Author + information
- Received July 31, 1996
- Revision received December 11, 1996
- Accepted December 16, 1996
- Published online April 1, 1997.
- Daniel L. Dries, MD, MPHAC,* (, )
- Yves D. Rosenberg, MD, MPHA,
- Myron A. Waclawiw, PhDB and
- Michael J. Domanski, MDA
- ↵*Dr. Daniel L. Dries, Clinical Trials Group, National Heart, Lung, and Blood Institute, II Rockledge Center, 6701 Rockledge Drive, MSC 7936, Bethesda, Maryland 20892-7936.
Objectives. The aims of this study were to describe the incidence and spectrum of thromboembolic events experienced by patients with moderate to severe left ventricular systolic dysfunction in normal sinus rhythm and to study the association between ejection fraction and thromboembolic risk.
Background. The annual incidence of thromboembolic events in patients with heart failure is estimated to range from 0.9% to 5.5%. Previous studies demonstrating a relation between worsening left ventricular systolic function and thromboembolic risk are difficult to interpret because of the prevalence of atrial fibrillation, an independent risk factor for thromboembolism, in the patients with a lower ejection fraction.
Methods. This is a retrospective analysis of the Studies of Left Ventricular Dysfunction prevention and treatment trials data base. Patients with atrial fibrillation were excluded, resulting in 6,378 participants in sinus rhythm at the time of randomization. Thromboembolic events include strokes, pulmonary emboli and peripheral emboli. Separate analyses were conducted in each gender because there was evidence of a significant interaction between ejection fraction and gender on the risk of thromboembolic events (p = 0.04).
Results. The overall annual incidence of thromboembolic events was 2.4% in women and 1.8% in men. On univariate analysis, a decline in ejection fraction was not associated with thromboembolic risk in women (relative risk [RR] per 10% decrease in ejection fraction 1.58, 95% confidence interval [CI] 1.10 to 2.26, p = 0.01), but not in men. On multivariate analysis, a decline in ejection fraction remained independently associated with thromboembolic risk in women (RR per 10% decrease 1.53, 95% CI 1.06 to 2.20, p = 0.02), but no relation was demonstrated in men.
Conclusions. In patients with left ventricular systolic dysfunction and sinus rhythm, the annual incidence of thromboembolic events is low. Ejection fraction appears to be independently associated with thromboembolic risk in women, but not in men.
(J Am Coll Cardiol 1997;29:1074–80)
© 1997 by the American College of Cardiology
In an overview of the available data (), the annual incidence of thromboembolic events in patients with heart failure ranges from 0.9 to 5.5 events per 100 patient-years, with the largest studies reporting an annual incidence of 2.4 to 2.6 events per 100 patient-years ([2–9]). The risk for clinical embolic complications has been linked to a low ejection fraction (EF) in several studies ([7, 10]). The biologic basis supporting this hypothesis relates to abnormal and sluggish flow patterns in dilated, poorly contracting cardiac chambers, which leads to mural thrombi as well as venous stasis. This may predispose to the development of deep venous thrombi and subsequent pulmonary emboli, as well as peripheral arterial emboli. The available studies suggest an increased risk of thromboembolic events in patients with worsening left ventricular systolic function. However, these data are difficult to interpret because of the prevalence of atrial fibrillation, an independent risk factor for thromboembolic events, in the participants with more severe left ventricular dysfunction. In their review of the available data related to thromboembolism in heart failure, Baker and Wright () emphasize that there is no published study that used multivariate analysis techniques to evaluate the independent effect of ventricular dysfunction on the risk for arterial thromboembolism. Although it is recognized that patients with concomitant left ventricular systolic dysfunction and atrial fibrillation comprise a group at high risk of embolic complications (), the incidence of thromboembolic events in patients with heart failure and sinus rhythm is less clear, as is the relation of EF to thromboembolic risk.
To address these issues, we conducted a retrospective data base analysis that consisted of the 6,378 patients entered into the Studies of Left Ventricular Dysfunction (SOLVD) treatment and prevention trials who, at the time of randomization, were in sinus rhythm. The purpose of this analysis was twofold: first, to describe the incidence and etiologic spectrum of embolic events for a cohort of patients with moderate to severe systolic dysfunction and sinus rhythm; and second, to study the hypothesis that EF is an independent risk factor for embolic events using multivariate analysis to adjust for potentially confounding variables that have been identified as independent predictors of thromboembolic risk (history of hypertension, previous stroke and diabetes) ([12, 13]), as well as the use of antiplatelet or anticoagulant agents.
1.1 Study design.
This retrospective data base analysis consisted of 6,378 patients in the SOLVD prevention and treatment trials. These studies were designed to test the hypothesis that enalapril, an angiotensin-converting enzyme inhibitor, would reduce mortality in patients with heart failure. There were 4,228 participants (62%) in the prevention trial, which consisted of participants with asymptomatic left ventricular dysfunction, and 2,569 participants (38%) in the treatment trial, which consisted of participants with symptomatic left ventricular dysfunction. The entry criteria in both trials required a documented EF ≤35%. The EF was measured using radionuclide angiography (68%), contrast angiography (11%) or two-dimensional echocardiography (21%), using Simpson’s rule to yield a calculated value. A more detailed description of the rationale, design and methodology of the SOLVD trial has been previously published ().
1.2 Definition of end points.
Deaths were classified by the principal investigators at each clinical site after a blinded review of the circumstances of each death. The main outcome variable for the present study was thromboembolic events, which included fatal or nonfatal cerebrovascular events, pulmonary emboli and peripheral emboli. Data regarding demographics, past medical history and current medication profiles were collected from each participant at the time of randomization. Patients with atrial fibrillation on the randomization electrocardiogram (ECG) were excluded, as were participants with missing baseline ECG data. This resulted in the exclusion of 419 participants (280 with missing data from the interpretation of the baseline ECG and 139 with atrial fibrillation). The final analysis, therefore, consisted of 6,378 participants, including 5,457 men (86%) and 921 women (14%).
1.3 Statistical analysis.
Group comparisons included the two-sample Student ttest for comparison of means, and the chi-square statistic for comparison of proportions. To satisfy the assumption of the independence of events, recurrent thromboembolic events in a participant after randomization were not included in the analysis. The prognostic significance of each study covariate on the time until a thromboembolic event was investigated using univariate and multivariate Cox proportional hazards models. A two-sided 95% confidence interval (CI) was constructed around the point estimate of relative risk (RR) associated with each study covariate. A p value ≤0.05 was considered statistically significant.
We tested for evidence of a significant two-way interaction between EF, and each of the other covariates in the final multivariate model. There was a significant interaction between gender and EF on the RR for a thromboembolic event (p = 0.04). Therefore, separate analyses were conducted for each gender. The variables, in addition to EF, that were studied in both univariate and multivariate analyses in both genders included: age, EF, history of hypertension, remote smoking history, diabetes, previous cerebrovascular event, use of enalapril, antiplatelet monotherapy, anticoagulant monotherapy and simultaneous use of antiplatelet and anticoagulant agents. Medication use was analyzed according to profiles obtained at the time of randomization. The statistical analyses were performed using SAS (Statistical Analysis System, version 6.07).
2.1 Baseline characteristics.
Baseline characteristics are presented according to gender in Table 1. Women were significantly older and had a higher prevalence of diabetes, and a greater proportion of them had a history of hypertension compared with their male counterparts. More men used aspirin monotherapy at baseline and had a remote history of tobacco use and a history of previous infarction compared with their female counterparts.
2.2 Distribution of embolic events according to gender.
Table 2describes the classification of the 340 first embolic events for the participants, with an average follow-up of 39.9 months. There were 278 events in the men for an overall incidence of 1.82 events per 100 participant-years of follow-up, and 62 events in the women for an overall incidence of 2.42 events per 100 participant-years of follow-up. In both genders, most thromboembolic events were the result of nonfatal cerebrovascular events. The distribution of thromboembolic events according to type was similar for men and women, except for a significantly greater proportion of pulmonary emboli in the women (24% vs. 14%, p = 0.01).
2.3 Incidence and crude RR stratified by EF quartile.
Table 3presents the overall incidence of thromboembolic events and unadjusted RR stratified according to EF quartiles. In women, there is a trend toward an increased incidence of thromboembolic events in the lower EF quartiles, but no such relation is noted in men. Moreover, women in the lowest two quartiles of EF demonstrated an incidence of thromboembolic events that was almost twice that of their male counterparts (3.80% vs. 2.01% for EF 11% to 20%; 4.20% vs. 1.96% for EF ≤10%). The unadjusted RR for a thromboembolic event was greater in women than in men for each quartile, although there was overlap of the corresponding 95% CIs for each quartile.
2.4 Univariate analysis.
As demonstrated in Table 4, EF was significantly related to the risk of a thromboembolic event in women, but not in men. In women, a 10% decrease in EF was associated with a 58% increased risk of an embolic events (RR 1.58, 95% CI 1.10 to 2.26, p = 0.01).
In men, other variables significantly associated with an increased risk of a thromboembolic event included a history of hypertension, diabetes and a history of a cerebrovascular event. Variables associated with a decreased risk of a thromboembolic event included use of enalapril and antiplatelet agents.
In women, univariate analysis demonstrated an increased risk of a thromboembolic event associated with diabetes. Similar to men, a decreased risk was associated with antiplatelet monotherapy for women.
2.5 Multivariate analysis.
These results are presented in Table 5. On multivariate analysis, a decline in EF remained independently associated with an increased risk of thromboembolic events in women, but no independent association was demonstrated for men. In women, a 10% decline in EF was associated with a 53% increased risk of an initial thromboembolic event during the period of follow-up, which averaged 39.9 participant-months (RR 1.53, 95% CI 1.06 to 2.20, p = 0.02). There was no association between EF and thromboembolic risk in men (RR 1.08, 95% CI 0.89 to 1.31, p = 0.42).
In men, other variables independently associated with the risk of a thromboembolic event included a 32% increased risk with a history of hypertension, a 64% increased risk with diabetes and a 131% increased risk in participants with a previous cerebrovascular event. The use of enalapril was associated with a 26% risk reduction, compared with participants receiving placebo. The use of antiplatelet agents at baseline was associated with a 23% risk reduction, although of borderline statistical significance (RR 0.77, 95% CI 0.59 to 1.00, p = 0.06).
In women, other variables associated with the risk of thromboembolic events on multivariate analysis included diabetes, which was associated with a 113% increased risk for thromboembolism, and antiplatelet monotherapy at baseline, which was associated with a 53% reduction in the risk of thromboembolism, compared with those not using antiplatelet agents at baseline.
2.6 Exclusion of pulmonary emboli.
The only significant difference between the genders in the distribution of a specific type of embolic event was the statistically significant greater proportion of events caused by pulmonary emboli in the women (24% vs. 14%, p = 0.01). When pulmonary embolism was excluded, there were 238 events in men and 48 events in women that were attributed to a cerebrovascular event or peripheral embolism.
Using these events as our definition of a thromboembolic event, identical univariate and multivariate analyses were conducted in the genders, and the results of the multivariate analysis are presented in Table 6. On univariate analysis, EF was significantly associated with the risk of thromboembolic events in women, but not in men. On multivariate analysis, there was a trend suggesting that a lower EF was associated with an increased thromboembolic risk in women (RR per 10% decrease in EF 1.47, 95% CI 0.97 to 2.23, p = 0.07), but in men there was no demonstration of an independent association between EF and thromboembolic risk (RR per 10% decrease in EF 0.99, 95% CI 0.81 to 1.22, p = 0.95).
3.1 Study findings.
This is the first large data base study of patients with heart failure that attempts to study whether EF is independently associated with thromboembolic risk for patients in sinus rhythm with moderate to severe left ventricular systolic dysfunction. The analysis demonstrates that the overall annual incidence of thromboembolic events is low in both genders (2.4% in men and 1.8% in women). Also, it appears that in women, but interestingly not in men, EF was independently associated with thromboembolic risk.
3.2 Gender differences in the association of EF and thromboembolic risk.
The finding of a significant interaction between gender and EF on thromboembolic risk was unexpected. Although the reasons for gender differences in the association of EF and the risk of a thromboembolic event cannot be determined with certainty from this study, several possible explanations are considered.
The only significant difference between the genders in the distribution of thromboembolic events was a greater proportion of pulmonary emboli in the women (24%) compared with the men (14%) (p = 0.01). Therefore, we considered the possibility that the gender differences might be statistically driven by an especially strong relation in women between a declining EF and risk of pulmonary emboli. Reasons for such a difference might include a greater degree of functional impairment limiting physical activity in women compared with men with a similar EF. The resulting inactivity might increase venous stasis, the development of deep venous thrombi and subsequent emboli to the pulmonary circulation. Therefore, we conducted a multivariate analysis adjusting for the same variables but excluding pulmonary emboli from our definition of a thromboembolic event. Although there was a reduction in the strength of the statistical association between EF and thromboembolic risk in women (p = 0.07), a trend persisted in the women, demonstrating an increased risk of thromboembolic events as EF fell, but again there was no association in men. The reduction in statistical significance in women (p = 0.07) may be the result of the reduced statistical power resulting from fewer overall events.
We included all cerebrovascular events in our definition of thromboembolic events, and we recognize that not all the cerebrovascular events were likely to be caused by a cardioembolic event. The possibility exists that there were gender differences in the proportion of cerebrovascular events that were the result of comorbid atherosclerotic disease or intracerebral hemorrhage, rather than cardiac emboli. Because these cerebrovascular events are not expected to be related to EF, a greater proportion of such events in men might make it difficult to statistically demonstrate a relation between EF and thromboembolic risk. The multivariate analysis adjusted for some of the baseline differences in the clinical variables, which were each previously demonstrated to independently predict thromboembolic risk ([12, 13]). Nonetheless, we cannot exclude the possibility of confounding from incomplete adjustment in important baseline differences between the genders, as well as residual confounding from the variables included in the multivariate model.
Although this is a retrospective analysis, the possibility exists that there may be important physiologic gender differences in the production of thromboembolism in heart failure. Important gender differences might exist in the clotting system and fibrinolytic system, the biology of the endothelium lining the ventricle or the influence of hormonal differences on the defense mechanisms against intraventricular or intravascular thrombosis.
3.3 Comparison with other studies.
This is the first study attempting to use multivariate analysis techniques to study the independent association between ventricular dysfunction and the risk of thromboembolism. Kyrle et al. () studied 38 patients with dilated cardiomyopathy and classified severity of left ventricular dysfunction according to echocardiographic fractional shortening (FS). Before the initiation of anticoagulation, the proportion experiencing an arterial or pulmonary embolism was 20% in those with mild dysfunction (FS 0.20 to 0.28), 47% in those with moderate dysfunction (FS 0.1 to 0.2) and 67% in the severe group (FS <0.1). There was no information in this study about the presence or absence of atrial fibrillation. Two of the largest trials, the Veterans Administration Heart Failure Trials (V-HeFT) I and II (), demonstrated an incidence of thromboembolic events—defined to include stroke, pulmonary emboli and peripheral emboli—of 2.7 per 100 patient-years in V-HeFT I and 2.1 per 100 patient-years in V-HeFT II. Patients experiencing a thromboembolic event had a lower EF compared with patients who did not (26.3% vs. 29.6%), but the difference was not statistically significant. This report did report separate incidence data for the participants with and without atrial fibrillation, but the results were paradoxic. For example, the incidence of thromboembolism in those participants with atrial fibrillation (3.8%) was less than that of those without it (5.6%). These results might be explained by differences in the use of anticoagulant agents as well as other baseline differences in the patients with atrial fibrillation. Segal et al. () reported a higher incidence of systemic emboli in patients with minimal to mild cardiac enlargement on chest roentgenogram (11%) compared with those with moderate to severe cardiac enlargement (16%), but the difference did not reach statistical significance. Yokota et al. () found that patients with mural thrombi had a lower EF (13.1%) than those without them (17.9%).
3.4 Study limitations.
The limitations of this study must be emphasized. The SOLVD trials were not designed to determine the incidence of thromboembolic complications in heart failure. This is a retrospective analysis of the large data base assembled from these studies. Criteria for the diagnosis of thromboembolic events were not established by protocol, and the events that were reported did not undergo central review for possible reclassification. Despite these limitations, we do not believe that there was significant underreporting of thromboembolic events, nor that there were important differences among the participating institutions in the diagnosis of stroke, transient ischemic attacks or pulmonary and peripheral emboli.
Medication use was analyzed based on reported regular use at the time of randomization, and it is likely that there were changes in some patients during the course of the studies. Treatment with antiplatelet or anticoagulant agents was not randomized or blinded, and a target prothrombin or international normalized ratio was not specified. Although we excluded participants with atrial fibrillation, this was assessed at baseline, and it is likely that some participants developed atrial fibrillation during the trials. There was a greater prevalence of hypertension in women, which can lead to the development of atrial fibrillation, so it is possible that a greater proportion of women might have subsequently developed atrial fibrillation compared with their male counterparts.
The presence of residual confounding from variables we attempted to adjust for is likely because some of these baseline characteristics were not assessed using a precise classification scheme. For example, we lacked information regarding the type and duration of diabetes, the duration and intensity of previous smoking habits, the duration and severity of previous hypertension, as well as the degree of anticoagulation achieved in the patients taking anticoagulant agents.
3.5 Clinical implications.
These data suggest that in patients with moderate to severe left ventricular dysfunction the incidence of thromboembolic events is low. Also, in women, but not in men, the severity of left ventricular dysfunction, quantified by the calculated EF, is independently associated with the risk for thromboembolic events. The low overall incidence of events raises the clinically important question of whether anticoagulation to prevent thromboembolic complications in these patients presents a favorable risk/benefit balance. The risk of serious bleeding among participants in the clinical trials assessing the efficacy of anticoagulation in atrial fibrillation was less than 1% per year (); however, the risk may be greater in a more generalized outpatient population ().
The annual incidence of thromboembolic events demonstrated by the women in the lowest two quartiles of EF (3.8% for EF 11% to 20%, 4.2% for EF ≤10%) approaches the estimated annual incidence of embolic complications in patients with atrial fibrillation, which is estimated to be 5% ([12, 19, 20]), and may identify a group with a more favorable risk/benefit ratio. Moreover, the efficacy of anticoagulation in heart failure to prevent thromboembolic events has not been demonstrated in any randomized, controlled clinical trial. Also, these data suggest that there may be important gender differences in the pathophysiologic mechanisms relating ventricular dysfunction to the risk of thromboembolism within the heart failure population, the specific nature of which remains to be further elucidated.
The overall incidence of thromboembolic events in patients with moderate to severe left ventricular dysfunction and sinus rhythm is low. Moreover, the degree of left ventricular dysfunction as quantified by EF appears to be independently associated with thromboembolic risk in women, but not in men. These data suggest the possible existence of physiologic gender differences in the production of thromboembolism in patients with significant left ventricular dysfunction.
- confidence interval
- ejection fraction
- fractional shortening
- relative risk
- Studies of Left Ventricular Dysfunction
- Veterans Administration Heart Failure Trial
- Received July 31, 1996.
- Revision received December 11, 1996.
- Accepted December 16, 1996.
- The American College of Cardiology
- Massumi RA,
- Rios JC,
- Gooch AS,
- Nutter D,
- De Vita VT,
- Datlow DW
- for the V-HeFT VA Cooperative Studies Group,
- Dunkman WB,
- Johnson GR,
- Carson PE,
- Bhat G,
- Farrel L,
- Cohn JN
- ↵Guidelines for the evaluation and management of heart failure: report of the American College of Cardiology/American Heart Association Task Force on Practice Guidelines. Circulation 1995;92:2764–84.
- ↵Risk factors for stroke and efficacy of antithrombotic therapy in atrial fibrillation: analysis of pooled data from five randomized controlled trials. Arch Intern Med 1994;154:1449–1457.
- Wolf PA,
- Abbott RD,
- Kannel WB