Author + information
- Received September 24, 1999
- Revision received January 20, 2000
- Accepted March 29, 2000
- Published online August 1, 2000.
- Steven E Reis, MD, FACC∗,*,
- Richard Holubkov, PhD∗,
- James B Young, MD, FACC†,
- B.G White, PhD‡,
- Jay N Cohn, MD, FACC§ and
- Arthur M Feldman, MD, PhD, FACC∗
- ↵*Reprint requests and correspondence: Dr. Steven E. Reis, Cardiovascular Institute, University of Pittsburgh Medical Center, 200 Lothrop Street, Pittsburgh, Pennsylvania 15213
This study sought to evaluate the effects of postmenopausal estrogen use on mortality in aging women with congestive heart failure (CHF).
The age-related increase in CHF mortality in women may be related to a menopause-associated increased incidence of coronary artery disease. In addition to inhibiting coronary atherosclerosis, estrogen may also have protective effects on cardiac myocytes independent of the coronary vasculature. We hypothesized that estrogen use is associated with improved survival in elderly women with CHF.
Associations between survival, estrogen use and patient characteristics were assessed in 1,134 women who were at least 50 years of age, had CHF and left ventricular ejection fraction (EF) ≤30% and were enrolled in one of three clinical trials of vesnarinone.
All-cause 12-month mortality was 15.0% among the 237 estrogen users versus 27.1% among the 897 estrogen nonusers (p = 0.004 for unadjusted comparison of survival). Similar results were observed for cardiac mortality. Regression analysis demonstrated that estrogen use was independently associated with improved survival (relative risk of mortality = 0.68, 95% confidence interval 0.48 to 0.96, p = 0.03). Advanced age, low EF, New York Heart Association class IV CHF, Caucasian race and abnormal serum creatinine, sodium, potassium and transaminase were independently associated with increased mortality.
Estrogen use among older women with CHF is associated with decreased overall and cardiac mortality.
Congestive heart failure (CHF) represents the leading hospital discharge diagnosis for the elderly. Because of limited inclusion of women, epidemiologic studies and clinical trials have failed to provide sufficient information about the outcome of older women with CHF and the influence of gender-specific cardiovascular physiology on CHF mortality. Although the age-dependent increased mortality in CHF patients may be related to an increased incidence of coronary artery disease (CAD) and ischemic cardiomyopathy, the effects of menopause-related estrogen loss on CHF in women are not known. Recent reports of gender-related differences in cardiovascular physiology provide speculative evidence that estrogen may have protective effects on the myocardium (1,2). We hypothesized that estrogen use among older women with CHF is associated with improved outcome.
The influence of estrogen use on CHF outcome among aging women was studied using a database of 1,134 women with symptomatic CHF over the age of 50 years who were enrolled in the:
1. Multicenter Vesnarinone Study, a randomized, double-blind, placebo-controlled trial of 60 and 120 mg of vesnarinone (Otsuka Pharmaceutical Company Ltd., Osaka, Japan; n = 54) (3);
2. Randomized Multiple Dose Study of the Chronic Administration of Vesnarinone, a randomized, double-blind study of 30 and 60 mg of vesnarinone (n = 323); or
3. Vesnarinone Trial (VesT), a randomized, double-blind, placebo-controlled study of 30 and 60 mg of vesnarinone (n = 757) (4).
The entry criteria for all studies were similar; subjects were at least 18 years of age, had symptomatic CHF in spite of traditional therapy and had a left ventricular ejection fraction (EF) of ≤30% by radionuclide ventriculography performed within two weeks of randomization. Subjects were excluded if they were women of childbearing potential; developed CHF within one year of childbirth; or had a history of myocardial infarction or cardiac surgery within the prior three months, severe or unstable angina, a reversible etiology for their myocardial failure, hemodynamically significant pericardial disease, ventricular fibrillation, implantible defibrillator, significant aortic or mitral stenosis or a life-limiting comorbidity or lupus. Additional exclusion criteria included concomitant therapy with ticlopidine, clozapine, encainide, flecainide, tocainide, use of intravenous inotropes at the time of screening or a history of agranulocytosis or drug-induced neutropenia.
Estrogen users were taking oral or transdermal estrogen at study entry alone or in combination with a progestin. Differences in distributions of continuous and ordered categorical variables between estrogen users and nonusers were assessed using the Kruskal-Wallis test; distributions of two-level and unordered categorical variables were compared using the Fisher exact test.
Median follow-up was 309 days. Rates of all-cause and cardiac mortality (i.e., sudden death, CHF or myocardial infarction) were calculated using the Kaplan-Meier method, with transplanted patients censored at the time of transplantation. For the 54 women in the Multicenter Vesnarinone Study, the three deaths were not classified by cardiac versus noncardiac, and, thus, this entire cohort was excluded from the cardiac mortality analyses. The log-rank test was used to compare unadjusted survival curves between estrogen user and nonuser groups.
The primary study outcome was the relative risk of mortality for estrogen users versus nonusers, adjusted for the effects of the study drug and other factors. To assess the associations of patient characteristics with mortality, univariate and multivariate Cox proportional hazards regression was used. Although drug dosage, specific trial, and all available clinical factors were used in the reported multivariate model (regardless of statistical significance), backwards elimination of nonsignificant terms did not affect the magnitude or significance of the estrogen use indicator.
Of the 1,134 women, 897 were estrogen nonusers, and 237 were estrogen users. Table 1 shows that estrogen nonusers were four years older than users and more likely to be non-Caucasian. The statistically significant differences observed between groups in EF, creatinine, potassium, sodium and aspartate aminotransferase (AST) were of a small magnitude.
Influence of estrogen on mortality
Figure 1 presents Kaplan-Meier survival curves by estrogen use. Unadjusted 12-month all-cause mortality rates for estrogen nonusers and users were 27.1% and 15.0%, respectively (p = 0.004). Similar analysis demonstrated that 12-month cardiac mortality (sudden death, CHF or myocardial infarction) was 25.6% and 14.0% for estrogen nonusers and users, respectively (p = 0.007). These findings were consistent within each of the vesnarinone studies. Analyses stratified for type of estrogen therapy demonstrated that the 186 unopposed estrogen users had a significant survival benefit compared with the 897 hormone nonusers (relative risk [RR] for death = 0.57, p = 0.004). ⇓ Survival for the 51 combined estrogen and progestin users was intermediate between that of unopposed estrogen users and hormone nonusers. However, we found no statistically significant difference in 12-month survival between women using unopposed estrogen and those using combination estrogen and progestin therapy (RR = 0.75, p = 0.42; Fig. 2). There was also no statistically significant difference in survival between combination estrogen plus progestin users and hormone nonusers (RR = 0.77, p = 0.41). Table 2 shows all deaths reported during follow-up (median 323 days) for the 1,080 women participating in trials in which cause of death was identified. These data demonstrate that, regardless of hormone use or hormone type, almost all deaths were classified as cardiac deaths, the majority of which were either sudden death or attributable to worsening CHF.
Univariate Cox regression analysis demonstrated that estrogen users had a significantly lower all-cause mortality than nonusers (Table 3; RR for users vs. nonusers: 0.61, p = 0.004). Other univariate predictors of death were age, EF, creatinine, potassium, sodium, AST, Caucasian race and New York Heart Association (NYHA) class IV CHF. There was a trend for increased mortality among women assigned to vesnarinone compared with placebo. After adjustment for these factors, estrogen use remained independently associated with lower mortality (adjusted RR: 0.68, p = 0.03). No evidence of an interaction between vesnarinone dose and estrogen use was found in the mortality analysis.
Mortality effects of estrogen in ischemic cardiomyopathy
Because estrogen has antiatherosclerotic properties, subgroup analysis was performed to evaluate its effects in women with and without CAD. Among the 521 women with clinically defined ischemic cardiomyopathy (i.e., abnormal coronary angiography or stress test or past myocardial infarction), there was a trend toward lower 12-month mortality in estrogen users (18.6%) versus nonusers (29.7%). A similar trend was seen among 596 women with idiopathic (i.e., nonischemic) cardiomyopathy (12-month mortality: 12.8% vs. 24.5%, estrogen users vs. nonusers). Therefore, estrogen use appears to be associated with lower mortality regardless of the etiology of cardiomyopathy.
Previous studies demonstrate that CHF is associated with higher mortality among the aging (5). In women, this has been attributed to an age-related increased incidence of ischemic cardiomyopathy. However, the clinical and physiologic effects of the absence of estrogen in older women with CHF have not been described. Our results suggest that, among aging women, estrogen users have an age-independent lower mortality than nonusers unrelated to the etiology of their cardiomyopathy.
Possible physiologic mechanisms for estrogen’s protective effects
Several mechanisms could account for beneficial effects of estrogen in CHF. First, estrogen has in vivo vasodilator properties (6,7). Because abnormal peripheral artery vasoconstriction is found in CHF patients and some vasodilators decrease CHF mortality, afterload reduction may provide a mechanism for the observation that postmenopausal estrogen use is associated with improved CHF outcome. An alternative hypothesis comes from reports that estrogen deficiency-induced osteoporosis is mediated by TNF alpha and that the menopause-associated increase in monocyte production of proinflammatory cytokines could be abrogated by estrogen (8,9). Proinflammatory cytokines are re-expressed by failing cardiac myocytes, elevated in plasma of CHF patients and are associated with increased CHF severity (10–12). Thus, estrogen-induced inhibition of cytokines is another potential mechanism for the observed association between postmenopausal estrogen use and lower mortality in women with CHF.
Another explanation is that estrogen may have decreased the incidence of fatal myocardial infarction by virtue of its atheroprotective effects. Although this is plausible in those with pre-existing CAD, it is an unlikely explanation in women without CAD. Our data show a trend toward improved short-term survival among estrogen users with an idiopathic (i.e., nonischemic) cardiomyopathy. Therefore, the observation that estrogen use is associated with lower mortality in older women with CHF may be independent of atherosclerosis. Interestingly, our data also show a trend towards estrogen-associated decreased mortality among women with ischemic cardiomyopathy. This finding appears to be contrary to the only reported large-scale prospective randomized, placebo-controlled trial of postmenopausal hormones, the Heart and Estrogen/progestin Replacement Study (HERS), which reported a trend towards a hormone-associated increased 12-month cardiac mortality in postmenopausal women with pre-existing CAD. However, few (9.5%) women in HERS had CHF before randomization, and a similar small percentage (8.7%) of HERS women were hospitalized for CHF during long-term follow-up (i.e., mean 4.1 years) (13). These demographics differ from this study, in which all subjects had significant symptomatic heart failure and less than half had CAD. Therefore, the discrepancy in the effects of hormone use between HERS and this study may be related to differences in patient population and not necessarily due to differences in study design. Another explanation for the discrepancy between studies is that the majority of hormone users in this study used unopposed estrogen, while HERS randomized women to a combined estrogen and progestin. Indeed, our results suggest a nonsignificant trend toward a progestin-associated attenuation of estrogen’s beneficial mortality effect in CHF. This is consistent with previously proposed explanations of the negative HERS findings; that is, progestin therapy might mitigate the cardioprotective effects of estrogen (14).
Our exploratory post hoc analysis pooled data from three clinical trials to obtain a sufficient population to test our hypothesis. Although this may be perceived as a limitation, the trials had similar entry criteria and comparable subject demographics with respect to clinically important factors, and multivariable modeling controlled for covariables, including a separate “study effect.” Another limitation was that the majority of women were receiving vesnarinone. However, after controlling for vesnarinone use and dosage, we found that estrogen use remained independently associated with improved survival. Finally, we could not control for the possibility that hormone users may have healthier lifestyles than nonusers (15), because socioeconomic status and other lifestyle measures were unavailable in the database. However, we hypothesize that any potential “lifestyle” bias between estrogen users and nonusers in our study is likely to be relatively small in magnitude because all subjects volunteered to participate in a randomized clinical trial of vesnarinone. Previous studies have demonstrated that subjects voluntarily participating in clinical trials tend to be highly conscious of their health. This awareness, together with the close monitoring that these patients receive while participating in a clinical trial, may lead to improved outcome among trial participants regardless of treatment assignment. In addition, estrogen use remained significantly associated with improved survival after we controlled for disease severity (NYHA class), which is likely to be a stronger predictor of CHF survival than is lifestyle. Nevertheless, the potential “lifestyle” bias suggests caution in the extrapolation of our findings to more general populations that may substantially differ from our study patients in terms of overall health or other important factors.
These results suggest that in older women with CHF, estrogen use is associated with lower overall and cardiac mortality that is independent of age, vesnarinone use, EF and CHF severity. This observation from a retrospective analysis of a large database of women with CHF generates a new hypothesis that needs to be tested by prospective large-scale randomized clinical trials.
☆ This study was sponsored, in part, by Otsuka America Pharmaceutical.
- serum aspartate aminotransferase
- coronary artery disease
- congestive heart failure
- ejection fraction
- Heart and Estrogen/progestin Replacement Study
- New York Heart Association
- relative risk
- Received September 24, 1999.
- Revision received January 20, 2000.
- Accepted March 29, 2000.
- American College of Cardiology
- Olivetti G,
- Giordano G,
- Corradi D,
- et al.
- Sudhir K,
- Chou T.M,
- Mullen W.L,
- et al.
- Reis S.E,
- Gloth S.T,
- Blumenthal R.S,
- et al.
- Pacifici R,
- Brown C,
- Puscheck E,
- et al.
- Torre-Amione G,
- Kapadia S,
- Benedict C,
- Oral H,
- Young J.B,
- Mann D.L
- Kubota T,
- McTiernan C.F,
- Frye C.S,
- et al.