Author + information
- Received May 10, 2001
- Accepted July 23, 2001
- Published online November 1, 2001.
- Brad G. Angeja, MD*,* (, )
- Michael G. Shlipak, MD, MPH*,†,‡,
- Alan S. Go, MD*,†∥,
- S.Claiborne Johnston, MD, PhD*,
- Paul D. Frederick, MBA, MPH¶,
- John G. Canto, MD, MSPH, FACC**,
- Hal V. Barron, MD, FACC*,#,
- Deborah Grady, MD, MPH*,†,‡,
- for the National Registry of Myocardial Infarction 3 Investigators
- ↵*Reprint requests and correspondence: Dr. Brad G. Angeja, Fellow, Division of Cardiology, University of California, San Francisco, Box 0124, Moffitt M1182, 505 Parnassus Ave., San Francisco, California 94143 USA
We examined the association of hormone therapy (HRT) with hemorrhagic and ischemic stroke among postmenopausal women with acute myocardial infarction (AMI).
Hemorrhagic and ischemic strokes are common complications of AMI, and women are at increased risk for hemorrhagic stroke after thrombolytic therapy. This risk may be related to female hormones.
Using data from the National Registry of Myocardial Infarction-3, we studied 114,724 women age 55 years or older admitted to the hospital for AMI, of whom 7,353 reported HRT use on admission. We determined rates of in-hospital hemorrhagic and ischemic stroke stratified by HRT use and estimated the independent association of HRT with each stroke type using multivariable logistic regression.
The HRT users were younger than non-users, had fewer risk factors for stroke including diabetes and prior stroke, and received more pharmacologic and invasive therapy including cardiac catheterization. A total of 2,152 (1.9%) in-hospital strokes occurred, with 442 (0.4%) hemorrhagic, 1,017 (0.9%) ischemic and 693 (0.6%) unspecified. Among HRT users and non-users, the rates of hemorrhagic stroke (0.40% vs. 0.42%, p = 1.00) and ischemic stroke (0.80% vs. 0.96%, p = 0.11) were similar. Among 13,328 women who received thrombolytic therapy, the rate of hemorrhagic stroke was not significantly different for users and non-users (1.6% vs. 2.1%, p = 0.22). After adjustment for baseline and treatment differences, HRT was not associated with hemorrhagic (odds ratio [OR], 0.88; 95% confidence intervals [CI], 0.58 to 1.35) or ischemic stroke (OR, 0.89; CI, 0.66 to 1.18).
Acute myocardial infarction is a high-risk setting for stroke among postmenopausal women, but HRT does not appear to modify that risk. Clinicians should not alter their approach to thrombolytic therapy based on HRT use.
Controversy surrounds the use of hormone therapy (HRT) to reduce cardiovascular events in postmenopausal women. Observational studies have found a beneficial effect on coronary heart disease events (1,2), but the first randomized trial to study the effect of HRT on coronary disease outcomes, the Heart and Estrogen/progestin Replacement Study (HERS), found no effect of HRT for secondary prevention of cardiac events (3). The association of HRT with stroke risk is also controversial. Although the use of estrogens for contraception has been associated with increased risk of ischemic and hemorrhagic stroke (4,5), longitudinal studies of HRT in postmenopausal women have been conflicting, with findings of increased (2,6), decreased (7)and no risk (8)for ischemic stroke. The few observational studies of hemorrhagic stroke have observed HRT to be associated with decreased risk (9,10). In the HERS trial, however, HRT had no effect on risk for the 215 cerebrovascular events that occurred (11). No study has examined the impact of HRT on stroke risk after acute myocardial infarction (AMI).
The risk of stroke is high after AMI, with an incidence of up to 2% at 30 days in unselected patients (12,13). The risk is highest in patients receiving thrombolytic therapy, due to an increased risk of intracranial hemorrhage, but ischemic strokes are also common (13,14). Risk factors for ischemic stroke after AMI include older age, diabetes, prior stroke or transient ischemic attack, previous angina and history of hypertension (15). Independent risk factors for hemorrhagic stroke after thrombolysis are older age, black race, prior stroke, elevated blood pressure (BP), lower body weight and, interestingly, female gender (16), raising the possibility that female hormones might have some impact on cerebrovascular disease or response to therapy. To address this possibility, we evaluated the association of HRT use at the time of hospital admission for AMI with hemorrhagic and ischemic stroke.
The third National Registry of Myocardial Infarction (NRMI-3) is a nationwide, prospective cohort study sponsored by Genentech, Inc. (South San Francisco, California), in which data are collected regarding clinical characteristics, treatment patterns and outcomes of patients presenting with AMI. The 1,674 participating hospitals are encouraged to enroll consecutive patients with AMI, regardless of treatments used or outcomes observed. Data are collected on prespecified forms by a study coordinator at each hospital and forwarded to a central agency for compilation (STATProbe, Inc., Lexington, Kentucky). Double key data entry and electronic data checks are routinely performed to ensure accuracy, consistency and completeness, which have compared favorably with the independently validated Cooperative Cardiovascular Project (17). Confirmation of AMI requires an International Classification of Diseasesversion 9 hospital discharge diagnosis code of 410.x1 and a patient history and clinical presentation suggestive of AMI combined with any of the following: a total creatine kinase (or MB fraction) at least twice the hospital’s upper limit of normal; electrocardiographic evidence of AMI; or alternative enzymatic, scintigraphic, echocardiographic or autopsy evidence indicative of AMI.
From April 1, 1998 to January 31, 2000, 404,106 men and women with AMI were enrolled in NRMI-3. Patients who were transferred to another hospital were excluded because of missing information on their vital status at hospital discharge (n = 86,205). Because NRMI-3 does not collect information on menopausal status, we limited our analysis to women age 55 years or older, which resulted in a cohort of 114,724 women with documented AMI.
The primary predictor of interest was use of HRT prior to hospital admission. Hormone therapy was defined as “the current use of estrogen, progestin, or estrogen/progestin for reasons other than contraception.” Women were coded as current users or non-users of HRT, but information on past use, dose, preparation, cyclic or non-cyclic administration, or duration of use was not available. Data were also collected on demographic characteristics (age, race and region of residence in the US), past medical history (diabetes, hypertension, prior stroke, prior AMI, congestive heart failure [CHF], percutaneous transluminal coronary angioplasty [PTCA], coronary artery bypass graft surgery [CABG], hypercholesterolemia and family history of coronary artery disease), presentation characteristics (presence of chest pain, systolic and diastolic BPs, heart rate, Killip class at initial evaluation, electrocardiogram findings and location of infarct) and treatments during the index hospitalization (medications and procedures).
The primary outcome was stroke occurring during the AMI hospitalization. The diagnosis of stroke required documentation of any new neurologic deficit by a treating physician. Radiographic imaging was not required for stroke diagnosis. For this analysis, strokes were divided into two classifications. Ischemic stroke was defined as “thromboembolic” or “thromboembolic with hemorrhagic conversion,” and hemorrhagic strokes were only primary “hemorrhagic” stroke types. Hemorrhagic strokes were verified with computed tomography or magnetic resonance imaging in 92.5% of reported cases.
Admission characteristics, in-hospital treatment, and outcomes of users and non-users of HRT were compared using the ttest for continuous variables and the chi-square test for categorical variables. Because of the large sample size in NRMI-3, differences with little clinical relevance may achieve statistical significance. Multivariable logistic regression was used to adjust for differences in baseline characteristics and treatments received in-hospital. A backward stepwise elimination procedure was used to select variables to remain in the final model with a significance threshold of p < 0.05. We conducted subgroup analyses based on decade of age and use of thrombolytic therapy by constructing multivariable models within these a priori subgroups. The presence of a statistical interaction of HRT use with age and thrombolytic use was assessed via an interaction term in the multivariable models for each stroke type.
In addition, propensity score methods were employed to adjust for residual selection bias (18). First, a multivariable logistic regression model was constructed with HRT use (yes/no) as the outcome variable and admission characteristics as the predictor variables. The coefficients from this analysis were used to calculate the likelihood of receiving HRT for each patient in the sample, whether they actually received HRT or not, and these likelihoods were grouped into deciles. These “propensity to receive HRT” deciles were then entered as predictor variables in the multivariable models, with each stroke subtype as the outcome. Thus, we attempted to adjust the association of HRT with stroke risk for differences in the propensity to receive HRT in order to reduce the influence of confounding. The predictive value of each model was assessed using the c statistic. The SAS statistical package was used for all statistical analyses (SAS 8.0, SAS Institute Inc., Cary, North Carolina).
Among the 114,724 women with AMI in NRMI-3 aged 55 years or older, 7,353 (6.4%) were current users of HRT at the time of hospital admission. The HRT users were younger, more likely to be white and to smoke, and more likely to have hypercholesterolemia or a family history of premature coronary artery disease, but less likely to have a history of diabetes, heart failure, prior MI or prior stroke than non-users (Table 1). At initial presentation, HRT users had slightly higher systolic BP and lower heart rates, were more likely to complain of chest pain and had fewer signs of heart failure (Killip Class II–IV).
After hospital admission, HRT users received more pharmacologic and invasive therapy than non-users. They were more likely to receive aspirin (87% vs. 79%), heparin (74% vs. 64%), and thrombolytic agents (19% vs. 12%, all p < 0.001). The HRT users were also more likely to undergo invasive cardiac procedures during hospitalization, including coronary angiography (65% vs. 42%), primary PTCA (9% vs. 6%) or any percutaneous coronary intervention (25% vs. 15%) and CABG (12% vs. 8%, all p < 0.001).
Hormone use and stroke
During the index hospitalization, stroke occurred in 2,152 women (1.9%, 95% confidence intervals [CI], 1.8% to 2.0%), including 442 (0.4%) hemorrhagic, 1,017 (0.9%) ischemic and 693 (0.6%) unspecified strokes. Although the overall stroke incidence was marginally lower for HRT users, we observed no statistically significant differences in the crude incidence of either hemorrhagic or ischemic stroke (Table 2). Among women receiving thrombolytic therapy, similar rates of hemorrhagic stroke were observed for users and non-users of HRT (1.6% vs. 2.1%, p = 0.22). In-hospital mortality was high among women suffering a stroke but was not different between HRT users and non-users (42% vs. 40%, p = 0.61).
In a multivariable model adjusting for differences between HRT users and non-users, HRT use was not significantly associated with the incidence of hemorrhagic stroke among all women (odds ratio [OR], 0.88; 95% CI, 0.58 to 1.35) or among those women receiving thrombolytic therapy (OR, 1.05; 95% CI, 0.65 to 1.69). Variables retained in the final model were history of diabetes, AMI, stroke or PTCA; chest pain on admission; systolic BP on admission; Killip Class > I; ST segment elevation; admission diagnosis “other”; any cardiac catheterization; use of thrombolytic therapy; and use of primary PTCA. The c statistic for the hemorrhagic model was 0.82 among all patients and 0.78 among patients treated with thrombolysis. Similarly, after adjustment, HRT use was not significantly associated with the incidence of ischemic stroke (OR, 0.89; 95% CI, 0.66 to 1.18). Variables retained in the final model were history of hypertension, CHF, stroke or CABG; chest pain on admission; systolic BP on admission; anterior AMI; ST segment elevation or depression; admission diagnosis of “unstable angina”; any cardiac catheterization; and any use of PTCA. The c statistic for the ischemic model was 0.65. There was no significant interaction between age and HRT use for ischemic stroke (p = 0.07) or for hemorrhagic stroke, either among all postmenopausal women (p = 0.70) or among those women receiving thrombolysis (p = 0.12).
In this large, community-based cohort, we found that stroke complicates 1 of every 50 AMI admissions among postmenopausal women and is associated with a 40% in-hospital mortality rate. There were no significant differences between current HRT users and non-users in the incidence of either hemorrhagic or ischemic stroke or in overall stroke mortality. Hormone therapy use did not appear to modify the risk of hemorrhagic stroke associated with thrombolytic therapy.
Prior studies of HRT and stroke
Previous observational studies of HRT in postmenopausal women have found conflicting results regarding its association with risk for stroke. Estimates of the association of HRT and ischemic stroke range from a 30% decreased risk (7)to a threefold increased risk (6), with other studies showing no association (8). A limitation of several studies is that ischemic and hemorrhagic strokes were not distinguished, despite their differences in pathophysiology. Two case-control studies found HRT to be associated with a 40% to 60% decreased risk for hemorrhagic stroke, but these studies based their findings on fewer than 100 cases with known HRT status (9,10).
Recent randomized trials of HRT for secondary prevention of cardiovascular disease have examined its effect on stroke risk. In the HERS trial, over 2,700 women with a known history of coronary artery disease were randomly assigned to receive either conjugated estrogen (0.625 mg/d) and medroxyprogesterone acetate (2.5 mg/d) or placebo and followed for a mean period of 4.1 years. The HRT users, compared with non-users, had a 2% higher absolute incidence of prespecified ischemic or hemorrhagic stroke or transient ischemic attack, but this difference was not statistically significant (7% vs. 5%, p = 0.20). No significant difference in hemorrhagic (n = 13) or ischemic (n = 128) strokes was observed (11). In the Women’s Estrogen for Stroke Trial (WEST), 652 postmenopausal women who had experienced a non-disabling stroke or transient ischemic attack were randomized to treatment with 17-beta estradiol (1 mg/d) or placebo with an average follow-up of 2.7 years. The primary endpoint of all-cause death or nonfatal stroke occurred in 27.6% of estradiol-treated patients and 27.7% of placebo-treated patients (p = 0.80), with stroke occurring in 16.8% and 17.4% (p = 0.83), respectively (19). Estradiol was associated with a more than two-fold increase in the risk of stroke at 6 months; death due to stroke had occurred in 12 women assigned to estradiol therapy but only three women assigned to placebo therapy. (Data presented by Lawrence M. Brass at the 26th International Stroke Conference of the American Stroke Association, February 2001.) These trials suggest that HRT does not protect against ischemic or hemorrhagic stroke and may be associated with an early increase in risk.
The present study has several unique strengths. First, to our knowledge, this is the largest cohort of postmenopausal women with AMI studied to date, with detailed data collected on over 114,000 women, over 7,000 of whom were HRT users. Second, although the period of observation in NRMI-3 is relatively short, extending only through the duration of the AMI hospitalization, the number of stroke events (2,152) exceeds that of any prior study, providing adequate power to observe clinically relevant associations between HRT and stroke after AMI. Detailed information was collected on the type of stroke in our cohort, and our ability to distinguish hemorrhagic and ischemic events extends the findings of previous observational studies. Finally, this is the first study to examine the association of HRT with stroke in the setting of AMI.
AMI: a high-risk period for stroke
Stroke is a relatively common occurrence in AMI whether thrombolytic therapy, primary angioplasty, or no reperfusion therapy is used. In the Global Use of Strategies to Open Occluded Coronary Arteries trial, ischemic stroke occurred in 1.4% of patients receiving thrombolytic therapy (12). The rate of stroke may be even higher in community settings. In NRMI-2, stroke occurred in 2.0% of patients receiving tissue plasminogen activator, more often in women than in men (3.3% vs. 1.4%, p < 0.001), but stroke also occurred in 0.9% of patients receiving no reperfusion therapy (14). Female gender was found to be an independent risk factor for hemorrhagic stroke after thrombolysis, increasing the risk by almost 60% compared with men (16).
Various mechanisms have been proposed by which female hormones might increase or decrease stroke risk. Hormone therapy increases risk for venous thromboembolic events about threefold (20). If the mechanisms of this increased risk is hypercoagulability, risk for embolic stroke might also be increased. In the Stroke Prevention in Atrial Fibrillation trials, women with atrial fibrillation, in whom emboli are the most likely cause of stroke, had a greater risk of stroke if they were HRT users (6). After an AMI, hypokinesis of the ventricular wall promotes thrombus formation and subsequent embolism, and any hypercoagulable effects of HRT in this setting might be expected to increase stroke risk. As a potential mechanism for hemorrhagic stroke risk, estrogen appears to modulate the endogenous fibrinolytic potential in women (21), and bleeding rates are higher in women than in men receiving thrombolytic therapy (22). However, in the present study, the lack of association between HRT and hemorrhagic stroke argues against any significant clinical impact of HRT that would explain these observed differences between men and women. That there was no increase in risk for hemorrhagic stroke among women taking HRT who received thrombolytic therapy should reassure clinicians who treat women with AMI.
Our study design is observational, and like all such studies, the results may be biased if the two groups of women are different in ways other than HRT use. Although we adjusted for certain characteristics associated with better health in women taking HRT using extensive multivariable models, residual confounding due to selection and adherence biases may still be present. We did use propensity score methods in an attempt to address this residual confounding. Outcomes were observed only until hospital discharge, and any strokes occurring after hospital discharge were not included in the analysis. Because over 75% of hemorrhagic strokes occur within 24 h of thrombolytic therapy (12), these two biases should have less impact on the observed association with hemorrhagic stroke.
No information was available on in-hospital use of HRT, so the short-term effects of HRT on stroke incidence or severity could not be assessed. Finally, HRT may be associated with early harm as seen in the HERS trial (3)and the WEST trial reviewed above, but information on the duration and formulation of HRT was not recorded in NRMI, so no comparisons between recent and long-term HRT use could be made.
Stroke is common among postmenopausal women hospitalized for AMI, occurring in about 1 in 50 patients. Among more than 114,000 such women, we found no evidence that use of HRT alters risk for ischemic or hemorrhagic stroke or stroke case fatality. Among over 13,000 women who received thrombolysis, there was no increased risk for hemorrhagic stroke among HRT users compared to non-users. Clinicians should not alter their approach to thrombolytic therapy in AMI based on HRT use.
☆ The National Registry of Myocardial Infarction-3 is supported by Genentech, Inc., South San Francisco, California. Dr. Angeja’s fellowship is in part supported by Genentech, Inc. Dr. Shlipak is supported by a Research Career Development Award from the Veterans Affairs Health Services Research and Development Service.
↵1 A complete listing of registry hospitals is available from STATProbe, Inc., Lexington, Kentucky.
- acute myocardial infarction
- blood pressure
- coronary artery bypass graft surgery
- congestive heart failure
- confidence intervals
- Heart and Estrogen/progestin Replacement Study
- hormone therapy
- National Registry of Myocardial Infarction
- odds ratio
- percutaneous transluminal coronary angioplasty
- Women’s Estrogen for Stroke Trial
- Received May 10, 2001.
- Accepted July 23, 2001.
- American College of Cardiology
- Hart R.G.,
- Pearce L.A.,
- McBride R.,
- Rothbart R.M.,
- Asinger R.W.
- Thrift A.G.,
- McNeil J.J.,
- Forbes A.,
- Donnan G.A.
- Simon J.A.,
- Hsia J.,
- Cauley J.A.,
- et al.
- Gore J.M.,
- Granger C.B.,
- Simoons M.L.,
- et al.
- Mahaffey K.W.,
- Granger C.B.,
- Sloan M.A.,
- et al.
- Every N.R.,
- Frederick P.D.,
- Robinson M.,
- Sugarman J.,
- Bowlby L.,
- Barron H.V.
- Viscoli C.M.,
- Brass M.B.,
- Kernan W.N.,
- Sarrel P.M.,
- Horwitz R.
- Gebara O.C.,
- Mittleman M.A.,
- Sutherland P.,
- et al.
- Berkowitz S.D.,
- Granger C.B.,
- Pieper K.S.,
- et al.