Sex Differences in Cardiac Arrest Survivors
Background Important sex differences in the epidemiology of sudden death and in the results of electrophysiological testing in survivors of cardiac arrest have been identified. These differences are currently poorly understood.
Methods and Results Three hundred fifty-five consecutive survivors of out-of-hospital cardiac arrest (84 women and 271 men) referred for electrophysiologically guided therapy were analyzed retrospectively for sex differences in underlying pathology and predictors of outcome. Women were significantly less likely to have underlying coronary artery disease than men (45% versus 80%) and more likely to have other forms of heart disease or structurally normal hearts (P<.0001). The mean left ventricular ejection fraction was higher in women (0.46±0.18 versus 0.41±0.18, P<.05), and women were more likely to have no inducible arrhythmia at baseline electrophysiological testing (46% versus 27%, P=.002), although when the patients were stratified by coronary artery disease status, these sex differences were no longer present. The independent predictors of outcome differed between men and women. In men, a left ventricular ejection fraction of <0.40 was the most powerful independent predictor of total (relative risk, 2.8; 95% CI, 1.6 to 5.0; P<.0001) and cardiac (relative risk, 6.3; 95% CI, 2.9 to 13.5; P<.0001) mortality. In contrast, the presence of coronary artery disease was the only independent predictor of total (relative risk, 4.5; 95% CI, 1.5 to 13.4; P=.003) and cardiac (relative risk, 4.4; 95% CI, 1.2 to 15.6; P=.012) mortality in women.
Conclusions Female survivors of cardiac arrest are less likely to have underlying coronary artery disease. The predictors of total and cardiac mortality differ between male and female survivors. Coronary artery disease status is the most important predictor in women, and impaired left ventricular function is the most important predictor in men.
Epidemiological studies1 2 3 4 and outcome data from clinical trials of acute myocardial infarction5 6 have identified sex differences in the incidence and risk factors for sudden death. Women are at decreased risk at all ages, constituting only 21% of ventricular fibrillation arrest victims in Seattle4 and 32% of sudden death victims in the Framingham study.1 Nevertheless, sudden death is still an important cause of cardiovascular mortality in women. In women, 64% of sudden deaths compared with 50% in men occur in persons without prior clinical evidence of coronary heart disease, and therefore, sudden death is often the first manifestation of heart disease.2 Efforts to prevent sudden death have been directed at the prevention of coronary artery disease, but the classic coronary risk factors do not appear to predict sudden death to the same degree in women as they do in men.1 2
There have also been reports of sex differences in the results of electrophysiological testing in survivors of cardiac arrest. Women have been reported to be less likely to have inducible ventricular arrhythmias during programmed ventricular stimulation.7 8 9 The causes of these sex differences in sudden death victims and cardiac arrest survivors are poorly understood. In an attempt to further understand these differences, we analyzed retrospectively a large referral population of 355 consecutive cardiac arrest survivors for sex differences in their underlying cardiac pathology, electrophysiological characteristics, and predictors of outcome.
The study was a retrospective analysis of 355 consecutive survivors of out-of-hospital cardiac arrest who were referred to the Cardiac Arrhythmia Service of the Massachusetts General Hospital for electrophysiologically guided therapy from January 1978 to December 1992. External electrical defibrillation was required for resuscitation in all cases. Patients were excluded if the cardiac arrest occurred during hospitalization or if subsequent evaluation revealed evidence of a new Q-wave myocardial infarction. There were 84 women and 271 men in the study population. The mean age (±SD) of the patient population was 57.3±13.7 years. Two hundred fifty-four patients (71.5%) had coronary artery disease (defined as ≥70% stenosis in one or more epicardial arteries and/or evidence of old myocardial infarction or ischemia) as their principal cardiac abnormality. The mean left ventricular ejection fraction (LVEF) was 41.8±18.6%, and 170 patients (47.9%) had an LVEF of <40%. The rhythm documented at cardiac arrest was ventricular fibrillation in 321 patients (90.4%), ventricular tachycardia in 30 patients (8.5%), and unknown in 4 patients.
After the patients had recovered from the presenting cardiac arrest, cardiac catheterization with coronary arteriography and left ventriculography was performed in all except 11 patients (3.1%). All patients underwent noninvasive assessment of left ventricular function. Three hundred thirty-three patients (94%) underwent baseline electrophysiological testing in the absence of antiarrhythmic drugs unless clinical instability made antiarrhythmic drug withdrawal impossible. In 22 patients (8.8%) with critical coronary artery stenoses and/or unstable ischemic symptoms, electrophysiological testing was deferred until after coronary revascularization. Initial electrophysiological testing was performed a mean of 3.9±6.0 days after resuscitation.
Electrophysiological studies were performed with a programmable stimulator with a constant current source delivering 2-ms rectangular pulses at fivefold diastolic threshold. The protocol for ventricular stimulation included introduction of single and double extrastimuli after pacing drive trains at a minimum of two basic cycle lengths (600 and 400 ms) and at two right ventricular sites in all patients. In studies before 1982, brief bursts of rapid ventricular pacing (5 to 10 beats at cycle lengths of 400 to 220 ms) were used. In 1982 and subsequent years, the protocol included triple ventricular extrastimuli instead of burst pacing. The end point of the stimulation protocol was the induction of sustained ventricular arrhythmia lasting more than 30 seconds or causing hemodynamic deterioration requiring intervention. Before 1982, the reproducible induction of nonsustained ventricular tachycardia (5 to 100 beats) was also used as an end point.
Coronary artery revascularization, with coronary artery bypass surgery (107 patients) or percutaneous coronary angioplasty (5 patients), was performed in 112 patients (31.5%) with evidence of significant coronary artery stenoses and reversible myocardial ischemia. Map-guided left ventricular aneurysmectomy was performed in 4 patients, and valve replacements were performed in another 4 patients.
Patients with inducible sustained ventricular arrhythmias underwent serial drug testing to identify an antiarrhythmic regimen that suppressed the induced ventricular arrhythmia. A favorable drug response was defined as the induction of fewer than 10 repetitive ventricular complexes in response to completion of the programmed stimulation protocol. Before the availability of the implantable cardioverter/defibrillator (ICD), patients were discharged on the drug regimen shown to prevent initiation of the ventricular arrhythmias induced at baseline electrophysiological study. If no drug suppressed the arrhythmia, drugs that rendered the arrhythmia more difficult to induce, slower in rate, or better tolerated were used. From 1983 on, placement of an ICD was recommended in patients who had persistently inducible sustained ventricular arrhythmias despite pharmacological and/or surgical therapy. Placement of an ICD was also recommended for patients who had no inducible sustained ventricular arrhythmia at baseline electrophysiological study and who did not have critical coronary artery disease requiring revascularization or another reversible cause for cardiac arrest.
Personal or telephone contact was made with all patients or their physicians or family members, and the circumstances of any deaths were investigated to determine probable cause. Sudden cardiac death was defined as death that occurred as a result of recurrent cardiac arrest (sudden unexpected circulatory collapse) or within 1 hour of the development of symptoms in a previously stable patient or an unwitnessed death in a patient known to be stable in the preceding 24 hours. Nonsudden cardiac death was defined as death due to progressive heart failure or recurrent myocardial infarction and/or with preceding symptoms of more than 1 hour in duration.
Baseline patient characteristics are expressed as mean±SD for continuous data and proportions for categorical data. Comparisons between men and women were made by unpaired t tests, χ2 analysis, or Fisher’s exact test as appropriate.
The effect of relevant covariates (predictors) on total mortality, cardiac mortality, and sudden death were evaluated in the entire population and in men and women separately by stepwise multivariate regression analysis using a Cox proportional-hazards model.10 Fifteen covariates (listed in Table 4⇓) were entered into the model. After initial screening of the univariate significance of the individual variables in the Cox model, relationships for which the probability value approached significance (P<.10) were investigated further within the framework of stepwise multiple Cox regression to examine their joint and relative effects on outcome (BMDP Statistical Software: P2L).11 The relative risk and 95% confidence interval for the regression parameter associated with each variable that reached significance (P<.05) in the final model were calculated. Kaplan-Meier survival curves for men and women were calculated with the program BMDP:P1L.12
There were 271 men and 84 women in the database. The clinical and electrophysiological characteristics of the men and women are presented in Table 1⇓. Several clinical features differed between the men and the women. Women were significantly less likely to have coronary disease as their underlying cardiac abnormality. There were 216 men (80%) with coronary artery disease compared with only 38 women (45%). Women were also less likely to have suffered a prior myocardial infarction (33% versus 47%). The mean LVEF was higher in the women at 0.46±0.18 compared with 0.41±0.18 in the men. The sexes did not differ with respect to age, past history of congestive heart failure, left ventricular aneurysm, antiarrhythmic drugs at cardiac arrest, or treatment with β-adrenergic antagonists.
Underlying Structural Heart Disease
Fig 1⇓ illustrates the proportion of various types of structural heart disease in the male and female cardiac arrest survivors. A large majority (80%) of the men had coronary artery disease as their principal diagnosis. The remaining 20% had primarily dilated cardiomyopathy (10%) and valvular heart disease (5%). In contrast, 55% of the women had nonischemic heart disease, including dilated cardiomyopathy (19%), valvular heart disease (13%), and coronary vasospasm (5%), which was uncommon in the men. Ten percent of women as opposed to 3% of men had structurally normal hearts. This difference in the distribution of underlying structural heart disease was significant when analyzed by χ2 analysis (P<.0001).
Electrophysiological Testing and Treatment Regimens
The results of baseline electrophysiological testing (Table 1⇑) differed significantly between men and women. Forty-six percent of the women versus 27% of the men had no inducible arrhythmia at baseline testing (P=.002). The greatest difference occurred in the subgroup with inducible sustained monomorphic ventricular tachycardia. Twenty-four percent of women versus 36% of men had this end point during programmed ventricular stimulation (P<.05). Men and women did not differ significantly in the distribution of results at predischarge electrophysiological testing and in the likelihood of antiarrhythmic drugs and/or surgery suppressing inducible ventricular arrhythmias. Similar proportions of men and women were discharged on antiarrhythmic drugs or β-adrenergic antagonists (Table 2⇓). Women were significantly less likely to undergo coronary artery bypass graft surgery (P<.05) during the hospitalization, and a higher percentage of women underwent ICD therapy, although this difference was not statistically significant.
Results Stratified by Coronary Artery Disease
Given the large sex difference in the proportion of patients with coronary artery disease, we stratified the population according to the presence or absence of underlying ischemic heart disease and again compared the men and the women in each subgroup (Table 3⇓). The men and women in the subgroup with coronary artery disease did not differ with respect to age, number of critically stenosed vessels at angiography, history of prior myocardial infarction, LVEF, or results of electrophysiological testing. The same proportion of women and men with coronary artery disease underwent coronary artery bypass surgery. In the subgroup without ischemic heart disease, women were still more likely to have no inducible arrhythmia at baseline electrophysiological testing, although this was only of borderline significance. There were no significant sex differences with respect to age and LVEF. The subgroup of patients without ischemic heart disease had a higher mean LVEF (0.49±0.19) and a lower mean age (48±17 years) than the patients with ischemic heart disease. Thus, the sex differences observed in LVEF and responses to electrophysiological testing in the entire population are due primarily to the sex difference in the prevalence of coronary artery disease.
Eighty-six deaths occurred over a median follow-up of 24.4 months: 19 deaths in women and 67 in men. The cumulative survival for the entire group at 5 years was 70% and at 10 years, 44%. The median follow-up did not differ significantly between men and women (23.5 months in men and 29.7 months in women). When analyzed by life-table survival analysis, men and women did not differ with respect to their overall survival, as seen in Fig 2⇓. Similarly, there was no sex difference in survival free of cardiac mortality or sudden death.
Predictors of Mortality
All variables analyzed as predictors of total, cardiac, and sudden death mortality are listed in Table 4⇓. The univariate significance for cardiac mortality for each of these predictors is also shown. The independent predictors from the multivariate Cox regression model and the relative risks and 95% confidence limits for the entire population are displayed in Table 5⇓. LVEF <0.40 was the most important independent predictor of total mortality, with a relative risk of 2.9 (95% CI, 1.8 to 4.6). Other independent predictors of total mortality included age (relative risk, 1.5 per decade of life; 95% CI, 1.2 to 1.8) and sustained monomorphic ventricular tachycardia at predischarge programmed ventricular stimulation (relative risk, 2.2; 95% CI, 1.4 to 3.7). ICD therapy was a negative predictor, with a relative risk of 0.5 (95% CI, 0.3 to 0.8).
LVEF <0.40 was again the most powerful predictor of cardiac mortality (relative risk, 4.6; 95% CI, 2.5 to 8.7). The presence of coronary artery disease and absence of an ICD were also independent predictors of cardiac mortality. The presence of an ICD was a powerful negative predictor of sudden death (relative risk, 0.18; 95% CI, 0.05 to 0.61), and sustained monomorphic ventricular tachycardia at predischarge programmed ventricular stimulation was a positive predictor of sudden death (relative risk, 3.6; 95% CI, 1.5 to 3.6). Of note, sex did not predict total, cardiac, or sudden death mortality in the univariate or multivariate analysis.
Sex-Specific Predictors of Outcome
Given the differences between men and women in their clinical and electrophysiological characteristics and underlying structural heart disease, we postulated that there may also be different predictors of outcome in women versus men. Table 6⇓ lists the results of the multivariate Cox model when the men and women were analyzed separately. The predictors of mortality in men for the most part resembled those in the total population. An LVEF of <0.40 was the most powerful independent predictor of cardiac and total mortality, with relative risks of 6.3 (95% CI, 2.9 to 13.5) and 2.8 (95% CI, 1.6 to 5.0), respectively. ICD therapy was the most important negative predictor of sudden death, with a relative risk of 0.05 (95% CI, 0.01 to 0.39). Coronary artery disease was not an independent predictor of cardiac mortality in the men, as it was in the total population. Discharge on antiarrhythmic drugs replaced ICD therapy and sustained monomorphic ventricular tachycardia at predischarge electrophysiological study as an independent predictor of total mortality in men. It is notable that antiarrhythmic drugs were associated with worsened overall survival, whereas ICD therapy was associated with improved survival from cardiac mortality and sudden death.
The results for women differed from those for the entire population and from those observed in men. In women, the presence of coronary artery disease was the only independent predictor in the multivariate Cox model of cardiac and total mortality, with relative risk of 4.4 (95% CI, 1.2 to 15.6) and 4.5 (95% CI, 1.5 to 13.4), respectively. Fifteen of the 19 deaths in women (79%) were in the subgroup with coronary artery disease. The smaller sample size of women is reflected in the wide CIs, and other predictors might have reached significance in a larger sample. This may explain, in part, why LVEF and ICD therapy were not found to be predictors in women as they were in men. The only predictor of sudden death in women was the presence of a left ventricular aneurysm, a marker for both a large myocardial infarction and coronary artery disease. Six of eight of the sudden deaths occurred in the patients with coronary artery disease. ICD therapy was not a predictor of sudden death in women, although the small number of sudden deaths in women limits the ability to accurately determine predictors of this outcome.
Surprisingly, LVEF was not even a univariate predictor of total mortality in women. It was a univariate predictor of cardiac mortality, but it was not as powerful as coronary artery disease and was not an independent predictor in the multivariate analysis. This finding did not change if LVEF was analyzed as a continuous variable or as a dichotomous variable with a cutoff at 0.30 versus 0.40. The independent effects of coronary artery disease and LVEF on survival in women are demonstrated in Fig 4⇓. Women were stratified by LVEF <0.40 or ≥0.40 and the presence or absence of coronary artery disease, and actuarial survival analysis was performed on the four groups. The two groups of women with coronary artery disease had a similar survival at 5 years (63% for those with an LVEF <0.40 versus 65% for those with an LVEF ≥0.40), which was worse than the 5-year survival in the two groups without coronary artery disease (100% for the low-LVEF group and 85% for the high-LVEF group).
In this referral population of 355 consecutive survivors of out-of-hospital cardiac arrest, sex differences exist in the underlying cardiac pathology and predictors of outcome. Most notably, women are a much more heterogeneous population than men. The distribution of underlying structural heart disease is not consistent with our current concept of cardiac arrest survivors.12 Female survivors of cardiac arrest are less likely to have underlying coronary artery disease and more likely to have other forms of heart disease or structurally normal hearts. This may explain, in part, why the classic risk factors for coronary artery disease do not predict sudden death as well in women in the Framingham Heart Study. Only hypertension and smoking conferred a long-term risk of sudden death in women, and to a lesser extent than in the men.1 Diabetes and hypercholesterolemia, which are strong risk factors for coronary artery disease in this population of women,13 were not associated with sudden death.1 Possible explanations for the lower incidence of sudden death in women and the lower frequency of coronary artery disease in female cardiac arrest survivors may be the lower incidence of coronary artery disease in women before age 65 years14 and/or an intrinsic protective effect against sudden death associated with being female in patients with coronary artery disease. The proportion of coronary heart disease deaths that were sudden was also lower in women (34% versus 46% in men) in the Framingham Study.1
The women in our population as a whole had higher LVEFs and were less likely to have inducible ventricular arrhythmias during programmed ventricular stimulation. This is not surprising, given the lower frequency of coronary artery disease and the larger percentage of structurally normal hearts in the women. The differences in LVEF and electrophysiological study results between men and women were no longer present once the patients were stratified by coronary artery disease status, except for the responses to electrophysiological testing in the subgroup without coronary artery disease. In this subgroup, women were less likely to have inducible sustained ventricular arrhythmias at baseline testing. This difference was of borderline statistical significance and may again be due to the higher percentage of structurally normal hearts in the women. These findings are in contrast to a smaller study8 of 13 women with coronary artery disease that found that women in this subgroup were less likely than men to have inducible ventricular arrhythmias. In our population, the results of electrophysiological testing in women with coronary artery disease were nearly identical to those observed in men.
Although total, cardiac, and sudden death mortality did not differ between men and women, the independent predictors of outcome were different. The predictors in men resembled those of the total population, which was expected, given the male predominance of the study population. One exception, however, was the presence of coronary artery disease, which was a significant predictor of cardiac mortality (P=.007) in the total population but was not a predictor in men. In contrast, the only independent predictor of cardiac mortality and total mortality in women was the presence of coronary artery disease. The women with coronary artery disease were 4.5 times more likely to die and 4.3 times more likely to die of a cardiac cause than women without coronary artery disease. Therefore, while coronary artery disease was less common in female cardiac arrest survivors, those survivors with coronary artery disease had a worse prognosis. Although the explanation for this finding is unknown, it is possible that women with coronary artery disease who present with a cardiac arrest may have a more aggressive form of the disease, which is associated with a poor prognosis. Conversely, the women without coronary artery disease had a surprisingly good prognosis (Fig 4⇑) even in the presence of impaired left ventricular function.
LVEF, which was the strongest independent predictor of total and cardiac mortality in men, did not have the same significance in the women. LVEF was not a significant univariate predictor of total mortality and was not an independent predictor of cardiac mortality in women once coronary artery disease status was accounted for. It is unclear why LVEF did not predict mortality to the same degree in female cardiac arrest survivors. Two studies5 15 have brought into question the importance of LVEF in prognosis after myocardial infarction in women. In the MILIS study, women were found to have a worse prognosis after myocardial infarction despite having had higher LVEFs at hospital discharge, suggesting that the prognostic information derived from this variable may differ between men and women.15 In a considerably larger population of post–myocardial infarction patients that included 538 women, Dittrich et al5 also found that the women had higher LVEFs. That study examined prognostic variables in men and women separately and found that female survivors and nonsurvivors were equally likely to have an LVEF of <0.40. The male nonsurvivors, however, were significantly more likely to have an LVEF of <0.40 than the survivors. Another study6 that examined this issue found that an LVEF of <0.40 after myocardial infarction was predictive of mortality in women and men but that frequent ventricular premature beats, which were a predictor in men, were not predictive of outcome in women.
Finally, ICD therapy, which was a negative predictor of cardiac mortality and a very powerful negative predictor of sudden death in men, was not a predictor of either end point in women. This may be due, in part, to the smaller number of each of these end points in the women (only 8 sudden deaths) and therefore insufficient power to detect ICD therapy as a predictor rather than a true sex difference. However, if the strength of the association in men (relative risk of 0.05) had also been present in women, it probably would have been detected.
The primary limitations of this study are its retrospective design, referral-based population, and smaller number of women in the study. Despite the nonrandom allocation of treatment, it is unlikely that significant bias occurred in the treatment of men versus women, since they did not differ significantly with respect to the major treatment modalities (Table 2⇑), with the exception of coronary artery bypass surgery, which was explained by the lower prevalence of coronary artery disease in women (Table 3⇑). This referral-based population may not be representative of the general population of sudden death victims or cardiac arrest survivors owing to referral bias. It is unlikely that women were selectively not referred for electrophysiological testing, given the similar percentage of women in this study (23.6%) compared with a population-based cohort in Seattle (21%).4 The mean age of this population was slightly lower at 57.3±13.7 versus 63.8±12.2 years in the survivors of out-of-hospital ventricular fibrillation arrest in Seattle. There may be referral bias against older patients, which may decrease the amount of coronary artery disease in the referral population, especially in women, and may have influenced our results. Finally, the small number of women (84) compared with men (271) limits our ability to detect all the independent predictors of outcome in women. However, the number is sufficiently large to provide information about the relative importance of the major predictors of outcome in women.
Significant differences exist in the underlying cardiac pathology of male and female cardiac arrest survivors. Women are less likely to have coronary artery disease; therefore, efforts directed at preventing coronary artery disease may not have the same impact on the rate of sudden death in women as in men. Further understanding of the mechanisms involved in cardiac arrests in women with nonischemic heart disease and with structurally normal hearts would advance our knowledge about sudden death considerably. Women were not routinely tested for coronary vasospasm, and provocative tests for long-QT syndrome were not routinely done. Given the greater proportion of women with these diagnoses, provocative testing may be worthwhile in women who survive a cardiac arrest. Various psychosocial factors3 may also play a role. Despite these differences in underlying pathology, the finding that men and women have similar total, cardiac, and sudden death mortality rates with standard12 electrophysiologically guided treatment regimens is reassuring.
This study contains important prognostic information that differs from our current understanding of cardiac arrest survivors. Female survivors who have coronary artery disease are more likely to die and to die of a cardiac cause independent of their LVEF. This observation raises the possibility that more aggressive use of revascularization and pharmacological therapy for ischemia may be appropriate in female survivors of cardiac arrest. Also, an LVEF ≥0.40 may not be as reassuring in women as it is in men. Alternatively, the group of women without ischemic heart disease have a better prognosis than previously thought. Finally, this study underscores the importance of examining women separately in studies of heart disease whenever possible as well as the risks of generalizing findings from a predominantly male population to women.
Reprint requests to Jeremy N. Ruskin, MD, Cardiac Arrhythmia Service, Massachusetts General Hospital, 32 Fruit St, Boston, MA 02114.
- Received June 1, 1995.
- Revision received October 19, 1995.
- Accepted October 23, 1995.
- Copyright © 1996 by American Heart Association
Cupples AL, Gagnon DR, Kannel WB. Long- and short-term risk of sudden coronary death. Circulation. 1992;85(suppl I):I-11-I-18.
Kannel WB, Schatzkin A. Sudden death: lessons from subsets in population studies. J Am Coll Cardiol. 1985;5:141B-149B.
Cobb LA, Weaver D, Fahrenbruch CE, Hallstrom AP, Copass MK. Community-based interventions for sudden cardiac death. Circulation. 1992;85(suppl I):I-98-I-102.
Moss AJ, Carleen E, and the Multicenter Postinfarction Research Group. Gender differences in the mortality risk associated with ventricular arrhythmias after myocardial infarction. In: Eaker ED, Packard B, Wenger NK, Clarkson TB, Tyroler HA, eds. Coronary Heart Disease in Women. Bethesda, Md: National Heart, Lung, and Blood Institute, National Institutes of Health; 1987:204-214.
Cox DR. Regression models and life-tables. J R Stat Soc B. 1972;34:187-220.
Dixon WJ, ed. BMDP Statistical Software Manual. San Francisco, Calif: University of California Press; 1992:825-866.
Eaker ED, Castelli WP. Coronary heart disease and its risk factors among women in the Framingham Study. In: Eaker ED, Packard B, Wenger NK, Clarkson TB, Tyroler HA, eds. Coronary Heart Disease in Women. Bethesda, Md: National Heart, Lung, and Blood Institute, National Institutes of Health; 1987.
Special Writing Group: Eaker ED, Chesebro JH, Sacks FM, Wenger NK, Whisnant JP, Winston M. AHA Medical/Scientific Statement: Special Report: cardiovascular disease in women. Circulation. 1993;88:1999-2009.
Tofler GH, Stone PH, Muller JE, Willich SN, Davis VG, Poole WK, Strauss HW, Willerson JT, Jaffe AS, Robertson T, Passamani E, and the MILIS Study Group. Effects of gender and race on prognosis after myocardial infarction: adverse prognosis in women, especially black women. J Am Coll Cardiol. 1987;9:473-482.