Author + information
- Received June 26, 1997
- Revision received November 21, 1997
- Accepted December 12, 1997
- Published online March 15, 1998.
- Ayman S Al-Khadra, MDA,
- Deeb N Salem, FACC, MDA,
- William M Rand, PhDA,
- James E Udelson, MD, FACCA,
- John J Smith, MD, PhD, FACCA and
- Marvin A Konstam, MD, FACCA,* ()
- ↵*Dr. Marvin A. Konstam, Box 108, Division of Cardiology, New England Medical Center Hospitals, 750 Washington Street, Boston, Massachusetts 02111.
Objectives. We sought to evaluate the relation between warfarin anticoagulation and survival and morbidity from cardiac disease in patients with left ventricular (LV) dysfunction.
Background. Warfarin anticoagulation plays a major role in the management of patients who have had a large myocardial infarction and in those with atrial fibrillation. However, its use in patients with LV systolic dysfunction has been controversial.
Methods. We reviewed data on warfarin use in 6,797 patients enrolled in the Studies of Left Ventricular Dysfunction (SOLVD) trial and analyzed the relation between warfarin use and all-cause mortality, as well as the combined end point of death or hospital admission for heart failure. We used Cox regression to adjust for differences in baseline characteristics and to test for the interaction between warfarin use and selected patient variables in relation to outcome.
Results. On multivariate analysis, use of warfarin was associated with a significant reduction in all-cause mortality (adjusted hazard ratio [HR] 0.76, 95% confidence interval [CI] 0.65 to 0.89, p = 0.0006) and in the risk of death or hospital admission for heart failure (HR 0.82, 95% CI 0.72 to 0.93, p = 0.0002). Risk reduction was observed when each trial or randomization arm was analyzed separately, as well as in both genders. It was not significantly influenced by the presence of atrial fibrillation, age, ejection fraction, New York Heart Association functional class or etiology.
Conclusions. In patients with LV systolic dysfunction, warfarin use is associated with improved survival and reduced morbidity. This association is primarily due to a reduction in cardiac events and does not appear to be limited to any particular subgroup.
The use of warfarin in patients with heart failure (HF) has been the subject of considerable controversy. Patients with severe left ventricular (LV) systolic dysfunction are thought to be at greater risk of developing arterial and venous thromboembolic complications. The increase in risk in this group of patients is multifactorial and is probably related to a combination of low cardiac output, aberrant blood flow dynamics and stasis, abnormal endocardial surface, left atrial enlargement and atrial fibrillation. Venous stasis may also occur and, when coupled with reduced activity, can contribute to the risk of venous thromboembolism. Recent evidence also suggests that patients with LV systolic dysfunction may be in a chronic hypercoagulable state [1–3]. Several population-based cohort studies have sought to quantitate the risk of arterial and venous thrombosis and embolism. Estimates of such risks varied among the different studies, but in general they were surprisingly low [4–6].
Despite these findings, several experts have recommended the use of anticoagulant agents in patients with HF [7, 8]to reduce thromboembolic events. In addition, warfarin use in patients with coronary artery disease, especially after a large myocardial infarction (MI), reduces the likelihood of nonfatal MI [9–12], and therefore may retard the progression of HF. In patients recovering from MI, warfarin therapy also reduces the risk of thromboembolic strokes .
The purpose of our analysis was to evaluate the relation between warfarin use and survival and cardiovascular morbidity in patients with LV systolic dysfunction enrolled in the Studies of Left Ventricular Dysfunction (SOLVD) trial.
The Studies of Left Ventricular Dysfunction trial was a multicenter, randomized, double-blind, placebo-controlled investigation that examined the effect of enalapril on morbidity and mortality in patients with LV systolic dysfunction, defined as an ejection fraction (EF) ≤0.35 [13, 14]. Patients receiving treatment other than angiotensin-converting enzyme inhibitors for symptoms of HF were enrolled in the treatment trial; patients without such symptoms were enrolled in the prevention trial. Patients were ineligible if they were >80 years of age or had any of the following: hemodynamically serious valvular heart disease requiring surgical treatment, unstable angina pectoris, angina thought to be severe enough to require revascularization procedures, MI during the previous month, severe pulmonary disease, serum creatinine level >177 mmol/liter (2 mg/dl) or any other disease that might substantially shorten survival or impede participation in a long-term trial. In total, 6,797 patients were enrolled in both trials. Ventricular function was assessed by contrast angiography, radionuclide ventriculography or two-dimensional echocardiography. The baseline study visit included a comprehensive review of the patients’ medical history, and patients were specifically asked whether they were taking certain medications regularly, including warfarin and those drugs listed in Table 1. After randomization to enalapril or placebo, patients were regularly evaluated, and their clinical status, including the development of HF, hospital admissions, adherence to study regimen and adverse effects, was recorded at each follow-up visit. For patients who died or were admitted to the hospital, the cause of death or the primary reason for hospital admission was evaluated and classified by an investigator who had no knowledge of the study treatment.
We reviewed data on the use of warfarin anticoagulation collected on the baseline data form. All patients in both trials with complete data were included in the analysis. The end points of all-cause mortality, death or hospital admission for HF, cardiovascular deaths, sudden death without antecedent worsening HF, death due to HF, fatal MI and noncardiac vascular causes of death were evaluated and correlated with warfarin use.
1.2 Statistical Analysis
The primary outcome measures were occurrence of death and length of survival (time from entry into the study to either death, end of study or loss to follow-up). In addition, hospital admission for unstable angina and nonfatal MI were evaluated. Risk factors of primary interest were use of anticoagulant agents, randomization to enalapril and trial (treatment or prevention). The chi-square test for independence was used to analyze unadjusted mortality rates and to calculate relative risk (RR) and 95% confidence intervals (CI). Kaplan-Meier survival analysis was used to examine duration of survival, with log-rank tests to compare survival analysis curves between subsets of patients.
Cox regression was used to adjust for potential influence of confounding factors on survival. These include age, gender, etiology, trial assignment, study drug randomization, EF, New York Heart Association functional class, history of atrial fibrillation, history of cerebrovascular disease and baseline drug therapy with antiplatelet agents, antiarrhythmic drugs or digitalis. Results are reported as the adjusted hazard ratio (HR) with 95% CIs and p values. Throughout, p < 0.05 was considered statistically significant. The computer statistical software package SPSS, version 7.5, was used for all analyses.
Based on the total number of patients analyzed and the ratio of warfarin users to nonusers (1:6.5), and assuming an end point incidence of 25% in the nonusers group, our analysis had 90% power to detect an RR reduction of 20%, with 95% confidence.
Of 6,797 patients enrolled (2,569 in the treatment trial and 4,228 in the prevention trial), 6,513 were considered in the final analysis. Two hundred eighty-four patients (4.2%) were excluded because of missing baseline data. Mean follow-up was 41.4 months in the treatment trial and 37.4 months in the prevention trial. The clinical characteristics of warfarin users and nonusers in the combined trial are shown in Table 1. Of the total group, 861 patients (13.2%) reported baseline warfarin use, whereas the remainder did not. Patients using warfarin were slightly younger and less frequently complained of angina. They also had a lower mean EF, worse functional class, more frequent atrial fibrillation, cerebrovascular disease and more frequent use of antiarrhythmic agents and digitalis. They were less likely to have an ischemic etiology and less likely to use antiplatelet agents or beta-blockers. The proportion of men and frequency of randomization to enalapril were not different between the two groups. These differences were consistent when the treatment and prevention trials were analyzed separately.
2.2 All-Cause Mortality and Causes of Death
There were 210 deaths among the patients who reported warfarin use, compared with 1,334 deaths among the patients who did not use warfarin (RR 1.03, 95% CI 0.91 to 1.17) (Table 2). Of those deaths, 177 in the warfarin use group and 1,181 in the nonuse group were due to cardiovascular disease (RR 0.99, 95% CI 0.86 to 1.14) (Table 3).
After adjusting for baseline differences, warfarin use was a highly significant predictor of favorable outcome (Tables 2 and 3). In the combined group, warfarin use was associated with an HR of 0.76 for all-cause mortality (95% CI 0.65 to 0.89, p = 0.0006) (Fig. 1). Mortality due to cardiovascular disease was also significantly reduced (HR 0.72, 95% CI 0.61 to 0.86, p = 0.0002) (Table 3). Hazard reduction was consistent among the various causes of cardiac death, although the reduction in fatal MI appeared to be more pronounced. There was no reduction in fatal noncardiac vascular events, including deaths due to stroke, pulmonary embolism or other vascular causes.
2.3 Death or Hospital Admission for HF
Among the patients who reported warfarin use, there were 306 deaths or hospital admissions for HF, compared with 1,910 deaths among those who did not use warfarin (RR 1.05, 95% CI 0.95 to 1.16) (Table 2). After adjusting for differences in baseline characteristics, warfarin use was associated with a significant reduction in this end point (HR 0.82, 95% CI 0.72 to 0.93, p = 0.002) (Fig. 2and Table 2). This benefit appeared to be driven primarily by a reduction in all-cause mortality.
2.4 Hospital Admission for Unstable Angina and Nonfatal MI
In the combined group, there were 185 admissions for unstable angina and 36 admissions for nonfatal MI among patients receiving warfarin compared with 1,407 admissions for unstable angina and 409 admissions for nonfatal MI among those not using warfarin (RR 0.86, 95% CI 0.75 to 0.99, p = 0.03 for unstable angina, and RR 0.58, 95% CI 0.41 to 0.81, p = 0.001 for nonfatal MI) (Table 4).
After adjusting for differences in baseline characteristics, warfarin use was associated with a 44% reduction in hospital admission for nonfatal MI (p = 0.002) and with a nonsignificant trend toward reducing admissions for unstable angina (Table 4).
2.5 Association between Warfarin Use and End Points in Different Subgroups
The relation between warfarin use and the various end points was analyzed by age, gender, etiology, EF, functional class, trial assignment (treatment and prevention), randomization to enalapril and the presence or absence of atrial fibrillation. Interaction terms were introduced into the Cox regression analysis, with adjustment for confounding variables. The presence of a significant interaction (p < 0.05) was considered to indicate dependence of the association between warfarin use and survival on the presence of or variation in any of these variables. There was no significant interaction between warfarin-associated mortality reduction, reduction in the end point of death or hospital admission for HF or reduction in hospital admission for nonfatal MI and any of these variables. The same was also true when both trials were analyzed separately. Table 5shows the HR for all-cause mortality in different subgroups when patients are classified by trial, randomization arm, etiology and presence or absence of atrial fibrillation. In all subgroups there is a consistent trend toward a reduction in this end point. A consistent pattern was also observed when the end point of death or hospital admission for HF was analyzed in these subgroups.
The use of warfarin in the management of patients with HF is controversial. Estimates vary regarding the risk of arterial and venous thromboembolism among these patients. In a group of patients with dilated cardiomyopathy, Fuster et al. estimated the incidence of arterial embolism to be 3.5 per 100 patient-years. Later estimates [5, 6, 8]varied from 1.4 to 42.4 per 100 patient-years, depending on the characteristics of the group studied (severity of LV dysfunction, LV size and the presence of atrial fibrillation), method of documenting events (clinical vs. autopsy) and referral base. For example, the highest estimate of thromboembolism came from a thrombosis referral center . Reports also differ regarding the degree of correlation between the severity of LV dysfunction, or other clinical variables, and the likelihood of arterial or pulmonary thromboembolism [6, 8].
This study is the first to examine the relation between warfarin use and long-term survival in a large cohort of patients selected primarily on the basis of LV systolic dysfunction. Because patients were not randomized to receive warfarin, significant differences in baseline characteristics between users and nonusers were present. Generally, warfarin users had less favorable baseline characteristics. We attempted to adjust for these differences by including relevant characteristics in a multivariate Cox regression model. The primary results of this analysis show that baseline warfarin use in patients with an EF ≤35% (with and without symptoms of HF) is associated with a significant reduction in all-cause mortality, in the combined end point of death or hospital admission for HF and in hospital admissions for acute coronary syndromes, primarily in the number of admissions for nonfatal MI. In addition, we found that this association was consistent in important clinical subgroups and that it was not influenced by age, gender, functional class, EF or presence of atrial fibrillation.
In patients recovering from acute MI, warfarin anticoagulation reduces mortality, fatal and nonfatal coronary events, pulmonary embolism and stroke, when started within 4 weeks from the onset of symptoms . Other studies in the chronic phase of coronary disease have suggested a similar reduction in coronary events, but not in overall mortality. In our analysis, the reduction in all-cause mortality was mainly driven by a reduction in cardiac mortality. This reduction, however, could not be attributed to one specific mode of cardiac death. We observed a significant reduction in fatal MI and in sudden death. The reduction in hospital admissions for nonfatal MI and unstable angina may have contributed to the reduction in deaths associated with worsening HF. Patients with HF may manifest a hypercoagulable state, as evidenced by increased plasma and blood viscosity, platelet activation and increased plasma levels of beta-thromboglobulin, thrombin–antithrombin complex, fibrinopeptide A, d-dimer and von Willibrand factor [1–3]. There is, however, no consistent association between the presence of these markers of hypercoagulability and either the clinical severity of HF, etiology, degree of neurohormonal activation or EF, and they do not clearly predict the likelihood of clinical events [1–3].
Our data did not show a reduction in noncardiac vascular deaths or hospital admissions for noncardiac vascular events in association with warfarin use. This negative finding is probably multifactorial: higher risk patients were more likely to receive anticoagulation, and in this group, the risk of fatal vascular events was small (<2%). In addition, unless the diagnosis is made before death, it is possible that many cases of fatal pulmonary embolism are labeled as sudden death or acute HF.
The findings from our analysis have significant implications with regard to the treatment of patients with LV systolic dysfunction. Benefit associated with warfarin use in this group of patients is not limited to a reduction in certain vascular events in high risk subgroups (e.g., those with atrial fibrillation) and is not limited to patients with symptoms of HF. Unfortunately, because of a lack of data on the intensity of anticoagulation, no recommendations can be made in that regard. Despite a lack of data from this study on the risks of anticoagulation, maintaining a therapeutic International Normalized Ratio in the range of 2 to 3 has been shown in multiple, large, randomized trials to be associated with <1.5% risk of major bleeding . It is unlikely that patients with LV dysfunction should require anticoagulation outside this range or would have a different complication profile associated with warfarin use.
3.1 Strengths and Limitations
Our analysis has important strengths, including a large sample size, prospective definition of end points and collection of data, long follow-up time, availability of data for most patients (>95%) and consistency of the results in different subgroups. However, it has limitations common to cohort studies, including its retrospective nature and lack of randomization. Data on the intensity of anticoagulation and bleeding complications of anticoagulation were not available, nor were follow-up data available, including warfarin use throughout the study period.
The use of warfarin anticoagulation in patients with LV systolic dysfunction is associated with improved survival and reduced morbidity. This association is not influenced by age, gender, etiology, functional class, presence or absence of atrial fibrillation, EF or use of enalapril. In the absence of randomized controlled data, the present findings support the use of warfarin in patients whose LVEF is ≤35%, including those in sinus rhythm.
This study was presented in part at the 45th Annual Scientific Session of the American College of Cardiology, Orlando, Florida, March 1996.
↵1 This study was supported by Grant MO1-RR00054 from the General Clinical Research Center, funded by the Division of Research Resources, National Institutes of Health, Bethesda, Maryland. The Studies of Left Ventricular Dysfunction (SOLVD) were funded by the National Heart, Lung, and Blood Institute, National Institutes of Health, Bethesda, Maryland, with a contribution from Merck, Sharpe and Dohme Pharmaceutical Company, West Point, Pennsylvania.
- confidence interval
- ejection fraction
- heart failure
- hazard ratio
- left ventricular
- myocardial infarction
- relative risk
- Studies of Left Ventricular Dysfunction
- Received June 26, 1997.
- Revision received November 21, 1997.
- Accepted December 12, 1997.
- The American College of Cardiology
- Jafri SM,
- Ozawa T,
- Mammen E,
- Levine TB,
- Johnson C,
- Goldstein S
- Yamamoto K,
- Ikeda U,
- Furuhashi K,
- Irokawa M,
- Nakayama TM,
- Shimada K
- ↵Dunkman WB, Johnson GR, Carson PE, Bhat G, Farrell L, Cohn J, for the V-HeFT Cooperative Studies Group. Incidence of thromboembolic events in congestive heart failure. Circulation 1993;87 Suppl VI:VI-94–101.
- Ciaccherie M,
- Castelli Q,
- Cecchi F,
- et al.
- Sherman D,
- Dyken M,
- Fisher M,
- et al.