Shlipak MG, Simon JA, Vittinghoff E, Lin F, Barrett-Connor E, Knopp RH, Levy RI, Hulley SB. Estrogen and Progestin, Lipoprotein(a), and the Risk of Recurrent Coronary Heart Disease Events After Menopause. JAMA. 2000;283(14):1845-1852. doi:10.1001/jama.283.14.1845
Author Affiliations: Departments of Medicine (Drs Shlipak and Simon), and Epidemiology and Biostatistics (Drs Simon, Vittinghoff, and Hulley and Ms Lin), University of California, San Francisco; General Internal Medicine Section, Veterans Affairs Medical Center, San Francisco (Drs Shlipak and Simon); Department of Family and Preventive Medicine, University of California, San Diego (Dr Barrett-Connor); Northwest Lipid Research Clinic and the University of Washington, Seattle (Dr Knopp); and American Home Products, Madison, NJ (Dr Levy).
Context Lipoprotein(a) [Lp(a)] has been identified as an independent risk factor
for coronary heart disease (CHD) events. However, few data exist on the clinical
importance of Lp(a) lowering for CHD prevention. Hormone therapy with estrogen
has been found to lower Lp(a) levels in women.
Objective To determine the relationships among treatment with estrogen and progestin,
serum Lp(a) levels, and subsequent CHD events in postmenopausal women.
Design and Setting The Heart and Estrogen/progestin Replacement Study (HERS), a randomized,
blinded, placebo-controlled secondary prevention trial conducted from January
1993 through July 1998 with a mean follow-up of 4.1 years at 20 centers.
Participants A total of 2763 postmenopausal women younger than 80 years with coronary
artery disease and an intact uterus. Mean age was 66.7 years.
Intervention Participants were randomly assigned to receive either conjugated equine
estrogens, 0.625 mg, plus medroxyprogesterone acetate, 2.5 mg, in 1 tablet
daily (n = 1380), or identical placebo (n = 1383).
Main Outcome Measures Lipoprotein(a) levels and CHD events (nonfatal myocardial infarction
and CHD death).
Results Increased baseline Lp(a) levels were associated with subsequent CHD
events among women in the placebo arm. After multivariate adjustment, women
in the second, third, and fourth quartiles of baseline Lp(a) level had relative
hazards (RHs) (compared with the first quartile) of 1.01 (95% confidence interval
[CI], 0.64-1.59), 1.31 (95% CI, 0.85-2.04), and 1.54 (95% CI, 0.99-2.39),
respectively, compared with women in the lowest quartile (P for trend = .03). Treatment with estrogen and progestin reduced mean
(SD) Lp(a) levels significantly (–5.8  mg/dL) (−0.20 [0.53]
µmol/L)compared with placebo (0.3  mg/dL) (0.01 [0.60] µmol/L)
(P<.001). In a randomized subgroup comparison,
women with low baseline Lp(a) levels had less benefit from estrogen and progestin
than women with high Lp(a) levels; the RH for women assigned to estrogen and
progestin compared with placebo were 1.49 (95% CI, 0.97-2.26) in the lowest
quartile and 1.05 (95% CI, 0.67-1.65), 0.78 (0.52-1.18), and 0.85 (0.58-1.25)
in the second, third, and fourth quartiles, respectively (P for interaction trend = .03).
Conclusions Our data suggest that Lp(a) is an independent risk factor for recurrent
CHD in postmenopausal women and that treatment with estrogen and progestin
lowers Lp(a) levels. Estrogen and progestin therapy appears to have a more
favorable effect (relative to placebo) in women with high initial Lp(a) levels
than in women with low levels. This apparent interaction needs confirmation
in other trials.
Lipoprotein(a) [Lp(a)] has been found to be an independent risk factor
for coronary heart disease (CHD) events in most1- 7
but not all8- 10
prospective studies of men without known coronary artery disease. Few prospective
studies have evaluated the importance of Lp(a) as a risk factor among women
or among persons with CHD.11 Furthermore, it
is unknown whether interventions directed at lowering Lp(a) levels will affect
subsequent risk of CHD. Lipoprotein(a) levels are not lowered by most lipid-lowering
medications, diet, or exercise.12,13
Among standard lipid-lowering therapies, only nicotinic acid has been shown
to reduce Lp(a) levels.14- 18
Recently, estrogen and the combination of estrogen and progestin have
been found to lower Lp(a) levels in postmenopausal women.18- 24
Because these studies have been conducted in women without CHD and without
assessment of CHD outcomes, the clinical importance of lowering Lp(a) levels
among women is unknown.
The Heart and Estrogen/progestin Replacement Study (HERS) was a randomized,
blinded, placebo-controlled trial of the effect of daily conjugated equine
estrogens, 0.625 mg, plus medroxyprogesterone acetate, 2.5 mg, on the rate
of new coronary events over a mean follow-up of 4.1 years in 2763 postmenopausal
women with known CHD. We evaluated the clinical importance of Lp(a) as a predictor
of CHD events among women with coronary artery disease; the efficacy of estrogen
and progestin in lowering Lp(a) levels in this population; and the relationship
among estrogen and progestin treatment, Lp(a) levels, and subsequent CHD events.
The design, methods, and primary results of the HERS study have been
published previously.25,26 Briefly,
HERS participants were postmenopausal women younger than 80 years who had
been previously diagnosed as having coronary artery disease and had not had
a hysterectomy. Coronary heart disease inclusion criteria required evidence
of 1 or more of the following: history of myocardial infarction (MI), coronary
artery bypass graft surgery, percutaneous coronary revascularization, or angiographic
evidence of at least a 50% occlusion of 1 or more major coronary arteries.
Exclusion criteria included a recent CHD event, New York Heart Association
class IV or severe class III congestive heart failure, serum triglyceride
levels greater than 3.39 mmol/L (300 mg/dL), recent use of any hormone therapy,
uncontrolled hypertension or diabetes mellitus, a disease (other than CHD)
judged likely to be fatal within 4 years, or intolerance to hormone replacement
therapy. At enrollment, participants were randomly assigned to receive a single
identical tablet containing either conjugated equine estrogen, 0.625 mg, and
medroxyprogesterone acetate, 2.5 mg, or placebo.
Demographic characteristics, health history, CHD risk factors, medication
use, and quality of life were assessed at baseline. Baseline levels of fasting
total cholesterol, low-density lipoprotein (LDL) cholesterol, high-density
lipoprotein (HDL) cholesterol, triglycerides, and Lp(a) were determined by
the Lipoprotein Analytical Laboratory at Johns Hopkins University Hospital
in compliance with standards set by the Centers for Disease Control and Prevention.
Subsequent lipid level measurements were performed at 1 year of follow-up.
Lipoprotein(a) levels were measured immunochemically with a sandwich
enzyme-linked immunosorbent assay that uses a monoclonal antibody to apolipoprotein(a)
[apo(a)] (antibody 1D1) as the capture antibody (Strategic Diagnostics, Newark,
Del). Capable of detecting all 11 Lp(a) isoforms, the antibody recognizes
an epitope in the kringle 4 domains of the apo(a) component of Lp(a), but
does not cross-react with plasminogen. The assay was calibrated by the manufacturer
with a reference pool supplied by the Northwest Lipid Research Clinic (Seattle,
Wash) that used purified Lp(a) as the standard. Lipoprotein(a) concentrations
are reported as milligrams per deciliter (micromoles per liter).27
Follow-up visits to the clinical centers occurred at 4-month intervals
over a mean of 4.1 years. All suspected outcome events were reported to the
HERS coordinating center and documentation was obtained for adjudication.
All events were classified independently by 2 coordinating center reviewers,
and discordant classifications were resolved in discussions between the reviewers.
Primary outcomes included nonfatal MI and CHD death, which included
sudden death within 1 hour of symptoms, unobserved death that occurred out
of the hospital in the absence of other known cause, or death due to coronary
revascularization procedure or congestive heart failure. Diagnosis of MI was
based on an algorithm that incorporated clinical symptoms, electrocardiographic
abnormalities, and cardiac enzyme levels.25
Secondary CHD outcomes included coronary artery bypass graft surgery, percutaneous
coronary revascularization, and hospitalization for unstable angina.
To evaluate the association of baseline Lp(a) levels with other known
predictors of CHD events, we divided the HERS cohort of women into quartiles
based on Lp(a) level at enrollment. Using analysis of variance and χ2 statistics, we compared demographic characteristics, CHD risk factors,
previous cardiovascular disease history, other lipid and lipoprotein levels,
and medical treatments among women in the 4 quartiles.
The association of baseline Lp(a) values with subsequent CHD events
was evaluated independently of estrogen and progestin treatment among HERS
participants randomized to the placebo arm of the study. We developed Cox
proportional hazards models using a forward stepwise procedure to include
variables that were associated with Lp(a) at P<.10.
Multivariable-adjusted relative hazards (RHs) for each quartile of baseline
Lp(a) level were calculated and compared with the lowest quartile. A χ2 test of trend was used to determine the relationship among quartiles.
When we repeated the analysis with a model that included all the variables
in Table 1, the point estimates
were virtually identical.
We determined the effect of hormone therapy on Lp(a) levels by calculating
the difference between measurements at baseline and at 1 year of follow-up.
Using a 2-tailed t test, we compared the difference
between participants in the 2 treatment arms. We repeated these comparisons
within each baseline Lp(a) level quartile and tested for linear trend in Lp(a)
level reduction across quartiles.
We evaluated the effect of changes in Lp(a) levels on CHD events after
year 1 for the women assigned to estrogen and progestin. For each woman in
the study, the difference in Lp(a) from enrollment to 1-year follow-up was
calculated and defined as Lp(a) level reduction;
subjects were then divided into quartiles of Lp(a) level reduction. Using
Cox proportional hazard models that included covariables from Table 1 significant at P<.10, baseline
Lp(a) values, variables representing the 1-year differences in lipid and lipoprotein
levels, and compliance with treatment assignment, we determined the association
between the reduction in Lp(a) level and CHD events.
To explore the presence of an interaction between hormonal therapy and
baseline Lp(a) values, we performed a stratified analysis by quartile of the
baseline Lp(a) value. Within each quartile, we compared outcomes among participants
assigned to estrogen and progestin or placebo using an unadjusted proportional
hazards model. We evaluated the effect of hormonal therapy on CHD events by
baseline Lp(a) quartile using a χ2 test for trend. We repeated
this analysis comparing women above and below the median value of Lp(a). Because
the effect of hormonal therapy on primary CHD events appeared to change over
time,26 we also stratified outcomes occurring
before and after the 1-year visits.
Because the 8% of our cohort that was African American had much higher
Lp(a) levels at baseline, we examined whether excluding this group from all
analyses would alter the conclusions and found that it did not. All analyses
were conducted using SAS, version 6.12.28 Two-tailed
tests with P<.05 were considered statistically
Among the 2763 participants in the HERS study, baseline Lp(a) measurements
were available for 2759 women. Levels of Lp(a) were skewed to the right (Figure 1); the median level was 25.3 mg/dL
(0.90 µmol/L), whereas the mean (SD) was 33.7 (32.6) mg/dL (1.20 [1.16]
We compared baseline characteristics of the women by quartile of Lp(a)
level (Table 1). No associations
between Lp(a) quartile and age or education level were observed. African American
women had higher mean Lp(a) levels than other women (58 mg/dL vs 32 mg/dL
[2.1 µmol/L vs 1.1 µmol/L]; P<.001);
more than half of the African American women were in the highest quartile
for the overall cohort. Women in the highest Lp(a) quartile were less likely
to be current smokers (P<.05).
Mean values of LDL and HDL cholesterol were greater in the higher Lp(a)
quartiles, whereas triglyceride levels were lower (P<0.01
for each comparison). Women in the highest Lp(a) quartile were more likely
to be receiving lipid-lowering agents. The significant associations between
Lp(a) quartile and the other lipid levels persisted after adjustment for differences
in the use of lipid lowering medications. Women in the highest Lp(a) quartile
more frequently used aspirin and calcium channel blockers (P<.05).
We examined the association of baseline Lp(a) quartile with subsequent
CHD events in women assigned to placebo. After adjustment for differences
in other baseline characteristics, the initial Lp(a) measurement was associated
with CHD risk (Table 2). Women
in the highest Lp(a) quartile had a 54% (95% confidence interval [CI], 0%-140%)
increased risk of primary CHD events compared with women in the lowest Lp(a)
quartile. Compared with the lowest Lp(a) quartile, the relative risk for revascularization
events was also significantly greater in the highest Lp(a) quartile (61%;
95% CI 10%-130%). We observed a similar increase in risk (although nonsignificant)
for unstable angina admissions and for the 2 components of primary CHD events
(CHD death and nonfatal MI).
Participants randomized to receive estrogen and progestin had significant
reductions in serum Lp(a) levels after 1 year of treatment compared with women
who were randomized to placebo (Figure 2).
The mean (SD) change in Lp(a) level was −5.8 (15) mg/dL (−0.2
[0.5] µmol/L) for women assigned to hormonal therapy, and 0.34 (17)
mg/dL (0.01 [0.6] µmol/L) for women assigned to placebo (P<.001). Although we observed a highly significant reduction associated
with estrogen and progestin in each quartile of baseline Lp(a) level, the
magnitude of this difference was greatest in the higher quartiles (P for trend, <.001).
We examined whether the reduction of Lp(a) levels after 1 year of estrogen
and progestin treatment was associated with a reduced risk of subsequent CHD
events. Cox regression analysis revealed no significant association, but inspection
of the data suggested a possible threshold effect. Among women assigned to
estrogen and progestin, those who were in the quartile of greatest Lp(a) level
reduction (>−8.8 mg/dL [>−0.31 µmol/L]) had a significantly
lower risk of CHD events compared with women who had smaller reductions in
Lp(a) level (RH, 0.62; 95% CI, 0.38-1.00 after adjusting for all risk covariates
significant at P<.10; compliance; and changes
in LDL cholesterol, HDL cholesterol, and triglycerides). When primary events
were analyzed separately, reduction in Lp(a) level appeared to be associated
with decreased risk for MI (RH, 0.46; 95% CI, 0.25-0.85), but not for CHD
death (RH, 0.99; 95% CI, 0.47-2.05). Changes in LDL cholesterol, HDL cholesterol,
and triglyceride levels at 1 year were not associated with decreased CHD risk,
either as continuous or quartile predictors.
There was no overall effect of estrogen and progestin on primary CHD
events (Table 3). However, we
did find evidence of a possible subgroup interaction. Baseline Lp(a) level
appeared to modify the effect of estrogen and progestin treatment on risk
of CHD events (P for interaction = .03). When divided
into quartiles of baseline Lp(a) levels, there was a 15% to 22% decreased
risk of CHD events (nonfatal MI and CHD death) among women with the highest
baseline levels of Lp(a) who were randomized to estrogen and progestin (compared
with those assigned to placebo) and a 49% increased risk among women with
the lowest baseline Lp(a) levels who were randomized to estrogen and progestin
(Table 3). Another way to look
at the same interaction is to compare women above and below the median baseline
Lp(a) value. Estrogen and progestin therapy was associated with an RH for
CHD events of 1.26 (95% CI, 0.93-1.71) for women with a baseline Lp(a) level
below the median value of 25.3 mg/dL (0.90 µmol/L), in contrast to an
RH of 0.82 (95% CI, 0.62-1.08) for women with baseline Lp(a) level above the
median (P for interaction = .04). Lipoprotein(a)
levels did not modify the association between hormone treatment and revascularization
or hospitalization for unstable angina.
In the HERS trial, there was an increased CHD risk associated with estrogen
and progestin in year 1 and a subsequent decreased CHD risk in years 3 through
5.26 We examined this time-dependent effect
stratified by baseline Lp(a) levels (above and below the median of 25.3 mg/dL
[0.90 µmol/L]) (Figure 3).
Among women with baseline Lp(a) levels below the median, we observed an increased
risk of CHD events during the first year among those assigned to estrogen
and progestin (RH, 2.10; 95% CI, 1.05-4.19), compared with those assigned
to placebo. During subsequent years, there was no decreased risk of CHD events
in women assigned to estrogen and progestin (RH, 1.10; 95% CI, 0.78-1.56).
Among women with baseline Lp(a) levels above the median, the RH associated
with hormonal therapy was 1.26 (95% CI, 0.75-2.10) during the first year,
decreasing to 0.68 (95% CI, 0.48-0.95) during subsequent years of follow-up.
Lipoprotein(a) is a unique molecule comprising a lipoprotein particle
resembling LDL cholesterol that is covalently bonded to apo(a), a large plasma
glycoprotein.29 The individual characteristics
of these 2 components are thought to be responsible for the apparent pathogenic
role of Lp(a), which has no known physiologic function. The LDL cholesterol
component likely contributes to atherogenesis, whereas apo(a), similar in
structure to plasminogen, may promote thrombosis. Thus Lp(a), which has been
isolated in the arterial wall at sites of atherosclerosis,30,31
may serve as a link between the pathogenic processes of atherosclerosis and
Investigators independently identified Lp(a) using both immunoelectrophoretic
techniques32 [called Lp(a)
lipoprotein] and lipoprotein electrophoresis (called pre-β-1-lipoprotein33 or sinking pre-β-lipoprotein).34
After further analysis found the 2 entities to be related,34- 36
epidemiological studies identified the association of the molecule with atherosclerosis
and CHD.35,37 Since then, multiple
prospective cohort studies have evaluated the relationship between Lp(a) and
CHD risk in persons without known heart disease. Although the majority of
studies have found an independent risk associated with Lp(a),1- 7
others, most notably the Physicians' Health Study, have found no association.8- 10
In HERS, 4 major findings pertaining to this lipoprotein were observed
among postmenopausal women with known coronary artery disease: (1) Lp(a) was
an independent risk factor for CHD events; (2) estrogen and progestin lowered
Lp(a) levels; (3) large reductions in Lp(a) levels were associated with decreased
risk for recurrent CHD events; and (4) baseline Lp(a) appeared to modify the
effect of estrogen and progestin treatment. The last finding is an interaction:
women with high initial Lp(a) levels had lower rates of CHD events in the
hormone therapy group compared with the placebo group, and those with low
initial Lp(a) levels had higher CHD rates in the hormone therapy group compared
with the placebo group.
Although multiple studies have evaluated the association of Lp(a) with
CHD in patients without known coronary artery disease, only 1 prior study
examined Lp(a) as a risk factor for recurrent CHD events.11
HERS demonstrates that Lp(a) is associated with recurrent CHD events in a
population of women with known coronary artery disease. This association was
independent of other significant predictors.26
Previous trials have demonstrated that estrogen (with or without progestins)
lowers Lp(a) levels in healthy postmenopausal women.19- 23
HERS confirmed the effect of estrogen and progestin on Lp(a) and demonstrated
that this effect also occurs among women with known CHD.
Because Lp(a) levels are not altered by most lipid-lowering drugs, lipid
trials have not examined whether lowering Lp(a) levels reduces risk of CHD.
The 1 drug that does lower Lp(a), nicotinic acid,14,16,17
reduced CHD and mortality rates when compared with placebo in the Coronary
Drug Project38; however, the relative contribution
of favorable changes in LDL and HDL cholesterol and Lp(a) levels to the mortality
reduction is unknown because neither HDL cholesterol nor Lp(a) level was measured
in that study.38 Thus, our study is the first
to have examined the association between Lp(a) level reduction and future
Inspection of our data suggested the possibility of an association between
Lp(a) level reduction and decreased CHD events within the active treatment
group that was independent of both baseline levels and changes in LDL and
HDL cholesterol and triglyceride levels; in fact, change in Lp(a) level was
the only 1 of the 4 lipid change variables that had any statistically significant
association with CHD risk. However, the association between Lp(a) level reduction
and CHD risk was not present over the whole range of values for Lp(a) level
change (as examined by Cox regression analysis). Only women with substantial
reductions in Lp(a) (>−8.8 mg/dL [>−0.31 µmol/L]) had a
significant decrease in risk for CHD events, and the risk reduction was observed
for MI but not for the CHD death outcome. Interpretation of this nonrandomized
association awaits future studies to address the possibility that it is due
to chance or unmeasured confounders.
Independent of the effects of baseline Lp(a) level and Lp(a) level reduction
on CHD events, we examined whether Lp(a) levels might modify the effect of
estrogen and progestin on CHD events. Such an interaction did appear to be
present—the effect of estrogen and progestin compared with placebo on
recurrent CHD events was significantly different among subgroups with low
or high Lp(a) levels. However, true interactions are rare,39
and the finding that estrogen and progestin decreased CHD risk among women
with elevated Lp(a) levels could be due to chance. In fact, among the many
possible subgroups evaluated by the HERS investigators, the number of significant
interactions (including the Lp(a) interaction) was roughly the proportion
expected from chance. On the other hand, the risk relationship between Lp(a)
level and CHD outcomes suggests a possible threshold at the median for this
group of women (25.3 mg/dL [0.90 µmol/L]); this finding, together with
the skewed distribution of Lp(a) levels in the population,40
is compatible with a true biological interaction.
The possible benefit of estrogen and progestin therapy to women with
elevated Lp(a) levels should be evaluated within the context of the primary
results of the HERS study.26 Overall, HERS
demonstrated that estrogen and progestin had neither a beneficial nor a harmful
effect on CHD events. There was a suggestion of an increased risk of CHD events
in the hormone treatment group during year 1 and a trend toward decreased
risk during years 3 through 5 of the trial. After stratifying HERS participants
by baseline Lp(a) levels above and below the median value, we found that the
increased CHD risk associated with hormone therapy in year 1 was primarily
observed in women with low Lp(a) levels. Furthermore, women with high initial
Lp(a) levels had a decreased CHD risk in years 2 through 5 that achieved nominal
criteria for statistical significance.
If confirmed by other studies, our findings could have implications
regarding screening and treatment for elevated levels of Lp(a). Although screening
for elevated Lp(a) levels has been considered to help identify individuals
at risk for CHD, the practice has not been recommended, in part because options
for lowering Lp(a) levels are limited.29 Estrogen
and progestin treatment clearly lowers Lp(a) levels, as do high doses of nicotinic
acid.14 However, policies on screening strategy
will require further studies of the potential benefits and harm of these treatments
in the context of the clinical situation and alternative treatments such as
statins for preventing CHD. In particular, any decision to begin estrogen
and progestin treatment in a woman with known coronary artery disease must
be made in the context of the overall null effect of the HERS trial and the
significantly increased risk of venous thromboembolic disease in women taking
estrogen and progestin.26,41
In conclusion, we found Lp(a) to be an independent and modifiable risk
factor for CHD events in postmenopausal women with known coronary artery disease.
Further studies, such as the ongoing HERS II follow-up study and the Women's
Health Initiative, will be helpful in clarifying the possible interaction
among hormone therapy, Lp(a) levels, and CHD risk.