Author + information
- Received August 15, 1996
- Revision received February 14, 1997
- Accepted February 26, 1997
- Published online June 1, 1997.
- Boris E Coronado, MDA,
- John L Griffith, PhDA,
- Joni R Beshansky, RN, MPHA and
- Harry P Selker, MD, MSPHA,* ()
- ↵*Dr. Harry P. Selker, Division of Clinical Care Research, New England Medical Center, Box 63, 750 Washington Street, Boston, Massachusetts 02111.
Objectives. This study sought to determine gender differences in hospital mortality in patients with acute cardiac ischemia.
Background. It is unclear why women experience higher mortality from acute myocardial infarction (AMI) than men and whether this applies to all patients with acute ischemia.
Methods. We analyzed data from a prospective multicenter study involving patients presenting to the emergency department (ED) with symptoms suggestive of acute ischemia.
Results. Of 10,783 patients, 5,221 (48.4%) were women. Mean age was 60.5 years for women and 56.9 for men (p < 0.001). Women had more hypertension (54.6% vs. 45.9%, p < 0.001) and diabetes (23.3% vs. 17.0%, p < 0.001) than men but fewer previous AMIs (21.1% vs. 28.9%, p < 0.001). Acute ischemia was confirmed in 1,090 women (20.8%) and 1,451 men (26.1%, p < 0.001), including AMI in 322 women (6.2%) and 572 men (10.3%, p < 0.001). Women with an AMI were in a higher Killip class than men: class I in 60.3% versus 72.2%, class II in 19.3% versus 16%, class III in 15.5% versus 8.7% and class IV in 5% versus 3.1%, respectively (p = 0.001). There was no significant difference in mortality from acute ischemia between genders (4.0% vs. 3.5%, p = 0.6), but there was a trend for higher AMI mortality in women (10.3% vs. 7.4%, p = 0.1). After controlling for age, diabetes, heart failure and presenting blood pressure, gender did not predict mortality from acute ischemia (odds ratio 0.9, 95% confidence interval 0.5 to 1.4, p = 0.5).
Conclusions. Among patients presenting to the ED with acute cardiac ischemia, gender does not appear to be an independent predictor of hospital mortality. The trend for higher mortality in women from AMI can be explained by their older age, greater frequency of diabetes and higher Killip class on presentation.
(J Am Coll Cardiol 1997;29:1490–6)
For many years, the exclusion of women from early clinical trials and the commonly held perception that cardiac ischemia occurred primarily in men prevented the identification of gender differences in the manifestations and clinical course of coronary artery disease (). However, with the recognition of cardiovascular disease as the number one cause of death in women in the United States ([2, 3]), considerable interest has grown in the research community in explaining gender differences in mortality after acute myocardial infarction (AMI). According to most studies, women’s mortality rates from AMI are higher than those seen in men, due in part to women’s older age at presentation and higher prevalence of comorbid conditions such as diabetes ([3–15]). However, the relative contribution of these and other factors are matters of ongoing controversy, which is compounded by the reported wide variations in methodology among published studies ().
The study of gender differences in mortality from acute cardiac ischemia should not be restricted to cases of AMI. Focusing exclusively on cases in which the ischemic process has culminated in necrosis may fail to account for coronary atherosclerotic plaque rupture and thrombus formation as the precipitants of other acute coronary syndromes ([17, 18]). Acute coronary insufficiency is a dynamic process where the occurrence and extent of myocardial necrosis depends, among other factors, on the balance between the duration of coronary occlusion, extent of collateral circulation and restoration of vessel patency. These interactions are highlighted by the proven effectiveness of pharmacologic and mechanical reperfusion therapies ([19–22]).
In the context of a clinical study, this approach must also include all patients who present with any symptom suggestive of cardiac ischemia and not be limited to those with chest pain, a symptom absent in up to 25% of patients with AMI ([23–26]). In fact, patients more likely to experience ischemia without chest pain, such as diabetic subjects and the elderly, represent a group at excess risk for fatal and nonfatal complications ([25–31]). Thus, broader inclusion criteria become particularly important in studying gender differences in mortality when women’s higher prevalence of diabetes and older age at presentation are taken into account ([1, 32]).
In the present study, to ascertain gender differences in hospital mortality from acute cardiac ischemia, we analyzed prospectively collected data on patients presenting to the emergency department (ED) with chest pain or other symptoms suggestive of acute ischemia, with particular attention to the influence of presenting clinical characteristics and comorbid conditions.
We analyzed data from the Acute Cardiac Ischemia Time Insensitive Predictive Instrument (ACI-TIPI) clinical trial (), a multicenter study where data were collected prospectively on 10,783 patients at 10 hospitals from April through December 1993.
1.1 Participating Hospitals.
The study sites included 10 different hospitals: Baystate Medical Center, Springfield, Massachusetts; Boston City Hospital, Boston University Medical Center, Boston, Massachusetts; Medical College of Virginia, Richmond, Virginia; Medical College of Wisconsin, Milwaukee, Wisconsin; New England Medical Center, Boston, Massachusetts; Newton-Wellesley Hospital, Newton, Massachusetts; Rhode Island Hospital, Providence, Rhode Island; University of Cincinnati Medical Center, Cincinnati, Ohio; and University of North Carolina Hospitals, Chapel Hill, North Carolina. These hospitals represented a range of settings and practice types, serving patients from different sociodemographic and ethnic backgrounds. Hospital sizes ranged from 350 to 1,019 beds, with yearly ED visits varying from 27,000 to 115,000.
The study included all consenting adults ≥30 years old who presented to the ED at any of the participating hospitals with symptoms suggestive of acute cardiac ischemia as described in the Imminent Myocardial Infarction Rotterdam (IMIR) study (). These symptoms included chest pain, chest pressure, left arm pain, jaw pain, upper abdominal pain, dizziness, nausea, vomiting or dyspnea. Patients in whom cardiac ischemia could not be excluded in the ED were also enrolled in the study, as was the case of those with dysnea and congestive heart failure.
Patients with chest pain of clear traumatic origin or other obvious noncardiac etiology were excluded, as were those transferred from other institutions and those in whom consent was declined or unobtainable by virtue of mental illness or language barrier. Additionally, patients <30 years old in whom cocaine abuse was suspected, although included in the original trial, were excluded from the present analysis because they were considered to represent a group of subjects with unique clinical and epidemiologic features. The inclusion rate for the trial was 92% for both women and men.
1.3 Data Collection.
Data were collected at the time of presentation to the ED and during the hospital period through direct patient interview and medical record review. Demographic and clinical information were recorded, including initial and 24-h follow-up electrocardiograms (ECGs) and serum creatine kinase, MB fraction (CK-MB), determinations.
1.4 Data Analysis.
Final diagnoses were assigned after review of the patients’ medical records and data collection forms by participating physician investigators at each study site. Interrater reliability was confirmed by 10% overreading of cases at the Coordinating Center.
Cardiac diagnoses were considered ischemic or nonischemic, and cardiac ischemia was in turn classified into AMI and unstable angina. AMI was confirmed according to the World Health Organization criteria on the basis of clinical presentation, diagnostic ECG changes and elevations in serum CK-MB levels () and further grouped according to Killip classification (). A diagnosis of unstable angina was based on clinical presentation, ECG changes and other confirmatory tests performed during or after the hospital period. Other diagnoses were assigned in a similar fashion after appropriate diagnostic investigation and thorough review of the patient’s medical record.
1.5 Statistical Methods.
All hypothesis tests were two-tailed. Analyses of gender differences in baseline demographic and clinical characteristics used the Student ttest for continuous variables and chi-square tests for dichotomous variables. Hospital mortality was defined as deaths occurring in the hospital at any moment from arrival to the ED through hospital discharge. Deaths occurring in the ED were counted as hospital deaths.
Unadjusted odds ratios and 95% confidence intervals were obtained for selected clinical factors judged to be associated with the outcome of interest. These variables were gender, age, history of diabetes, history of hypertension, history of previous infarction, congestive heart failure (defined as pulmonary congestion or edema, or both, on the initial chest roentgenogram obtained in the ED), blood pressure on presentation (expressed as a previously validated function for hospital mortality from acute cardiac ischemia with linear and quadratic components []) and ethnicity. The unadjusted level of significance of this last variable was determined by chi-square distribution, with ethnic groups classified as white, black, hispanic or “other.”
A multivariable logistic regression model (SAS version 6.1) was constructed to predict hospital mortality among patients with acute cardiac ischemia. Gender was forced into the model as a primary study variable (univariate p = 0.4), with statistical adjustment made for hospital procedures and for unmeasured effects across study sites. Adjustment for hospital procedures (thrombolysis, heart catheterization, coronary angioplasty, coronary artery bypass graft surgery) was made by forcing into the model interventions that exhibited unadjusted statistical significance for the outcome of interest (p < 0.15). Study sites with similar unadjusted p values were also forced into the model.
Clinical variables with statistical significance by univariate analysis (p < 0.15) were chosen as candidates for stepwise selection into the model. These variables included age, history of diabetes, congestive heart failure on the initial chest radiograph and blood pressure.
A second logistic regression model was constructed to predict hospital mortality among the subgroup of patients with AMI, following the same procedure. Age, diabetes, congestive heart failure on presentation and blood pressure were again chosen for stepwise selection based on the same level of unadjusted statistical significance.
2.1 Presenting Patient Characteristics.
There were important clinical and demographic differences at baseline between women and men. As shown in Table 1, both genders were represented equally in the study sample, with 48.4% women and 51.6% men. Additionally, women were significantly older as a group and more frequently had a history of diabetes and hypertension. Men, however, more commonly elicited a history of previous AMI.
There was substantial diversity in ethnic distribution between genders (Table 1). Although whites comprised the majority in both genders, women had a greater proportion of blacks than men. Additionally, comorbidities were distributed differently between ethnic groups, with a history of hypertension elicited in 58.4% of blacks, 46% of whites, 49% of hispanics, and 41% of patients from other ethnic backgrounds (chi-square, p = 0.001). Diabetes was present in 22.3% of blacks, 21.2% of hispanics, 19% of whites, and 17% other groups (p = 0.001), and previous AMI in 29.7% of whites, 19% of hispanics, 17.8% of blacks, and 15.7% other groups (p = 0.001).
There were small yet statistically significant differences in chief complaint between women and men (chi-square distribution, p = 0.01). For instance, 67% of women presented with chest pain, as opposed to 70% of men. Conversely, dyspnea was present in 21% of women and in 19% of men. Other symptoms, such as abdominal pain, diaphoresis, nausea, vomiting, dizziness and back pain were distributed similarly between genders.
2.2 Final Diagnosis.
As seen in Table 2, acute cardiac ischemia was diagnosed less frequently in women than in men (20.8% vs. 26.1%). Both types of acute ischemia were more frequent in men, reaching statistical significance for AMI and borderline significance for unstable angina.
Nonischemic heart disease, which included conditions such as heart failure, arrhythmias and acute pericarditis, among others, was found at similar rates in both genders (36.3% in women and 36.4% in men). However, noncardiac diagnoses were significantly more common in women (42.2% vs. 36.6%) and included a wide range of conditions, such as respiratory infections and musculoskeletal disorders.
Although AMIs were more common in men, women had significantly more severe infarctions as reflected by higher Killip classes (Table 3). Women had fewer Killip class I AMIs than men and more Killip class II, III and IV AMIs. Despite these differences, peak CK-MB serum levels among AMI patients were significantly lower in women than in men (152.5 vs. 219 IU/ml, p = 0.02).
Among all patients with cardiac ischemia, women received thrombolytic agents less frequently than men and underwent fewer invasive procedures, including operation (Table 4). All these comparisons reached statistical significance, except for coronary artery bypass graft surgery among patients with AMI, which was performed at a similar rate in both genders.
2.3 Hospital Mortality.
There was no significant difference in hospital mortality between women and men among all patients with acute cardiac ischemia or among those with unstable angina. However, among patients with AMI, there was a trend for higher mortality in women (p = 0.1). No significant differences were found when gender comparisons were made within each Killip class (Table 5), but there was a statistical trend for higher female mortality in patients in Killip class III AMI (p = 0.1).
After determining unadjusted odds ratios for selected clinical variables (Table 6), we constructed a logistic regression model to identify independent predictors of hospital mortality due to acute cardiac ischemia (n = 2,541) after controlling for hospital interventions (Table 4) and variations across study sites. Gender was forced into the model as the variable of interest, and the remaining variables with unadjusted statistical significance (p < 0.15) were chosen as candidates for stepwise selection. These included age, diabetes mellitus, blood pressure and congestive heart failure on the initial chest radiograph (Table 7). Of note, the lack of statistical significance of gender by univariate analysis (p = 0.4) persisted after multivariable testing (p = 0.5). In the multivariable model, factors identified as significant predictors of hospital mortality from acute cardiac ischemia included age, diabetes, presenting systolic blood pressure and radiographic evidence of heart failure. As indicated in the Methods section, systolic blood pressure was entered into the model using a previously validated variable with linear and quadratic components (), and, although statistically significant, no odds ratio is reported because of its nonlinearity. The receiver operating characteristic curve of this regression model was 0.80.
To more specifically test the univariate trend for higher hospital mortality from AMI for women, we constructed a second logistic regression model predicting mortality using only patients with proved AMI (n = 894) (Table 7). The same variables were again chosen for stepwise selection based on their univariate p values. Ultimately, the same factors that predicted mortality in all patients with acute cardiac ischemia also predicted hospital death in patients with AMI (age, diabetes, congestive heart failure, blood pressure), and the statistical trend for higher mortality in women noted by univariate analysis (p = 0.1) was no longer present in this multivariable analysis (p = 0.9). The receiver operating characteristic curve area for this model was 0.77.
The first regression model was subsequently applied to those patients with unstable angina with the purpose of testing the validity of the identified predictors of mortality in this subgroup. All variables identified as independent predictors of hospital mortality in the logistic regression models also exhibited positive regression coefficients, with a receiver operating characteristic curve area of 0.85.
3.1 Epidemiologic and Methodologic Issues.
For many years, ischemic heart disease was considered a condition that primarily affected men, exhibiting a benign course when present in women. This notion was reinforced by data from the Framingham study indicating that women more commonly experience angina pectoris as the first manifestation of coronary disease, leading to AMI less frequently than in men ([38, 39]). However, concern about ischemic heart disease in women as a public health problem became apparent after it became recognized as the leading cause of death in the United States in both genders ([2, 3]) and after the total number of cardiovascular deaths in women was noted to exceed those in men ().
Epidemiologic and clinical gender differences may contribute to the different manifestations and outcomes in women and men. Although ischemic heart disease is more prevalent in men, the incidence of coronary events in women increases after menopause, so that by age 75 the incidence of AMI in both genders becomes similar (). Furthermore, certain conditions known to be more common in women, such as diabetes, not only contribute to the development of atherosclerosis, but also are independently associated with adverse acute and long-term outcomes.
The interpretation of available data in this regard has been difficult because of variability in the selection of clinical factors and differences in methodology across studies. In a recent review by Vaccarino et al. () on gender differences in mortality after AMI, formal meta-analysis was not feasable because of the heterogeneity across studies in the type and number of factors analyzed and in the methods used for covariate adjustment. Furthermore, whereas some studies were retrospective ([8, 9, 41]), others only included patients with proved AMI in critical care settings ([7, 8, 15]) or excluded patients with advanced age or with other contraindications for thrombolysis ([9, 41]).
Such differences in methodology have led to wide variation in the interpretation of reported results. Most studies have reported higher unadjusted mortality rates after AMI in women ([4–7, 9–15, 42–44]); however, some studies report that gender differences decrease or disappear after adjustment for age alone ([4, 5, 8, 15, 42–45]), whereas others report that these differences persist ([7–10]). Furthermore, in other studies, mortality rates remained higher in women than in men despite additional adjustment for diabetes, congestive heart failure on admission, previous infarction, hypertension, infarct location and peak CK levels ([10, 14]). Although variable inclusion criteria and heterogeneity in covariate adjustment may account for these discrepancies (), other factors have been implicated in explaining gender differences in mortality after AMI, such as women’s longer delays in seeking medical treatment ([9, 46]) and lower access to invasive procedures ([15–17, 43, 47, 48]). However, this latter finding is in contrast to our study, where women’s lower access to invasive procedures (Table 4) did not significantly affect mortality even after multivariable adjustment.
Patient inclusion is a key methodologic issue in studying acute cardiac ischemia given these discrepancies. Many studies of acute mortality from acute cardiac ischemia do not include all patients, regardless of symptoms, who present to the ED. A study sample relevant to the practicing physician must include all patients presenting with any symptom suggestive of acute cardiac ischemia, as was the case in the present study. This is particularly important considering that 13% to 25% of patients with AMI do not experience chest pain ([23–26]).
We did not restrict the present study to patients with AMI. From the clinician’s perspective, it is often unclear whether myocardial necrosis has already occurred at the time of presentation or whether the ischemic process has not reached that stage. Furthermore, in many instances prompt intervention may relieve coronary occlusion, thereby limiting or even averting permanent myocardial damage.
3.2 Predictors of Mortality.
In the present study we found that the same clinical factors that predicted mortality from acute cardiac ischemia also predicted mortality in the subgroup of patients with AMI. These findings are most likely due to the low mortality rates observed in patients with angina pectoris, with the effect of mortality from AMI driving the regression model for the larger group of patients with acute cardiac ischemia. However, the regression model performed well when applied to only patients with unstable angina, indicating that the same factors predictive of hospital death for all patients with acute ischemia are also applicable to this subgroup.
Among patients with AMI, women experienced more severe infarctions as reflected by higher Killip class. However, within each Killip class there were no significant differences in mortality between genders. Why do women have more severe infarctions? Although many factors are likely, several observations from the present study may provide some insight. The Killip classification provides useful mortality prognostic information that correlates with incremental degrees of left ventricular dysfunction (). Although women more frequently presented with heart failure than men, this may not necessarily mean that they experience larger infarcts with greater loss of myocardial mass; rather, they may be more prone to develop pulmonary edema for reasons that remain unclear. In fact, one study reported higher rates of pulmonary congestion in women with AMI despite higher left ventricular ejection fractions than in their male counterparts ([5, 9, 49]), an observation supported by women’s lower peak serum CK-MB in the present study. The greater tendency of women to experience diastolic dysfunction during ischemia may account for their higher rates of pulmonary congestion out of proportion to infarct size, but other factors, such as the site of coronary occlusion, duration of ischemia, development of collateral circulation and baseline ventricular function, must be taken into account. Uniform timing in the measurement of serum CK-MB from time of presentation coupled with prospective ECG and echocardiographic data would allow more precise correlation of gender differences in ventricular function during acute ischemia with ultimate infarct size.
Although there was substantial ethnic diversity in our study, race was not independently associated with hospital mortality. Despite differences in the distribution in comorbidity among ethnic groups (e.g., the higher prevalence of hypertension and diabetes among blacks), our findings suggest that the gender differences in hospital mortality for patients with AMI were due to differences in the distribution of clinical factors and that ethnicity had no independent effect.
3.3 Study Limitations.
This study has several limitations: 1) Despite broad inclusion criteria, patient enrollment may have been limited by the clinician’s suspicion of cardiac ischemia. Despite the adequate performance of IMIR criteria (), a small but undetermined number of patients with cardiac ischemia may still have gone undetected. 2) We attempted to include a wide range of hospital types, but we did not include any truly rural sites in our study. How our findings or the findings in other studies apply to rural settings remains to be determined. However, there appears to be no reason why our findings should not apply to patients in more rural settings. 3) The exclusion of patients with a language barrier may have led to an underrepresentation of certain ethnic groups with different risks for hospital mortality or other complications. However, the general availability of interpreters across study sites and high overall study inclusion rate (92%) made this an issue of unlikely significance.
Although questions remain with respect to some of the pathophysiologic features of acute mortality in women and men with AMI or acute cardiac ischemia in general, gender itself does not appear to be an independent predictor of hospital death among ED patients with symptoms suggestive of acute cardiac ischemia after controlling for baseline characteristics, comorbidity and hospital interventions. The slightly higher mortality rates seen in women with AMI appears to be accounted for by their higher degree of pulmonary congestion at presentation, older age and greater incidence of comorbid conditions. These findings are in contrast to other studies, where gender differences persisted after controlling for these and other clinical factors. Future studies need to pay particular attention to patient inclusion criteria, clinical study setting and study design to help construct a yet more detailed understanding of gender-related factors in the presentation and outcomes of acute cardiac ischemia.
☆ This study was supported by Grants T32HS00060 and RO1HS07360 from the Agency for Health Care Policy and Research, Rockville, Maryland and Grant RO1HL53900 from the National Heart, Lung, and Blood Institute, National Institutes of Health, Bethesda, Maryland.
- acute myocardial infarction
- creatine kinase, MB fraction
- electrocardiogram, electrocardiographic
- emergency department
- Imminent Myocardial Infarction Rotterdam study
- Received August 15, 1996.
- Revision received February 14, 1997.
- Accepted February 26, 1997.
- The American College of Cardiology
- Wenger NK
- ↵U.S. Bureau of the Census. Statistical Abstracts of the United States: 1990 (110th ed.) Washington (DC): US Bureau of the Census, 1990.
- ↵National Center for Health Statistics. Vital Statistics of the United States, 1988, Vol II, Part A, DHHS Publication No. (PHS) 91-1101. Washington (DC): US Government Printing Office, 1991.
- Greenland P,
- Reicher-Reiss H,
- Goldbourt U,
- Behar S
- Goldberg RJ,
- Gorak EJ,
- Yarzebski J,
- et al.
- White HD,
- Barbash GI,
- Modan M,
- et al.
- Vaccarino V,
- Krumholz HM,
- Berkman LF,
- Horwitz RI
- Chesebro JH,
- Foster V
- Falk E
- Fibrinolytic Therapy Trialists (FTT) Collaborative Group
- Kinlen LJ
- Jacoby RM,
- Nesto RW
- Soler NG,
- Bennett MA,
- Pentecost BL,
- Fitzgerald MG,
- Malins JM
- for the TIPI Working Group,
- Selker HP,
- Beshansky JR,
- Griffith JL
- Van der Does E,
- Lubsen J,
- Pool J
- ↵American Heart Association. 1992 Heart and Stroke Facts. Chicago; American Heart Association, 1991.
- Henning R,
- Lundman T
- Krumholz HM,
- Douglas PS,
- Lauer MS,
- Pastyernak RC
- Hammar N,
- Larsen FF,
- Sandberg E,
- Alfredsson L,
- Theorell T
- Tofler GH,
- Stone PH,
- Muller JE,
- et al.