Author + information
- Received July 24, 2000
- Revision received February 2, 2001
- Accepted March 28, 2001
- Published online July 1, 2001.
- Karen P Alexander, MD, FACCa,* (, )
- L.Kristin Newby, MD, FACC†,
- Anne S Hellkamp, MS†,
- Robert A Harrington, MD, FACC†,
- Eric D Peterson, MD, MPH, FACCa,
- Steve Kopecky, MD, FACC‡,
- Antoly Langer, MD, FACC§,
- Patrick O’Gara, MD, FACC∥,
- Christopher M O’Connor, MD, FACC†,
- Robert N Daly, MS, PhD¶,
- Robert M Califf, MD, FACC†,
- Steven Khan, MD, FACC# and
- Valentin Fuster, MD, FACC∗∗
- ↵*Reprint requests and correspondence:
Dr. Karen P. Alexander, Duke University Medical Center, Box 3411, Durham, North Carolina 27710
This study explored the association between the initiation of hormone replacement therapy (HRT) and early cardiac events (<1 year) in women with a recent myocardial infarction (MI).
Observational studies have linked postmenopausal hormone use with a reduced risk of death from heart disease. However, a recent randomized trial of HRT found no long-term benefit, primarily due to an increase in cardiac events in the first year.
The Coumadin Aspirin Reinfarction Study (CARS) database contains information on HRT use and menopausal status for women with a recent MI. We classified the 1,857 postmenopausal women in CARS as prior/current HRT users if they took HRT before enrollment, new users if they began HRT during the study period or never users. We assessed the incidence of cardiac events (death, MI, unstable angina [UA]) during follow-up.
In our cohort, 28% (n = 524) used HRT at some point. Of these, 21% (n = 111) began HRT after their MI. New users had a higher incidence of death/MI/UA (41% vs. 28%, p = 0.001) during follow-up than never users, largely due to a higher incidence of UA (39% vs. 20%, p = 0.001). After adjustment, new users still had a significantly higher risk of death/MI/UA than never users during follow-up (relative risk [RR] = 1.44 [1.05–1.99]). Prior/current users had no excess risk of the composite end point after adjustment. Users of estrogen/progestin had a lower incidence of death/MI/UA during follow-up than users of estrogen only (RR = 0.56 [0.37–0.85]).
Postmenopausal women who initiated HRT after a recent MI had an increased risk of cardiac events largely due to excess UA during follow-up.
A cardioprotective effect of estrogen in premenopausal women has been postulated to explain the natural delay in the onset of coronary artery disease (CAD) in women (1). Protective effects of hormone replacement therapy (HRT) have also been reported in numerous observational studies of primary (2–4)and secondary prevention of cardiovascular disease in postmenopausal women (5,6).
However, these observational studies were confounded by “prevention bias.” Specifically, women who took HRT were often younger and had lower cardiovascular risk profiles, than those who did not (7–9). The first randomized trial of HRT for secondary prevention of cardiovascular disease in postmenopausal women (Heart and Estrogen/progestin Replacement Study [HERS]) showed no difference in cardiac events or mortality at four years, but did show an unexpected significant increase in myocardial infarction (MI) during the first year after therapy initiation (10). This phenomenon may have been due to an increased early risk from therapy or a random fluctuation in event rates. Because the HERS study drug was a combination of estrogen and progestin, either hormone may have contributed to the increase in infarction rates. Overall, questions remain regarding how estrogen and progestin modulate cardiovascular risk.
Our study examined the association between the initiation of HRT (estrogen alone or in combination with progestin) and the subsequent risk of cardiovascular events in the secondary prevention cohort enrolled in the Coumadin Aspirin Reinfarction Study (CARS) (11). The postmenopausal women in CARS were uniquely suited for our analysis because they were a post-MI cohort who had frequent study visits for drug monitoring where all medications were recorded. As such, our study provides information about the timing and impact of initiation of HRT in a group of postmenopausal women who had recently suffered an MI.
Our analysis population consisted of 1,857 postmenopausal women enrolled in CARS (11). The CARS population consisted of 8,803 post-MI patients from the U.S. and Canada who were treated daily with either 160 mg or 80 mg of aspirin and 1 mg or 3 mg of warfarin. Patients were enrolled 3 to 21 days after an acute MI, defined as elevated myocardial enzyme concentrations, with either typical chest pain or electrocardiographic (ECG) changes. The CARS inclusion criteria required that women were either postmenopausal or surgically sterilized. To ensure we had selected a postmenopausal population, we included in our analysis only those women who were either age ≥50 years, or who used HRT.
Use of HRT was left to the discretion of the patient and physician and included both estrogen only and estrogen/progestin preparations. At the time of enrollment, HRT use in the past two years was recorded on the case report form (yes/no). Patients were followed at 1, 2, 3, 4, 6 and 12 weeks following study enrollment and then every 3 months until the study’s conclusion. At follow-up visits, patients were asked to list in free-text form all medications they were currently taking. All case report forms were monitored against source documents to ensure accuracy. We searched the data for all possible estrogen/progestin medications to identify which patients were taking HRT and the follow-up visit at which that therapy first appeared in their medication list (see the Appendix). We separated our population of postmenopausal women into three categories. Prior/currentusers were those who reported use of HRT at the time of randomization or within the prior two years. New userswere those who did not use HRT prior to randomization but reported use during follow-up. Never userswere those who had no recorded use. The HRT type was reported as topical or oral, and combination or estrogen alone. Both the timing of initiation and the duration of HRT use are reported.
Clinical end points
The primary end point in the overall CARS study was the composite of reinfarction (ischemic pain with raised creatine kinase-MB fraction or new Q-waves), nonfatal ischemic stroke, or cardiovascular death. Secondary end points in CARS included all-cause mortality, silent MI, unstable angina (UA) requiring hospitalization, transient ischemic attack (TIA) and systemic embolization. An independent committee unaware of treatment assignment adjudicated all suspected end points. The primary end point for our analysis was the composite of cardiovascular death, reinfarction and UA requiring hospitalization. Unstable angina was defined as rest or crescendo angina requiring hospitalization for management with no or a nondiagnostic enzyme rise and without the development of pathologic Q-waves on the ECG. We further looked at the impact of estrogen use on the individual components of the triple end point and on subsequent use of revascularization.
Statistical analysis and model adjustment
Descriptive statistics summarizing baseline characteristics, hormone use, index MI variables and event rates are presented as percentages for discrete variables, and medians and first and third quartiles for continuous variables. Comparisons between groups were made using chi-square tests for discrete variables and Wilcoxon rank-sum tests for continuous variables, and they were considered significant at a level of p < 0.01.
Unadjusted rates of the composite and its components are shown by HRT use category. Kaplan-Meier curves for survival free from death/MI/UA during follow-up are presented by HRT use category. Because UA was the most frequent cardiac event, we looked at its occurrence at two time intervals post-MI: 0 to 6 months and 6 to 12 months, and in relation to the timing of revascularization procedures if they occurred.
A Cox proportional hazards survival model for time to death or MI was developed for all CARS patients (n = 8,803). Based on univariate tests and clinical relevance, 27 candidate predictors were tested in the model: age, race, gender, weight, height, current or past smoker, family history, heart rate, systolic blood pressure (BP), diastolic BP, ejection fraction (EF), congestive heart failure, hypertension, diabetes, hypercholesterolemia, peripheral vascular disease (PVD), renal insufficiency, previous angina, prior percutaneous transluminal coronary angioplasty (PTCA), prior coronary artery bypass graft surgery, prior MI, prior stroke or TIA, estrogen use, atrial fibrillation, resuscitated cardiac arrest, anterior MI and randomized treatment. Missing values were assigned their predicted value from a linear model of baseline variables. Of the candidate predictors, age, previous angina, congestive heart failure, current smoker, EF, hypertension, prior MI, PVD, prior stroke or TIA, race and weight were found to be significant predictors (p < 0.05). We then applied these results to our 1,857 postmenopausal women by developing Cox proportional hazards survival models for death, MI and the composite of death and MI, which included the foregoing 11 predictors as well as randomized treatment and HRT (prior/current, new, or never). Both EF and weight were included as nonlinear terms. Interaction between randomized treatment and HRT was also included. A Cox proportional hazards survival model for time to UA was also developed starting with all of the 27 candidate predictors listed above, except for gender. The final UA model included age, history of angina, atrial fibrillation, systolic BP, prior PTCA and HRT use.
Hazard ratios with 95% confidence intervals (CI) describe the likelihood of having the composite of death/MI/UA and its components among prior/current or new users compared with never users of HRT after adjusting for baseline differences. Finally, we examined rates of death/MI/UA by type of HRT (estrogen only vs. estrogen/progestin) and by use category (new vs. prior) both before and after adjustment using the death/MI model.
Estrogen use patterns
The CARS trial enrolled 8,803 post-MI patients of whom 1,857 (21%) were postmenopausal women. Of these, 413 (22%) took HRT at randomization or within the prior two years and were classified as prior/current users. One-hundred eleven (6%) began taking HRT during study follow-up and were classified as new users. The remaining 1,333 women (72%) never took HRT.
Overall, the majority of prior/current and new users were taking oral estrogen (67% and 69%, respectively) although approximately one quarter of both groups were taking oral combination therapies (28% of prior users and 27% of new users) with very few women using topical therapies of any kind (Table 1). Therapy onset for 83% of the new users was before one-year follow-up, with 40% beginning therapy within three months of their enrolling MI. Very few prior/current or new users discontinued HRT within one year of their enrolling MI (7.8% and 12.5%, respectively). Median follow-up for our population was 15 months (range 8 to 23 months).
Prior/current users and new users of HRT were younger than never users of HRT and were more likely to be white and college educated. They were more likely to be smokers, but less likely to have diabetes than never users (Table 2). The three groups were fairly evenly matched in terms of their cardiac history, with the exception of less congestive heart failure among the new users. New users of HRT also had higher use of beta-blockers during follow-up than never users (61% vs. 51%), but there were no significant differences in other medication use.
There were no differences in location of index infarct, Q-wave infarcts or the use of thrombolytics across estrogen use groups (Table 3). Both prior/current and new HRT users were more likely than never HRT users to undergo cardiac catheterization during their hospitalization. At cardiac catheterization, never users of HRT had a trend toward more three-vessel CAD, and prior/current and new users of HRT were more likely to have one- or two-vessel disease. Both prior/current and new users of HRT were also more likely to undergo direct angioplasty or angioplasty during follow-up than never users.
Cardiac event rates
Several differences were seen in cardiac event rates among the different HRT user groups at follow-up. First, new users of HRT were significantly more likely to experience the triple end point of death/MI/UA during follow-up than either prior users or never users of HRT (Fig. 1; Table 4). This was largely accounted for by new users of HRT having significantly more UA than never users of HRT (39% vs. 20%, p = 0.001) (Table 4). New users were also more likely than never users to undergo a PTCA procedure (23% vs. 10%, p = 0.001). However, they were significantly less likely to have suffered a reinfarction (4% vs. 9%, p = 0.03) or to have died (0% vs. 6%, p = 0.001) than women who never used HRT (Table 4).
Compared with never users, prior/current users of HRT were also significantly less likely to have died (2% vs. 6%, p = 0.001), but there was no significant difference in their rate of a subsequent MI (8% vs. 9%, p = 0.4) or UA (24% vs. 20%, p = 0.07). Like new users, prior/current users were more likely to have undergone a PTCA procedure during follow-up than never users (15% vs. 10%, p = 0.004) (Table 4). We investigated the timing of the UA end points and PTCA, and we found that the majority of UA events occurred prior to any percutaneous intervention. The PTCA preceded the occurrence of UA in <5% of the study population and <3% of new users of HRT.
The highest incidence of UA events was within the first three months after the index MI; however, these events occurred throughout the follow-up period. Unstable angina was consistently more frequent among new users of HRT for both the first six month interval (new vs. never relative risk [RR] 1.88, 95% CI = 1.28–2.76) and the interval beginning six months after the index infarction (new vs. never RR 2.03, 95% CI = 1.12–3.68). There were no end points in new users that occurred prior to the initiation of HRT.
To determine whether treatment with the anticoagulant warfarin could attenuate the rate of UA in HRT users, we examined cardiac event rates in all of our estrogen subgroups by their CARS treatment assignment. Assignment among the three CARS treatment arms was evenly balanced across HRT use subgroups. As in the main CARS trial in which there was no overall treatment effect, we found no significant treatment effect of warfarin within our HRT use subgroups in the incidence of cardiac events.
Event rates after adjustment
Because of significant differences in baseline characteristics among HRT user groups, we adjusted outcomes using our risk-adjustment models. After adjustment, the hazard ratio for the likelihood of death/MI/UA among new users compared with never users of HRT remained significantly higher (RR = 1.44, 95% CI = 1.05–1.99), but was not increased in prior users of HRT. The adjusted hazard ratio for the likelihood of UA among new users of HRT was also higher than never users (RR = 1.86 [95% CI = 1.32–2.61]) (Table 5). Prior/current users had no significant increased risk of UA or MI compared with never users but did have a significant decrease in the risk of death. Because of the small numbers of death and MI among new users, no adjustment was performed.
Estrogen versus estrogen plus progestin
We also examined whether HRT type (estrogen only or estrogen/progestin combination) impacted on cardiac events. When all post-MI users were combined (prior/current still using, and new users, n = 354), a significant difference was seen between the estrogen/progestin and the estrogen-only users’ composite end point (20% vs. 35%, p = 0.002) and UA one-year event rates (21% vs. 35%, p = 0.013). New users whose HRT was estrogen only (n = 70) had a 48% occurrence of the composite end point compared with 23% among estrogen/progestin users (n = 26) (p = 0.014), and a 44% occurrence of UA compared with 19% among estrogen/progestin users (p = 0.02). Prior/current users of estrogen-only therapy also had a trend toward higher UA rates than prior/current users of combination therapy (31% vs. 22%, p = 0.15). After adjustment for baseline differences, the hazard ratio for users of estrogen/progestin compared with estrogen only for the composite end point was 0.56 (0.37–0.85) among all users, 0.61 (0.38–0.99) among prior/current users and 0.45 (0.20–1.01) among new users.
In this retrospective analysis of the CARS data, we found a significant increase in the composite of death/MI/UA, and a nearly twofold increased risk of UA among postmenopausal women who initiated HRT soon after an MI. There was an even greater increased risk among subjects using estrogen only than among those using the combination of estrogen and progestin. However, women who were on HRT prior to their index MI (prior/current HRT users) did not have a significantly increased risk for UA or MI with continued use, and they had a lower risk of death compared with women who never used HRT.
Although the balance of prior evidence has suggested a protective effect of HRT, several other studies have suggested a possible increase in cardiac events such as UA and MI. In the early 1960s, 10 mg of estrogen was given to men within three months of MI, but this resulted in an unexpected number of deaths early in the study. Investigators concluded that estrogen should not be started within the first three months following infarction (12). In the Coronary Drug Project, men were randomized to a number of therapeutic strategies, including two doses of estrogen (2.5 mg, 5 mg) or placebo to assess the effects on cholesterol and cardiac events. Both estrogen arms were discontinued early, with a higher rate of MI seen among men in the 5-mg estrogen arm (13,14). Also, the Framingham investigators initially reported an increase in cardiac events with use of HRT (15), and it was only after they removed chest pain from the combined end point that they were able to show a reduction in cardiac events (16). Finally, it has long been known that oral contraceptives increase the risk of MI, especially in older women who smoke (17).
The HERS trial was the first randomized study to show that women with known coronary artery disease who were randomized to conjugated estrogen plus progestin had a higher rate of MI in the year following initiation of treatment than those taking placebo. The association of HRT use in HERS with UA events alone has not been reported, but there was a slight, but not significant, early increase in the composite end point of UA or revascularization. The HERS study also found an increase in venous thrombosis (RR 2.7, 95% CI = 1.4–5.0), which focused questions about estrogen’s effects on the coagulation cascade (10,18). This nearly threefold increase in venous thrombosis with HRT has been previously reported in four observational reports and one randomized study (19–23). Our study adds to the evidence, suggesting a potential prothrombotic effect of estrogen particularly when initiated early after MI.
The causal relationship between HRT and early ischemic events is yet to be fully explored, but it may relate to prothrombotic or procoagulant properties. Estrogen use was associated with lower fibrinogen levels compared with placebo in the Postmenopausal Estrogen/Progestin Intervention (PEPI) trial, a phenomenon that should lower thrombotic risk (24). The response of factor VII, proteins C and S and antithrombin III to HRT has been variable across a number of small studies, leaving this question unanswered (25–29). The recently completed Estrogen Replacement and Atherosclerosis (ERA) trial found no effect of estrogen or combination therapy on regression of mild atherosclerosis but did find an associated increase in C-reactive protein levels and other potentially prothrombic molecules (30). Prior studies have also found C-reactive protein, associated with plaque instability and increased thrombotic risk, to be elevated in patients taking HRT (31,32). Therefore, although the exact nature of the thrombotic risk with HRT remains unclear, there is growing evidence to support its existence.
Unstable angina was significantly higher among new users of HRT than among the other groups. The absolute number of death and MI end points was small among new users (total of four events), which limits conclusions about significance. The use of HRT in our population was not randomized, so as with the other observational studies of HRT use, women in our group who used HRT were younger, with fewer risk factors and less three-vessel disease compared with never users (7–9). This “prevention bias” might explain a real lower relative risk of death and MI in HRT users as seen in prior series and in our population of prior/current users (2). New users also took more beta-blockers after MI, which lowered their risk of death, and this suggests other unmeasured treatment differences. Also, HRT may be acting through several mechanisms in its modulation of cardiovascular end points: one mechanism resulting in the increase in UA and another in the reduction of death and MI. For example, an increase in thrombotic risk may precipitate UA, whereas the action of HRT on vasomotor tone may reduce death and MI. However, although the small numbers preclude firm conclusions regarding death and MI among new users, the excess of UA in this group is clearly seen.
During our follow-up period, there were few MIs and no deaths among the new users of HRT. Therefore, our ability to detect clear associations between HRT use and the end points of death and MI is diminished. The end point of “unstable angina requiring hospitalization” was broadly defined, although adjudicated by a clinical end point committee. We were able to distinguish between opposed and unopposed users of estrogen, but we had to combine different estrogen formulations into a single HRT category. We were unable to show a diminution in long-term risk after initiation of HRT use owing to our relatively short period of follow-up, although the lack of association between prior/current HRT use and these events is encouraging.
In this secondary prevention population where HRT use was not randomized, we have shown an increased risk of the combined end point of death/MI/UA among postmenopausal women who initiated HRT use soon after MI. This was primarily due to the higher rate of UA. This finding adds to the growing concern regarding the potential risks of HRT when used in patients with known cardiovascular disease.
The authors thank Ms. Tracey Dryden for her excellent editorial assistance.
Hormone Replacement Drugs Listed in Free Text
|HRT Typelegend||Free Text Names||No. of Patients|
|Conjugated estrogen||Conjugated estrogen||424|
|Topical estrogen||Estrogen patch||39|
↵legend HRT = hormone replacement therapy.
☆ Supported in part by DuPont Pharmaceuticals.
- blood pressure
- Coumadin Aspirin Reinfarction Study
- confidence interval
- electrocardiographic or electrocardiogram
- ejection fraction
- Estrogen Replacement and Atherosclerosis trial
- Heart and Estrogen/progestin Replacement Study
- hormone replacement therapy
- myocardial infarction
- Postmenopausal Estrogen/Progestin Intervention trial
- percutaneous transluminal coronary angioplasty
- peripheral vascular disease
- relative risk
- transient ischemic attack
- unstable angina
- Received July 24, 2000.
- Revision received February 2, 2001.
- Accepted March 28, 2001.
- American College of Cardiology
- Palmer J,
- Rosenberg L,
- Shapiro S
- O’Brien J.E,
- Peterson E.D,
- Keeler G.P,
- et al.
- Rodstrom K,
- Bengtsson C,
- Lissner L,
- Bjorkelund C
- Matthews K.A,
- Kuller L.H,
- Wing R.R,
- Meilahn E.N,
- Plantinga P
- Eaker E.D,
- Castelli W.P
- Gutthann S.P,
- Garcia Rodriguez L.S,
- Castellague J,
- Oliart A.D
- Scarabin P.Y,
- Vissac A.M,
- Kirzin J.M,
- et al.
- Scarabin P.Y,
- Alhene-Gelas M,
- Plu-Bureau G,
- Taisne P,
- Ahger R,
- Aiach M
- Ridker P.M,
- Buring J.E,
- Shih J,
- Mattias M,
- Hennekens C.H
- Ridker P.M,
- Hennekens C.H,
- Rifai N,
- Buring J.E,
- Manson J.E