cancer. J Natl Cancer Inst, 93: 1133-40, 2001.

Russo, J., and Russo, I. H. Toward a physiological approach to breast
cancer prevention. Cancer Epidemiol Biomarkers Prev, 3: 353-64, 1994.

192

 

122.

123.

124.

125.

126.

127.

128.

129.

130.

131.

132.

133.

Hoover, R. N., and Troisi, R. J. Understanding mechanisms of breast
cancer prevention. J Natl Cancer Inst, 93: 1119-20, 2001 .

ISI Web of Knowledge: http://www.isiwebofknowledgecoml.

Pathak, D. R., Speizer, F. E., Willett, W. C., Rosner, B., and Lipnick, R. J.
Parity and breast cancer risk: possible effect on age at diagnosis. Int J
Cancer, 37: 21-5, 1986.

Clavel-Chapelon, F ., Launoy, G., Auquier, A., Gairard, B., Bremond, A.,
Piana, L., Lansac, J., and Renaud, R. Reproductive factors and breast
cancer risk. Effect of age at diagnosis. Ann Epidemiol, 5: 315-20, 1995.

Potischman, N., Swanson, C. A., Siiteri, P., and Hoover, R. N. Reversal of
relation between body mass and endogenous estrogen concentrations
with menopausal status. J Natl Cancer Inst, 88: 756-8, 1996.

*1! :-----=- — "-3—

 

Albrektsen, G., Heuch, l., and Kvale, G. Does gender of offspring modify
the time-related effects of a pregnancy on breast cancer risk? Int J
Cancer, 86: 595-7, 2000.

Troisi, R., lnnes, K. E., Roberts, J. M., and Hoover, R. N. Preeclampsia
and maternal breast cancer risk by offspring gender: do elevated
androgen concentrations play a role? Br J Cancer, 97: 688-90, 2007.

Redman, C. W., and Sargent, I. L. Latest advances in understanding
preeclampsia. Science, 308: 1592-4, 2005.

Samadi, A. R., Mayberry, R. M., Zaidi, A. A., Pleasant, J. C., McGhee, N.,
Jr., and Rice, R. J. Maternal hypertension and associated pregnancy
complications among African-American and other women in the United
States. Obstet Gynecol, 87: 557-63, 1996.

Zhang, J., Meikle, S., and Trumble, A. Severe maternal morbidity
associated with hypertensive disorders in pregnancy in the United States.
Hypertens Pregnancy, 22: 203-12, 2003.

MacKay, A. P., Berg, C. J., and Atrash, H. K. Pregnancy-related mortality
from preeclampsia and eclampsia. Obstet Gynecol, 97: 533-8, 2001.

Sibai, B. M., Caritis, S. N., Thom, E., Klebanoff, M., McNellis, D., Rocco,
L., Paul, R. H., Romero, R., Witter, F., Rosen, M., and et al. Prevention of
preeclampsia with low-dose aspirin in healthy, nulliparous pregnant
women. The National Institute of Child Health and Human Development

193

 

134.

135.

136.

137.

138.

139.

140.

141.

142.

143.

144.

Network of Maternal-Fetal Medicine Units. N Engl J Med, 329: 1213-8,
1993.

Roberts, J. M., Pearson, G., Cutler, J., and Lindheimer, M. Summary of
the NHLBI Working Group on Research on Hypertension During
Pregnancy. Hypertension, 41: 437-45, 2003.

Roberts, J. M., and Lain, K. Y. Recent Insights into the pathogenesis of
pre-eclampsia. Placenta, 23: 359-72, 2002.

Duckitt, K., and Harrington, D. Risk factors for pre-eclampsia at antenatal
booking: systematic review of controlled studies. ij, 330: 565, 2005.

Terry, M. B., Perrin, M., Salaﬁa, C. M., Zhang, F. F., Neugut, A. l.,
Teitelbaum, S. L., Britton, J., and Gammon, M. D. Preeclampsia,

pregnancy-related hypertension, and breast cancer risk. Am J Epidemiol,
165: 1007-14, 2007.

Vatten, L. J., Forman, M. R., Nilsen, T. l., Barrett, J. C., and Romundstad,
P. R. The negative association between pre-eclampsia and breast cancer
risk may depend on the offspring's gender. Br J Cancer, 96: 1436-8, 2007.

Calderon-Margalit, R., Friedlander, Y., Yanetz, R., Deutsch, L., Perrin, M.
C., Kleinhaus, K., Tiram, E., Harlap, S., and Paltiel, O. Preeclampsia and
subsequent risk of cancer: update from the Jerusalem Perinatal Study. Am
J Obstet Gynecol, 2008.

Aagaard-Tillery, K. M., Stoddard, G. J., Holmgren, C., Lacoursiere, D. Y.,
Fraser, A., Mineau, G. P., and Varner, M. W. Preeclampsia and
subsequent risk of cancer in Utah. Am J Obstet Gynecol, 195: 691-9,
2006.

Simpson, E. R., and MacDonald, P. C. Endocrine physiology of the
placenta. Annu Rev Physiol, 43: 163-88, 1981.

Acromite, M. T., Mantzoros, C. S., Leach, R. E., Hurwitz, J., and Dorey, L.
G. Androgens in preeclampsia. Am J Obstet Gynecol, 180: 60-3, 1999.

Masse, J., Forest, J. C., Moutquin, J. M., Marcoux, S., Brideau, N. A., and
Belanger, M. A prospective study of several potential biologic markers for
early prediction of the development of preeclampsia. Am J Obstet
Gynecol, 169: 501-8, 1993.

Ranta, T., Stenman, U. H., Unnerus, H. A., Rossi, J., and Seppala, M.
Maternal plasma prolactin levels in preeclampsia. Obstet Gynecol, 55:
428-30, 1980.

194

 

145.

146.

147.

148.

149.

150.

151.

152.

153.

154.

Rosing, U., and Carlstrom, K. Serum levels of unconjugated and total
oestrogens and dehydroepiandrosterone, progesterone and urinary
oestriol excretion in pre-eclampsia. Gynecol Obstet Invest, 18: 199-205,
1984.

Stamilio, D. M., Sehdev, H. M., Morgan, M. A., Propert, K., and Macones,
G. A. Can antenatal clinical and biochemical markers predict the
development of severe preeclampsia? Am J Obstet Gynecol, 182: 589-94,
2000.

Salamalekis, E., Bakas, P., Vitoratos, N., Eleptheriadis, M., and Creatsas,
G. Androgen levels in the third trimester of pregnancy in patients with
preeclampsia. Eur J Obstet Gynecol Reprod Biol, 126: 16-9, 2006.

Wald, N. J., and Morris, J. K. Multiple marker second trimester serum
screening for pre-eclampsia. J Med Screen, 8: 65-8, 2001.

Troisi, R., Vatten, L., Hoover, R. N., Roberts, J. M., Cole, B. F., and
Potischman, N. Maternal androgen and estrogen concentrations are not
associated with blood pressure changes in uncomplicated pregnancies.
Cancer Epidemiol Biomarkers Prev, 15: 2013-5, 2006.

Bartha, J. L., Romero-Carmona, R., Torrejon-Cardoso, R., and Comino-
Delgado, R. Insulin, insulin-like growth factor-1, and insulin resistance in
women with pregnancy-induced hypertension. Am J Obstet Gynecol, 187:
735-40, 2002.

Grobman, W. A., and Kazer, R. R. Serum insulin, insulin-like growth
factor-I, and insulin-like growth factor binding protein-1 in women who
develop preeclampsia. Obstet Gynecol, 97: 521-6, 2001.

Hubinette, A., Lichtenstein, P., Brismar, K., Vatten, L., Jacobsen, G.,
Ekbom, A., and Cnattingius, S. Serum insulin-like growth factors in normal
pregnancy and in pregnancies complicated by preeclampsia. Acta Obstet
Gynecol Scand, 82: 1004-9, 2003.

Ingec, M., Gursoy, H. G., Yildiz, L., Kumtepe, Y., and Kadanali, 8. Serum
levels of insulin, IGF-1, and IGFBP-1 in pre-eclampsia and eclampsia. Int
J Gynaecol Obstet, 84: 214-9, 2004.

Ning, Y., Williams, M. A., Vadachkoria, S., Muy-Rivera, M., Frederick, I.
O., and Luthy, D. A. Maternal plasma concentrations of insulinlike growth
factor-1 and insulinlike growth factor-binding protein-1 in early pregnancy
and subsequent risk of preeclampsia. Clin Biochem, 37: 968-73, 2004.

195

155.

156.

157.

158.

159.

160.

161.

162.

163.

164.

165.

Vatten, L. J., Holly, J. M., Gunnell, D., and Tretli, S. Nested case-control
study of the association of circulating levels of serum insulin-like growth
factor I and insulin-like growth factor binding protein 3 with breast cancer
in young women in Norway. Cancer Epidemiol Biomarkers Prev, 17: 2097-
100, 2008.

Schrocksnadel, H., Daxenbichler, G., Artner, E., Steckel-Berger, G., and
Dapunt, 0. Tumor markers in hypertensive disorders of pregnancy.
Gynecol Obstet Invest, 35: 204-8, 1993.

Raty, R., Koskinen, P., Alanen, A., lrjala, K., Matinlauri, I., and Ekblad, U.
Prediction of pre-eclampsia with maternal mid-trimester total renin, inhibin
A, AFP and free beta-hCG levels. Prenat Diagn, 19: 122-7, 1999.

Waller, D. K., Lustig, L. S., Cunningham, G. C., Feuchtbaum, L. B., and
Hook, E. B. The association between maternal serum alpha-fetoprotein
and preterm birth, small for gestational age infants, preeclampsia, and
placental complications. Obstet Gynecol, 88: 816-22, 1996.

Gago-Dominguez, M., Castelao, J. E., Pike, M. C., Sevanian, A., and
Haile, R. W. Role of lipid peroxidation in the epidemiology and prevention
of breast cancer. Cancer Epidemiol Biomarkers Prev, 14: 2829-39, 2005.

Boudreau, N., and Myers, C. Breast cancer-induced angiogenesis:
multiple mechanisms and the role of the microenvironment. Breast Cancer
Res, 5: 140-6, 2003.

Mutter, W. P., and Karumanchi, S. A. Molecular mechanisms of
preeclampsia. Microvasc Res, 75: 1-8, 2008.

Thomas, H. V., Murphy, M. F., Key, T. J., Fentiman, l. S., Allen, D. S., and
Kinlen, L. J. Pregnancy and menstrual hormone levels in mothers of twins
compared to mothers of singletons. Ann Hum Biol, 25: 69-75, 1998.

Batra, S., Sjoberg, N. O., and Aberg, A. Human placental lactogen,
estradiol-17beta, and progesterone levels in the third trimester and their
respective values for detecting twin pregnancy. Am J Obstet Gynecol,
131: 69-72, 1978.

Martin, J. A., Hamilton, B. E., Sutton, P. 0., Ventura, S. J., Menacker, F.,
Kirmeyer, S., and Mathews, T. J. Births: Final data for 2006. National Vital
Statistics Reports, 57: 1-104, 2009.

Callahan, T. L., Hall, J. E., Ettner, S. L., Christiansen, C. L., Greene, M. F .,
and Crowley, W. F., Jr. The economic impact of multiple-gestation

196

166.

167.

168.

169.

170.

171.

172.

173.

174.

175.

pregnancies and the contribution of assisted-reproduction techniques to
their incidence. N Engl J Med, 331: 244-9, 1994.

Blondel, B., and Kaminski, M. Trends in the occurrence, determinants, and
consequences of multiple births. Semin Perinatol, 26: 239-49, 2002.

Blondel, B., Kogan, M. 0., Alexander, G. R., Dattani, N., Kramer, M. S.,
Macfarlane, A., and Wen, S. W. The impact of the increasing number of
multiple births on the rates of preterm birth and low birthweight: an
international study. Am J Public Health, 92: 1323-30, 2002.

Sibai, B. M., Hauth, J., Caritis, S., Lindheimer, M. D., MacPherson, C.,
Klebanoff, M., VanDorsten, J. P., Landon, M., Miodovnik, M., Paul, R.,
Meis, P., Thurnau, G., Dombrowski, M., Roberts, J., and McNellis, D.
Hypertensive disorders in twin versus singleton gestations. National
Institute of Child Health and Human Development Network of Maternal-
Fetal Medicine Units. Am J Obstet Gynecol, 182: 938-42, 2000.

Schwartz, D. B., Daoud, Y., Zazula, P., Goyert, G., Bronsteen, R., Wright,
0., and Copes, J. Gestational diabetes mellitus: metabolic and blood

glucose parameters in singleton versus twin pregnancies. Am J Obstet
Gynecol, 181: 912-4, 1999.

Bortolus, R., Parazzini, F., Chatenoud, L., Benzi, G., Bianchi, M. M., and
Marini, A. The epidemiology of multiple births. Hum Reprod Update, 5:
179-87, 1999.

Endres, L., and Wilkins, I. Epidemiology and biology of multiple gestations.
Clin Perinatol, 32: 301-14, v, 2005.

Lichtenstein, P., Olausson, P. O., and Kallen, A. J. Twin births to mothers
who are twins: a registry based study. ij, 312: 879-81, 1996.

Neale, R. E., Darlington, 8., Murphy, M. F., Silcocks, P. B., Purdie, D. M.,
and Talback, M. The effects of twins, parity and age at first birth on cancer
risk in Swedish women. Twin Res Hum Genet, 8: 156-62, 2005.

Ji, J., Forsti, A., Sundquist, J., and Hemminki, K. Risks of breast,
endometrial, and ovarian cancers after twin births. Endocr Relat Cancer,
14: 703-11, 2007.

Olsen, J., and Storm, H. Pregnancy experience in women who later
developed oestrogen-related cancers (Denmark). Cancer Causes Control,
9: 653-7, 1998.

197

176.

177.

178.

179.

180.

181.

182.

183.

184.

185.

Trapp, M., Kato, K., Bohnet, H. G., Gerhard, I., Weise, H. C., and
Leidenberger, F. Human placental lactogen and unconjugated estriol
concentrations in twin pregnancy: monitoring of fetal development in

intrauterine growth retardation and single intrauterine fetal death. Am J
Obstet Gynecol, 155: 1027-31, 1986.

Wald, N. J., and Densem, J. W. Maternal serum free alpha-human
chorionic gonadotrophin levels in twin pregnancies: implications for
screening for Down's syndrome. Prenat Diagn, 14: 717-9, 1994.

Murphy, M., Key, T., Wang, D., Moore, J., Clark, G., and Allen, D. Re:
”Multiple births and maternal risk of breast cancer". Am J Epidemiol, 132:
199-201, 1990.

Burkman, R. T., Tang, M. T., Malone, K. E., Marchbanks, P. A.,
McDonald, J. A., Folger, S. G., Norman, S. A., Strom, B. L., Bernstein, L.,
Ursin, G., Weiss, L. K., Daling, J. R., Simon, M. S., and Spirtas, R.
Infertility drugs and the risk of breast cancer: ﬁndings from the National
Institute of Child Health and Human Development Women's Contraceptive
and Reproductive Experiences Study. Fertil Steril, 79: 844-51, 2003.

Breast cancer and breastfeeding: collaborative reanalysis of individual
data from 47 epidemiological studies in 30 countries, including 50302
women with breast cancer and 96973 women without the disease. Lancet,
360: 187-95, 2002.

Geraghty, S. R., Pinney, S. M., Sethuraman, G., Roy-Chaudhury, A., and
Kalkwarf, H. J. Breast milk feeding rates of mothers of multiples compared
to mothers of singletons. Ambul Pediatr, 4: 226-31, 2004.

Diagnosis and classiﬁcation of diabetes mellitus. Diabetes Care, 32 Suppl
1: 862-7, 2009.

Dabelea, D., Snell-Bergeon, J. K., Hartsﬁeld, C. L., Bischoff, K. J.,
Hamman, R. F., and McDufﬁe, R. S. Increasing prevalence of gestational
diabetes mellitus (GDM) over time and by birth cohort: Kaiser Permanente
of Colorado GDM Screening Program. Diabetes Care, 28: 579-84, 2005.

Lawrence, J. M., Contreras, R., Chen, W., and Sacks, D. A. Trends in the
prevalence of preexisting diabetes and gestational diabetes mellitus

among a racially/ethnically diverse population of pregnant women, 1999-
2005. Diabetes Care, 31: 899-904, 2008.

Chu, S. Y., Callaghan, W. M., Kim, S. Y., Schmid, C. H., Lau, J., England,

L. J., and Dietz, P. M. Maternal obesity and risk of gestational diabetes
mellitus. Diabetes Care, 30: 2070-6, 2007.

198

186.

187.

188.

189.

190.

191.

192.

193.

194.

195.

196.

197.

Buchanan, T. A., and Xiang, A. H. Gestational diabetes mellitus. J Clin
Invest, 115: 485-91, 2005.

Dawson, S. l. Long-term risk of malignant neoplasm associated with
gestational glucose intolerance. Cancer, 100: 149-55, 2004.

Xue, F., and Michels, K. B. Diabetes, metabolic syndrome, and breast
cancer: a review of the current evidence. Am J Clin Nutr, 86: 5823-35,
2007.

Wolf, l., Sadetzki, S., Catane, R., Karasik, A., and Kaufman, B. Diabetes
mellitus and breast cancer. Lancet Oncol, 6: 103-11, 2005.

Kim, 0, Newton, K. M., and Knopp, R. H. Gestational diabetes and the
incidence of type 2 diabetes: a systematic review. Diabetes Care, 25:
1862-8, 2002.

Ballard-Barbash, R. Anthropometry and breast cancer. Body size--a
moving target. Cancer, 74: 1090-100, 1994.

Kumar, N. B., Lyman, G. H., Allen, K., Cox, C. E., and Schapira, D. V.
Timing of weight gain and breast cancer risk. Cancer, 76: 243-9, 1995.

Friedenreich, C. M. Review of anthropometric factors and breast cancer
risk. Eur J Cancer Prev, 10: 15-32, 2001.

Han, D., Nie, J., Bonner, M. R., McCann, S. E., Muti, P., Trevisan, M.,
Ramirez-Marrero, F. A., Vito, D., and Freudenheim, J. L. Lifetime adult
weight gain, central adiposity, and the risk of pre- and postmenopausal
breast cancer in the Western New York exposures and breast cancer
study. Int J Cancer, 119: 2931-7, 2006.

Peterson, N. B., Huang, Y., Newcomb, P. A., Titus-Ernstoff, L., Trentham-
Dietz, A., Anic, G., and Egan, K. M. Childbearing recency and modiﬁers of
premenopausal breast cancer risk. Cancer Epidemiol Biomarkers Prev,
17: 3284-7, 2008.

Rossner, S., and Ohlin, A. Pregnancy as a risk factor for obesity: lessons
from the Stockholm Pregnancy and Weight Development Study. Obes
Res, 3 Suppl 2: 2673-2755, 1995.

Linne, Y., Dye, L., Barkeling, B., and Rossner, S. Long-term weight
development in women: a 15-year follow-up of the effects of pregnancy.
Obes Res, 12: 1166-78, 2004.

199

198.

199.

200.

201.

202.

203.

204.

205.

206.

207.

208.

Amorim, A. R., Rossner, S., Neovius, M., Lourenco, P. M., and Linne, Y.
Does excess pregnancy weight gain constitute a major risk for increasing
long-term BMI? Obesity (Silver Spring), 15: 1278-86, 2007.

Butte, N. F., Ellis, K. J., Wong, W. W., Hopkinson, J. M., and Smith, E. 0.
Composition of gestational weight gain impacts maternal fat retention and
infant birth weight. Am J Obstet Gynecol, 189: 1423-32, 2003.

Petridou, E., Katsouyanni, K., Hsieh, C. C., Antsaklis, A., and
Trichopoulos, D. Diet, pregnancy estrogens and their possible relevance
to cancer risk in the offspring. Oncology, 49: 127-32, 1992.

FaupeI-Badger, J. M., Hoover, R. N., Potischman, N., Roberts, J. M., and
Troisi, R. Pregnancy weight gain is not associated with maternal or mixed

umbilical cord estrogen and androgen concentrations. Cancer Causes
Control, 2008.

Kaijser, M., Jacobsen, G., Granath, F., Cnattingius, S., and Ekbom, A.
Maternal age, anthropometrics and pregnancy oestriol. Paediatr Perinat
Epidemiol, 16: 149-53, 2002.

Lagiou, P., Lagiou, A., Samoli, E., Hsieh, C. C., Adami, H. O., and
Trichopoulos, D. Diet during pregnancy and levels of maternal pregnancy

hormones in relation to the risk of breast cancer in the offspring. Eur J
Cancer Prev, 15: 20-6, 2006.

de Assis, S., Wang, M., Goel, S., Foxworth, A., Helferich, W., and Hilakivi-
Clarke, L. Excessive weight gain during pregnancy increases carcinogen-

induced mammary tumorigenesis in Sprague-Dawley and lean and obese
Zucker rats. J Nutr, 136: 998-1004, 2006.

Kramer, M. S., McLean, F. H., Boyd, M. E., and Usher, R. H. The validity
of gestational age estimation by menstrual dating in term, preterm, and
postterm gestations. Jama, 260: 3306-8, 1988.

Owen, P., Donnet, M. L., Ogston, S. A., Christie, A. D., Howie, P. W., and
Patel, N. B. Standards for ultrasound fetal growth velocity. Br J Obstet
Gynaecol, 103: 60-9, 1996.

Villar, J., and Belizan, J. M. The timing factor in the pathophysiology of the
intrauterine growth retardation syndrome. Obstet Gynecol Surv, 37: 499-
506, 1982.

Paneth, N. Invited commentary: the hidden population in perinatal
epidemiology. Am J Epidemiol, 167: 793-6; author reply 797-8, 2008.

200

209.

210.

211.

212.

213.

214.

' 215.

216.

217.

218.

219.

220.

Zhang, J., and Bowes, W. A., Jr. Birth-weight-for-gestationaI-age patterns
by race, sex, and parity in the United States population. Obstet Gynecol,
86: 200-8, 1995.

Oken, E., Kleinman, K. P., Rich-Edwards, J., and Gillman, M. W. A nearly
continuous measure of birth weight for gestational age using a United
States national reference. BMC Pediatr, 3: 6, 2003.

Kramer, M. S. Intrauterine growth and gestational duration determinants.
Pediatrics, 80: 502-11, 1987.

Romo, A., Carceller, R., and Tobajas, J. Intrauterine growth retardation
(IUGR): epidemiology and etiology. Pediatr Endocrinol Rev, 6 Suppl 3:
332-6, 2009.

Xiao, R., Sorensen, T. K., Williams, M. A., and Luthy, D. A. Influence of
pre-eclampsia on fetal growth. J Matern Fetal Neonatal Med, 13: 157-62,
2003.

Peck, J. D., Hulka, B. S., Savitz, D. A., Baird, D., Poole, C., and
Richardson, B. E. Accuracy of fetal growth indicators as surrogate

measures of steroid hormone levels during pregnancy. Am J Epidemiol,
157: 258-66, 2003.

Pike, M. C., Spicer, D. V., Dahmoush, L., and Press, M. F. Estrogens,
progestogens, normal breast cell proliferation, and breast cancer risk.
Epidemiol Rev, 15: 17-35, 1993.

Hankinson, S. E., Colditz, G. A., and Willett, W. C. Towards an integrated
model for breast cancer etiology: the lifelong interplay of genes, lifestyle,
and hormones. Breast Cancer Res, 6: 213-8, 2004.

Yetter, J. F., 3rd. Examination of the placenta. Am Fam Physician, 57:
1045-54, 1998.

Wingo, P. A., Newsome, K., Marks, J. S., Calle, E. E., and Parker, S. L.
The risk of breast cancer following spontaneous or induced abortion.
Cancer Causes Control, 8: 93-108, 1997.

Beral, V., Bull, 0., Doll, R., Peto, R., and Reeves, G. Breast cancer and
abortion: collaborative reanalysis of data from 53 epidemiological studies,
including 83?000 women with breast cancer from 16 countries. Lancet,
363: 1007-16, 2004.

Lapillonne, H., Golsteyn, R. M., and Lapillonne, A. Duration of pregnancy
and risk of breast cancer. Lancet, 353: 2075, 1999.

201

221.

222.

223.

224.

225.

226.

227.

228.

229.

230.

231.

NCHS. National Center for Health Statistics, ﬁnal natality data. From
www.marchofdimes.com/peristats. 2009.

Savitz, D. A., Dole, N., Herring, A. H., Kaczor, D., Murphy, J., Siega-Riz,
A. M., Thorp, J. M., Jr., and MacDonald, T. L. Should spontaneous and
medically indicated preterm births be separated for studying aetiology?
Paediatr Perinat Epidemiol, 19: 97-105, 2005.

Goldenberg, R. L., Culhane, J. F., lams, J. D., and Romero, R.
Epidemiology and causes of preterm birth. Lancet, 371: 75-84, 2008.

Braunstein, G. Endocrine Changes in Pregnancy. In Williams Textbook of
Endocrinology (edited by Henry Kronenberg, Shlomo Melmed, Kenneth
Polonsky, P. Reed Larsen). , 2003.

Mazor, M., Hershkovitz, R., Chaim, W., Levy, J., Sharony, Y., Leiberrnan,
J. R., and Glezerman, M. Human preterm birth is associated with systemic
and local changes in progesterone/17 beta-estradiol ratios. Am J Obstet
Gynecol, 171: 231-6, 1994.

Obiekwe, B. C., and Chard, T. Human chorionic gonadotropin levels in
maternal blood in late pregnancy: relation to birthweight, sex and condition
of the infant at birth. Br J Obstet Gynaecol, 89: 543-6, 1982.

Chen, R. J., Lin, Y. H., and Huang, S. C. Fetal sex and maternal alpha-
fetoprotein concentration at late normal singleton pregnancies. Acta
Obstet Gynecol Scand, 73: 192-4, 1994.

Bremme, K., Eneroth, P., Nilsson, B., Marsk, L., and Hagenfeldt, L.
Outcome of pregnancy in relation to maternal serum alpha-fetoprotein
levels in the second trimester. An evaluation of a screening program and a
longitudinal follow-up. Gynecol Obstet Invest, 26: 191-205, 1988.

Lagiou, P., Samoli, E., Lagiou, A., Lambe, M., Trichopoulos, D., Adami, H.
O., and Hsieh, C. C. Levels and correlates of alpha-fetoprotein in normal
pregnancies among Caucasian and Chinese women. Eur J Cancer Prev,
16: 178-83, 2007.

Richardson, B. E., Hulka, B. S., Peck, J. L., Hughes, C. L., van den Berg,
B. J., Christianson, R. E., and Calvin, J. A. Levels of maternal serum

alpha-fetoprotein (AFP) in pregnant women and subsequent breast cancer

risk. Am J Epidemiol, 148: 719-27, 1998.

Melbye, M., Wohlfahrt, J., Lei, U., Norgaard-Pedersen, B., Mouridsen, H.
T., Lambe, M., and Michels, K. B. alpha-fetoprotein levels in maternal

202

232.

233.

234.

235.

236.

237.

238.

239.

240.

serum during pregnancy and maternal breast cancer incidence. J Natl
Cancer Inst, 92: 1001-5, 2000.

Lukanova, A., Toniolo, P., Zeleniuch-Jacquotte, A., Grankvist, K., Wulff,
M., Arslan, A. A., Afanasyeva, Y., Johansson, R., Lenner, P., Hallmans,
G., Wadell, G., and Lundin, E. Insulin-like growth factor I in pregnancy and
maternal risk of breast cancer. Cancer Epidemiol Biomarkers Prev, 15:
2489-93, 2006.

Lukanova, A., Andersson, R., Wulff, M., Zeleniuch-Jacquotte, A.,
Grankvist, K., Dossus, L., Afanasyeva, Y., Johansson, R., Arslan, A. A.,
Lenner, P., Wadell, G., Hallmans, G., Toniolo, P., and Lundin, E. Human
chorionic gonadotropin and alpha-fetoprotein concentrations in pregnancy
and maternal risk of breast cancer: a nested case-control study. Am J
Epidemiol, 168: 1284-91, 2008.

Peck, J. D., Hulka, B. S., Poole, C., Savitz, D. A., Baird, D., and
Richardson, B. E. Steroid hormone levels during pregnancy and incidence
of maternal breast cancer. Cancer Epidemiol Biomarkers Prev, 11: 361-8,
2002.

Faupel-Badger, J. M., Hsieh, C. C., Troisi, R., Lagiou, P., and Potischman,
N. Plasma volume expansion in pregnancy: implications for biomarkers in
population studies. Cancer Epidemiol Biomarkers Prev, 16: 1720-3, 2007.

Lawlor, D. A., Smith, G. D., and Ebrahim, S. Hyperinsulinaemia and
increased risk of breast cancer: ﬁndings from the British Women's Heart
and Health Study. Cancer Causes Control, 15: 267-75, 2004.

Hsieh, C. C., Goldman, M., Pavia, M., Trichopoulos, D., Petridou, E.,
Ekbom, A., and Adami, H. O. Re: "The relation between multiple births
and maternal risk of breast cancer" and "multiple births and maternal risk
of breast cancer". Am J Epidemiol, 139: 445-7, 1994.

La Vecchia, C., Negri, E., Braga, C., and Fanceschi, S. Multiple births and
breast cancer. Int J Cancer, 68: 553-4, 1996.

Talamini, R., Franceschi, S., Favero, A., Negri, E., Parazzini, F., and La
Vecchia, C. Selected medical conditions and risk of breast cancer. Br J
Cancer, 75: 1699-703, 1997.

Forstall, R. Population of Counties by Decennial Census: 1900 to 1990.
Population Division. US Bureau of the Census, Washington, DC,
http://www. census. gov/population/cencounts/mi1 90090. txt, 1 995

203

241.

242.

243.

244.

245.

246.

247.

248.

249.

250.

251.

US. Cancer Statistics Working Group. United States Cancer Statistics:
2004 Incidence and Mortality. Atlanta: US. Department of Health and
Human Services, Centers for Disease Control and Prevention and
National Cancer Institute; 2007.

MDCSS. Metropolitan Detroit Cancer Surveillance System. Website:
http://www.karmanos.org/epid/history.html.

NCI. Surveillance, Epidemiology. and End Results Program Brochure.
National Cancer Institute. NIH Publication No. 05-4772. September 2005.

Weiss, L. K., Burkman, R. T., Cushing-Haugen, K. L., Voigt, L. F., Simon,
M. S., Daling, J. R., Norman, S. A., Bernstein, L., Ursin, G., Marchbanks,
P. A., Strom, B. L., Berlin, J. A., Weber, A. L., Doody, D. R., Wingo, P. A.,
McDonald, J. A., Malone, K. E., Folger, S. G., and Spirtas, R. Hormone
replacement therapy regimens and breast cancer risk(1). Obstet Gynecol,
100: 1 148-58, 2002.

Simon, M. S., Mueller, B. A., Deapen, D., and Copeland, G. A comparison
of record linkage yield for health research using different variable sets.
Breast Cancer Res Treat, 89: 107-10, 2005.

Barnholtz-Sloan, J. S., Tainsky, M. A., Abrams, J., Severson, R. K.,
Qureshi, F., Jacques, S. M., Levin, N., and Schwartz, A. G. Ethnic
differences in survival among women with ovarian carcinoma. Cancer, 94:
1886-93, 2002.

Michigan Department of Community Health. Cancer Reporting Manual:
Michigan Cancer Surveillance Program. Vital Records and Health Data
Development Section (2003).

Personal communication, Glenn Copeland, Director, Division for Vital
Records and Health Statistics, Michigan Department of Community
Health, July 2006.

Ventura, S. J., Martin, J. A., Curtin, S. C., and Mathews, T. J. Births: ﬁnal
data for 1997. Natl Vital Stat Rep, 47: 1-96, 1999.

Zhu, B. P., and Le, T. Effect of interpregnancy interVal on infant low birth
weight: a retrospective cohort study using the Michigan Maternally Linked
Birth Database. Matern Child Health J, 7: 169-78, 2003.

Abrevaya, J. Trends and determinants of second-pregnancy smoking
among young-adult mothers who smoked during their ﬁrst pregnancy.
Nicotine Tob Res, 10: 951-7, 2008.

204

252.

253.

254.

255.

256.

257.

258.

259.

260.

261.

262.

Ventura, S. J., Martin, J. A., Curtin, S.‘C., Menacker, F., and Hamilton, B.
E. Births: ﬁnal data for 1999. Natl Vital Stat Rep, 49: 1-100, 2001.

Northam, S., and Knapp, T. R. The reliability and validity of birth
certiﬁcates. J Obstet Gynecol Neonatal Nurs, 35: 3—12, 2006.

Baumeister, L., Marchi, K., Pearl, M., Williams, R., and Braveman, P. The
validity of information on "race" and "Hispanic ethnicity" in California birth
certiﬁcate data. Health Serv Res, 35: 869-83, 2000.

Piper, J. M., Mitchel, E. F., Jr., Snowden, M., Hall, C., Adams, M., and
Taylor, P. Validation of 1989 Tennessee birth certiﬁcates using maternal
and newborn hospital records. Am J Epidemiol, 137: 758-68, 1993.

Buescher, P. A., Taylor, K. R, Davis, M. H., and Bowling, J. M. The quality
of the new birth certiﬁcate data: a validation study in North Carolina. Am J
Public Health, 83: 1163-5, 1993.

Dobie, S. A., Baldwin, L. M., Rosenblatt, R. A., Fordyce, M. A., Andrilla, C.
H., and Hart, L. G. How well do birth certiﬁcates describe the pregnancies

they report? The Washington State experience with low—risk pregnancies.
Matern Child Health J, 2: 145-54, 1998.

Roohan, P. J., Josberger, R. E., Acar, J., Dabir, P., Feder, H. M., and
Gagliano, P. J. Validation of birth certiﬁcate data in New York State. J
Community Health, 28: 335-46, 2003.

Adams, M. Validity of birth certiﬁcate data for the outcome of the previous
pregnancy, Georgia, 1980-1995. Am J Epidemiol, 154: 883-8, 2001.

DiGiuseppe, D. L., Aron, D. C., Ranbom, L., Harper, D. L., and Rosenthal,
G. E. Reliability of birth certificate data: a multi-hospital comparison to
medical records information. Matern Child Health J, 6: 169-79, 2002.

Lydon-Rochelle, M. T., Holt, V. L., Cardenas, V., Nelson, J. C., Easterling,
T. R., Gardella, C., and Callaghan, W. M. The reporting of pre-existing
maternal medical conditions and complications of pregnancy on birth
certiﬁcates and in hospital discharge data. Am J Obstet Gynecol, 193:
125-34, 2005.

Green, D. C., Moore, J. M., Adams, M. M., Berg, C. J., Wilcox, L. S., and
McCarthy, B. J. Are we underestimating rates of vaginal birth after
previous cesarean birth? The validity of delivery methods from birth
certiﬁcates. Am J Epidemiol, 147: 581-6, 1998.

205

263.

264.

265.

266.

267.

268.

269.

270.

271.

272.

273.

274.

Clark, K., Fu, C. M., and Burnett, C. Accuracy of birth certiﬁcate data
regarding the amount, timing, and adequacy of prenatal care using
prenatal clinic medical records as referents. Am J Epidemiol, 145: 68-71,
1997.

Zollinger, T. W., Przybylski, M. J., and Gamache, R. E. Reliability of
Indiana birth certiﬁcate data compared to medical records. Ann Epidemiol,
16: 1-10, 2006.

Emery, E. 8., 3rd, Eaton, A., Grether, J. K., and Nelson, K. B. Assessment
of gestational age using birth certiﬁcate data compared with medical
record data. Paediatr Perinat Epidemiol, 11: 313-21, 1997.

Reichman, N. E., and Hade, E. M. Validation of birth certiﬁcate data. A
study of women in New Jersey's HealthStart program. Ann Epidemiol, 11 :
186-93, 2001 .

Wier, M. L., Pearl, M., and Kharrazi, M. Gestational age estimation on
United States Iivebirth certiﬁcates: a historical overview. Paediatr Perinat
Epidemiol, 21 Suppl 2: 4-12, 2007.

Pearl, M., Wier, M. L., and Kharrazi, M. Assessing the quality of last
menstrual period date on California birth records. Paediatr Perinat
Epidemiol, 21 Suppl 2: 50-61, 2007.

Schachter, J. Geographical Mobility: 1990-1995, Special Studies. Current
Population Reports. US Department of Commerce: Economics and
Statistics Administration. US Census Bureau. September 2000.

Arias, E., Curtin, L. R., Wei, R., and Anderson, R. N. US. decennial life
tables for 1999-2001, United States life tables. Natl Vital Stat Rep, 57: 1-
36, 2008.

Alexander, G. R., Himes, J. H., Kaufman, R. B., Mor, J., and Kogan, M. A
United States national reference for fetal growth. Obstet Gynecol, 87: 163-
8, 1996.

Parker, J. D., and Schoendorf, K. C. Implications of cleaning gestational
age data. Paediatr Perinat Epidemiol, 16: 181-7, 2002.

Kalache, A., Maguire, A., and Thompson, S. G. Age at last full-term
pregnancy and risk of breast cancer. Lancet, 341: 33-6, 1993.

Tavani, A., Gallus, S., La Vecchia, C., Negri, E., Montella, M., Dal Maso,

L., and Franceschi, S. Risk factors for breast cancer in women under 40
years. Eur J Cancer, 35: 1361-7, 1999.

206

275.

276.

277.

278.

279.

280.

281.

282.

Althuis, M. D., Brogan, D. D., Coates, R. J., Daling, J. R., Gammon, M. D.,
Malone, K. E., Schoenberg, J. B., and Brinton, L. A. Breast cancers

among very young premenopausal women (United States). Cancer
Causes Control, 14: 151-60, 2003.

Layde, P. M., Webster, L. A., Baughman, A. L., Wingo, P. A., Rubin, G. L.,
and Ory, H. W. The independent associations of parity, age at ﬁrst full
term pregnancy, and duration of breastfeeding with the risk of breast
cancer. Cancer and Steroid Hormone Study Group. J Clin Epidemiol, 42:
963-73, 1989.

Rosenberg, L. U., Magnusson, C., Lindstrom, E., Wedren, S., Hall, P., and
Dickman, P. W. Menopausal hormone therapy and other breast cancer
risk factors in relation to the risk of different histological subtypes of breast
cancer: a case-control study. Breast Cancer Res, 8: R11, 2006.

Beaber, E. F., Holt, V. L., Malone, K. E., Porter, P. L., Daling, J. R., and Li,
C. l. Reproductive factors, age at maximum height, and risk of three

histologic types of breast cancer. Cancer Epidemiol Biomarkers Prev, 17:
3427-34, 2008.

Li, C. l., Malone, K. E., Porter, P. L., Weiss, N. S., Tang, M. T., and Daling,
J. R. Reproductive and anthropometric factors in relation to the risk of

lobular and ductal breast carcinoma among women 65-79 years of age. Int
J Cancer, 107: 647-51, 2003.

Ursin, G., Bernstein, L., Lord, S. J., Karim, R., Deapen, 0., Press, M. F.,
Daling, J. R., Norman, S. A., Liff, J. M., Marchbanks, P. A., Folger, S. G.,
Simon, M. S., Strom, B. L., Burkman, R. T., Weiss, L. K., and Spirtas, R.
Reproductive factors and subtypes of breast cancer deﬁned by hormone
receptor and histology. Br J Cancer, 93: 364-71, 2005.

Wohlfahrt, J., Mouridsen, H., Andersen, P. K., and Melbye, M.
Reproductive risk factors for breast cancer by receptor status, histology,
Iaterality and location. Int J Cancer, 81: 49-55, 1999.

Li, C. l., Daling, J. R., Malone, K. E., Bernstein, L., Marchbanks, P. A., Liff,
J. M., Strom, B. L., Simon, M. S., Press, M. F., McDonald, J. A., Ursin, G.,
Burkman, R. T., Deapen, D., and Spirtas, R. Relationship between
established breast cancer risk factors and risk of seven different histologic
types of invasive breast cancer. Cancer Epidemiol Biomarkers Prev, 15:
946-54, 2006.

207

 

 

284.

285.

286

287

28

A)

283.

284.

285.

286.

287.

288.

289.

290.

291.

292.

293.

Granstrom, C., Sundquist, J., and Hemminki, K. Population attributable
risks for breast cancer in Swedish women by morphological type. Breast
Cancer Res Treat, 111: 559-68, 2008.

Garcia-Closas, M., Brinton, L. A., Lissowska, J., Chatterjee, N.,
Peplonska, B., Anderson, W. F., Szeszenia-Dabrowska, N., Bardin-
Mikolajczak, A., Zatonski, W., Blair, A., Kalaylioglu, Z., Rymkiewicz, G.,
Mazepa-Sikora, D., Kordek, R., Lukaszek, S., and Sherman, M. E.
Established breast cancer risk factors by clinically important tumour
characteristics. Br J Cancer, 95: 123-9, 2006.

Ewertz, M., and Duffy, 8. W. Risk of breast cancer in relation to
reproductive factors in Denmark. Br J Cancer, 58: 99-104, 1988.

Li, C. l., Littman, A. J., and White, E. Relationship between age maximum 1
height is attained, age at menarche, and age at ﬁrst full-term birth and 5.
breast cancer risk. Cancer Epidemiol Biomarkers Prev, 16: 2144-9, 2007. L__ j.‘

ClaveI-Chapelon, F ., and Gerber, M. Reproductive factors and breast
cancer risk. Do they differ according to age at diagnosis? Breast Cancer
Res Treat, 72: 107-15, 2002.

Li, C. l., Malone, K. E., Daling, J. R., Potter, J. D., Bernstein, L.,
Marchbanks, P. A., Strom, B. L., Simon, M. 8., Press, M. F., Ursin, G.,
Burkman, R. T., Folger, S. G., Norman, 8., McDonald, J. A., and Spirtas,
R. Timing of menarche and ﬁrst full-term birth in relation to breast cancer
risk. Am J Epidemiol, 167: 230-9, 2008.

Mathews, T. J., and Hamilton, B. E. Mean age of mother, 1970-2000. Natl
Vital Stat Rep, 51: 1-13, 2002.

Arpino, G., Bardou, V. J., Clark, G. M., and Elledge, R. M. lnﬁltrating
lobular carcinoma of the breast: tumor characteristics and clinical
outcome. Breast Cancer Res, 6: R149-56, 2004.

Pathak, D. R. Dual effect of ﬁrst full term pregnancy on breast cancer risk:
empirical evidence and postulated underlying biology. Cancer Causes
Control, 13: 295-8, 2002.

Nguyen, R. H., and Wilcox, A. J. Terms in reproductive and perinatal
epidemiology: 2. Perinatal terms. J Epidemiol Community Health, 59:
1019-21, 2005.

Jacobson, H. l., Lemanski, N., Narendran, A., Agarvval, A., Bennett, J. A.,
and Andersen, T. T. Hormones of pregnancy, alpha-feto protein, and
reduction of breast cancer risk. Adv Exp Med Biol, 617: 477-84, 2008.

208

 

294.

295.

296.

297.

Caughey, A. B., Stotland, N. E., Washington, A. E., and Escobar, G. J.
Who is at risk for prolonged and postterm pregnancy? Am J Obstet
Gynecol, 200: 683 e1-5, 2009.

Michels, K. B., Terry, K. L., and Willett, W. C. Longitudinal study on the
role of body size in premenopausal breast cancer. Arch Intern Med, 166:
2395-402, 2006.

Shea, K. M., Wilcox, A. J., and Little, R. E. Postterrn delivery: a challenge
for epidemiologic research. Epidemiology, 9: 199-204, 1998.

Ananth, C. V. Menstrual versus clinical estimate of gestational age dating

in the United States: temporal trends and variability in indices of perinatal
outcomes. Paediatr Perinat Epidemiol, 21 Suppl 2: 22-30, 2007.

209

 

   

"Illl'lslillilllilllil IIIlllilllllllllllllIllillllllll'ﬁs
3 1293 03062 9112