Management of Heart Failure IV. Anticoagulation for Patients With Heart Failure Due to Left Ventricular Systolic Dysfunction David W. Baker, MD, MPH, Richard F.

Wright, MD

Objective.\p=m-\Thisarticle reviews the incidence of arterial thromboembolism in patients with heart failure who are not receiving anticoagulants. We also examine whether more severe ventricular dysfunction increases this incidence and the efficacy and risks of anticoagulation for patients in sinus rhythm. Data Sources.\p=m-\English-languagestudies referenced in MEDLINE or EMBASE (January 1966 to September 1993) were reviewed. We used the search terms heart failure, congestive; congestive heart failure; heart failure; cardiac failure; and dilated cardiomyopathy in conjunction with the terms anticoagulation, cerebrovascular disorders, stroke, and thromboembolism. Study Selection.\p=m-\Allstudies with separate data for patients with chronic heart

failure not receiving anticoagulants were included. Articles addressing valvular heart disease or heart failure secondary to acute myocardial infarction or Chagas' disease were excluded. Studies of the occurrence of left ventricular mural thrombi were also reviewed. Data Extraction and Synthesis.\p=m-\Inclusionand exclusion criteria, prevalence of atrial fibrillation, mean follow-up, and the occurrence of arterial thromboembolic events were extracted. If the incidence was not given, this was estimated using the proportion of patients with events divided by the mean follow-up. Conclusion.\p=m-\Theincidence of arterial thromboembolism ranged from 0.9 to 5.5 events per 100 patient-years, with the largest studies reporting incidence of 2.0% and 2.4%. Findings regarding the relationship between ventricular function and thromboembolic events are contradictory. No controlled trial has assessed the efficacy or risks of anticoagulation for patients with heart failure and sinus rhythm, and reported efficacy in case series ranged from 0% to 100%. Until adequate studies are performed, anticoagulation should be discouraged for patients with heart failure who are in sinus rhythm. (JAMA. 1994;272:1614-1618)

From the Health Sciences Program, RAND, Santa Monica, Calif (Dr Baker); Division of General Internal Medicine, Harbor-UCLA Medical Center, Torrance, Calif (Dr Baker); Pacific Heart Institute, Santa Monica, Calif (Dr Wright); and University of California-Los Angeles, School of Medicine (Dr Wright). Dr Baker is now with the Division of General Medicine, Grady Memorial Hospital, Emory University School of Medicine, Atlanta, Ga. The views expressed herein are those of the authors and do not reflect the position of the Agency for Health Care Policy and Research, the US Public Health Service, or the US Department of Health and Human Services. Reprint requests to Pacific Heart Institute, 2001 Santa Monica Blvd, Suite 280, Santa Monica, CA 90404

(Dr Wright).

ARTERIAL thromboemboli in patients with heart failure due to left ventricular systolic dysfunction have been reported to occur with an incidence of 0.9 to 42.4 events per 100 patient-years.110 Low car¬ diac output, aberrant flow through di¬ lated cardiac chambers, and poor con¬ tractility may all predispose to cardiac mural thrombi and subsequent emboli. However, other factors contribute to the incidence of thromboemboli. The preva¬ lence of atrial fibrillation in these stud¬ ies was 12% to 36%, and many of the reported strokes may have been caused

Downloaded From: http://jamanetwork.com/pdfaccess.ashx?url=/data/journals/jama/9269/ on 01/16/2017

by comorbid atherosclerotic cerebrovas¬ cular disease in patients with ischemie cardiomyopathy. There has never been a controlled trial assessing the efficacy of anticoagulation for patients with heart failure outside the hospital setting.11 Despite the uncertainties about the true incidence of cardiogeni e emboli and the efficacy of treatment, many authors

recommend the routine use of antico¬ agulation for patients with heart fail¬ ure, or they suggest that anticoagulation be "considered" for this condition even in the absence of atrial fibrillation, systemic embolism, or left ventricular thrombus on echocardiogram.8·1219 Al¬ though the prevalence of anticoagulant use for patients with heart failure in normal sinus rhythm in community prac¬ tice has not been

reported, anticoagu¬ widely used at academic cen¬ ters. Sixteen percent of patients in the Studies of Left-Ventricular Dysfunction (SOLVD) treatment trial,20 and 20% of patients in the Veterans Affairs Coop¬ lants

are

erative Studies Vasodilator-Heart Fail¬ Trials (V-HeFT I and II) were tak¬ ing anticoagulants when they entered the trials, although the reasons why they were taking this medication were not specified. We reviewed the literature to determine the incidence of thromboem¬ bolic events, whether this incidence in¬ creases with more severe ventricular dysfunction, and the benefits and risks of anticoagulation in patients with heart failure. Except where indicated, the term "anticoagulation" refers to the use of coumarin-derivative drugs rather than ure

antiplatelet agents.

METHODS We searched the MEDLINE data¬ base (January 1966 to September 1993) using the Medical Subject Headings an-

ticoagulation, cerebrovascular disor¬ ders, stroke, and thromboembolism and the set of terms heart failure, conges¬ tive; congestive heart failure; heart fail¬ ure; cardiac failure; and dilated cardiomyopathy. No clinical trials were found

among these citations. A total of 11 ar¬ ticles dealt specifically with the topic of anticoagulation for heart failure. We used the references from these articles along with textbooks and review articles to identify another 40 articles for re¬ view. In addition, we reviewed the V-

HeFT trials,9·21·22 the CONSENSUS trial,23 and the SOLVD treatment trial.20 All studies of patients with chronic

heart failure due to left ventricular sys¬ tolic dysfunction were included. Studies of patients with valvular heart disease or heart failure secondary to acute myo¬ cardial infarction or Chagas' disease were excluded. Studies were reviewed for patient inclusion criteria, demograph¬ ics, mean left ventricular ejection frac¬ tion, the proportion of patients with atrial fibrillation, the mean follow-up, and the reported incidence of arterial embolism. If the incidence per 100 pa¬ tient-years was not reported, we esti¬ mated this from the proportion (p) of patients who had arterial thromboem¬ boli and the mean follow-up period (t) using the formula:

incidence=[-ln(l-p)]/t.

A weighted average ofthe incidence was calculated using the inverse of the vari¬ ance as the weight for each study. Only

patients

not

receiving anticoagulants

included. In addition, we reviewed studies that evaluated whether the in¬ cidence of arterial embolism varied with the severity of heart failure or left ven¬ tricular dysfunction, studies that re¬ ported data on the effectiveness of anticoagulation in patients with heart failure, and studies of the risks of anticoagulawere

tion in

patients with heart failure.

RESULTS The Incidence of Arterial Thromboembolism A total of 11 studies provided quanti¬ tative information on the incidence of ar¬ terial thromboembolism in patients with heart failure who were not receiving an¬ ticoagulants and who did not have docu¬ mented left ventricular thrombi (Table).110 All studies followed up patients prospec¬ tively except for the autopsy study by Roberts et al7 and the study by Kyrie et al8 on retrospective analysis of thromboembolic events in patients referred to the "thrombosis service." Five studies di¬ rectly reported the incidence of arterial thromboembolic events, and for three oth¬ ers we estimated the incidence using the number of events and the mean follow-up

period. The studies by Massumi et al,1 Segal et al,3 and Hatle et al4 reported the proportion of patients who had suffered

embolism, but the mean followup period was not reported and the in¬ cidence of embolie event could not be cal¬ culated. In the remaining studies, the in¬ cidence ranged from 0.9 to 5.5 per 100 patient-years, except for the study by Kyrie et al8 (42.4 per 100 patient-years). The weighted average of the reported incidence was 1.9 per 100 patient-years. In addition to these studies, the CON¬ SENSUS trial reported a 2.3 per 100 pa¬ an arterial

tient-years incidence offatal stroke,23 and an abstract on patients enrolled in the SOLVD treatment and prevention trials stated

a

total thromboembolic risk (in¬

cluding pulmonary embolism) of 1.6 per

100 patient-years.24 Two studies reported separate data for patients with and without atrial fi¬ brillation who were not anticoagulated. Fuster et al2 found that 33% of patients with atrial fibrillation and 14% of pa¬ tients without atrial fibrillation had em¬ bolie events during the follow-up period (mean, 11 years). None of these patients was anticoagulated. In the V-HeFT I and II trials 3.0% and 3.8%, respectively, of patients with atrial fibrillation had stroke or thromboembolism, and 5.0% and 5.6% Of those without atrial fibril¬ lation had events. However, this part of the analysis was not controlled for the use of anticoagulants or antiplatelet agents, which could explain the para¬ doxical results. Because of their large size, the rig¬ orous data collection system, and the heterogeneous origins of heart failure included, the incidence ofarterial throm¬ boembolism reported in the V-HeFT I and II studies is probably the most ac¬ curate for a general population of pa¬ tients with heart failure (2.4 and 2.0 per 100 patient-years, respectively). Inter¬ pretation of these data is still limited by the high prevalence of atrial fibrillation (16% and 15%, respectively) and the un¬ known proportion of strokes that re¬ sulted from comorbid cerebrovascular disease rather than cardiogenic emboli. All strokes were considered "possibly thromboembolic," and transient ische¬ mie attacks were counted as strokes. Strokes accounted for 83% and 95% of all the thromboembolic events in the two studies. The similar incidence reported in other studies support the generalizability ofthe V-HeFt findings. The only study with a markedly different incidence of emboli was by Kyrie et al,8 which estimated the incidence of embolism at 42.4 events per 100 patient-years. The patient population for this study was selected from referrals to the "thrombosis service," so it is likely

Downloaded From: http://jamanetwork.com/pdfaccess.ashx?url=/data/journals/jama/9269/ on 01/16/2017

that these patients were at much higher risk ofthromboembolic events than a gen¬ eral population of patients with heart fail¬ ure. In addition, both "definite" and "prob¬ able" embolie events were included, and 70% of arterial emboli were classified as

"probable."

The studies that enrolled only patients with idiopathic dilated cardiomyopathy reported incidence similar to that in the V-HeFT trials. Fuster et al2 evaluated 104 patients seen at the Mayo Clinic between 1960 and 1973 and reported an incidence of 3.5 arterial embolie events per 100 patient-years. Almost one half of the patients died within the first 2 years, so the study population may have had more severe or more complicated heart failure than a typical patient popu¬ lation. Ciaccheri et al5 found a some¬ what lower incidence than Fuster et al2 (1.4 per 100 patient-years), which may be attributable to the lower prevalence of atrial fibrillation in the studyby Ciac¬ cheri et al (12% compared with 23% in the series by Fuster et al2). The CONSENSUS study,23 which in¬ cluded only patients with New York Heart Association (NYHA) class IV heart failure, reported on fatal strokes but not the overall incidence of arterial embolism. Half of those enrolled had atrial fibrillation, but only one third of all patients were receiving anticoagu¬ lants at the start of the trial. Thus, it appears that at least one sixth of all patients in the trial had atrial fibrilla¬ tion that was not treated with antico¬ agulants. Three of 253 patients suffered a fatal stroke during the average followup period of 6 months, for an estimated incidence of 2.3 fatal strokes per 100 patient-years. The case fatality rate for stroke has been estimated at 30%,25 so the total stroke incidence is likely to be approximately three times the fatal stroke rate. It was not specified how many of the patients who suffered a stroke also had atrial fibrillation that was not treated with anticoagulation. Is the Risk of Arterial Thromboembolism Greater in Patients With More Severe Ventricular Dysfunction? We found no study that used multivariate techniques to evaluate the inde¬ pendent effect ofventricular dysfunction on the risk of arterial thromboembolism. Segal et al3 stratified study patients by their heart size on chest roentgenogram. Of patients with minimal or mild cardiac enlargement, 11% suffered systemic em¬ boli compared with 16% of those with moderate to severe enlargement (not sig¬ nificant). Kyrie et al8 also investigated the association between arterial embo¬ lism and the severity of left ventricular

Incidence of Arterial Thromboembolism ¡ Patients With

Congestive

Heart FaiUre Not

Receiving Anticoagulants* Arterial Thromboembolism

Study Years

Study Massumi et al1

1960-1963

Fuster et al2

1960-1973

et al3

1961-1978

Hatle et al"

1962-1972

Segai

Atrial Inclusion Criteria

Idiopathic dilated cardiomyopathy Idiopathic dilated cardiomyopathy Idiopathic dilated cardiomyopathy Signs of myocardial

No.

Age,

50

NR

y

Male, %

Mean EF

80

Fibrillation, 30

62

NR

23

Mean %

Follow-up,

y

% of Patients

NR

115

NR

62

NR

19

NR

106

NR

71

NR

36

NR

126

54

58

56

131

45

38

49

Incidence

(per 100 Patient-years) NR

18.0

3.5

12

NR NR

disease

Ciaccheri et al5

1980-1987

Gottdiener et al6

NR

Idiopathic dilated cardiomyopathy,

3.5

4.5

2.0

10.0

5.3t

11.04

2.6t

1.1

42.0

42.4

2.3

NR

2.4

clinical CHF

LV dilation and diffuse

100

NR

NR

NR

25

hypokinesis, no

Roberts et al7

Kyrie

1959-1981

NR

et al8

LV thrombus

Chart review and autopsy study for cases of idiopathic dilated cardiomyopathy

CHF, cardiomegaly,

or

89

decrease fractional

shortening on echocardiogram;

referred to thrombosis service Dunkman et al9

1980-1985

CHF and evidence of cardiac dilation

642

58

1986-1991

CHF and evidence of cardiac dilation

804

61

100

0.29

15

2.6

NR

1988-1991

Chronic CHF

229

62

61

0.27

NR§

2.0

1.1

(V-HeFT I) Dunkman et al9

(V-HeFT II) Katz et al10

0.9t

*EF indicates ejection fraction; NR, not reported; CHF, congestive heart failure; and LV, left ventricle. tEstimated incidence based on the mean duration of follow-up. ^Includes both definite and probable cases of arterial emboli. §The overall prevalence reported was 13%, but patients with atrial fibrillation were usually anticoagulated and therefore excluded from this analysis.

dysfunction using the echocardiographic fractional shortening as an index. The risk of embolism for patients with se¬ vere,

moderate,

or

mild left ventricular

dysfunction was 67% (six ofnine patients), 47% (nine of 19 patients), and 20% (two of 10 patients), respectively. Ciaccheri et al6 found that the 12 patients in their series who suffered an arterial emboli had a mean left ventricular ejection frac¬ tion of 0.24 compared with a mean ejec¬ tion fraction of 0.30 for those without

thromboembolic complications (not sig¬ nificant). There was no difference in NYHA scores between groups. The V-HeFT studies suggest that pa¬ tients with more severe heart failure have a higher risk of thromboembolic events.9 Patients who had thromboem¬ bolic events had lower ejection fraction and peak exercise oxygen consumption than patients who did not (29.6% vs 26.3% and 14.2 vs 12.6 mL/kgVmin1, respectively), although only the differ¬ ence in oxygen consumption was statis¬ tically significant. The high incidence of fatal stroke in the CONSENSUS trial, which only enrolled patients with se¬ vere heart failure, may also indicate that the risk of emboli increases with wors¬ ening heart failure. However, this is con¬ founded by the higher prevalence of atrial fibrillation in this study. Other studies have examined the re¬ lationship between ventricular function

and the prevalence of left ventricular mural thrombi. Patients with mural thrombi appear to be at increased risk for embolie events,10·26·27 so the occur¬ rence of mural thrombi may be a useful intermediate outcome. Yokota et al28 found that patients with a cardiac throm¬ bus had a lower fractional shortening and a lower ejection fraction than pa¬ tients without a thrombus (13.1% vs

17.9%, P