,9

A Z r .
.. mu.
...»...th ,...
. Gila. .
. I} .. a. .
n... r

         

    

           

 

 

   

.- V
v? 1‘ I.)
. a! Ami“...

3L: , ... Mar, ..
A . flywwmamnuwfbumwm‘mfimfm-

         

ma, a;

,n; 1 .
_, «W V
a

.3. .u

 

 

    

 

  

 

 

 

    

 

 

 

  
  
 

 

. . .. . nip.)
. . I . . .31me
' n A . . .
. u -A n v . , v u. l
r ‘ h. . . . _ 2 q
. . . l. .
m I u , . .
~ .. .._ . z . ‘
. . .A v ‘
. , . . .
a . . . .. . n . . . ;
.4 H. ‘ .. .
h 1 . . u. . ‘
.. y . L. ,
m: . ., . . . V v. than».
‘ ; . . - 3.9.}? 5&5 4
u . V “Pfizfluiiizlsfitifiw
4 . . A . . . . 50.11.13? y...
. .. J .1. . y ._ ...
. n . ‘7 .

 

 

.- an

£1.31 flu
vulva

.‘In.

   

      

       

 

.z&a§g .fiéafifi‘ m érw4§$¢~§w§§ an? .. 7 , , ‘ .. ._ dammfi

.. . .. .. z ...

 

Treats

llllllll\lllllllllllllll‘ll‘lillllhlllllllllllll

3 1293 01772

This is to certify that the

thesis entitled
HORMONE REPLACEMENT THERAPY AND CORONARY HEART
DISEASE IN WOMEN: AN EPIDEMIOLOGIC REVIEW AND
DESCRIPTIVE STUDY OF POST MYOCARDIAL INFARCTION
WOMEN

presented by

Kimberly R. Barber

has been accepted towards fulfillment
of the requirements for

M.S. degree in Epidemiology

 

7 \~
~, 2

Major professor

Date / // 1261/ fig

0-7639 MS U is an Affirmative Action/Equal Opportunity Institution

 

 

 

LIBRARY
Michigan State
University

 

 

 

O ---~-o .

PLACE IN RETURN BOX to remove this checkout from your record.
TO AVOID FINE-3 return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

  
 
 

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

108 mm“

HORMONE REPLACEMENT THERAPY AND CORONARY HEART DISEASE IN
WOMEN
AN EPIDEMIOLOGIC REVIEW AND DESCRIPTIV E STUDY
OF POST MYOCARDIAL INFARCTION WOMEN

By

Kimberly R. Barber

A THESIS

Submitted to
Michigan State University
in partial fulfillment of the requirements
for the decree of

MASTER OF SCIENCE

Department of Epidemiology
College of Human Medicine

1998

ABSTRACT
HORMONE REPLACEMENT THERAPY AND CORONARY HEART DISEASE IN
WOMEN
AN EPIDEMIOLOGIC REVIEW AND DESCRIPTIVE STUDY
OF POST MYOCARDIAL INFARCTION WOMEN
By

Kimberly R. Barber

Coronary heart disease (CHD) is the leading cause of death among women in the United
States. A promising prevention for CHD in older women is hormone replacement therapy
(HRT). Epidemiologic evidence suggests that HRT decreases the risk of myocardial
infarction (MI) by up to 50%. Despite established benefits, HRT is under utilized among
elderly women. Physician prescribing behavior may be low for a cardioprotective use of
HRT. This study examined prevalence of HRT use among a series of women hospitalized
for MI and the frequency of having discussed HRT with a physician. Women were either
interviewed post discharge or in-hospital. A total of 156 women responded With 23%
using HRT, while 54% reported never having discussed HRT with a physician. Only one
women reported discussion of HRT with her cardiologist. The current study supports
observations that HRT use among older women is low and cardiologist initiated discussion

remains low.

ACKNOWLEDGMENT

This thesis, as well as a clear understanding of epidemiologic concepts, would not have
been accomplished without the time and effort put forth by Professor Claudia Holzman.
Her insight and assistance have been invaluable throughout the process and I thank her for

her support and patience.

iii

LIST OF TABLES

Table 1 - CHD Deaths Attributable to Specific CHI) Risk Factors .......................... 4

Table 2 - Studies on Prevalence of HRT Use
Among Women 45 years and Older .......................................................... 10

Table 3 - Selected Characteristics Distinguishing

HRT Users from Nonusers ........................................................................ 14
Table 4 - Overview of the Impact of Hormone Use on CHD .................................... 24
Table 5 - Reasons MICH Women Were Not Interviewed ........................................ 53
Table 6 - Precision Based on Hypothesized Estimates

for HRT Use /Discussion ........................................................................... 56
Table 7 - Precision Based on Observed Estimates

for HRT Use Discussion ........................................................................... 59
Table 8 - Demographics of MICH Women .............................................................. 60
Table 9 - Demographics of Study Patients by Phase Status ...................................... 62
Table 10 - Health Behavior Data .............................................................................. 64
Table 11- Hormone Use by Phase Status .................................................................. 66

Table 12 - Proportions of Selected Use and Nonuse
Characteristics According to Sample Phase Status .................................. 69

iv

Study Population ......................................................................................... 48

Human Subjects and Informed Consent ............................................ 50
Study Sample ............................................................................................... 51
Data collection ................................................................................. 53
Plan of Analysis ........................................................................................... 56
Time Line .................................................................................................... 57
CHAPTER 3
RESULTS ............................................................................................................... 59
Demographics .............................................................................................. 60
Health Behavior ........................................................................................... 64
Hormone Use and Discussion ....................................................................... 66
CHAPTER 4
DISCUSSION ......................................................................................................... 7O
MICH Data Findings .................................................................................... 72
Conclusions .................................................................................................. 76
Strengths and Weaknesses ............................................................................ 77
Approaches to Strengthen Existing Database Use ......................................... 79
Future Directions for Research ..................................................................... 80
LIST OF REFERENCES ......................................................................................... 82
APPENDIX ............................................................................................................. 88

LIST OF TABLES

Table l - CHD Deaths Attributable to Specific CHD Risk Factors .......................... 4

Table 2 - Studies on Prevalence of HRT Use
Among Women 45 years and Older ......................................................... 10

Table 3 - Selected Characteristics Distinguishing

HRT Users from Nonusers ....................................................................... 14
Table 4 - Overview of the Impact of Hormone Use on CHD .................................... 25
Table 5 - Reasons MICH Women Were Not Interviewed ........................................ 54
Table 6 - Precision Based on Hypothesized Estimates

for HRT Use /Discussion .......................................................................... 57
Table 7 - Precision Based on Observed Estimates

for HRT Use /Discussion .......................................................................... 60
Table 8 - Demographics of MICH Women .............................................................. 61
Table 9 - Demographics of Study Patients by Phase Status ...................................... 63
Table 10 - Health Behavior Data .............................................................................. 65
Table l 1- Hormone Use by Phase Status .................................................................. 67

Table 12 - Proportions of Selected Use and Nonuse
Characteristics According to Sample Phase Status ................................... 70

vi

LIST OF FIGURES

Figure l - Time line of Subject Identification

and HRT Data Collection ....................................................................... 49
Figure 2 - MICH Study Sample F low Chart ............................................................. 51
Figure 3 - Interview Flowchart Regarding HRT ....................................................... 55

vii

INTRODUCTION

Coronary heart disease (CHD) is a major public health concern in the United States. It is
the leading cause of disability and death among women. A promising preventive
intervention for CHD risk in older women is hormone replacement therapy (HRT).
Epidemiologic evidence from research conducted over the last two decades suggests that
HRT decreases the risk of myocardial infarction (MI), the major subgroup of CHD, by up
to 50%. More recent studies also suggest that hormone use may be an effective secondary

preventive measure which decreases the risk of recurring MI.

This thesis begins by examining the literature over the past 15 years that deals with the
burden of CHD among women, patterns of HRT use, and the impact of HRT on MI risk.
It then presents data from a recently assembled series of community-recruited MI patients,
who reported on their frequency of HRT use and their history of HRT discussion with a
health care professional. The thesis ends with a discussion on the strengths and
weaknesses of utilizing an existing database to answer questions on HRT use among

community women.

Questions specifically addressed by this thesis are:

0

What does the current literature reveal about HRT utilization and its impact on
CHD in older women?

What is the prevalence of HRT use in a series of community recruited MI patients?
What proportion of women have ever discussed HRT With a health care
professional?

What characteristics are associated with women who are on HRT at the time of
their MI?

What are the strengths and weaknesses of the existing MICH database to address
these questions?

What are the lessons learned from accessing the MICH database for research?

Answers to these questions will be useful for present and future assessments of trends in

HRT use and for guiding physicians who provide information to older women on the use

of hormone therapy.

Chapter 1

BACKGROUND

Incidence/Prevalence of CHD in Women

Coronary heart disease (CHD) continues to be the leading cause of death in the United
States. The US mortality rate for heart disease in 1995 was 289.5 per 100,000. ' The age-
adjusted incidence rate of initial acute myocardial infarction (MI) in women was 137 per
100,000 in 1990.2 Although the incidence of MI, as well as all CHD, is lower in women
than in men until age 85, this disease has a major impact on the morbidity and mortality of
postmenopausal women. The CHD mortality rate (per 100,000) in 1995 was 318.9 for
women 55 to 59 years, 532.5 for 60 to 64 years, 820.4 for 65 to 69 years, and was

1,323.4 for 70 to 74 years (accounting for over 350,000 deaths per year). The prevalence
of CHD during the postmenopausal years is high-- 13% among women 45-65 years,
increasing to 33% among women over 65 years. 3 This contributes to a great deal of the
nation’s morbidity costs. In 1993, 875,000 females diagnosed with CHD were discharged

from short-stay hospitals.3

The risk of CHD among women varies with ethnicity. Afiican Americans have a higher MI
mortality rate than Caucasians. In 1992, CHD age-adjusted death rates (per 100,000) were

64.1 for Caucasian women and 85.0 for African-American women (a 32.6% higher risk

4

for the African Americans).3 Compared to Caucasian women, African-American women
have a higher prevalence of CHD risk factors. Over 40% of African-American females are
diagnosed with hypertension (HTN), 30% have high cholesterol, 45% are obese, and over
20% have diabetes.4 The prevalence of these risk factors in Caucasian women is 20% for
hypertension, 20% for high cholesterol, 20% for obesity, and 10% for diabetes. These
statistics are troubling since many M1 deaths (30% — 40%) are attributed to hypertension

and high cholesterol (Table 1).

Table 1: CHD Deaths Attributable to Specific CHD Risk Factors*

 

       
   

 

 

 

 

 

 

 

 

Risk CHD # %
Factor attributed of deaths
deaths
Smoking, 148,879
Carat/Former
Chol 253,194
_>_ 5.2 mmoIIL
High BP 171,121
SBP>140
Obesity 190,456 32.1 n
>110%desirble
No 205,254 34.6
exercise
Diabetes 77,709 1 3. 1
Total 593,111

 

 

 

 

*Prevention of Myocardial Infarction. Manson JE, et a1. Ed. NY, Oxford University Press 1996

Prognosis following an MI differs depending on gender and ethnicity. F orty-four percent

of women who experience an M1 die within the first year compared to 27% of men.1

5

Based on findings from the Framingham Heart Study and others since, within 6 years of an
M1: 31% of women will experience a subsequent attack; 34% will develop angina; 18%
will have a stroke; 20% will be disabled with heart failure; and 6% will experience a
sudden death" Partly because women tend to be older when the MI occurs, they are

twice as likely as men to die within the first few weeks. It has been speculated that the
decreased survival in women may be due to their increased mean age at the time of
infarction, and there are studies which report similar survival rates for men and women
after age adjustment.“2 However, in another study it was demonstrated that excess
mortality in women was not a function of age at the time of infarction. After a women
develops CHD, case fatality rates exceed those of men. 7 Prognosis following an MI is
worse for African-American women compared with Caucasian women. The case-fatality
rate (per 10,000) for African-American women has been reported at 47.5 versus 35.5 in
Caucasian women.8 The differential may in part be explained by the higher prevalence of
multiple risk factors in African-American women. A postmenopausal woman currently has
a 31% lifetime mortality risk from CHD, in contrast to a 2.8% mortality risk from hip

fracture or breast cancer.5

Trends
Coronary heart disease mortality rates have been declining significantly over the past few
decades. Data from the Framingham Heart Study have shown for women a 51% decline in

mortality from CHD during the 1970's and a further 25% decline between 1985 and 1990.4

6

Despite the role risk factors play in the incidence of disease, the substantial decline in
mortality has recently been attributed to advancements in medical management" A
‘period’ effect, such as medical advances in heart surgery, impacts changes in the
outcome, such as CHD mortality, across multiple age cohorts. Declining rates of CHD-
specific mortality disappear for very elderly women and have slowed considerably for
women in general. A period effect would explain the former lack of declining mortality,
whereby costly medical advances are reserved for those less advanced in age but do
impact more than one birth cohort. The declines in mortality observed over the last decade
do appear for each postmenopausal birth cohort.7 However, recent years have observed

half the decline in CHD mortality rates of earlier years.

Given that CHD is a major contributor to morbidity and mortality in older women,
primary and secondary prevention measures impacting incidence and prognosis are
important. This is especially relevant as the population shifts to a larger proportion of
older women. By the year 2015, the proportion of all women in the United States who will
be greater than 45 years of age will increase to 45%.lo Because life expectancy is greater
for women today than for women 50 years ago, there is a substantial number of years that

may be impacted by CHD preventive measures.

Hormone Replacement Therapy

Prevalence

One such preventive measure being examined for its impact on CHD in women is hormone
replacement therapy (HRT). Hormone replacement has been a commonly prescribed
therapy for postmenopausal women, extensively used for the relief of menopausal
symptoms. During the mid—1980's hormone therapy became indicated for the prevention of
osteoporosis and more recently considered as a therapy to impact coronary heart disease
risk. Currently, HRT is prescribed for approximately 30% of women experiencing
menopausal symptoms who range in age from 40 to 60 years.1 "'3 Short-term utilization
(less than 5 years) is most common in women seeking menopausal symptom relief which

can quite often be obtained in a year or two. '4

An older but smaller proportion of women are also prescribed HRT for decreasing or
preventing osteoporosis. Currently, 15 to 20 million women are at risk of bone fracture
and are candidates for preventing or slowing the disease with hormone therapy.'5 Such
women are commonly older, aged 55 to 85 years. Hormone replacement has been shown

to both prevent and partially reverse osteoporosis disease.”18

Despite the established benefits to HRT use in the postmenopausal years, only half of the

women prescribed therapy continue use beyond one year. A survey conducted at a

Massachusetts Women’s Health Center in 1987 demonstrated that 30% of HRT

8

prescriptions are never filled and 20% of women stopped treatment within 9 months. '9
Factors affecting the use of HRT included complications, fears, and mistrust of the
medical profession. Breakthrough bleeding, cramping, and bloating occurs in about 10%
of HRT users and is associated with high dropout rates.20 For many, achieving a
therapeutic dose that is consistent with benefits but associated with little or no

complications requires additional return visits and often results in noncompliance.

Some women may decline HRT use because it has been linked to breast and endometrial
cancers. The association between breast cancer risk and HRT has been investigated
extensively. Meta-analyses including 39 observational studies and 1 clinical trial estimate
no increased risk of breast cancer for those who are on HRT less than 5 years. 2'

However, epidemiologic studies do suggest a 20% to 30% increase in risk of breast cancer
for long-term current users of HRT compared to women who have never used HRT
(RR=1.25).22 The relationship of hormone use and endometrial cancer has long been
established. Estrogen alone significantly increases severe uterine hyperplasia in about 1/3rd
of women. The relative risk for endometrial cancer among estrogen users is reported to
be 8.22.21 These risks are based on estrogen use alone, whereas current hormone
replacement for women with natural menopause consists of a combined estrogen and
progesterone therapy which attenuates the risk of endometrial cancer (RR=1.3 and
RR=0.9, CI not reported) but may also attenuate the benefit (RR=1.2 and RR=0.5).20

Other studies suggest the degree of CHD protection may be as great as that of estrogen

9

(RR=0.53, CI:0.30-0.87).23 When the side effects associated with use and the fear of
cancer are included in reasons for noncompliance, from 20% up to 40% of women
prescribed HRT either never begin or cease treatment an average of 8 months after
initiation.24 Physician knowledge and attitudes about HRT also play a role in women’s
use of HRT. It is estimated that only 30% of postmenopausal women in the US have
ever been prescribed HRT and less than 40% of those with a prescription continue it

after 1 year.”

Estrogen therapy accounts for a majority (over 80%) of hormone therapy use and is

the most ofien measured formulation in the research on HRT. Estrogen combined with
progestin has accounted for less than 20% of HRT use. Table 2 lists the studies which
estimate prevalence of HRT among older, postmenopausal subjects. Estimates of HRT
use, which may vary according to a study’s time period, the sample age range, or study

location, range from 6% up to 32% in studies from the previous decade.

10

 

 

 

aw.

coa—

quO can E8» mv :25? 9.2.3 8: kg be 85.30...— ae 835m ”u 03:.

11

The highest reported prevalence of HRT use has been observed among groups
younger than 65 years and among women of higher socio-economic status (SES).
Prevalence was measured in a 1992 random sample of 2000 Finnish women who
returned a mailed survey (response rate: 84%). The majority (81%) reported natural
menopause and prevalence of HRT use was 22%. 26 A similar finding was observed in
a population sample of 4 southern US. communities during 1991. Personal interviews It?“
were conducted on 4958 postmenopausal women of whom 79% reported natural I
menopause. Prevalence of current HRT use was 21%. 27 Both of these studies

measured HRT prevalence in postmenopausal women less than 65 years of age and

 

therefore may be an overestimate of the true prevalence among older women in the

general population.

Postmenopausal use of estrogen was assessed in 1057 older women aged 50 to 79
years in a 1984-1987 cross-sectional study.28 Subjects were a community sample of
upper-middle class, Caucasian women first recruited for a cohort study 10 years prior.
Prevalence of estrogen use was 31% (n=328), and a small proportion of these (8%,
n=86) were taking combination therapy. Prevalence of HRT use differed by age and
was highest among women 65-69 years. Use among women in their 70th decade was
25%. The lowest frequency was for those on combination therapy. Only 3.7% of
women in their 60's and 2% of women in their 70's reported current use of combined

therapy.

In a study of 9704 older women, age 65 years and greater, from a multi-center

12

population-based study during 1986-1988, HRT prevalence was less than 15%.29 The
study targeted non African-American women for inclusion in research on risk factors
for osteoporosis from population lists of four major US. cities (Baltimore,
Minneapolis, Portland, and a Pittsburgh valley). This study included predominately
women with naturally induced menopause (84%). Overall, 13.7% of women reported
current use of oral estrogen (10.9% took estrogen alone, 2.8% took combined).
Current use declined sharply with age from 17% for those age 65-69, to 10% for those
75-79, and to 4% for women 85+. Because women volunteered themselves for the
study, they are likely to have been healthier than the general population. Consequently,
these prevalence rates may be overestimates of use among women in the general

population.

The above studies on HRT prevalence among older women were conducted on
homogeneous populations who may or may not reflect the general population of older
women. To address this issue, patterns of estrogen utilization were examined within
the Duke Established Populations for Epidemiologic Studies of the Elderly (EPESE),
an existing database which conducted baseline interviews in 1987.30 The Duke EPESE
population was a random sample of 28,000 persons over the age of 65 who lived in 5
adjacent counties in the Piedmont region of North Carolina. The study provided a
racially-balanced, well-sampled community population of older women. The principle
outcomes were self reported current and past use of estrogen at baseline interviewing.
Sampling identified 5226 men and women for enrollment and interviews were

completed by 4162 (80%) participants, of whom 2688 (65%) were women. Prevalence

13

of current use was 6.1% while 18.5% reported past use in the 2688 women

responding. Estrogen use for more than 2 years was significantly more common among
current users than among those reporting past use. The authors suggest that their
findings do not reflect a regional phenomenon , as estrogen use in the southeast United
States is comparable with use in other parts of the country. However, the studies cited,
as well as their own, were conducted many years ago (1980's) when formulations and
dosages were dramatically different than today and may not accurately reflect current

use patterns.

User Characteristics

Hormone users have been distinguished from non-users by selected characteristics
identified in many studies. These include differences in health and in lifestyle factors.
There are a few factors which have consistently been observed across different samples
studied and include socio-economic status and health behavior. Women who use
hormones after menopause are more likely to be Caucasian, educated, upper middle

class, and lean. These factors themselves place women at lower risk for heart disease.

The type of characteristics detected can differ depending on how the data was
gathered. A major concern is that women who take hormones are by definition
compliant, and compliance itself has been shown to decrease a subject’s risk of CHD
mortality.“ The difficulty with identifying characteristics associated with HRT use in
observational studies is that information on why women were prescribed hormones or

why they may not have been offered hormones is nonexistent. We are therefore unable

14
to contrast these groups and obtain a true picture. Table 3 lists the characteristics most
commonly distinguishing hormone users from an excellent review by Scalley and
Benrich, 1993.

Table 3: Selected Characteristics Distinguishing HRT Users from Nonusers*

Demographics Higher education
Higher income
Caucasian race

Health behavior More active
More likely under physician care

Attitudes / Beliefs Views menopause as medical condition
Less concerned about medical risks, side
effects, or Stigma
Views symptom relief as reason
for HRT use

 

 

‘Adapted front Scalley EK and Benrich .18. An overvrew Ofestrogen replacement therapy In
postmenopausal women. J Women’s Health 1993; 2(3): 289-94.

Studies conducted on the characteristics of use representing a wide range of

observational study designs are examined in greater detail below.

15

National Surveys

A national household survey on estrogen use with data obtained from 5 biennial cross-
sectional surveys examined selection-to-use factors between 1981 and 1990. 32 Among
the 13,186 persons surveyed, 3,279 were women 40-64 years of age and form the
basis of this analysis. Regression analysis controlling for age was employed.
Postmenopausal estrogen use was positively associated with income, with LDL
cholesterol levels, with exercising weekly, and with yearly cholesterol checks(p=
<0.005). A positive trend with education was suggested but not significant (p=0.25).
An inverse association with estrogen use was observed for smoking and body mass
index (p<0.005). Only a small proportion of women used HRT and had experienced
natural menopause (3.5%, CI 0.7-6.2). Although women reporting surgical menopause
made up approximately 29% of the women each survey, the majority of HRT users
were those with surgical menopause (80%). Thus the group status may have
determined other factors associated with HRT use in this study. Women seeking or
agreeing to have hysterectomies may have different characteristics, such as SES,
educational level, or health care seeking behaviors when compared to those who do
not have a hysterectomy. Cholesterol checks may be an indicator for recent healthcare

encounters which are required for HRT initiation and surveillance.

Population Based Surveys

The characteristics of hormone use were examined in 1984 using sociodemographic
and health-related information obtained in a telephone survey of 2 1 37 women aged 40-

52 years, randomly selected from a driver’s license list of Pittsburgh-area women

16

(response rate of 89%).'2 Overall, 6% of those surveyed reported current hormone
use. Hormone users tended to be older and thinner than nonusers. Only 813 women of
the sample were postmenopausal (38%) with 27% of the total representing
hysterectomized women and 1 1% of women with natural menopause. In separate
analysis of women having natural menopause (n=23 l), HRT use (5%, n=1 1) was
associated with higher education, lower BMI, and greater alcohol consumption. The
post (natural) menopausal women of this sample were relatively young (68% < 50
years) and the numbers small. Yet the factors associated with use are consistent with

characteristics associated with use from other studies.

Report of current use was examined during 1980 among 1057 women aged 38-82
years from a southern California community.28 Current users were more likely to have
had surgically-induced menopause, to have been past users of estrogen, and were
significantly younger (66 years vs 69 years). Age-adjusted characteristics were
analyzed. Estrogen users had statistically significantly lower weight, diastolic blood
pressure, and fasting glucose levels in comparison with nonusers. However, mean
differences between the groups were of minimal clinical difference; BMI difference of
0.85%, a blood pressure difference of 2.4 Hg/d, and a glucose difference of 0.06 mg.
Other CHD risk factors (e. g. hypertension, previous MI, family history, cholesterol
levels, and triglyceride levels) were equivalent regardless of HRT use or nonuse. This
sample of women were all from a similar SES background and biologic variables did
not differ in Significantly important ways between HRT users and non-users, with the

exception of hysterectomy status.

17

Medical Facility Surveys

A 1990 study on osteoporosis examined characteristics of HRT use in a homogeneous
group of Caucasian women who attended multi-site health clinics (N =9704).29 The
sample consisted of women aged 65 years and older predominately with naturally-
induced menopause (84%). Estrogen use was more common among women who had
higher education (p< 0.001), who were thinner (p< 0.001 ), and who participated in
sports or recreational exercise (p< 0.001). Estrogen use was associated with surgical
menopause (p< 0.001), with a diagnosis of osteoporosis (p< 0.001), and with alcohol
use (p< 0.001). A difference of 20% to 40% was observed for each of these factors

between current users and non-users.

The above study stratified women according to surgical or natural menopause. When
women were stratified by route to menopause, age (OR 0.68, CI 67-69), educational
level (OR 0.60, CI .49-.74), obesity (0.58, CI .49-.69), and alcohol use (OR 1.4, CI
11-16) were associated with current HRT use non-differentially in both strata.
However, only among women experiencing natural menopause, were there positive
association between HRT use and a diagnosis of osteoporosis (OR 2.58, CI 2.4-3.4) )

and increased physical activity (OR 1.59, CI 1.3-1.8).

The previously mentioned study (EPESE) on estrogen utilization among a racially
balanced random sample of 28,000 elderly persons also found differences between
users and nonusers.30 For demographic variables, estrogen use was negatively

associated with age (OR: 0.88, CI: 085-092), with number of children (OR: 0.09,

l8

CI:0.05—0.l7), with rural residence (OR:0.3, CI: 0.2-0.6), and with Medicaid
coverage (OR:0.02, CI:0.00-0. 14). A positive association was observed for use with
income (OR:6.3, CI: 2.2-19.9), being married (OR:2.8, CI:1.8-4.4), and private
insurance coverage (OR:6.4, cCI: 3.3-12.1). For medical and behavioral variables,
estrogen use was negatively associated with functional status (OR:0.7, CI: 0.6-0.9),
and the presence of obesity (OR:0.2, CI:0.l-O.9). Estrogen use was positively
associated with alcohol consumption (OR: 5.6, CI:3.6-8.8), cigarette smoking
(OR:5.6, CI: 3.6-5.9) and calcium supplementation (OR: 3.7, CI: 3.3-5.9). Especially
strong was the association between ethnicity and estrogen use. Caucasians were 10
times more likely than African Americans to be current users. Nine percent of
Caucasians (n=90) reported current use and another 270 (24%) were past users. The
corresponding percentages for African-American women were 1% (n=12) current use
and 10% (n=122) for past use. Population-based studies on elderly minorities (EPESE
study) confirmed low usage patterns showing minorities to be Significantly less likely

to report use of hormone therapy.30

Although several studies have reported that hormone users were healthier and had
greater access to healthcare,'2‘29'30 a few studies do not confirm these findings. Several
studies have found that smoking, drinking alcohol, and consuming a high fat diet are
more common among hormone users. 2830'” In the Nurses’ Heath Study, HRT users
tended to be slightly younger with lower body weight but also reported higher fat
consumption and higher rates of abnormal cholesterol levels, two risk factors that play

a major role in CHD. 33

19

The majority of women on HRT (80% - 90%) are those taking estrogen alone which is
common for women who have had a hysterectomy. A concern is that women with a
hysterectomy differ from women who go through natural menopause. The major
difference may be that women having the surgery are a subgroup who come under
routine medical care. Studies have found that women under routine gynecologic care
are more likely to be estrogen users. 34'” However, an encounter with a health care
provider does not ensure that hormone replacement will be addressed. Although a
plausible explanation for ethnic differences in HRT use is physician access, there are
studies which refute this explanation. A population-based study in Philadelphia of
women 45- years and over observed that African-American women were more likely
to be managed for diabetes and therefore more likely to have frequent physician visits,
but were half as likely to have ever used hormone therapy.22 Perhaps the very
conditions that bring African-American women into clinics (e. g., diabetes) are also
conditions in which clinicians are more reluctant to recommend HRT use. African
Americans may be less likely to have access or to use HRT when no chronic illness is
Present. However, even minority women having contact with a health care provider on
a regular basis have been shown to have significantly lower rates of HRT utilization.”
Attitudes and Behavior
Only a few studies have examined the receptiveness of women to HRT as a preventive
01‘ management therapy for heart disease. The importance placed on hormone therapy
by Women was examined during 1993 with a postal questionnaire to 1649 women aged
20-69 years.36 The sample was random and taken from 8 general practice listings in

northeast England. A 75% response rate was achieved. The majority of women listed

20

HRT promotion as ‘not important’. However, few (9%) named heart disease as a
threat to their health and fewer (7%) realized the role HRT may have in heart disease
prevention. Even when stratified for age, no trend in the fear of heart disease
appeared, although the level of knowledge about risk factors for CHD was very high

(> 85% correctly identified CHD risk factors).

A survey of 600 Stanford women graduates, median age 50, confirmed these findings
for United States women.37 Although knowledge of potential long term benefits and
risks of hormone therapy was high, only 27% rated the benefit on heart disease to be
most important. However, these women under estimated their risk of heart disease.
When asked what illness they feared most, 3 times as many feared breast cancer as
feared heart disease (48% vs 16%). Although these women may or may not have
underestimated their risk of CHD, their fears could explain the importance placed on

the motivation for preventing disease.

Experiences with menopause, as well as attitudes, can impact the probability of
hormone use. For those women who experience debilitating symptoms associated with
menopause, the need to seek relief provides the opportunity for discussing HRT and
heart disease prevention. Women not experiencing debilitating symptoms (as well as
healthy elderly women in general) and women without a primary care physician may
not have the opportunity provided to them to discuss the risks and benefits of HRT for

MI risk reduction.

21

Trends

The few studies that have examined changes in HRT use over time have found that
prevalence is increasing somewhat among post menopausal women. A study
conducted during the 1980s examined estrogen use among 2215 post menopausal
females aged 40-64 years sampled from 5 biennial cross-sectional household surveys
and observed increasing prevalence over time.32 Among all women surveyed, the
prevalence of estrogen use had doubled from 5.3% (CI: 3.2-7.4) in 1981/82 to 10.9%
(CI: 7.5-14.4) in 1989/90. The majority of estrogen use was among women with
surgical menopause (80% of users each survey). Among women with natural
menopause, estrogen use also increased from 1.5% (Cl: 004-29) to 3.5% (CI: 0.7-

6.2) resulting in a 57% increase over time.

The use of estrogen and its duration has increased significantly during the past decade
worldwide. A series of cross-sectional studies (repeated at 2-year intervals) of women
from Finland aged 45 to 65 years reported a fivefold increase in use from less than 5%
in 1976 to nearly 20% by 1989.'4 Regional and characteristic differences were
observed. The greatest trend in HRT prevalence was within the metropolitan area of
Helsinki (from 6% in 1979 to 30% use in 1989). Different from the earlier surveys was
the increased use of HRT in 1989 among postmenopausal women and those in rural
areas. The first group to adopt HRT were highly educated women. Its use then spread
to other educational groups with increases in prevalence seen among those of primary
and secondary schooling with each succeeding survey (from 5% or less among all

educational groups in 1979 to over 15% in 1989). For all surveyed periods, the highest

22

prevalence by age was among those 50 to 54 years.

In South Australia, there was a 40% increase between a 1991 survey and a 1993
survey in the number of women taking HRT.38 Current HRT use significantly
increased from 13.6% (CI: 11.3-15.9) in 1991 to 19% (CI: 16.2-21.8) in 1993. The
highest rates were for those in the 50 to 54 years age group (46.2% current use and

57.8% ever use).

Studies have also observed an increased trend in the number of women remaining on
HRT for long-term use. In the United States during the 1970's only 20% of HRT users
continued use for 5 or more years. While in 1992, the proportion remaining on HRT
for 5 years or more was 31%.39 Other countries also report increases in duration of
use. The Finland study observed that for HRT users in 1989, 32% had used it for up to
1 year, 43% between 1 and 5 years, and 25% more than 5 years.'4 These percentages
were significantly greater than each previous cross-sectional survey. Of those 60 years
and older, 37% had remained on HRT for more than 5 years. In the South Australian
study, a significant increase (p<0.01) in the median duration of therapy was observed,

from 38 months in 1991 to 54 months in 1993.38

Two developments directly impacting on HRT use in postmenopausal women of older
age occurred during the mid to late 19805 and may have led to a resurgence of HRT
use. These were the addition of progestin to attenuate the risk of endometrial cancer,

and a 1984 National Institutes of Health Consensus Conference advocating routine

23

estrogen replacement to prevent bone loss.28

Although there has been an increased trend in HRT utilization, it is still under utilized
among women who might benefit. There are more than 30 million postmenopausal
women in the United States. 4° Life expectancy for women reaching menopause is 28
years or l/3rd of a woman’s life span. At present only 2 to 3 million women in the

United States use HRT. 2°

The Effect of Hormone Replacement on Coronary Heart Disease

Substantial evidence from epidemiologic studies supports the cardio-protective effect
of estrogen. Table 4 lists the studies conducted on HRT and M1 or CHD. A 30-40%
reduction in risk of MI has been observed among women remaining on HRT for 5
years (short-term users), the reduction appears even lower with long-term use(5-10
years).33"°'4' It is clear that the majority of HRT research has been conducted with
estrogen exposure rather than combined therapy. This may be mainly due to the high
prevalence of hysterectomized women among hormone users who only need to be on
estrogen. The table also highlights a benefit observed in several studies among past
users of HRT, although the magnitude is less than that of current use. In addition,
these studies suggest a protective effect in several different outcomes; CHD events, MI

events, and total mortality.

i...

038% 55 3.55 «583292 coach»?
5 ..a C S
545 Foam: <2 .\. SE :2

.33 £615.
$929 «3an

w... my

.3 V 2 x»,

.5 63. 8
Host 8?:

as.
$8.15 Sumo §3 .32 Em x:

 

55 a. 8: 2.2.5: ... .22....— o... ... £355 3. 03.:

25
Epidemiologic Studies
However, the evidence of a protective effect of HRT risk of MI in postmenopausal
women is not conclusive, as there are currently no results from randomized clinical
trials. Two large trials have been implemented (W HI and PEPI) with results still a few
years off. The evidence that we do have on the relationship of HRT use and MI covers
a wide spectrum of study designs--from cross sectional to cohort. Caution is advised in
the interpretation of the findings on HRT and CHD because of biases inherent in the
design and with the analysis of the data. Examples include selection biases from a
‘healthy user’ effect and from a ‘physician referral’ effect, confounding from
HRT/CHD associated factors, and effect modification. Findings on HRT and CHD risk
are examined in light of the various threats to validity for each type of study design.
The following is a review of the strengths and weaknesses of each study type in HRT

research and a summary of the evidence from these studies.

Cross-Sectional Studies

One of the earliest examinations of HRT and CHD in post menopausal women was
conducted on the Framingham Offspring cohort during 1983 through 1987.42 A cross-
sectional design was employed measuring HRT use and lipid profiles in 938
postmenopausal women. Self-reported HRT use was obtained by a physician-
administered questionnaire. The use of estrogen was significantly associated with
increased levels of HDL-C (p=.04) but not with LDL-C levels. Combined therapy was
also significantly associated with increased HDL-C (p=.03) , as well as, with decreased

LDL-C levels (p=.04) and total cholesterol (p=.02). Other similar studies have

26

observed a significantly lower LDL level in HRT users.27 In this cross-sectional
comparison, as with studies of similar design, there is no means for determining
whether exposed subjects (users) represent a more health-conscious and compliant
group than non-users. It is well established that people who are health-conscious (e. g.,
physically active, good eating habits, non-smoking) are at decreased risk of CHD. In
addition, people who are compliant, even with placebos, reduce their risk of CHD due
to these and related behaviors.CDP In the case of health and compliant behavior leading
women to request or be prescribed HRT, the positive effect on CHD would be
spuriously attributed to the exposure (HRT) where measured cross-sectionally. For
the above study, age of the postmenopausal women ranged from 31 years to 69 years.
Given this wide range in ages, exposure status at the time of study measurement may

have little to do with exposure at the time the disease process began.

Several cross-sectional designs have been conducted to examine HRT exposure and
angiographically-documented CHD.43 Exposure was determined by self-report at the
time of radiologic exam. Although angiographically determined atherosclerosis is a
more definitive means of identifying predisposition to CHD than a single lipid
measurement, it is Still a proxy measure of mediating pathology. Inherent in the design
of all cross-sectional studies is the problem of ‘temporality’ or an inability to determine
whether HRT use preceded atherosclerosis, particularly subclinical disease. In addition,
the tendency for healthy women to self-select toward HRT use can not be determined

from this design.

27

Case - Control Studies

Case-Control studies are those which identify existing cases of M1 or CHD and
retrospectively determine exposure to HRT. Ideally, these studies should focus on all
newly diagnosed (incident) cases, include deceased cases as well as those alive, and
include controls that have had the same opportunity to be exposed as the cases.
Hospital-based studies, having clearly defined diagnostic criteria, provide a good
means of identifying the incident cases of M1 for a certain geographic area. In addition,
medical information is readily available for each subject. However, these studies can
provide inaccurate estimates due to the inability to identify and include fatal cases
occurring from sudden Ml’s. For women, 39% of initial Mls are fatal without
symptoms, and thus may not be included in hospital-based studies.8 Another source of
bias is the misclassification of cases as controls. There may be many patients in the
hospital, or even in the population at large, who suffer silent Ml’s, and would therefore
be identified as a control if included in the study. Both types of misclassification
classification can lead to inaccurate estimates of the impact of HRT on risk of MI and
may explain the various effect estimates (RRs from 0.6 to 0.9) found in the many

case-control studies. ”'44'45

Self-selection bias according to exposure factors (a problem for all observational
designs) can also occur. Women who are healthier in general and more involved in
their medical decisions may be more likely to seek out a physician to request the use of
HRT for protective benefits. The bias may also occur regarding professional factors.

Physicians may be more likely to suggest and prescribe HRT to women who are

28

generally in good health than to those who suffer from diabetes, hypertension, venous
thrombosis, or obesity.“ Self-selection biases are confounding and can spuriously

produce a protective effect that is much larger than truly exists.

To test whether a reduction in MI in HRT users may have been overestimated because
of a greater tendency for women at low risk to use estrogens, a case-control matched
study was conducted in l993.'l The authors selected first MI cases among
Massachusetts women aged 45-69 years in which each of 858 cases was age-matched
with a control from the same geographic area. Important correlates of estrogen use
(early menopause, smoking, parental MI hx, HT, diabetes, education, leanness, and
exercise level) and M1 were controlled by conditional logistic regression. The
estimated relative risk for ever use of estrogen was 0.9 (CI: 0.7-1.2) and the estimate
decreased with increasing duration of use (RR=0.6, CI: 0.4-l.1for 5 or more years).
The association with long-term duration was stronger for recent use (RR=0.5, CI: 0.3-
1.1)than for past users, but not significant. The results suggest that estrogen use may
reduce the risk of first M 1. However, this reduction is related to duration and recency
of use, and estimates may be smaller than previously Shown by similar studies. In
addition, due to the retrospective design of the above study, selection bias from ill
women tending to quit, or be removed from, therapy may spuriously produce a larger

estimate of benefit for HRT users.

Unlike women with a hysterectomy, women experiencing natural menopause require

estrogen therapy combined with progestin to attenuate the risk of uterine cancer. The

29

effect of combined therapy on risk of MI cannot be assumed from estrogen therapy.
The addition of progestogens may reduce or eliminate the cardioprotective effect
gained by elevations in HDL and lowering of LDL that is associated with estrogen use
alone.“45 Therefore, studies have targeted the exposure of combined therapy and
examined its impact on MI. If estrogen has a true beneficial affect on MI risk then it is
expected that combined therapy may also impact on risk, since combination therapy

does include estrogen.

A population based case-control study among those enrolled in a group health plan
examined this relationship.44 Cases were 502 postmenopausal women who had
sustained an incident fatal or nonfatal MI between 1986 through 1990. Controls were a
stratified random sample of 1193 non-MI females matched to the cases by age and
interviewed in the same calendar year as the diagnosed case. Telephone interviews
were conducted with survivors (and with proxies of mortality cases) and subjects were
classified into 1 of 3 groups: nonusers (reference group), estrogen users, and
combined-therapy users. The main purpose of the study was to address whether
combined HRT increased the risk of MI which previous studies had suggested due to
adverse lipid effects. Cases were similar to controls on mean age, length of
membership in the HMO, height, weight, number of pregnancies, and proportion with
hysterectomy. However, the controls were a significantly healthier group with lower
cholesterol and blood pressure levels, fewer diabetics, fewer hypertensives, and fewer
smokers. Both estrogen and combined hormone use were associated with a decreased

risk of MI in unadjusted analyses (RR=0.63, CI 0.44-0.89 and RR= 0.47, Cl 027-08

30

respectively). After adjustment for major coronary risk factors (age, diabetes, smoking,
HTN), the relative risk for combined HRT moved closer to the null effect (0.69) and
both CIS included 1.0. The strengths of this study include: 1) the use of population-
based controls rather than hospital-based controls, which more accurately reflects the
underlying exposure distribution in the source population; 2) complete case
identification which accounted for fatal, as well as, non fatal Mls avoiding a differential
due to missed cases; and 3) the use of incident cases only so that those with obvious
underlying disease processes do not obscure the estimates. In addition, all subjects
were members of the same HMO for a similar duration and so had equal access to
medical care. Limitations of the design that deserve attention are that: l) bias may still
be introduced with physician and/or patient self-selection to the exposure--patients
seeking hormones may also be more aggressive in their lifestyle behaviors and
physicians choosing to prescribe hormones may be more aggressive in other prevention
strategies; 2) duration for combined-therapy users was relatively short--averaging less
than 2 years for all subjects, exposures of three years or more may be required before
an impact is detected; and 3) women who were ‘not’ current users served as the
reference group. If women who stop therapy are more likely to be ill, they may also be
more likely to suffer an adverse event. The results observed may not be a true estimate

of the risk due to exposure.

Cohort Smdies

In prospective cohort studies, subjects free of CHD are classified according to

hormone exposure status and followed several years for the occurrence of an M1.

31

Confidence in the effect estimate is higher with this type of design because exposure is
ascertained before clinical disease develops. However, there are biases inherent in this
design as well. The loss of patients from follow-up (attrition) due to lack of interest,
migration, or death from other causes is particularly problematic. In addition, changes
in exposure status can occur over time making it difficult to classify subjects correctly
at the end of the study. In the case of hormone therapy it is necessary to provide
thorough follow-up to establish exposure as to current use, past use, or nonuse and to
achieve a Situation where all exposure groups have an equal probability of disease
detection. For HRT, it is suggested that the greatest benefit occurs with long-term use

of 5-10 years ” and follow-up may require 10 years to observe an effect.

There have been many cohort studies conducted to examine HRT and heart disease. In
general, a ‘cohort’ describes any designated group of persons who are followed over
time. This is distinguished from an ‘inception’ cohort which refers to a group identified
at an early, uniform point prior to the development of symptoms. A concern for HRT
studies is the ability, or inability, to determine disease status at the time of recruitment.
There may be underlying heart disease present at menopause when women make the
decision about HRT use. Unless a cohort is followed fi‘om menarche through
menopause, it is difficult to determine whether the disease process at postmenopausal
status has already begun or whether women with disease select to HRT use. Studies
have addressed this concern by eliminating women from the analysis who have a
history of CHD or poor lipid levels and recalculating risk estimates. The Lipids

Research Clinics (LRC) Program identified cohort of 2270 women whose status of

32

CVD and level of cholesterol were determined at baseline. They followed subjects an
average of 8.5 years.47 The investigators then analyzed the risk of death according to
exposure within the subgroups of women with existing CVD disease and with elevated
lipid levels. The overall age-adjusted relative risk (RR) of CVD death in estrogen
users compared with nonusers was 0.34 (CI: 0.12-0.81). After exclusion of women
with prevalent CVD, the relative risk in estrogen users compared to nonusers was
similar but of less magnitude -- 0.42 (CI: 0.0-1.0). Because this estimate is similar to
the estimate prior to exclusion but has a confidence interval much wider and including
1, the difference may be due to decreased power from a reduced sample Size. The
number of deaths for the group after exclusion was very small, in users (n=1) and in
nonusers (n=6). However, even with a larger sample Size the bias due to underlying
disease status is still a concern. Previous history of CVD was measured by self report
and by baseline lipid level, which may not accurately identify those with or without
disease. No current studies on HRT have examined the CHD association utilizing an
inception cohort and bias due to the failure of determining underlying disease remains a

concern in the interpretation of findings.

The above study also highlights the epidemiologic concerns of effect modification and
confounding. Effect modification occurs when the magnitude of an association
between a causal agent and a disease differs according to the level of a third variable.
In the above example, estrogen use was more protective against CVD in
hyperlipidemic women (RR: 0.21, CI:0.0-0.5) than in women from the target

population (RR:0.34, CI: 0.12-0.81). Among women selected because of elevated

33

lipid levels, the negative association of subsequent CVD mortality and estrogen use
was most pronounced. However, the women with estrogen-induced
hypertriglyceridemia may not be at increased risk of CVD, as their high lipidemic
counterparts would be. Initial analysis of women according to total cholesterol
suggested a significant reduction in mortality for estrogen users. When HDL level was
substituted for total cholesterol in the multi-variate model, the estrogen association
was no longer significant (p=0.29) suggesting that the reduction in mortality may be

mediated by the HDL-C fraction.

A Confounder is a variable that is causally related to the disease outcome and is
associated with, but not a consequence of, the exposure. The LRC study allowed for
the interpretation of confounding through statistical control of variables most
commonly characterized for HRT users--education, exercise, and type of menopause.
Education is associated with healthier lifestyles and lower risk of mortality m is
associated with greater involvement with one’s own medical care, including seeking
and using hormone therapy. The CVD mortality rates of the LRC study were
calculated separately for those with less than a high school education, those who were
high school graduates, and those with some college education. If education was
confounding the relationship with CVD we would expect the association between
estrogen and CVD to disappear. In this case, at each education level, estrogen users
had significantly lower (age adjusted) cardiovascular death rate than nonusers. The
CVD death rates (per 10,000) for those with less than a high school education was 14

(users) vs 40 (nonusers); for high school graduates, 16.5 (users) vs 37 (nonusers); and

34

for those with some college, 9 (users) vs 23 (nonusers). Exercise and hysterectomy
may also confound the association between estrogen and CVD risk and were included
as covariates in the LRC multi-variate model. A negative association between estrogen

use and CVD mortality remained regardless of exercise level or type of menopause.

Another example that has been suggested is that frequent contact with physicians
increases the probability of disease detection and confounds the association between
HRT and MI risk . In other words, those with an opportunity for exposure to HRT are
more likely to be healthy because conditions such as high cholesterol are kept under
control. If this were true and an over estimation of the relative risk was observed then
controlling for the confounder would result in a higher (or null) estimate. To further
assess whether frequent medical care might account for HRT’S benefit in
postmenopausal estrogen users, a multi-variate analysis was conducted on 48,470
participants of the Nurses’ Health Study.33 Researchers followed postmenopausal
women who were free from heart disease at baseline for 10 years. In an analysis limited
to women who reported having visited a physician in the previous year, results were
similar to those for the sample as a whole. The age-adjusted relative risks for CHD
remained low-- 0.45 (95% CI: 0.31-0.66) for current users and 0.79 (95% CI: 0.60-

1.05) for past users.

That type of menopause, access to health care, smoking, and risk factors for CVD do
not confound the association between estrogen use and CHD risk is supported by the

Nurses Health Study.33 This cohort consisted of 48,470 postmenopausal professional

35

nurses and followed them for over 20 years with less than 8% attrition. The study
observed a reduced risk of MI among women reporting current estrogen use, as well
as, women reporting past long-term use of 5 years or more. There were 405 cases of
non-fatal MI and coronary deaths. After adjustment for age, HTN, diabetes, and
smoking the relative risk of CHD associated with estrogen use was 0.56 (CI: 0.40-
0.80). The CHD risk was significantly reduced among both those with hysterectomy
(RR=0.4, CI:0.22-0.73) and those with natural menopause (RR=0.62, CI:0.39-0.97).
The association was similar in analyses limited to women who had recently visited a
physician (RR=0.45, CI: 0.31-0.66) and in those without a history of smoking,
diabetes, HTN, and hypercholesteremia (RR=0.53, CI: 0.31-0.91). The Nurses’
Health Study is the largest and most internally valid study conducted on the effect of

estrogen and risk of CHD.

Though the majority of studies have evaluated CHD risk with estrogen use alone,
studies, including the PEPI trial, have examined combination therapy with progestin.
Preliminary results suggest that combined therapy provides equal reductions in CHD
risk and, with some dosing plans, a more favorable effect on HDL. In a recent update
to the Nurses’ Health Study on 59,337 women with 16 years of follow up, women
taking estrogen combined with progestin showed decreased risk of CHD compared
with non users (RR= 0.39, CI: 0.19-0.78).'3 A prospective cohort study utilizing
prescription record linkage followed the entire female population of HRT users in a
Swedish region for 5 to 8 years to assess the effects of combined hormone therapy.

For the 23,174 women age 35 years and older, combined therapy reduced the risk of

36

first MI (RR= 0.53, CI: 0.30-0.87). The M1 risk was reduced only among women with
at least one year of combined therapy.23 The risk of MI in those filling prescriptions
during a 3-year period was compared with the risk of MI estimated for the general
population. Confidence in the estimates is reduced due to potential confounding. A
small random postal survey of the general population observed a less healthy group
than the HRT users regarding body weight, level of exercise, and lower education.
The confounding from unmeasured lifestyle factors may also be great. In addition,
rrrisclassification may be substantial. Use was defined as anyone filling at least one
prescription during a 3-year period. Noncompliers may have been included in the
cohort and women from the general population with exposure before and after the 3-
year period may have been included in the nonexposed comparison group. Such

misclassification could spuriously decrease the protective effect.

More recent cohort studies have also examined the impact of HRT on recurrent MI
risk in women with pre-existing CHD. In a retrospective cohort study of 726 women
(mean age=66) who had survived a first MI between 1980 and 1991, a marked
reduction in reinfarctions was observed for those on estrogen therapy.48 Adjusting for
age and time since first infarction, the relative risk was 0.64 (CI: 0.32-1.3) for current
users and 0.9 (Cl: 062-13) for past users. Other studies found that in women with
established CHD the risk reduction of recurrent MI may be as great as 80%. 43 The
benefit of HRT may be greatest in these populations who have experienced a cardiac
event. In women who have previously experienced an M1, the relative risk may be as

low as 0.2 for estrogen users.49

37

Caution is required in the interpretation of observational study results for many
reasons, including those of selection bias, effect modification, and confounding.
Cohort studies, as well as others, may suffer from self-selection which can bias the
estimate in either direction-- where healthy persons who are at lower risk of CHD seek
out HRT, where ill persons refuse or are not recommended HRT by a physician, or
where those at increased risk of CHD death (i.e., smokers or those with family history
of CHD) are recommended HRT. Given the magnitude of the difference in CHD
mortality risk between estrogen user and nonuser, it seems less likely that an
unidentified factor could completely confound and explain the association. However,
self selection of HRT use remains a major concern in the interpretation of the HRT and

CHD risk association from observational studies.

Mechanism of Action

The biological mechanism for HRT’S hypothesized impact on MI is chiefly due to the
hormone’s effect on risk factors for CHD. In animal studies, estrogen prevents low
density lipoprotein cholesterol @DL-C) buildup in coronary arteries.5°‘5' Excess LDL-
C is a known risk factor for heart disease in humans. Although the effect of lipoprotein
levels has a substantial impact on CHD risk, other factors also play a role. Additional
studies in animals demonstrate that estrogen has a vasodilating effect that can improve
cardio-performance in ischemic heart disease.”54 These animal studies are all well-
controlled randomized designs which introduce little, if no internal biases when
blinded. All but one of the above studies were blinded. However, it is unclear how

generalizable animal models are to similar human designs.

38

In-vitro studies have examined human arterial responses to estrogen exposure.
Atherosclerosis-free arteries were removed from patients undergoing heart
transplantation, precontracted with a prostaglandin, and exposed to estradiol-17 B or
control. The estradiol induced significant relaxation in coronary arteries pre-contracted
with thromboxane compared to control solution (82% vs 12% p< 0.01).55 Coronary
artery rings obtained from explanted hearts during cardiac transplantation were
precontracted with prostaglandin and exposed to either estradiol or control substance.
Results showed that estradiol induced a Significant relaxation of the smooth muscle
cells (85% vs 10%, p< 0.05).56 These studies support the biologic relationship but
suffer from low external validity. Explanted tissues may not react similarly to those of
endogenous organs. However, if the relationship is true, we would expect these results

in explanted tissues.

In vivo studies of humans suggest that HRT modifies CHD risk through lipid profile
changes, thrombolytic changes, and vasomotor effects. Half of the reduction in risk of
MI has been shown to be due to 10% to 15% decreases in LDL-C and corresponding
increases in HDL-C while using HRT.”58 Other studies have speculated that 25% to
50% of the risk reduction is due to the impact of HRT on vasomotor parameters such

as vessel wall elasticity and cholesterol oxidation in the endothelium.“59

Studies on estrogen use in menopausal women have consistently shown it to favorably
effect the lipid profile. The LDL-C levels is decreased 10%-15% and HDL-C increased

10% with use. A cross-sectional examination of the association of HRT with plasma

39

lipid levels was conducted on 15,800 men and women 45 to 64 years of age of the
Atherosclerosis Risk in Communities (ARC) study.27 Current estrogen users had
significantly higher HDL-C and lower levels of LDL-C than non users (p< 0.001). The
PEPI trial (a large multi center randomized trial) was conducted in 1994 and examined
effects of HRT on lipid profiles in women prospectively. Participants were blinded and
randomized to estrogen, combination therapy with progestin, or placebo and followed
for up to 10 years. Preliminary results have demonstrated significant increases in HDL
and significant decreases in LDL, as well as, decreases in fibrinogen levels (a
contributing factor in atherosclerosis which is associated with CHD). In addition, the
combination therapy was found to elevate HDL-C levels Without increasing the risk of
endometrial hyperplasia." The results of this study, a randomized controlled trial,
suggest strong support for a protective effect of HRT on MI through its positive

impact on CHD risk factors. It should be emphasized that this evidence from
randomized trial results is on HRT and intermediaries of CHD disease. Clinical trials

on disease outcomes have yet to be reported.

Strength of Evidence

The following paragraphs include six criteria commonly used to assess a causal
relationship. The evidence for HRT as protective of CHD is considered in the context
of each of these criteria.

1) Strength of the Association: The relative risks for M1 among HRT users has been

40

as low as 0.30 and consistently less than 0.80 across the various designs of
observational studies. The stronger study designs, cohort studies, have reported
relative risks of CHD mortality in HRT users of 0.34 (CI: 0.12-0.81), 47 relative risks

of MI in HRT users of 0.59 (CI: 0.42-0.82), '7 to 0.56 (CI: 0.40-0.80).22

2) Consistency of the Association: Protective effects of HRT use on the risk of CHD
have been observed among different samples of women, from fairly healthy
professionals (N urses’ Health Study) to retired community populations (Rancho
Bernardo, CA Study). A protective effect due to HRT use has been observed when the
analysis is confined to subgroups of women with differing underlying CHD risk
factors. For women healthy at baseline, the relative risk of MI among HRT users is
0.40 (CI: 012-082).22 For women who access a physician regularly, the relative risk
of MI among HRT users is 0.45 (Cl: 031-066), and even among women with pre-
existing heart disease, an HRT protective effect is observed with a relative risk of 0.64

(CI: 0.32-1.31“8

3) Dose-Response Relationship: The HRT and MI relationship shows substantial
evidence for a dose-response effect. The longer women remain on HRT the lower their
risk of MI. For those on HRT at least 5 years a 30% to 40% reduction in MI risk has
been observed compared to those on HRT for a year or two."°, while a 50% to 80%
reduction is observed for those with long-term HRT use of 5 to 10 years.” In addition,
the association is significantly Stronger for recent use (RR: 0.50) than for past use (P<

0.05).‘l The combination therapies also support the dose-response relationship as one

41

would expect if a causal link is present the combination therapies would also Show a
benefit, although not as strong due to the presence of progesterone. A protective effect
has also been observed for combination therapies with a relative risk of 0.53 (CI: 0.30-

0.87).23

4) Temporality: The direction of the association must be from cause to effect to argue
for a causal relationship. The inability to detect subclinical CHD disease often makes
this criteria difficult to meet. The HRT and MI question has been examined in
prospective studies on cohorts who were premenopausal (women in their thirties) and
followed them for over 20 years. The relative risks observed have been low (RR: 0.40
and 0.50) and it is unlikely that coronary heart disease existed in many women so
young. Therefore, effect estimates most likely suggest that the protection of HRT did
occur prior to the disease process. And in women with pre-existing disease the

protection from subsequent reinfarction that was observed is also temporally correct.

5) Specificity of Association: This criterion refers to the situation that a factor in the
causal pathway is a singular cause (or protection) of a single disease process. The
requirement to meet this criterion is less strict than the other 5 criteria because in the
real world a disease can be caused by multiple factors and a single factor can cause or
lead to multiple diseases. An example of a true causal factor for disease which is not
specific is smoking. That the HRT and MI link does not meet this criterion strictly is
mainly due to the multi-faceted aspects of heart disease, which can be attributed to

several causal pathways (i.e., smoking, diabetes, HTN, family history). In addition,

42

HRT can impact on other disease processes such as osteoporosis and menopausal
illnesses. Because HRT is implicated in reducing risk through its effect on
atherosclerosis and lipid profiles, it may be a sufficient but not necessary benefit in risk

reduction. This does not by itself weaken the causal argument.

6) Biologic Plausibility/Coherence: Additional support for a causal link exists if
biological plausibility is present in terms of the current knowledge about the exposure
and the disease. This is certainly true with HRT and CHD. Use of HRT has been
shown in clinical trials to positively impact the lipid profile, a major risk factor for
CHD. Women on HRT significantly decrease their LDL-C and total cholesterol levels
and increase their HDL-C levels.“l Because high LDL-C levels lead to atherosclerotic
plaque formation and atherosclerosis leads to the MI event, it is biologically plausible
that HRT impacts positively on MI risk. Also, when a woman’s estrogen levels are
high (e.g., pre menopausal), there exists significant gender differences in the incidence
of CHD. These gender differences lessen or disappear for women in the post
menopausal years, suggesting a link between intrinsic estrogen and CHD risk.
Therefore, it is plausible that extrinsic estrogen may play a similar protective role in
CHD risk. The CHD link is further strengthened by animal models which show that
estrogen treatment prevents LDL-C buildup in coronary arteries and has a positive

vasodilating effect on arteries.

Coherence of the evidence is provided by the protective effect of HRT exposure

observed in different groups. A protective effect is suggested for those reporting no

43

history of CHD and in those with a history of CHD,” 33 in younger postmenopausal
women and in older women,l7 and in women of countries other than the United

States.23

A preponderance of evidence on HRT and MI over the last ten years suggests a
significant impact due to HRT exposure and the association meets 5 of the 6 criterion
for a causal relationship. This Strongly supports the view that HRT is protective
against MI events. The consistency of findings is especially apparent in the cohort
studies which have produced a summary relative risk of 0.50 (CI: 04-05) in a meta-
analysis study.44 The relative risk of CHD in estrogen users compared with never users
in several recent meta-analyses was consistently below unity (0.55 to 0.65) in users
regardless of the endpoints investigated-~fatal or nonfatal MI, fatal or nonfatal CHD,
coronary stenosis, or sudden death?" 4“ However, it should be emphasized that an
over-riding problem of the evidence for a protective effect of HRT on CHD is that of
self-selection and its threat to the validity of reported findings. Selection bias is not
addressed by the above criterion but could effect any study that is not a randomized

design.

Even though a publication bias toward accepting studies which show an affect may
exist for even peer-reviewed journals, the epidemiologic studies reported cover a wide
range of designs which have consistently demonstrated reductions in risk of CHD
among women using HRT. Table 4 (page 27) presented an overview of principal

studies in the last ten years regarding estrogen’s impact on CHD. A 30% - 50%

44

decreased risk of MI has been observed for current users of HRT and a smaller but still
significant decreased risk of 15% to 35% for past users.“7’”’23'33 It is unlikely that the
protective effect is entirely explained by design flaws, selection biases, or confounding.
Evidence from multiple epidemiologic studies argues favorably for HRT as a

preventative therapy in reducing the risk of MI in postmenopausal women.

45

Thesis Rationale

The relationship between HRT and CHD is a major public health issue. Prevention of
later-life disability in the management of health for older women is important. Because
the majority of HRT use has been recommended to women who have had
hysterectomies and to women seeking relief of menopausal symptoms (60%-80%),
older women in the general population who do not fall into these categories may be
less aware of the benefits of HRT for CHD risk reduction. However, they may also be
more likely to reject hormone therapy because of complications or fears of cancer.
Previous studies report that, on average, 30%-40% of postmenopausal women are
prescribed HRT and only half of those prescribed HRT remain on therapy beyond 1
year.” 2" Prevalence of HRT use among women 2 65 years in non-select populations
is less than 14% and decreases with age. The prevalence of HRT use among a

population of women who have experienced MI remains unclear.

Support by health professionals regarding the utilization of HRT in older women and
especially in women who have coronary heart disease may be changing. Physicians are
now being encouraged to consider HRT as a prophylaxis in the management of MI.
The 1996 US Preventive Health Task Force recommends counseling all
perimenopausal and postmenopausal women regarding the risks and benefits of HRT."
The American College of Physicians in 1992 made a general recommendation to
consider preventive hormone therapy for women who have or are at risk of CHD.62
Physicians from the Cardiovascular Division of Harvard Medical School advocate

cardiologists take an active role in educating all postmenopausal women with CHD of

46

the benefits/risks of hormone replacement.63 Interestingly, a survey by Ferguson, et a1
addressing women’s decisions for using estrogen revealed that the most important
reason for starting estrogen was a recommendation by a physician.34 With recent
support of professional medical groups advocating counseling women at risk of CHD
about HRT, studies are needed to update estimates of HRT use and physician
discussion of HRT use.

Thesis Objectives

This thesis describes HRT use and discussion of HRT use with a health care provider
among women who presented to community hospitals for myocardial infarction. The
initial objectives were to estimate the prevalence of HRT use (and HRT discussion
with a physician) among a series of population-based MI patients, to contrast the
obtained prevalence rates with those of previous studies, and to compare prevalence
rates between ethnic groups. However, these objectives required modification
following problems with data collection which resulted in a failure to reach the
projected sample size, thus limiting study power and generalizability. In light of this,
the objectives were modified to a descriptive analysis of the sample data for frequency

of HRT use and of discussion with a physician.

Chapter 2

METHODS

The MICH Study

Purpose

The parent study for this thesis research is the Michigan Inter-Institutional
Collaborative Heart Study (MICH) which recruited patients from five hospitals of the
central Michigan area of Genesee and Saginaw Counties. The MICH study is
implemented by a committee at Michigan State University, College of Human
Medicine. The purpose of the MICH study was to collect key data on AMI patients
during hospitalization and one year following the AMI event. Initial study aims were to
assess quality of medical care for women and minorities, cost effectiveness of various

treatments, and the impact of AMI on quality of life and return to work.

Geographic Setting

Genesee County has approximately 370,000 residents, 51% of who are female and
18% African American or other minority."4 Saginaw County has approximately
250,000 residents and a similar demographic composition to that of Genesee County.
The MICH study is a population-based case series of acute myocardial infarction

(AMI) hospital patients recruited from these two counties.

47

48
Eligibility and case definition
The diagnosis of myocardial infarction was confirmed by electrocardiographic(EKG)
results (Q or OS finding + ST elevation or T-Wave inversion), by the typical symptoms
of shortness of breath, nausea, dizziness, severe discomfort in the anterior chest, and
by an elevation of enzymes (elevation of CK and MB fraction > 1.2 times the upper
limit of normal or elevation of SGOT or LDH > 2 times upper limit of normal) within

48 hours of symptom onset.

All non-transferred patients meeting the criteria at presentation of chest pain, high
enzymes, and positive EKG were included in the MICH study. The study criteria for
Phase 1 patients were: any one typical symptom of MI plus elevated serum enzymes
§_n_(_1_ EKG confirmation. The study criteria for Phase 11 patients were: elevated enzymes
plus evidence of MI in the patient chart by symptoms gr EKG confirmation. Patients
excluded were those who were transferred from a non-MICH hospital, met enzyme
criteria due to a scheduled surgery, met enzyme criteria due to angina and/or PTCA or
CABG, or met enzyme criteria but were hospitalized due to other causes (e. g.,
pneumonia hypoxia, pancreatitis, or car accident). A total of 621 women met the MI

criteria for MICH (407 from Phase I and 214 from Phase 11)

Recruitment
Figure l is the time line of MICH implementation. Subjects were enrolled beginning in
January 1994 until October 1997 (patient recruitment ceased during 1996 to conduct

follow-up interviews). Recruitment occurred in two distinct phases. Phase 1

49

Phase I 1994 - 1996

 

January March May July Sept December

Subject identification 1994 - 1995
and enrollment _ ..................................

1996

Followupinterviews _.....__.__.........._____

1996 - 98

Closeout interviews

Phase II 1997

 

January March May July Sept December

 

Subject identification

and enrolhnent _ .................... _

Hospitalinterviews _..............________

Follow up interviews _ . . _ . . _ .

Figure l - Time line of Subject Identification and Data Collection

50

began January 1, 1994 and recruited patients through April 30, 1995. Between May 1,
1995 and March 1, 1997 research activity focused on obtaining follow-up data. Phase
2 began additional patient recruitment during March 1, 1997 through September 30,
1997. The MICH study was approved by the University Committee on Research in
Human Subjects (U CRIHS). Trained nurses identified eligible patients in-hospital and
approached them for study consent. Potential participants were approached in the
hospital prior to discharge, at which time they were informed about the MICH study
and about subsequent interviews, both in the hospital and during follow-up after
discharge. All eligible patients consenting to study inclusion were then interviewed
prior to discharge. The MICH assistants conducted all chart abstraction on patients
within the study and collected follow-up data for Phase 1 participants from subsequent
3-month, 6-month, and 9-month mail surveys and a close-out (l to 3 year) telephone

interview. Phase 2 follow-up included a 6-week post discharge telephone interview.

Data Collection

Figure 2 outlines women enrolled and interviewed for each phase of MICH. In
addition to interview data, information was abstracted from patient charts by MICH
nurses. The data collected included medical histories, hospitalization course,
laboratory test results, medications, and procedures preformed. Women were
interviewed in-hospital regarding demographic, health, lifestyle, and clinical
information. The data was collected and entered into a MICH database. Data cleaning

and subsequent data analyses were conducted by MICH assistants.

1994 Phase I

1997 Phase II

 

 

Subjects Meeting

Criteria 407

 

 

 

Subjects Meeting

Criteria 214

 

 

 

Abstracted Data

407 (100%)

 

 

Abstracted Data

214(100%)

 

 

 

 

Hospital Interviews

124 (31%)

 

 

Hospital interviews

87 (41%)

 

Figure 2 - MICH Study Sample Flow Chart

52

Thesis Study

Purpose

Limited information exists on HRT use among women with preexisting heart disease.
The purpose of the thesis research was to describe the utilization of HRT among
women experiencing an acute MI and the frequency of having discussed HRT with a
physician. All postrnenopauasal women recruited into the MICH study were eligible

for the thesis research on HRT.

Eligibility

All MICH study women who were postmenopausal or who had a hysterectomy were
eligible for the HRT study. Study participants were postmenopausal women recruited
into the MICH study and completing follow-up telephone interviews during Phase 1 or

in-hospital interviews during phase 2.

Human Subjects and Informed Consent

The original Institutional Review Board approval for the MICH consent form was not
altered because it was consistent with collection of follow-up information obtained
from MICH patients. Questions on HRT were added to the existing MICH interview
instrument. This addition was formally approved by the University Committee on

Research in Human Subjects (U CRIHS) on 3/27/97.

53
Recruitment
There were 124 Phase I women with completed in-hospital interviews. Follow-up
telephone interviews which included questions on HRT (conducted 1 to 3 years post
discharge) were completed for 68 women (56% of those with hospital interview). For
Phase II recruitment, 214 women met study criteria. A total of 87 (41% of those
meeting criteria) Phase H women completed in-hospital interviews which included
questions on HRT. In some cases, interviews were conducted shortly after discharge.
Table 5 lists the various reasons that MICH women were not interviewed during the

in-hospital and follow-up surveys.

Table 5: Reasons MICH Women Were Not Interviewed

Not meet criteria 00 00
Refused 00 5 (04) 60 (12)
Deceased 00 11 (09) 62 (12)
No answer 00 12 (10) NA
Disconnect/wrong # 00 28 (23) 00
Discharg/unknown 00 00 79 (15)
Unable to interview 6 (04) 00 74 (14)

Reasons not 273 (67) 00 00
specified

Total non-response 283 (70) 56 (45) 127 (59)

Completed 124 (31) 68 (55) 87 (41)
interviews

 

54

This study includes 156 women; 69 who were interviewed at follow-up in Phase I and

87 who were interviewed close to discharge in Phase 11.

Data Collection

At the follow-up interview (Phase I), a telephone survey was conducted by MICH
nurses and medical graduate students regarding follow up information on lifestyle
changes, health, symptoms, and HRT use. For Phase 11 participants, questions
regarding HRT use and discussion with a physician were obtained during the in-
hospital interview. Validated questions on HRT utilization and participant
characteristics were obtained from the published literature. Questions specific to
physician recommendation were developed myself. The survey was then pilot tested
on 50 selected older women by telephone survey. The HRT questionnaire consisted of
a 3-page structured survey with forced choice and open-ended questions regarding
whether a physician ever recommended HRT to the patient, the specialty of those
physicians who had recommended HRT, HRT use at the time of the heart attack,
current use and compliance patterns, past HRT use and compliance patterns, reasons
for noncompliance, and whether a physician had ever recommended withdrang from
HRT. Figure 3 shows a flowchart outlining HRT questions. These questions (with a
slight variation in the query to reflect use ‘at the time of the heart attack’) were those
used for Phase I women at the follow-up interview 1 to 3 years post discharge and for
Phase 11 women during the in-hospital interview. The Appendix provides the

instruments used to survey subjects about their exposure to HRT.

55

Were you taking HRT at the time of your index heart attack? m

l

E CS_._-.l Have you ever taken HRT in the past?

 

I

Physician ever discuss HRT with you? —>

 

v

 

 

1' Yes No
Who recommended you take HRT?
Family physician Gynecologist Cardiologist Other DK
Did a physician recommend you stop taking HRT?
m ’ @ou take your HRT medication regularly? 5 CS."

 

1

Why not? Cost / not helpful / worried of

V

cancer

/ side eflects
Who? Family physician Gynecologist Cardiologist Other

1

Are you currently menstruating? 4
Yes/No

 

 

Have you ever had breast cancer? ¢
Yes/No

1

Have you had a hysterectomy?
Y es/No

 

 

FIGURE 3: INTERVIEW FLOWCHART REGARDING HRT

56
Plan of Analysis
Demographic characteristics of study participants were grouped into categorical
variables such as age (<45, 45-54, 55-64, 65-74, 75+), ethnicity (Caucasian, African
American, Other), insurance status (Private, Medicare, Medicaid, HMO, None),
income (<20,000, 20,001-40,000, >40,000), and education (< HS, HS degree, Post
HS schooling). Smoking status and family history of CHD were expressed as
dichotomous variables, present or absent. The main variables of interest, HRT use and

discussion of HRT were scaled as dichotomized variables, yes or no.

Precision was calculated for estimates of HRT use and discussion of use with a
healthcare professional from Phase I follow-up interviews. A calculation for the
precision that could be achieved with a hypothesized point estimate of 12% for HRT
use and 30% HRT discussion and a projected sample size of 200, assuming a 95%

confidence level was performed (Table 6).

Table 6: Precision Based on Hypothesized Estimates for HRT Use /Discussion

 

Parameter

Projected
Sample Size

Precision i

 

57

As time progressed we realized that the sample size was considerably less than the 200
projected initially and we prepared for increasing the sample Size by incorporating the
HRT questions into the Phase II in-hospital interview rather than waiting for additional
follow-up interviews. When it was discovered that the sample size of both Phases
would fall short of the 200 projected, we anticipated precision to be somewhat poorer.

This is particularly true for data examined separately by Phase.

Time Line

An initial request was made to incorporate the hormone replacement therapy query
into the MICH research project in January 19, 1996. The HRT related questions were
developed and submitted to the MICH coordinator for revisions which was approved
at a subsequent MICH meeting on March 15, 1996. An invitation for additional
investigators interested in HRT to be co-investigators was put forth and three were
identified. Additional questions from the new investigators were incorporated and a
second revision submitted to the MICH committee was made on April 19, 1996. A
request for pilot testing was made and then conducted on a convenience sample of 10
women. Results were presented at the June 21, 1996 meeting. At this time,
recruitment of patients for Phase I was completed and the HRT questions could not be
added to the Phase I in-hospital interview. A final vote on the questions for inclusion
was made in December of 1996, at which time it was determined that the hour long
telephone MICH instrument was too lengthy and HRT questions (as well as many
others) had to be minimized. The HRT questions remaining were those essential to my

inquiry (i.e., use, discussion with health provider, who recommended HRT, and

58

reasons for noncompliance). The Phase I close-out interviews were begun in February

of 1997 and the HRT questions were included.

Phase 11 patient recruitment was begun on March 1, 1997. The HRT questions were
revised to be incorporated into the in-hospital interview. Contacting women for HRT
questioning prior to discharge would increase our sample Size from that achieved at
Phase I follow-up. Recruitment was scheduled for March through September 30,
1997 with an anticipated enrollment of over 200 women. However, recruitment
stalled, ceasing at one location during the month of July, resulting in a total of 87
women interviewed in-hospital. Chart abstraction by medical volunteers and previous
MICH recruiters was conducted from October 1997 through April 1998 with data

entry completed in June 1998.

Chapter 3

RESULTS

Final Precision Levels

The combined total from both Phases resulted in 156 female participants with a
prevalence of HRT use of 22% and prevalence of discussing HRT with a health care
provider of 46%. The resulting precision for these point estimates is decreased from
that anticipated and values are listed in Table 7. The precision values by phase are
reported separately in Table 7 as well. The resulting precision by phase was

considerably compromised due to the smaller sample sizes.

Table 7: Precision Based on Observed Estimates for HRT Use lDiscussion

 

59

60

Demographics

Of the 621 women enrolled into the MICH study with chart data abstracted, 156
completed interviews on HRT questions. Table 8 compares demographic data on the
466 women with chart data only compared to the 156 woman interviewed for HRT

(Phase I follow up= 69, Phase II: 87).

Table 8: Demographics of MICH Women

 

 

 

 

 

 

Primary Insurance
Private 98 (21) 55 (35)
Medicare 331 (71) 67 (43)
Medicaid 19 (04) 06 (04)
None 14 (03) 07 (05)
Missing 5 (01) 21 (13)
Ethnicity
Caucasian 401 (86) 134 (86)
AfricanAm 56 (12) 17(11)
Other 9 (02) 5 (03)
Age II: 67.9 yrs It: 66 yrs
44 and under 14 (03)
45 thru 54 33 (07) 11 (07)
55thru 64 79(17) 21 (14)
65 thru 74 122 (26) 43 (28)
75 and over 168 (36) 35 (22)
Missing 56 (12) 46 (29)
00

 

Women interviewed were less likely to be on Medicare and slightly more likely to have

private insurance than women not interviewed (35% vs 21%, p=.0005). The younger

 

61

mean age of interviewed women may explain differences in insurance coverage ( u =
68 yrs vs II: 66 yrs). Differences between the two groups may reflect a tendency for
those of higher economic status and younger age to respond to surveys or reflect that
older women are more likely to die before interviews can be obtained. The ethnic

distribution was similar between the groups.

Demographics of study participants were examined according to their Phase status.
Variables available for both Phases included age, ethnicity, and employment.
Education, income, and marital status were available depending on the Phase. Table 9

outlines the demographic data according to Phase status.

62

Table 9: Demographics of Study Patients by Phase Status

 

 

 

 

 

 

 

Betrayaphie ' . I, ' Phaael'fo'ta‘lF69 » Phase HMW : .1':
Age (11) 64.2 yrs 68.9 yrs
range: 42 - 88 range: 28 - 92
Ethnicity
Caucasian 62 (90) 73 (84)
African American 6 (09) ll (13)
Other 1 (01) 1 (01)
Missing 00 2 (02)
Employment
Employed 22 (32) 18 (21)
Homemaker 20 (29) 18 (21)
Retired 21 (30) 37 (43)
Un-employed 6 (09) 10 (l 1)
Missing 00 4 (04)
Education
< H. S. Degree -NA-- 26 (30)
Have H.S. Degree 36 (41)
Some College 17 (20)
Assoc Degree or 06 (07)
more 2 (02)
IinsSing
Income
< $12,000/yr 17 (25) -NA-
3121: to $19,999k 12 (18)
$20k to $39, 999k 23 (34)
$40k to $59,999k 8 (12)
$60k or more 5 (07)
Missing 3 (04)
Marital Status
Married -NA- 37 (43)
Divorced/Separated 9 (10)
Widowed 37 (43)
Single/never 3 (03)
married 1 (01)
Missing

 

 

 

 

 

NA= not available

63

There were ethnic differences between the two Phase groups. Phase I women had
fewer African Americans then Caucasians (9% AA vs 13% AA) but this difference was
not statistically significant (p=0.6). Phase I women were more likely to be employed
(32% vs 21%) while Phase 11 women were more likely to be retired (43% vs 30%).
Again, these differences were not statistically significant (p=0.2). A majority (71%) of
the Phase 11 women had a high school degree or less and 56% were single. F orty-three
percent of Phase I women reported an annual income of less than $20,000. Although
the mean ages were similar, Phase 11 women included those as young as 28 years while

Phase I had a 42-year-old as their youngest.

Health Behaviors

Descriptors for health behaviors were examined between the two Phases. Table 10
presents the health behavior variables of smoking, history of MI, family history of MI,
level of activity, and reported presence of diabetes. The variables differed depending

on Phase status.

Table 10: Health Behavior Data

    

 

 

 

 

Told have Diabetes I Family Hx MI*
Yes 18 (26) 3 Mother 32 (37)
No 50 (73) Father 36 (41)
Don ’t know 1 (01) Don ’t know 11 (13)
Chol Checked 66 (96) History MI
Told Chol High 44 (64) Self 23 (26)
Smoke Post MI Smoke Prior MI
Yes 5 (07) Yes 48 (55)
No 63 (91) No 39 (45)
Physically Active Physically
Yes 37 (54) Active 49 (56)
No 31 (45) Yes 38 (44)
N0

 

 

 

 

* multiple responses and do not add to 100%

Information describing health behavior for the two phases were as follows; more than
half (55%) of Phase 11 women interviewed in-hospital reported current smoking and
7% of Phase I interviewed post-MI reported current smoking. The difference is likely
due to the timing of the question. Phase I participants were asked about smoking 1 to

3 years following their MI and Phase II were asked about smoking just prior to their

65

MI. Twenty-six percent of Phase 1 women reported having diabetes and 64% reported
having high cholesterol. For Phase 11 women 77% reported either a mother or father
or both as having heart disease. The percent reporting that they were physically active

at the time of the interview were similar between the two groups (54% and 56%).

66

Hormone Use and Discussion

Women in both Phases were questioned about HRT use at the time of MI. Of the
Phase I women who reported HRT use at follow-up (n= 16), 7 (44%) had been placed
on HRT post-MI discharge. Nine (75%) of the 12 women on HRT at time of MI
remained on it at the time of follow-up interview and 3 (25%) had quit taking HRT

afier the MI. Table l 1 provides firrther data on hormone activity.

Table 11: Hormone Use by Phase Status

 

 

 

 

 

 

 

 

 

HRT Use
At follow-up 16 (23) -NA--
At time of MI 12 (17) 23 (26) 35 (22)
Prior to M1 -NA-- 16 (19)
Ever Discussed HRT 27 (39) 44 (50) 71 (46)
(with BC provider)
Recommend by:
Family Phys 1 1 (16) 26 (30) 37 (24)
Gynecologist 0 l4 (l6) 14 (O9)
Cardiologist 0 0 0
Other doctor 1 (01) 0 l (.01)
Does not apply 53 (77) 47 (54) 100 (64)
Advised Start HRT
(Following MI, n=8)
By Family Phys 3 (3 7) --NA--
By Gynecologist 2 (25)
By Cardiologist 1 (12)
Other doctor 2 (25)
Hysterectomy 25 (36) 34 (39) 59 (38)
Still Experiencing 7 (10) 8 (09) 15 (10)
Periods

 

 

 

 

67

The HRT data were examined by Phase. Prevalence of HRT use in Phase I was 17%
(CI: 8 - 26) and 39% (CI: 27.5-50.5) for discussion. Prevalence of HRT use in Phase
11 was 26% (CI: 16.8-35.2) and 50% (CI:35.5-60.5) for discussion. Precision was
poorer for each phase than for the combined total due to the smaller sample size.
There was a lower prevalence of HRT use in Phase I than in Phase II which may be
attributed to differences in when and how the data were obtained. The proportion of
women having a hysterectomy were similar in the two groups (3 7% and 39%).
However, more Phase 11 women reported hormone use at the time of MI (26% vs
17%). There were more Phase 11 women reporting recommendation of HRT use by a

gynecologist (16% vs 0) and a family physician ( 16% vs 30%) than Phase I women.

The combined Phases total 156 women responding to questions on HRT. The
prevalence on HRT use for the combined data (n=35) was 23 % (95% CI: 16.4 - 29.5)
and prevalence on history of HRT discussion with health care provider (n=72) was
46% (95% CI: 38.2 - 53.8). The prevalence for the combined data, as well as

physicians recommending HRT, is listed in Table 1 1.

Characteristics distinguishing HRT users from nonusers in the study sample were
examined by Phase and for the total sample. Table 12 lists the demographic, medical,
and health behavior variables available to compare users from nonusers. Although the
numbers are small there were some interesting findings. Users overall were more likely
to be younger (62 yrs vs 66 yrs, p=.21), to have private insurance (49% vs 32%,

p=.10), to be employed (34% vs 23%, p=.27) but not significantly so. They were

68

significantly more likely to have had a hysterectomy (63% vs 32%, p.002). The HRT
use by ethnicity differed for the two phases. The Phase I group had no African
Americans reporting use of HRT at the time of MI while the Phase 11 group had a
greater proportion of African Americans among HRT users (17%) than among non
users (1 1%). Information on education and family history was unavailable for Phase I.
The Phase II data showed that users were more likely to have a degree beyond high
school (39% vs 21%, p=.16) and to have a family history of MI (48% vs 39%, p=.61)

but these differences were not statistically significant.

69

 

swim . 9.3.2
as S EC 3. A5 2
£ch a: : 80 e

WEB“? N
Ea as
:83. 88:52

scam 8:: nae-m a “.5283. 83.30220 oasez ...... a: 3.8.8 ... at 3:288.— fi 2.:

Chapter 4

DISCUSSION

Coronary heart disease is a major contributor to morbidity and mortality in older
women.1 The prevalence of CHD during the postmenopausal years is high--33%
among women over the age of 65 years.3 African-American women are particularly
vulnerable. Compared to Caucasian women, African Americans have a much higher
prevalence of CHD risk factors such as hypertension and high cholesterol.“ They are
also at higher risk of death due to CHD than Caucasian women.l Although recent
years have observed declines in CHD mortality rates, a postmenopausal woman

currently has a 31% lifetime mortality risk from CHD.5

Hormone therapy is commonly prescribed for women experiencing symptoms of
menopause. Among women 40 to 60 years of age, 30% to 40% are prescribed HRT.
”“2 However, more than 60% of those with an HRT prescription discontinue use
within one year. 25 Unlike women who utilize hormone replacement for the relief of
menopausal symptoms, older postmenopausal women on HRT for CHD or
osteoporosis are in an age group that, as a whole, are more concerned about breast
cancer 37 and object more to HRT side effects than younger women.34 Prevalence of
HRT use among postmenopausal women greater than 60 years of age is only
12%,29'30‘“ and a mere 6% among minority women.3O Although research shows that

women who have taken HRT in the past are more likely to take it again when older,

70

71
the prevalence of HRT use in older women remains half that of pen-menopausal
women.”29 In addition, a majority of prevalence studies on current and past HRT use

Show half as many older women reporting current use as reporting past use. ”33:41“

A preponderance of evidence suggest that HRT is effective as a preventative therapy in
reducing the risk of MI in postmenopausal women. A 30% to 50% decreased risk of
MI has been observed for current users of HRT and a smaller but still significant
decreased risk of 15% to 35% for past users."7'”‘23'33 Although it does not explain the
full impact of risk reduction, HRT use postmenopausally has been linked to higher
SES, 32' 3° higher education, '2‘ 29 greater awareness of health issues,” and better access
to medical care 32 in many studies and selection bias cannot be fully ruled out.
However, other studies have observed no differences between HRT users and

nonusers in SES, 28 education}2 and access to medical carefm

Because half of all postmenopausal women are at risk of osteoporosis15 and a third are
at increased risk of CHD,3 a 30% to 50% prevalence of HRT use among older women
would not be unexpected. However, the literature emphasizes that HRT prevalence
among older, postmenopausal women, who are at the greatest risk of MI and of
osteoporosis, is currently very low (6% to 14% among those 2 65 yrs). The prevalence
of HRT use was 17% among a select group of women who were members of a health

organization in Puget Sound and had previously experienced an M1.48

72

MICH Data Findings

Prevalence

Among the 156 women of this study sample, less than half (46%) had ever discussed
HRT with a health care provider. Because women with a hysterectomy accounted for
almost 40%, few women in the study with natural menopause had ever discussed HRT
with a physician. Although the level of HRT discussion is relatively low, reported use
of HRT at the time of MI (22%) is higher than that of other studies involving
postmenopausal women. In studies with women reporting a Similar frequency of
hysterectomy and in studies including only women older than 65 years, prevalence of

HRT use was less than 15%.”'“

Unique to this study is that all women had experienced an MI. Compared to a
historical cohort study conducted in Puget Sound involving only women with M1,
prevalence of HRT use in our sample was slightly higher (22% vs 17% 48). However,
the confidence intervals of our sample do include the estimate of the Puget Sound
study and these two studies may be similar in their prevalence. The two studies were
similar regarding type of menopause (3 7% with hysterectomy) and ethnicity (86% &
92% Caucasian). The two studies differed in other ways. The Puget sound study
included only postmenopausal women and had fewer than 9% under the age of 56
years. The MICH sample was younger and included 32% of women under the age of

55 years.

Phase 11 women reported a 31% greater prevalence of HRT use at the time of MI than

73

Phase I women and a 20% greater prevalence of HRT discussion. This difference may
be explained by random sampling or differences in recall. Although both groups were
asked similar questions about HRT use at the time of MI, they were queried at
different time points. Phase I women were asked to remember 1 to 3 years in the past
whether they were on HRT at the time of their MI and Phase 11 women were asked
about their HRT use immediately following their index MI. The recency of the event
could explain more women remembering and thereby reporting their prescription use,
or 3) selection factors- - Women responding to the Phase I follow-up survey may be
inherently different than those responding in Phase II who were queried in-hospital.
They are a select group who survived long enough to be contacted 1 to 3 years post
MI. While Phase 11 women were contacted near the time of their MI. Survivorship may
be positively or negatively associated with the use of HRT. The use of HRT may
improve the probability of survival or may be associated with other risk lowering
factors. Alternatively, higher risk women placed on HRT may be less likely to survive
long term. The Phase I prevalence of HRT use is lower than the 22% to 35% of other
studies including women in their 40's.”'33‘47 The Phase II HRT prevalence of 26% is
Similar to other studies including women of Similar age and frequency of hysterectomy.
However, the decreased precision of our estimate

(:5 9.2) prevents accurate conclusions. The differing methods of measurement, time
frame differences in data collection, and small sample size make it difficult to compare

or combine women from Phases I and H.

74

Characteristics

Although it reflects a select group of women willing and/or able to respond to the
surveys, it is interesting to note the descriptive findings of this sample. The current
study was quite heterogenous on SES and demographic factors. Half of respondents
depended on Medicare and a third had private insurance. There were similar
proportions reporting less than a high school degree (29%) and greater than a high
school degree (26%). Similar proportions of older age groups were represented (20%
55-64 y, 27% 65-74 y, 37% 2 75y). The employment status most commonly cited was
retirement (3 7%) and of those not retired, equal proportions were either employed or a
homemaker (25%). A third (3 8%) reported having a hysterectomy. This heterogeneity
has greater representation for the population than other studies confining

generalizability to select groups of a single ethnicity or a single socio-economic status.

Frequency of HRT Discussion

Of concern is the large number of women (54%) reporting having never discussed
HRT with a health care provider. Previous studies suggest that this high percentage of
non discussion is explained by ‘non-opportunity’ or the lack of access to a primary
care or cardiac physician. However, all the women in the current study had
experienced an acute MI and were under the care of a cardiologist at some time during
their hospital stay and 57% queried 2-3 years post discharge remained under the care
of a cardiologist. It may be that at the time of the heart attack, the immediacy of
treatment decisions and the concerns of rehabilitation over shadow the opportunity to

discuss HRT during the hospital stay. However, when analyzing just the Phase I data

75

where women were interviewed one to three years post M1, the prevalence of never

having discussed HRT with a health-care provider was still high (59%).

Although the cardiologist’s role is important in influencing risk reduction following an
M1, their involvement in discussing HRT for heart disease prevention remains low. Of
the 8 (12%) Phase I women who discussed HRT use after the MI event, only 1 woman
reported that a cardiologist discussed the issue with her. A nationwide inquiry of 1268
general practice physicians conducted by The Medical Research Council in the United
Kingdom reported only half of physicians (57%) prescribed estrogen for the prevention
of heart disease. Wk“ 9' Studies have shown that patient willingness to consider
hormone replacement is very much influenced by physician recommendation.”
However, the current study showed few cardiologists involved in the discussion of
HRT benefits and risks. This suggests missed opportunities to improve the health of at-
risk postmenopausal women by way of risk reductions in osteoporosis and/or heart

disease.

Characteristics Distinguishing HRT Users

Although the numbers are small, there are several observations of the current study
consistent with that of the Puget Sound study among women experiencing MI. Users
among the total group of 156 women were more likely to be younger, to have private
insurance, to be employed, and to have had a hysterectomy than were nonusers. The
data on education and history of M1 were limited to the Phase 11 interview. Users were

more likely to have a degree beyond high school and to have a family history of MI.

 

76

These findings were also consistent with the observations from the Puget Sound study.
In addition, both studies found that users were less likely to have a history of MI
(MICH sample) or history of heart failure (Puget sound) suggesting that women with
existing disease are less likely to be placed on HRT following hospitalization for M1.
These findings suggest the possibility of selection factors associated with HRT use

among women experiencing an MI.

Limitations

Our study sample suffered considerably from low numbers. Although over 600 MICH
women were identified prospectively at the time of MI, only 25% (n=156) completed
HRT interviews. Exposure (HRT use at time of MI) was reported by only 35 women.
This small sample size and low frequency of HRT use resulted in a lower precision (:
6.6) than expected with hypothesized numbers. For these reasons, findings on HRT

use may not be generalizable to all women hospitalized with M1.

Conclusions

Use of HRT in elderly women remains especially low (6%-18%). This low prevalence
of use was previously identified with women’s concerns regarding endometrial and
breast cancer, as well as, a commonly held belief by health care providers that HRT
increased the risk of thrombosis. However, studies reported in the last five years have
shown that breast cancer risk is not increased with HRT use of 5 to 10 years and that
the risk of endometrial cancer is attenuated with combination therapy

(estrogen/progestin) for women with intact uteri.“ In addition, animal studies on

77

vasomotor effects and human studies using MRI imaging have shown no increase in

thrombotic events for those exposed to HRT. 52:53

The overwhelming evidence strongly suggests that long-term use of HRT is safe and
impacts favorably on MI risk for peri and postmenopausal women. However,
prevalence of use remains low. The trend appears to be toward an increase in HRT use
among older women, but prevalence remains extremely low. Prescription rates of
HRT for women with menopausal symptoms remained the same between 1989 and
1994 (40%, 38%), while the number of postmenopausal women without symptoms
who reported use of HRT doubled during the same time period (3% to 7%).67 The
current study supports observations that HRT use among older women is low and that

most cardiologists are not discussing HRT risks and benefits.

Strengths and Weakness of Current Database

The opportunity to access an existing database in its initial implementation phase for
hypothesis generation and testing, in principle was a good one. Benefits included the
saving of personal resources in designing and conducting all aspects of a prospective
study. Benefits also included access to expertise in areas of hypothesis generation,
question formulation, instrument construction, interviewer training, and telephone
interviewing. Other benefits included professional contacts with others in the field,
involvement in related grant and investigative projects, and experience at orally

presenting research at a national level forum.

 

 

78

However, problems for the current research were evident throughout the project. Time
delays existed, from incorporating new questions into the existing instrument to
obtaining the data necessary for review. Much of the delays experienced are inherent in
any collaboration where all members are provided opportunity and time to consider,
discuss, and provide input on new questions for research. Because all collaborators
were professionals, many of them clinicians, receiving feedback sometimes took
months. Obtaining consensus on approval from all co-investigators required waiting
on signatures which often took several weeks at each step of the process (added
questions, revised questions, hypothesis, study design, and analysis). These approvals
did not displace institutional review board approval through UCRIHS which together

took many months to obtain full approval.

Because the HRT research was conducted through an existing database that was being
implemented at the time of incorporation of study questions, this investigator
underestimated the time involved for collecting final interviews, conducting chart
abstraction, and entering and cleaning data into the database. At the same time a
process for data access from investigators outside the MICH committee was being
defined. All data requests were to be presented to the principal MSU- MICH
investigators. These requests required a specific hypothesis, background and rationale,
identification of the questions by interview labels (form name, question number, and
test group), operational definition of all key variables requested and the evidence for
those definitions, a description of the analytic approach, and calculation of power for

answering the proposed question. The MICH data supervisor and statistician had

 

79

oversight approval authority for each step of the process. Any deficiencies were
returned to the investigator for corrections. Upon final approval at the MSU site, the
request was then to be sent to the other community MICH investigators for approval.
This process was being developed at the same time that data was being requested for
this descriptive study. Although an efficient, streamlined data request process has since
been worked out, this created additional time delays for requesting and obtaining data

on HRT questions.

The MICH population was ideal for the study question of HRT use among women
experiencing MI. Cases of MI patients were being prospectively recruited and
questioned in-hospital about health and lifestyle behaviors. However, in hindsight,
incorporating new questions for a timely research project into a study that had already

begun, and was in the middle of implementation, was less than ideal.

Approaches to Strengthen Existing Database Use

The benefits of using an existing database for conducting research far outweighs that
of implementing research from scratch when acquisition of data is timely. The MICH
database has observed a decrease in turn around time due to a streamlined data-request
process and the use of electronic transfer of variables to other investigators. Having a
single coordinator responsible for data entry, analysis, and data release is important for
outside investigators to have the data they need and to understand the particulars of
each variable (coding, value labeling, definitions, etc). The value of community

involvement through multi-site approval of new research use and site rotation of the

80

monthly meetings can not be under estimated. The continued use and success of a
collaborative research project depends on such activities because of the volunteer

nature of community investigator time.

Future Directions for Research

The merits of determining HRT prevalence among post menopausal women or among
women with CHD risks is important for understanding trends in heart disease
prevention strategies. Due to the low interview response rate (41 %) and resulting
small sample size of the current study, this descriptive study lacked precision in
estimating HRT prevalence in women suffering from MI. In future research, women
should be interviewed prior to discharge from the hospital and then followed within
one year post-MI to decrease attrition rates. The study would also require following all
discharged subjects for mortality and subsequent medical record data collection for

prescription use and/or proxy interviews.

The following is a list of suggested studies that may build on other descriptive
research;

1) Physician Survey- - assessing the attitudes of cardiologists on the merits of HRT for
older women and those suffering from heart disease. A random selection of
cardiologists practicing throughout the United States could be surveyed on their
reasons for or against counseling women who present with symptoms of heart disease,
2) Community Surveillance - - examining changes in HRT use of women with M1 over

time. Because use patterns are changing rapidly, previously collected data is not

 

 

81

informative, except to document change. Therefore, it is important to continually
update prevalence studies and conduct prospective studies with periodic assessment of
HRT use over time, and 3) Community Survey- -The attitudes of women regarding
HRT utilization and their ability to access a system likely to provide important
information on the risks and benefits of the therapy. Although one community survey
including women up to 65-years found that HRT was not a high priority for
prevention, few women in that study and in others correctly perceived their true risk of
heart disease.” 37' 3‘ Determining the receptiveness of postmenopausal women to HRT

for CHD prevention would help target information for public health campaigns.

10.

LIST OF REFERENCES

Rosenberg HM, Ventura SJ, Maurer JD, et a1. Births and Deaths: United
States, 1995. Monthly vital statistics report; v 45(3):suppl 2. Hyattsville,
Maryland: National Center for Health Statistics. 1996.

Goldberg RJ, Gorak EJ, Yarzebski J, et al. A communitywide perspective of
sex differences and temporal trends in the incidence and survival rates after
acute myocardial infarction and out-of-hospital deaths caused by coronary
heart disease. Circulation 1993;87: 1947-53.

American Heart Association, Heart and Stroke Facts: 1926 Statistical
Supplement. American Heart Association, United States 1995.

Manson JE, Tosteson H, Ridker PM, Satterfield S, et al. The primary
prevention of myocardial infarction. NEJM 1992, ;326(21):1406-16.

Manson JE, Gaziano JM, Ridker PM, et a1. Myocardial infarction:
epidemiologic overview. In Prevention of Myocardial Infarction. Manson JE,
Ridker PM, Gaziano JM, and Hennekens (eds). Oxford University Press, New
York. 1996: 3-31.

Orencia A, Bailey K, Yawn BP, Kottke TE. Effect of gender on long-term
outcome of angina pectoris and myocardial infarction/sudden unexpected
death. JAMA. 1993; 269:2392-2402.

Lerner DL and Kannel WB. Patterns of coronary heart disease morbidity and
mortality in the sexes: A 26-year follow-up of the Framingham population.
Progress in Cardiology. American Heart J 1986; 11 1(2):3 83-90.

Eaker ED, Chesebro JH, Sacks FM, Wenger NK, and Wisnant JP.
Cardiovascular disease in women: special report. Circulation 1993;
88(4): 1999-2006.

Hunink MG, Goldman L, Tosteson NA, Mittlemean MA, Goldman L,
Tosteson NA, Mittleman MA, Goldman PA, Williams LW, Tsevat J, and
Weinstein MC. The recent decline in mortality from coronary heart disease,
1980-1990: the effect of secular trends in risk factors and treatment. JAMA
1997;277:535-42.

US Senate Special Committee on Aging. Aging America: Trends and

82

ll.

12.

13.

14.

15.

16.

l7.

l8.

19.

20.

21.

22.

83

Projections. Washington, DC: US Dept of Health and Human Services; 1988.

Rosenberg L, Palmer JR, and Shapiro S. A case-control study of myocardial
infarction in relation to use of estrogen supplements. Am J Epidemiol
1993;137: 54-63.

Egeland GM, Matthews KA, Kuller LH, et a1. Characteristics of
noncontraceptive hormone users. Preventive Med 1988; l 7:4.3-1 l.

Grodstein F, Manson J E, and Stampfer MJ. Postmenopausal hormone therapy.
In Prevention of Myocardial Infarction. Manson JE, Ridker PM, Gaziano JM,
and Hennekens (eds). Oxford University Press, New York. 1996: 413-430.

Topo P, Klaukka T, Hemminki E, and Uutela A. Use of hormone replacement
therapy in 1976-1989 by 45-65 year old Finnish women. J Epidemology and
Community Health 1991; 45:275-80.

Christiansen C. The different routes of administration and the effect of
hormone replacemnt therapy on osteoporosis. Fertility & Sterility 1994; 64(6)
Sup 2:152-55.

Hunt K, Vessey M, McPherson K. Mortality in a cohort of longterm users of
hormone replacement therapy: an updated analysis. Br J Obstet Gynacol 1990;
97: 1080-89.

Henderson BE, Paganini-Hill A, and Ross RK. Decreased mortality in users of
estrogen therapy. Arch Int Med 1991; 20:47-63.

Udoff L, Langenberg P, and Adashi EY. Combined continuous hormone
replacement therapy: a critical review. Obstet Gynecol 1995; 86:306-16.

Ravnikar VA. Compliance with hormone therapy. Am J Obstet Gynecol 1987;
156:1332-1334.

Thompson W. Estrogen replacement therapy in practice: Trends and issues.
Am J Obstet Gynecol 1995;173:990-993.

Grady D, Rubin SM, Petitti DB, F ox CS, Black D, Ettinger B, Emster VL, and
Cummings SR. Hormone therapy to prevent disease and prolong life in
postmenopausal women. Ann Intern Med 1992; 117: 1016-37.

Pettiti DB. Coronary heart disease and estrogen replacement therapy: can
compliance bias explain the results of observational studies? Ann Epidemiol
1994; 4:115-18.

23.

24.

25.

26.

27.

28.

29.

30.

31.

32.

33.

34.

84

Falkeborn M, Persson I, Adami HO, Bergstrom R, Baker E, Mohsen HLR, and
Naessen T. The risk of acute myocardial infarction after oestrogen and
oestrogen-progestogen replacement. Br J Obstet Gynacol 1992; 99:821-28.

Stumpf PG and Trolice MP. Compliance problems with hormone replacement
therapy. Obstet & Gyn Clinics of NA 1994; 21(2): 219-28.

Hammond CB. Women’s concerns with hormone replacement therapy-
compliance issues. Fertil Steril 1994; 62(Supp 2):157S-l608.

Hemminki E, Malin M, and Topo P. Selection to posmenopausal therapy by
women’s characteristics. J Clin Epidemiol. 1993; 46(3):211-19.

Nebulsi AA, Folsom AR, White A, Patsch W, Heiss G, Wu K, and Szklo M.
Association of horrnone-replacement therapy with various cardiovascular risk
factors in postmenopausal women. NEJM 1993; 328( 1 5):1069-75.

Barrett-Connor E, Wingard DL, and Criqui MH. Postmenopausal estrogen use
and heart disease risk factors in the 1980's. JAMA 1989; 261 :2095-2100.

Cauley JA, Cummings SR, Black DM, Mascioli SR, and Seeley DG.
Prevalence and determinants of estrogen replacement therapy in elderly
women. Am J Obstet Gynecol 1990; 163:1438-1444.

Handa VL, Landerman R, Hanlon JT, Harris T, and Cohen HJ. Do older
women use estrogen replacement? Data from the Duke Established Populations
for Epidemiologic Studies of the Elderly (EPESE). J Am Geriatr Soc 1996;
44:1-6.

The Coronary Drug Project Research Group. Influence of adherence to
treatment and response of cholesterol on mortality in the Coronary Drug
Project. NEJM 1980;303:1038-41.

Derby CA, Hume AL, Barbour MM, et a1. Correlates of postmenopausal
estrogen use and trends through the 1980s in two southeastern New England
communitites. Am J Epidemiol 1993; 47:435-42.

Stampfer MJ, Colditz GA, Willett WC, Manson JE, Rosner B, Speizer, FE,
and Hennekens CH. Postmenopausal estrogen therapy and cardiovascular
disease: ten year follow-up from the Nurses’ Health Study. NEJM 1991;
325:756-62.

Ferguson KJ, Curtis H, and Johnson S. Estrogen replacement therapy; a survey
of women’s knowlege and attitudes. Arch Intern Med 1989; 1492133-36.

35.

36.

37.

38.

39.

40.

41.

42.

43.

44.

45.

46.

85

Harris RB, Laws A, Reddy VM, King A, and Haskell WL. Are women using
postmenopausal estrogens? a community survey. Am J Public Health 1990;
80(10):]266- 68.

Griffiths F. Women’s health concerns: is the promotion of hormone
replacement therapy for prevention important to women? Family Practice
1995; 12(1): 54-59.

Pilote L and Hlatky M. Attitudes of women toward hormone therapy and
prevention of heart disease. Am Heart J 1995; 129: 123 7-8.

MacLennan AH, Taylor AW, and Wilson DH. Changes in the use of hormone
replacement therapy in South Australia. Med J Aust 1995; 1622420-22.

Brett KM and Madans JH. Use of postemnopausal hormone replacement
therapyzestimates from a nationally representative cohort study. Am J
Epidemiol 1997; l45(6):536-45.

Kennedy DL, Baum C, Forbes MB. Noncontraceptive estrogens and
progestins: use patterns over time. Obstet Gynecol 1985;65:441-446.

Postmenopausal Estrogen/Progestin Interventions (PEPI) Trial Writing Group.
Effects of estrogen/progestin regimens on heart disease risk factors in
postmenopausal women. JAMA 1995; 273:199-203.

Vaziri SM, Evans JC, Larson MG, and Wilson PW. The impact of female
hormone usage on the lipid profile: The Framingham Offspring Study. Arch
Intern Med. 1993;153:2200- 06.

Sullivan JM, Zwaag RV, Hughes JP, Maddock V, Kroetz FW, Ramanathan
KB, and Mirvis DM. Estrogen replacement and coronary artery disease: effect
on survival in postmenopausal women. Arch Intern Med. 1990;150:2557-62.

Psaty BM, Heckbert SR, Arkins D, Lemaitre R, Koepsell TD, Wahl PW,
Siscovick DS, and Wagner EH. The risk of myocardial infarction associated

with the combined use of estrogens and progestins in postmenopausal women.
Arch Intern Med. 1994;154:1333-39.

Psaty BM, Heckbert SR, Atkins D, Siscovick DS, Koepsell TD, Wahl PW,
Longstreth WT, Weiss NS, Wagner EH, Prentice R, and Furberg CD. A
review of the association of estrogens and progestins with cardiovascular
disease in postemenopausal women. Arch Intern Med. 1993;153:1421-27.

Kadri AZ. Attitudes to HRT. The Practitioner 1990;234:222-24.

47.

48.

49.

50.

50.

51.

52.

53.

54.

55.

56.

57.

86

Bush TL, Barrett-Connor E, Cowan LD, Criquie MH, Wallace RB, Suchindran
CM, Tyroler HA, and Riflcind BM. Cardiovascular mortality and
noncontraceptive use of estrogen in women: results from the Lipid Research
Clinics program Follow-up Study. Circulation 1987;75(6):1 102-09.

Newton KM, LaCroix AZ, McKnight B, Knopp RH, Siscovick DS, Heckbert
SR, and Wiess NS. Estrogen replacement therapy and prognosis after first
myocardial infarction. Am J Epidemiol 1997; 145:269-77.

Samsioe G. Cardioprotection by estrogenszimplications of observational
studies. Intern] J Fertil & Menopausal Studies 1994;39(Suppl. l):20-27.

Adams MR, Kaplan JR, Manuck SB, et a1. Inhibition of coromary artery
atherosclerosis by 17-[3 estradiol in ovariectomized monkeys. Arteriosclerosis
1990; 10:1051-57.

Hough JL and Zilversmith DB. Effect of beta estradiol on aortic cholesterol
content and metabolism in cholesterol fed rabbits. Atherosclerosis 1986; 6:57-
63.

Jiang C, Sorrel PM, Lindsay DC, Poole-Wilson PA, and collins P.
Endothelium-independent relaxation of rabbit coronary artery by 17-13
oestradiol in vitro. Br J Pharmacol 1991; 104:1033-07.

Jiang C, Sorrel Pm, Poole-Wilson PA, and Collins P. Acute effect of 17-0
estradiol on rabbit coronary artery contractile responses to endothelin- 1. Am J
Physiol 1992; 263:H2715.

Chang W-C, Nakao J, Orimo J, and Muroto SI. Stimulation of prostacyclin
biosynthetic activity by estradiol in rat aortic smooth muscle cells in culture.
Biochem Biophys Acta 1980; 619: 107-18.

Chester AH, J iang C, Borland JA, Yacoub MH, and Collins P. Oestrogen
relaxes human epicardial coronary arteries through non-endothelium-dependent
mechanisms. Cor Artery Dis 1995; 6:417-22.

Mugge A, Matthias R, Matthias B, Kuhn M, and Lichtlen PR. Endothelium
independent relaxation of human coronary arteries by 17-[3 oestradiol in vitro.
Cardiovascular Res 1993; 27: 1939-42.

Walsh BW, Schiff E, Rosner B, Greenberg L, Ravnikar V, and Sacks FM.
Effects of postmenopausal estrogen replacement on the concentrations and
metabolism of plasma lipoproteins. NEJM 1991; 325:1196-204.

Sempos CT, Cleeman JI, Carroll MD, et al. Prevalence of high blood

58.

59.

60.

61.

62.

63.

64.

65.

66.

87

cholesterol among US adults: an update based on guidelines from the Second
Report of the National Cholesterol Education Program Adult Treatment Panel.
JAMA 1993;269:3009-14.

Working Group on Primary Prevention of Hypertension. Report of the
National High Blood Pressure Education Program Working Group on Primary
Prevention of Hypertension. Arch Intern Med 1993;153:186-208.

Cooper RS and Ford E. Comparability of risk factors for coronary heart
disease among Blacks and Whites in the NHANES-I epidemiologic follow-up
study. Ann Epidemiol.1992; 2(5):637-45.

Spangler JG. Primary Care Clinics in Office Practice: women’s health. Primary
Care 1997; 24(1):205

American College of Physicians. Guidelines for counseling postmenopausal
women about preventive hormone therapy. Ann Intern Med 1992; 117:1039-
41.

Michigan Employment Security Commission. Michigan Statistical Abstract-

1996 Edition (Detroit, Michigan: 1996). The University of Michigan Press

(Ann Arbor, MI). Office of the State Registrar and Division of Health
Statistics. Michigan Health Statistics. Michigan Department of Public Health.
Lansing, Michigan 1995.

Ginsburg GS and Douglas PS. Why cardiologists should be interested in
estrogen. Am J Cardiology 1996; 78: 559-61.

Manolio TA, Furberg CD, Shemanski L, Psaty BM, O’Leary DH, Tracy RP,
Bush TL (for the CH8 Collaborative Research Group). Associations of
postmenopausal estrogen use with cardiovascular disease and its risk factors in
older women. Circulation 1993; 88(part I):2163-2171.

Von Muhlen DG, Kritz-Silverstein D, and Barrett-Connor E. A community-
based study of menopause symptoms and estrogen replacement in older
women. Maturitas 1995; 22(2):71-78.

Stafford RS, Saglam D, Causino N, and Blumenthal D. Low rates of hormone
replacement in visits to US primary care physicians. Am J Obstet & Gynecol
1997; 177(2): 381-87.

‘0

‘. I'l‘

0.1!!! 0... i 1.414! t 6.f'."‘-

ll 1 ..I lg'i

I i}.l$tl"lil’l\, ’51,. 1"] III It. |. Fin, I! II’. .31'1.’ I.Il\' It! 1' :I ' - 'P ' "r

-‘I’I'O’I’r‘f’tiih

V. L BRA S

nrcuran Stare UNI
lllllllllllllllllllllllllllllllll
31293017

I

12

RIE
llllll
445

 

ill

.7