University of Southern California Dissertations and Theses

Postmenopausal hormone therapy, risk of heart disease and total mortality among women in the California Teachers Study

(USC Thesis Other)

Postmenopausal hormone therapy, risk of heart disease and total mortality among women in the California Teachers Study

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content

POSTMENOPAUSAL HORMONE THERAPY,
RISK OF HEART DISEASE AND TOTAL MORTALITY AMONG
WOMEN IN THE CALIFORNIA TEACHERS STUDY

by
Yuan Liu

A Thesis Presented to the
FACULTY OF THE GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
MASTER OF SCIENCE
(BIOSTATISTICS)

December 2007

Copyright 2007 Yuan Liu

ii

Dedication

To my parents, Chunhu Liu & Hongyun Zhou,
for their support and understanding

iii

Acknowledgements

This work was done under the direction and supervision of my guidance committee
chair, Dr. Daniel Stram. I would like to give my enormous gratitude and appreciation
to Dr. Stram for his invaluable guidance and support provided throughout the course
of my research and the writing of my thesis. Meanwhile, I would like to express my
appreciation to Dr. Giske Ursin, for her insight and suggestions throughout my
research and in the preparation of this manuscript. I would also like to extend my
gratitude to my academic advisor, Dr. Stanley Azen, for his comprehensive guidance
throughout my graduate study at USC, and for his advice in the writing of my thesis.

I would like to give my most sincere appreciation to my boyfriend, Zhenqi He, for his
understanding and support. Special appreciation also goes to my dear friends, Wen
Mao, Jia Yao, Dawei Jia, and Rui Chang, for their company and support throughout
my life at the United States.

iv

Table of Contents

Dedication ii
Acknowledgements iii
List of Tables v
List of Figures vi
Abstract vii
Introduction 1
Methods 4
Results 9
Discussion 20
Bibliography 28

v

List of Tables

Table 1: Selected Baseline Characteristics in Relation to Age Groups 10
in 46,156 Women Eligible for the Analysis of Total Mortality

Table 2: Selected Baseline Characteristics in Relation to Hormone 12
Therapy in 46,156 Women Eligible for the Analysis of Total Mortality

Table 3: Multivariate Hazard Ratios (HRs) and 95% Confidence 16
Intervals (CIs) for the Association between Hormone Therapy and
Total Mortality by Age at Questionnaire

Table 4: Multivariate Hazard Ratios (HRs) and 95% Confidence 17
Intervals (CIs) for the Association between Hormone Therapy and
Death of CHD by Age at Questionnaire

Table 5: Remaining Effects of Types of Hormone, Duration of Estrogen 19
Use, and Age Starts Hormone Therapy in Relation to Total Mortality

Table 6: Comparison of the Sample Size in the Women’s Health Initiative 23
and the California Teachers Study

vi

List of Figures

Figure 1: Comparison of Hazard Ratios (95% Confidence Intervals) 22
for Total Mortality in the Women’s Health Initiative and the
California Teachers Study

Figure 2: Comparison of Hazard Ratios (95% Confidence Intervals) 23
for Incidence of CHD in the Women’s Health Initiative and Death
from CHD in the California Teachers Study

vii

Abstract

Postmenopausal estrogen use has been strongly and consistently associated with
reduced risk of coronary heart disease (CHD) in multiple observational studies. In
contrast, clinical trials have shown no benefits and even suggested an increased risk of
CHD. We aimed to clarify this relationship in our large cohort study – the California
Teachers Study (CTS).

Among 46,156 eligible women, 5,808 (12.6%) women died overall and 1,054 (2.3%)
women died from heart attack by December 31, 2004. Risk of CHD death and total
mortality was reduced in younger women with current hormone therapy. Compared
with women who never used hormones, the relative risk of total mortality associated
with current hormone therapy in women aged 55-59 was 0.61 (95% confidence
interval=0.48-0.77). The preventive effect of hormone therapy on total mortality
diminished as age increased with a significant trend (p<0.0001). Older hormone users
and never users were at the same risk of all-cause death.

1

Introduction

The cardioprotective effect of postmenopausal estrogen use in women with and
without existing coronary heart disease (CHD) has been recognized in many
observational studies (2). Since CHD is the most common and most deadly disease in
women, any significant reduction in CHD risk would benefit all women and decrease
total mortality. A meta-analysis published in 1998 including data from 25
observational studies reported a 30% lower risk of CHD in estrogen users compared to
nonusers (2). Similar reduction in CHD risk was also reported for users of estrogen
plus progestin in that paper. The findings from observational studies are consistent and
biologically plausible (13, 26). There are several possible mechanisms for the
protective effects of hormone therapy, while the most widely recognized one is the
favorable changes in lipoproteins – LDL and HDL. The Postmenopausal
Estrogen/Progestin Interventions (PEPI) study, a large randomized trial of hormone
effect on heart disease risk factors, found that hormone therapy significantly reduced
LDL cholesterol and increased HDL cholesterol compared to placebo (27).

A 20-45% decreased risk of all-cause mortality in estrogen users compared to
nonusers has also been reported by some observational studies (8, 9, 16). The
reduction of total mortality is largely due to reductions in cardiovascular disease. In
contrast, the overall cancer mortality has been non-significantly decreased in women

2

who are long-term users of hormone therapy (8, 16). Beneficial effects of estrogen use
depend on the duration of hormone therapy. Total mortality decreases with increasing
duration of use, and current users have lower overall death rate than past users (16).

However, clinical trials have shown no benefits and even increased risk of CHD with
hormone use (15, 19, 22, 28). A meta-analysis of 22 small clinical trials was published
in 1997 and reported that the calculated odds ratio for women taking hormones was
1.39 (95% CI=0.48-3.95) compared to women not taking hormones (15). Results from
a secondary prevention trial - the Heart and Estrogen-Progestin Replacement Study
(HERS) showed the lack of significant difference in myocardial infarction and CHD
death between estrogen plus progestin group and placebo group (HR=0.99, 95%
CI=0.80-1.22) (19). The Women’s Health Initiative (WHI), which was a large
randomized clinical trial in postmenopausal women, also reported no protection effect
in either conjugated equine estrogens (CEE) trial or CEE plus medroxyprogesterone
acetate (CEE + MPA) trial (18, 22). The WHI recently found that the relative risk of
CHD in women who used postmenopausal hormone therapy compared to nonusers
was not always the same; rather, the effect of hormone use was modified by women’s
age and years since menopause (24). Women who were randomized to receive
estrogen at age 50-69 tended to have reduced CHD risk compared with increased risk
among women who were older at age 70-79. A similar trend was observed for total
mortality. Moreover, at the youngest age, an apparent but nonsignificant decrease in

3

total mortality was noted in both trials of CEE and CEE+MPA with the relative risk
around 0.70. The protective effect of hormone use on total mortality became
significant in young women (aged 50-59) if we combined data from both trials (CEE
and CEE+MPA) (HR=0.70, 95% CI=0.51-0.96).

Despite potential biases in observational studies which would be expected to enhance
the observed cardioprotective effect, such as “a healthy woman effect” and
compliance bias (2), the reasons for discrepancies about the effect of hormone therapy
are still unclear. We investigated the association between postmenopausal estrogen use
and total mortality, death from CHD in a large prospective cohort study - the
California Teachers Study (CTS). We sought to clarify whether the effect of hormone
therapy on CHD and total mortality depended on age of current use and on age at start
of use, and to directly compare observational study results to that from the WHI.

4

Methods

Study population
A detailed description of California Teachers Study is available elsewhere (3). Briefly,
the California Teachers Study is a prospective cohort study of 133,479 current and
retired female California public school teachers and administrators who participated in
the California State Teachers Retirement System and returned mailed questionnaires
in 1995-1996. The baseline questionnaire addressed a wide variety of issues related to
women’s health including menstrual and reproductive events, use of hormone therapy
and oral contraceptives, family and personal history of diseases, physical activity,
smoking, alcohol use, etc. The study was approved by the Institutional Review Boards
at the University of Southern California, the University of California, Irvine, and the
California State Health Department.

For this analysis, we excluded in sequence women who were ineligible for mortality
follow-up (n=66, 65 had a foreign address and 1 had died before the study started),
who died with unknown cause of death (n=22), who were younger than 55 years or
older than 94 years at baseline (n=73,609), who had incomplete information on use of
hormones (n=1,365). Women with missing data on prior history of heart attack
(n=357), smoking status (n=409), pack-years of smoking (n=1,884), drinking status

5

(n=3,141), physical activity in the last three years (n=439), diet (n=2,621), body mass
index (BMI) or a BMI outside the range of 16 to 55 kg/m2 which was considered
implausible (n=3,410), were also excluded. As a result, data on 46,156 women were
available for the analysis of total mortality.

Follow up and outcomes
Women contributed person-days to the analysis, starting from the date of completion
of the baseline questionnaire, which was distributed by mail in 1995 and returned by
participants in 1995-1996, until the date of death, date of move outside of California
or December 31, 2004, whichever occurred earliest. We censored individuals at the
time of a move out of California. Deaths were identified by annual linkage with the
California state death file and the Social Security Administration death master file and
by relative reports. In 2006, information was supplemented from our annual linkages
with a search of the National Death Index (NDI) of all women from whom we had no
confirmation that they were alive, by either direct contact or linkage follow-up, since
the end of 2003. Causes of death were obtained from the California state death file and
for out-of state deaths from a NDI search of known descendants.

The main outcomes for the analyses were total mortality (defined as deaths of all
causes excluding external ones: e.g. car accidents) and CHD death (defined as deaths
caused by ischemic heart disease, ICD10: I20-I50).

6

Risk factors
All risk factors included in the analyses as potential confounders are based on self-
reported data from the baseline questionnaire.

Personal risk factors. Race was divided into six categories: White; African-American;
Hispanic; Native American; Asian/Pacific Islander; and other/mixed/not specified.
BMI was calculated from height and weight and categorized into five groups: <18
kg/m
2
; 18-<22.5 kg/m
2
; 22.5-<25 kg/m
2
; 25-<30 kg/m
2
; and ≥ 30 kg/m
2
.

Prior history of disease factors. Personal previous heart attack, stroke, cancer (skin
cancer not included), and diabetes were categorized as yes or no.

Lifestyle factors. Smoking status was divided into never, past smokers, and current
smokers. Total pack-years of smoking was analyzed as a continuous variable. Alcohol
consumption was measured in grams per day and women were grouped into never
drinkers; past drinkers; current drinkers < 20 g/d; and current drinkers ≥ 20 g/d. The
physical activity reflected a participant’s strenuous and moderate activities during a
week in the three years preceding the baseline questionnaire and was kept as a
continuous variable as hours/week/year. Daily dietary calories, calories from fat,
calories from protein, and cholesterol were calculated from the self-reported dietary
intake during the past year.

7

Statistical analysis
Multivariable Cox proportional hazards regression models were used to evaluate
associations between use of hormone therapy and total mortality, with age (in days) to
define the time scale (6, 25). Hazard ratios, presented as relative risks, and
corresponding 95% confidence intervals were estimated. Age at questionnaire was
included in all models. The analysis only included women who were 55 years or older
to ensure that they were postmenopausal over all or most of the follow-up period.
Their baseline age were categorized into six groups (55-59, 60-64, 65-69, 70-74, 75-
84, and 85-94). Mean age of each age categories were used as a continuous linear
variable in the models to test for trend. Participants provided information on their
postmenopausal hormone therapy and were categorized into never users, former users,
and current users. We included interaction terms for age at questionnaire and current
or past use of hormones to investigate whether age at questionnaire modified the effect
of hormone use on total mortality and death from CHD.

Models were stratified by race, and adjusted by potential confounders: BMI, smoking,
total pack-years, alcohol consumption, physical activity during past three years,
dietary intake, and prior history of heart attack, stroke, cancer, and diabetes. These
potential confounders were chosen on the basis of previously reported covariates, as
well as variables found to have statistically significant (p ≤0.05) associations with total
mortality in this study.

8

Based on the models mentioned above, effects of types of hormone used, duration of
estrogen use, and age at which hormone therapy was started were estimated as well,
using likelihood ratio tests. Types of hormone used had four categories: none, estrogen
only, estrogen plus progestin, estrogen/estrogen plus progestin/both/mixed. Duration
of estrogen use was self-reported and was investigated as a categorical factor (never
users, ≤ 4 years, 4-7.5 years, 7.5-17 years, and >17 years). Age at start of estrogen use
was self-reported and categorized into 5 groups: never users, <45, 45-54, 55-64, and
≥ 65. We also included interaction terms for age at start of estrogen use and current or
past hormone therapy in the model to investigate the modifier effect of age at start
estrogen use on total mortality.

Similar analyses were performed on the relationship between hormone therapy and
death from CHD.

We used statistical significance level 0.05 for all analyses.

9

Results

Baseline characteristics
Selected risk factors, behavioral and family history characteristics and death rates
stratified by age are presented in Table 1. Death and death due to CHD increased
markedly with increasing age (death: 3.1% in 55-59 years old vs. 63.5% in 85-94;
death due to CHD: 0.3% in 55-59 vs. 18.2% in 85-94). Besides, the percentage of
women who had prior history of heart disease, diabetes, and cancer was higher in
older women. Other risk factors of death such as smoking and alcohol decreased as
age increased. Older women were more likely to be non-smokers and non-drinkers.

Table 2 presents characteristics of risk factors and death rates stratified by hormone
use. During follow-up, deaths were reported in 17.4% of non-hormone users and
18.0% of former users, compared to only 8.2% of current users. Similarly, deaths from
CHD were less in current hormone users (1.3%) than women who never (3.5%) or
previously (3.4%) used hormone therapy. More women who reported current use of
hormone therapy had a normal BMI (18 kg/m
2
- 30 kg/m
2
); fewer were current
smokers; and more were current drinkers than non-users or past hormone users. There
were no significant differences in the physical activity and daily dietary intake among
the three groups. 2.1% of current hormone users had a prior history of heart attack,
compared with 2.9% of non-hormone users and 3.5% of previous hormone users.

10

Table 1. Selected Baseline Characteristics in Relation to Age Groups in 46,156
Women Eligible for the Analysis of Total Mortality
Age Group
55-59*
n=11674
60-64*
n=9782
65-69*
n=9228
70-74*
n=6870
75-84*
n=7255
85-94*
n=1347

Hormone Use and Mortality
Hormone therapy
Never
2219
(19.0)
2142
(21.9)
2528
(27.4)
1974
(28.7)
2310
(31.8)
648
(48.1)
Former 1411 (12.1) 1504 (15.4) 1814 (19.7) 1852 (27.0) 2430 (33.5) 455 (33.8)
Current 8044 (68.9) 6136 (62.7) 4886 (52.9) 3044 (44.3) 2515 (34.7) 244 (18.1)

Death
No
11307
(96.9)
9247
(94.5)
8403
(91.1)
5855
(85.2)
5044
(69.5)
492
(36.5)
Yes 367 (3.1) 535 (5.5) 825 (8.9) 1015 (14.8) 2211 (30.5) 855 (63.5)

CHD death
No
11641
(99.7)
9732
(99.5)
9137
(99.0)
6708
(97.6)
6782
(93.5)
1102
(81.8)
Yes 33 (0.3) 50 (0.5) 91 (1.0) 162 (2.4) 473 (6.5) 245 (18.2)

Risk Factors
BMI†
<18 121 (1.0) 107 (1.1) 108 (1.2) 99 (2.1) 151 (2.1) 53 (3.9)
18-22.5 3627 (31.1) 2782 (28.4) 2583 (28.0) 1923 (28.0) 2260 (31.2) 555 (41.2)
22.5-25 2765 (23.7) 2320 (23.7) 2279 (24.7) 1779 (25.9) 1897 (26.2) 357 (26.5)
25-30 3236 (27.7) 2941 (30.1) 2834 (30.7) 2117 (30.8) 2175 (30.0) 305 (22.6)
>30 1925 (16.5) 1632 (16.7) 1424 (15.4) 952 (13.9) 772 (10.6) 77 (5.7)

Smoking
Never 6773 (58.0) 5542 (56.7) 5289 (57.3) 4213 (61.3) 4761 (65.6) 1024 (76.0)
Former 4120 (35.3) 3574 (36.5) 3365 (36.5) 2286 (33.3) 2203 (30.4) 298 (22.1)
Current 781 (6.7) 666 (6.8) 574 (6.2) 371 (5.4) 291 (4.0) 25 (1.9)

Alcohol
Never 1897 (16.2) 1737 (17.8) 1835 (19.9) 1560 (22.7) 1989 (27.4) 483 (35.9)
Former 1554 (13.3) 1255 (12.8) 1262 (13.7) 922 (13.4) 1141 (15.7) 247 (18.3)
Current,
<20 g/d
7031 (60.2) 5573 (57.0) 5052 (54.8) 3715 (54.1) 3536 (48.7) 544 (40.4)
Current,
≥20 g/d
1192 (10.2) 1217 (12.4) 1079 (11.7) 673 (9.8) 589 (8.1) 73 (5.4)

11

Table 1. Continued
Age Group
55-59*
n=11674
60-64*
n=9782
65-69*
n=9228
70-74*
n=6870
75-84*
n=7255
85-94*
n=1347

Medical History
Prior heart attack
No
11591
(99.3)
9646
(98.6)
8996
(97.5)
6615
(96.3)
6871
(94.7)
1241
(92.1)
Yes 83 (0.7) 136 (1.4) 232 (2.5) 255 (3.7) 384 (5.3) 106 (7.9)

Prior stroke
No 11584(99.2) 9659 (98.7) 9090 (98.5) 6688 (97.4) 6947 (95.8) 1258 (93.4)
Yes 90 (0.8) 123 (1.3) 138 (1.5) 182 (2.6) 308 (4.2) 89 (6.6)

Prior cancer
No 10418(89.2) 8477 (86.7) 7755 (84.0) 5534 (80.6) 5533 (76.3) 1015(75.4)
Yes 1256 (10.8) 1305 (13.3) 1473 (16.0) 1336 (19.4) 1722 (23.7) 332 (24.6)

Prior diabetes
No 11334 (97.1) 9399 (96.1) 8792 (95.3) 6501 (94.6) 6862 (94.6) 1293(96.0)
Yes 340 (2.9) 383 (3.9) 436 (4.7) 369 (5.4) 393 (5.4) 54 (4.0)

Behavioral Factors
Physical
activity‡,
hrs/wk/yr
3.5 ± 4.0 4.1 ± 4.4 4.2 ± 4.4 3.9 ± 4.4 3.1 ± 4.0 1.7 ± 3.1

Dietary
calories, kcal
1574.4 ±
540.0
1475.5 ±
499.0
1457.7 ±
473.9
1438.9 ±
475.0
1429.0 ±
482.4
1428.3 ±
496.6

Dietary
calories from
fat, kcal/d
496.2 ±
238.4
465.6 ±
222.9
455.2 ±
212.5
449.9 ±
217.3
453.2 ±
216.4
479.3 ±
221.8

Dietary
calories from
protein,
kcal/d
247.4 ±
94.0
233.5 ±
89.1
234.7 ±
87.7
233.6 ±
88.6
231.9 ±
90.5
230.5 ± 94.0

Daily dietary
cholesterol,
mg
193.7 ±
108.0
184.5 ±
106.1
183.3 ±
106.6
181.8 ±
106.5
184.1 ±
108.1
197.0 ±
118.1
*Data presented as No. (%) for categorical variables, or mean ± SD for continuous variables.
†BMI: body mass index, calculated as weight in kilograms divided by height in meters squared.
‡Physical activity: combination of strenuous and moderate activities in past three years.
Note: All comparisons across age groups were significant (p<0.0001).

12

Table 2. Selected Baseline Characteristics in Relation to Hormone Therapy in 46,156
Women Eligible for the Analysis of Total Mortality
Hormone Therapy
Never*
n=11821
Former*
n=9466
Current*
n=24869

Freq. (%) Freq. (%) p-value
a
Freq. (%) p-value
a

Mortality
Death
No 9760 (82.6) 7765 (82.0) 22823 (91.8)
Yes 2061 (17.4) 1701 (18.0)
0.31
2046 (8.2)
<.0001

Death of CHD
No 11410 (96.5) 9147 (96.6) 24545 (98.7)
Yes 411 (3.5) 319 (3.4)
0.67
324 (1.3)
<.0001

Risk Factors
BMI†
<18 199 (1.7) 137 (1.4) 303 (1.2)
18-22.5 3239 (27.4) 2652 (28.0) 7839 (31.5)
22.5-25 2695 (22.8) 2242 (23.7) 6460 (26.0)
25-30 3526 (29.8) 2970 (31.4) 7112 (28.6)
>30 2162 (18.3) 1465 (15.5)
<.0001
3155 (12.7)
<.0001

Smoking
Never 7445 (63.0) 5627 (59.4) 14530 (58.4)
Former 3533 (29.9) 3292 (34.8) 9021 (36.3)
Current 843 (7.1) 547 (5.8)
<.0001
1318 (5.3)
<.0001

Alcohol
Never 3021 (25.6) 2139 (22.6) 4341 (17.5)
Former 1766 (14.9) 1433 (15.1) 3182 (12.8)
Current, <20 g/d 5911 (50.0) 4918 (52.0) 14622 (58.8)
Current, ≥20 g/d 1123 (9.5) 976 (10.3)
<.0001
2724 (10.9)
<.0001

Medical History
Prior heart attack
No 11476 (97.1) 9138 (96.5) 24346 (97.9)
Yes 345 (2.9) 328 (3.5)
0.02
523 (2.1)
<.0001

Prior stroke
No 11555 (97.7) 9197 (97.2) 24474 (98.4)
Yes 266 (2.3) 269 (2.8)
0.006
395 (1.6)
<.0001

Prior cancer
No 9475 (80.2) 6667 (70.4) 22590 (90.8)
Yes 2346 (19.8) 2799 (29.6)
<.0001
2279 (9.2)
<.0001

13

Table 2. Continued
Hormone Therapy
Never*
n=11821
Former*
n=9466
Current*
n=24869

Mean ± SD p-value
a
Mean ± SD p-value
a

Prior diabetes
No 11128 (94.1) 8989 (95.0) 24064 (96.8)
Yes 693 (5.9) 477 (5.0)
0.009
805 (3.2)
<.0001

Behavioral Factors
Physical Activity‡,
hrs/wk/yr
3.6 ± 4.3 3.5 ± 4.1 0.35 3.8 ± 4.2 <.0001

Daily dietary calories,
kcal
1469.6 ± 520.4 1465.1 ± 498.1 0.52 1495.8 ± 493.8 <.0001

Dietary calories from
fat, kcal/d
469.7 ± 232.9 459.4 ± 221.2 0.001 469.3 ± 220.3 0.87

Dietary calories from
protein, kcal/d
235.9 ± 94.3 235.3 ± 91.2 0.64 238.1 ± 88.6 0.03

Dietary cholesterol,
mg/d
189.9 ± 114.6 185.4 ± 109.4 0.003 185.3 ± 103.2 0.002
*Data presented as No. (%) for categorical variables, or mean ± SD for continuous variables.
a
p-value: compare former user vs. non-users and current uses vs. non-users, calculated using chi-square
for categorical variables, t-test for continuous variables.
†BMI: body mass index, calculated as weight in kilograms divided by height in meters squared.
‡Physical activity: combination of strenuous and moderate activities in past three years

14

Overall effects of hormone therapy
Consistent with previous observational studies, hormone therapy reduced overall risk
of total mortality and CHD death (Table 3, 4). Women who currently took hormone
therapy were 0.82 times as likely to die as women who never used hormone (95%
CI=0.77-0.87) in the Cox regression analyses adjusted for the confounders described
above (Table 3). There was also a significant protective effect of previous hormone
therapy (HR=0.91, 95% CI=0.85-0.97), although the effect was not as strong as
current therapy. However, while current therapy decreased risk of CHD by 19% (95%
CI=6%-31%), the association between previous therapy and risk of CHD death was
not significant (Table 4).

Effects of hormone therapy by age
More detailed results were presented in Table 3 and 4 as we analyzed the association
between death and hormone therapy by age. Age was an effect modifier in their
relationship. For current hormone therapy, HRs for total mortality were 0.61, 0.60,
0.77, 0.85, 0.92, 0.97 across six age groups, which suggested that the effect of
hormone therapy was diminished with increasing age. Test for interaction between age
at questionnaire and current use were statistically significant (p=0.0002) and the test
for trend in the RR estimate across age groups was also strongly statistically
significant (p<0.0001). Furthermore, the additional effect of current hormone therapy
was only statistically significant in some age groups. Among younger women, current

15

postmenopausal hormone therapy was significantly protective. However, for women
who were older than 75 and still took hormone therapy, the risk of death was not
significantly reduced (75-84: 0.92(0.83-1.02); 85-94: 0.97(0.80-1.17)). In contrast to
the overall HR for previous hormone therapy, none of the age groups (except 75-84)
showed a statistically significant protective effect among women who received
previous hormone therapy. Besides, the trend of past hormone use across age groups
was not significant either.

For risk of CHD death, the same pattern was seen with a test for trend in the RRs of
current users with increasing age significant (p=0.005). However, the protective
effects of current hormone therapy were only significant in women aged 55-59
(HR=0.36, 95% CI=0.17-0.74). Moreover, in older women (aged 85-94), the effects of
hormone therapy were not protective any more (HR=1.01, 95% CI=0.71-1.43). Past
use of hormone had no significant effects on death due to CHD in women of all age
groups.

16

Table 3. Multivariate Hazard Ratios (HRs) and 95% Confidence Intervals (CIs) for
the Association between Hormone Therapy and Total Mortality by Age at
Questionnaire
Hormone Therapy
Never Former Current

No. of deaths No. of deaths HR (95% CI)* No. of deaths HR (95% CI)*

Overall† 2061 1721
0.91
(0.85-0.97)
2046
0.82
(0.77-0.87)

Age Group
55-59 99 64 0.96
(0.70-1.32)
204 0.61
(0.48-0.77)
60-64 164 115 0.98
(0.78-1.25)
256 0.60
(0.49-0.74)
65-69 273 187 0.93
(0.77-1.12)
365 0.77
(0.66-0.91)
70-74 322 306 0.96
(0.82-1.12)
387 0.85
(0.73-0.99)
75-84 780 742 0.88
(0.79-0.97)
689 0.92
(0.83-1.02)
85-94 423 287 0.91
(0.79-1.06)
145 0.97
(0.80-1.17)

P for trend‡ 0.46 <0.0001
*Cox regression models included age at questionnaire, status of hormone therapy and their interaction,
stratified by race, adjusted for BMI, smoking, total pack-years, alcohol consumption, physical activity
during past three years, dietary intake, and prior history of heart attack, stroke, cancer, and diabetes.
† Overall estimates using models mentioned above without interaction between age and hormone
therapy.
‡ Test for trend (interaction) using mean age of each group as continuous variable. Cox regression
models stratified by race, and adjusted for all potential confounders.

17

Table 4. Multivariate Hazard Ratios (HRs) and 95% Confidence Intervals (CIs) for
the Association between Hormone Therapy and Death of CHD by Age at
Questionnaire
Hormone Therapy
Never Former Current

No. of CHD
deaths
No. of CHD
deaths
HR (95% CI) No. of CHD
deaths
HR (95% CI)

Overall† 411 319
0.88
(0.76-1.02)
324
0.81
(0.69-0.94)

Age Group
55-59 13 4 0.48
(0.16-1.47)
16 0.36
(0.17-0.74)
60-64 17 8 0.72
(0.31-1.66)
25 0.58
(0.31-1.08)
65-69 32 23 1.04
(0.61-1.78)
36 0.66
(0.41-1.06)
70-74 56 49 0.92
(0.62-1.35)
57 0.72
(0.50-1.04)
75-84 166 160 0.93
(0.75-1.15)
147 0.92
(0.74-1.15)
85-94 127 75 0.82
(0.61-1.09)
43 1.01
(0.71-1.43)

P for trend‡ 0.98 0.005
*Cox regression models included age at questionnaire, status of hormone therapy and their interaction,
stratified by race, adjusted for BMI, smoking, total pack-years, alcohol consumption, physical activity
during past three years, dietary intake, and prior history of heart attack, stroke, cancer, and diabetes.
† Overall estimates using models mentioned above without interaction between age and hormone
therapy.
‡ Test for trend (interaction) using mean age of each group as continuous variable. Cox regression
models stratified by race, and adjusted for all potential confounders.

Additional analyses
Other factors of hormone use which may affect the relationship with total mortality
such as types of hormone used, duration of therapy, and age at start of hormone
therapy were assessed in the model, respectively. However, none of these factors were

18

significant (p>0.05), based on the model which already included current /previous
hormone therapy and its interaction with age (Table 5). The effects of estrogen and
estrogen plus progestin were not significantly different in the reduction of risk of total
mortality. Different from age at questionnaire, age at start of hormone therapy did not
seem to be important for the effect of hormone on death. Even women who started at
their late 60’s were not at significant higher risk than women who started at younger
age (<45). In addition, there was no statistically significant modifier effect of age at
start of estrogen use on total mortality (p=0.65).

19

Table 5. Remaining Effects of Types of Hormone, Duration of Estrogen Use, and Age
Starts Hormone Therapy in Relation to Total Mortality

Frequency (%)
Remaining effect
HR (95% CI)*
p-value†
Types of hormone 0.12
No 12431 (26.9) 1.0
E 15361 (33.3) 0.93
(0.81-1.08)

E+P 11572 (25.1) 0.98
(0.83-1.16)

E/E+P/both/mixed 6792 (14.7) 0.86
(0.73-1.02)

Duration of estrogen use 0.65
Never 12431 (27.9) 1.0
≤ 4 yrs 11007 (24.7) 0.90
(0.78-1.05)

4-7.5 yrs 5396 (12.1) 0.95
(0.80-1.13)

7.5-17 yrs 9332 (21.0) 0.93
(0.79-1.09)

>17 yrs 6385 (14.3) 0.96
(0.81-1.16)

Age starts hormone therapy 0.61
Never 12431 (27.1) 1.0
<45 6521 (14.2) 0.96
(0.82-1.13)

45-54 18137 (39.6) 0.91
(0.78-1.06)

55-64 7339 (16.0) 0.92
(0.78-1.08)

≥ 65 1436 (3.1) 0.96
(0.80-1.15)

Abbreviations: E, estrogen; P, progestin; HR, hazard ratio; CI, confidence interval.
*Hazard ratios for types of hormone, duration of estrogen use, and age starts hormone therapy,
respectively, adding to Cox regression models with age at questionnaire, hormone therapy status, and
their interaction, stratified by race, adjusted for BMI, smoking, total pack-years, alcohol consumption,
physical activity during past three years, dietary intake, and prior history of heart attack, stroke, cancer,
and diabetes
†p-value obtained from likelihood ratio tests comparing models with and without the factor.

20

Discussion

Our analyses suggest that the protective effect of hormone therapy is most strongly
related to current use rather than either duration of use or past use. Postmenopausal
women benefit from current hormone therapy for the total mortality and death caused
by CHD. However, there was no statistically significant effect of previous use of
hormone on mortality. In other words, the effect of hormone does not last for a long
time. This is similar to the results from the Nurse Health study which included about
18 years of observation of postmenopausal women (12). In that study, current use
decreased overall risk of total mortality about 25%, compared with no effect of past
use.

Even in women who currently receive hormone therapy, the protective effect differs
among various age groups (p=0.0002 for interaction between age and current use).
The relation between hormone therapy and risk of death is modified by baseline age.
There is a significant trend of the protective effect with increasing age. Death is
significantly decreased by 25%-40% in younger women. Although older women
(women aged ≥ 75) who take hormone therapy are not at higher risk of death
compared to never users, there is no significant protection by the therapy either.

21

For death caused by CHD, we observed the same pattern as for total mortality: hazard
ratios increased with increasing age (p-value for interaction: 0.08). Even though
overall hazard ratio seems to be reduced significantly by hormone therapy, the effect
does not exist after we conducted the analyses within different age groups.

Although our results are consistent with previous observational studies in that the
overall risk of death or CHD is decreased by postmenopausal hormone therapy, results
are not supported by clinical trials; we obtained results comparable to that from the
WHI after we analyzed the hormone effect in different age groups. Age seems to be an
effect modifier in the relation between hormone therapy and mortality (Figure 1, 2).
The WHI suggested that total mortality was significantly reduced in women aged 50-
59 years, which is also supported by our results. Thus, the protective effect of
hormone therapy on risk of death is probably true for younger women. In addition,
there is a similar trend of hazard ratios with increasing age in both studies. However,
our study is inconsistent with the WHI in the hormone effect in older women.
Although not significant, the WHI reported a non-significant increased risk of total
mortality in older women, compared with decreased non-significant risk in our results.
This discrepancy may be explained by the power of these two studies. Comparison of
the number of participant in each age group was summarized in Table 6. The sample
size of hormone users is larger than the WHI in all age groups and follow-up is longer
in the California Teachers Study (5.2 years for the WHI vs. 8 years for the CTS).

22

There are also many other possible reasons for the discrepancies in the effect on older
women. For example, since the HRs for older women were not significant in either of
these two studies, inclusion of some potential confounders may change the HRs as
well. There may be simply no effect of hormone therapy, no matter protective or
harmful, on women at older age.

Figure 1. Comparison of Hazard Ratios (95% Confidence Intervals) for Total
Mortality in the Women’s Health Initiative and the California Teachers Study
Comparison of total mortality
0.4
0.7
1
1.3
45 55 65 75 85 95
CTS: age at questionnaire
WHI: age at randomization
Hazard ratio
WHI
Teachers

23

Figure 2. Comparison of Hazard Ratios (95% Confidence Intervals) for Incidence of
CHD in the Women’s Health Initiative and Death from CHD in the California
Teachers Study
Comparison of CHD
0
0.5
1
1.5
45 55 65 75 85 95
CTS: age at questionnaire
WHI: age at randomization
Hazard ratio
WHI
Teachers

Table 6. Comparison of the Sample Size in the Women’s Health Initiative and the
California Teachers Study
50-59* 60-69* 70-79*

Hormone†
Non-
Hormone‡
Hormone
Non-
Hormone
Hormone
Non-
Hormone
Total No.
WHI 4476 4356 6240 6122 3100 3053
Teachers 8044 2219 11022 4670 4825 3386
No. of Deaths
WHI 69 95 240 225 237 208
Teachers 204 99 621 437 786 711
*The same age categories as the WHI, for California Teachers Study, 50-59: 55-59 age at questionnaire;
60-69: combination of 60-64 and 65-69; 70-79: combination of 70-74 and 75-84, excluding women >79
years.
†Hormone group: hormone therapy group in the WHI; current hormone users in the California Teachers
Study.
‡Non-Hormone group: placebo group in the WHI; never users in the California Teachers Study.

24

There are two ages which are important in the hormone therapy: the age that a woman
begins to use hormone and the current age. In the WHI, age at randomization was not
only women’s current age, but also the age they began to receive therapy in the trial if
they were in the CEE or CEE+MPA group. That is, these two ages were the same in
the WHI analyses. Therefore, it was impossible to distinguish the modifier effect of
current age from the age that a woman begins therapy. In contrast, our analyses
separated these two ages. Age at questionnaire was considered as the modifier in the
relation between hormone therapy and total mortality; then age that a woman begins to
use estrogen was added to the model, but its influence to the protective effect was not
significant. Moreover, our results also suggested that age that a woman begins to use
estrogen did not function as a modifier as age at questionnaire. As a result, we believe
that it is the current age of taking hormone, rather than the age that a woman begins
hormone therapy that matters and modify the effect on total mortality.

Years since menopause and hormone therapy was suggested as an important modifier
of the hormone effect by some observational studies (7, 21), as well as the WHI.
Unfortunately, we cannot construct this variable well enough to conduct statistical
analysis.

There are several potential sources of bias that might account for the protective effect
seen in the observational studies: selection bias, compliance bias, and follow-up

25

bias(2). Our study is also subject to these biases, but their influence on our results is
decreased by the study design and statistical analysis to a certain extent. “A healthy
woman effect” was pointed out by Elizabeth Barrett-Connor: women taking hormones
are more educated and of higher social class, more compliant, healthier, and have
more favorable lifestyle than nonusers (1). Education and social class are strongly and
inversely associated with the risk of CHD (20). Since participants in our study were all
public school teachers or administrators in California, our sample is more
homogeneous in some aspects: they all have a minimum level of education and
maintain a more similar socioeconomic status than women of general population.
There is therefore less possibility that education and social class would be potential
confounders in the relation of hormone therapy and risk of death from CHD.

It has been reported that most women prescribed postmenopausal hormones are not
compliant (14). Compliance plays an important role in the relation between hormone
therapy and risk of death. Even in the placebo group, good compliance has been
shown to reduce the risk of CHD 40%-60% in randomized clinical trials (4, 10, 17). In
observational studies, compliance level in hormone users and nonusers are not the
same: women taking hormones tend to be more compliant than women not taking
hormones. Thus, compliance bias could attribute to the benefit of postmenopausal
hormone therapy suggested by observational studies. However, women in our study
are more homogeneous than many other observational studies, and therefore, the

26

compliance level for hormone users and nonusers may be comparable to each other. In
addition, our overall reduction in the risk of total mortality and CHD death is about
20%, which is much less than the 50% attributed to such compliance bias. The fact
that our results about the relation between hormone and total mortality in younger
women are consistent with the WHI, which has less compliance bias, also indicates
that compliance bias is well controlled in our study and should not affect the
effectiveness of our findings.

The baseline characteristics in the CTS suggested that current hormone users were
healthier and had more favorable lifestyles. More current users were in the normal
range of BMI (18 kg/m
2
-30kg/m
2
), and fewer of them had prior history of heart attack,
stroke, cancer, and diabetes. Women taking hormones were less likely to be current
smokers, although many are ex-smokers. A majority of women who were hormone
users were also current drinkers, and drinking was found to be protective in many
mortality studies (5, 11, 23). Although it seemed that there were no differences in the
daily dietary calories and fat intake among groups, the average level of physical
activity was higher in hormone users than nonusers and past users. Since these
elements may be risk factors of death and CHD, we adjusted for all of them in our
statistical models. There may be some residual confounding effects of these factors,
but we think it is unlikely that our main results would be altered by their influence.

27

In summary, our study suggested that the health consequences of hormone therapy
may vary by current age of taking hormone. Younger women reduced their risk of
total mortality by hormone use, with a significant trend of relative risks as age
increases. Only current users, rather than past users of hormone therapy would benefit
from its protective effect. Our results clarified some of the discrepancies between
previous observational studies and clinical trials, and distinguished the modifier effect
of current age from the effect of the age that a woman starts therapy.

28

Bibliography

1. Barrett-Connor E. Postmenopausal estrogen and prevention bias. Ann. Intern. Med.
1991;115:455-456

2. Barrett-Connor E, Grady D. Hormone replacement therapy, heart disease, and
other considerations. Annu Rev Public Health. 1998;19:55-72

3. Bernstein L, Allen M, Anton-Culver H, et al. High breast cancer incidence rates
among California teachers: results from the California Teachers Study. Cancer
Causes Control. 2002;13:625-635

4. Coronary Drug Project Research Group. Influence of adherence to treatment and
response of cholesterol on mortality in the Coronary Drug Project. N. Engl. J. Med.
1980;303:1038-1041

5. Corrao, G., Rubbiati, L., Bagnardi, V., Zambon, A., & Poikolainen, K. Alcohol
and coronary heart disease: A meta-analysis. Addiction. 2000;95:1505–1523.

6. Cox D. Regression models and life tables. J R Stat Soc Ser B. 1972;34:187-220

7. Dwyer KM, Nordstrom CK, Bairy Merz CN, Dwyer JH. Carotid wall thickness
and years since bilateral oophorectomy. Am J Epidemiol. 2002;156:438-444

8. Ettinger B, Friedman GD, Bush T, Quesenberry CP Jr. Reduced mortality
associated with long-term postmenopausal estrogen therapy. Obstet. Gynecol.
1996; 87:6-12

9. Folsom AR, Mink PJ. Sellers TA, Hong C-P, Zheng W, et al. Hormonal
replacement therapy and morbidity and mortality in a prospective study of
postmenopausal women. Am. J. Public. Health. 1995;85:1128-1132

10. Gallagher EJ, Viscoli CM, Horwitz RI. The relationship of treatment adherence to
the risk of death after myocardial infarction in women. JAMA. 1993;270:742-744

11. Gmel, G., Gutjahr, E., & Rehm, J. How stable is the risk curve between alcohol
and all-cause mortality and what factors influence the shape? A precision-
weighted hierarchical meta-analysis. European Journal of Epidemiology.
2003;18:631–642.

29

12. Grodstein F, Stampfer MJ, Colditz GA, Willett WC, Manson JE, et al.
Postmenopausal hormone therapy and mortality. N. Eng. J. Med. 1997;336:1769-
1775

13. Guetta V, Cannon RO III. Cardiovascular effects of estrogen and lipid-lowering
therapies in postmenopausal women. Circulation. 1996;93:1928-1937

14. Hemminki E, Brambilla DJ, McKinlay SM, Posner JG. Use of estrogens among
middle-aged Massachusetts women. Ann. Pharmacother. (DICP) 1991;25:418-423

15. Hemminki E, McPherson K. Impact of postmenopausal hormone therapy on
cardiovascular events and cancer: pooled data from clinical trials. Br. Med. J.
1997;315:149-153

16. Henderson BE, Paganini-Hill A, Ross RK. Decreased mortality in users of
estrogen replacement therapy. Arch. Intern. Med. 1991;151:75-78

17. Horwitz RI, Viscoli CM, Berkman L, Donaldson RM, Horwitz SM, et al.
Treatment adherence and risk of death after a myocardial infarction. Lancet.
1990;336:542-545

18. Hsia J, Langer D, Manson JE, et al. Conjugated equine estrogens and the risk of
coronary heart disease. Arch Intern Med. 2006;166:357-365

19. Hulley S, Grady D, Bush T, et al. Randomized trial of estrogen plus progestin of
secondary prevention of coronary heart disease in postmenopausal women. JAMA.
1998;280:605-613

20. Iribarren C, Luepker RV, McGovern PG, Arnett DK, et al. Twelve-year trends in
cardiovascular disease risk factors in the Minnesota heart survey: Are
socioeconomic differences widening? Arch. Intern. Med. 1997;157:873-881

21. Mack WJ, Slater CC, Xiang M, Shoupe D, Lobo RA, Hodis HN. Elevated
subclinical atherosclerosis associated with oophorectomy is related to time since
menopause rather than type of menopause. Fertil Steril. 2004;82:391-397

22. Manson JE, Hsia J, Johnson KC, et al; Women’s Health Initiative Investigators.
Estrogen plus progestin and risk of coronary heart disease. N Engl J Med. 2003;
349:523-534

30

23. Rehm, J., Gutjahr, E., & Gmel, G. Alcohol and all-cause mortality: A pooled
analysis. Contemporary Drug Problems. 2001;28:337–361.

24. Rossouw JE, Prentice RL, Manson JE, et al. Postmenopausal hormone therapy and
risk of cardiovascular disease by age and years since menopause. JAMA.
2007;297(13):1465-1476

25. Thiebaut AC, Benichou J. Choice of time-scale in Cox’s model analysis of
epidemiological cohort data: a simulation study. Stat Med. 2004;23:3803-3820

26. Tikkanen MJ. Mechanisms of cardiovascular protection by postmenopausal
hormone replacement therapy. Cardiovasc. Risk Factors. 1993;3:138-143

27. Writing group for the PEPI Trial. Effects of estrogen or estrogen/progestin
regimens on heart disease risk factors in postmenopausal women. JAMA.
1995;273:199-208

28. Writing Group for the Women’s Health Initiative Investigators. Risks and benefits
of estrogen plus progestin in healthy postmenopausal women. JAMA.
2002;288:321-333

Abstract (if available)

Abstract Postmenopausal estrogen use has been strongly and consistently associated with reduced risk of coronary heart disease (CHD) in multiple observational studies. In contrast, clinical trials have shown no benefits and even suggested an increased risk of CHD. We aimed to clarify this relationship in our large cohort study -- the California Teachers Study (CTS).

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Hormone therapy timing hypothesis and atherosclerosis

PDF

Physical activity and sex hormone levels in postmenopausal women

PDF

The effects of hormone therapy on carotid artery intima-media thickness and serum lipids by ApoE4 genotype

PDF

Sex hormones and atherosclerosis in postmenopausal women

PDF

A randomized controlled trial of hormone therapy on postmenopausal women’s quality of life

PDF

Effect of soy isoflavones on anthropometric and metabolic measurements in postmenopausal women

PDF

Hormonal and genetic risk factors of endometrial cancer and trends in incidence and survival of adult acute lymphoblastic leukemia

PDF

Effect of estradiol on circulating levels of inflammatory cytokines in postmenopausal women

PDF

Effects of post-menopausal hormone therapy on arterial stiffness in the ELITE trial

PDF

Association between cardiovascular risk factors and coronary CT measures of coronary atherosclerosis in healthy postmenopausal women

PDF

The relationship of resistin, leptin, adiponectin, and ghrelin with bone mineral density in healthy postmenopausal women: longitudinal analysis

PDF

Bone mineral density is associated with carotid atherosclerosis in healthy postmenopausal women: a longitudinal analysis of randomized clinical trial data

PDF

Effects of testosterone and growth hormone supplementation therapies on quality of life in older men: exploratory findings from the HORMA study

PDF

Personal hair dye use and risk of B-cell non-Hodgkin’s lymphomas among adult women in Los Angeles County

PDF

Association between endogenous sex hormone levels and cognition: the Women’s Isoflavone Soy Health (WISH) trial

PDF

Applications of multiple imputations in survival analysis

PDF

Obesity paradox in acute heart failure decompensation

PDF

A cross-sectional examination: impact of diet and physical activity on plasma glucose metabolism, insulin sensitivity and pancreatic β-cell function in Hispanic women with histories of gestationa...

PDF

The role of genetic ancestry in estimation of the risk of age-related degeneration (AMD) in the Los Angeles Latino population

PDF

Exploring the relationship between menopausal hot flushes and Alzheimer's disease biomarkers: a cross-sectional analysis in postmenopausal women

Asset Metadata

Creator Liu, Yuan (author)

Core Title Postmenopausal hormone therapy, risk of heart disease and total mortality among women in the California Teachers Study

School Keck School of Medicine

Degree Master of Science

Degree Program Biostatistics

Publication Date 10/24/2009

Defense Date 10/22/2007

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

Tag CHD,hormone therapy,mortality,OAI-PMH Harvest

Place Name California (states), USA (countries)

Language English

Advisor Stram, Daniel O. (committee chair), Azen, Stanley Paul (committee member), Ursin, Giske (committee member)

Creator Email yuanliu@usc.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-m884

Unique identifier UC1331906

Identifier etd-Liu-20071024 (filename),usctheses-m40 (legacy collection record id),usctheses-c127-556839 (legacy record id),usctheses-m884 (legacy record id)

Legacy Identifier etd-Liu-20071024.pdf

Dmrecord 556839

Document Type Thesis

Rights Liu, Yuan

Type texts

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

Repository Name Libraries, University of Southern California

Repository Location Los Angeles, California

Repository Email cisadmin@lib.usc.edu