Author + information
- Received May 8, 2001
- Revision received September 21, 2001
- Accepted October 19, 2001
- Published online January 16, 2002.
- Robert Kelly, MD, MRCPI∗,
- Anthony Staines, PhD, MRCPI∥,
- Ron MacWalter, FRCP†,
- Peter Stonebridge, MCh, FRCS‡,
- Hugh Tunstall-Pedoe, MD, FRCP§ and
- Allan D Struthers, MD, FRCP∗,* ()
- ↵*Reprint requests and correspondence:
Prof. Allan D. Struthers, Department of Clinical Pharmacology, Ninewells Hospital, Dundee, Scotland DD1 9SY, United Kingdom.
Objectives We sought to determine the prevalence of treatable left ventricular (LV) systolic dysfunction (LVSD) in patients who present with their first noncardiac vascular episode.
Background Screening for LV dysfunction in patients who present with their first stroke (cerebrovascular accident), their first transient ischemic attack (TIA) or their first manifestation of peripheral vascular disease (PVD) may represent a golden opportunity to identify treatable LV dysfunction, and so their known high incidence of sudden cardiac death may be reduced.
Methods Participating in this study were 522 (75%) of 700 consecutive patients (302 patients with stroke, TIA or PVD and 220 age- and gender-matched control subjects). Each underwent a full clinical assessment, 12-lead electrocardiography and two-dimensional echocardiography. Left ventricular dysfunction was defined as LV ejection fraction ≤40%.
Results Seventy-two (28%) patients with vascular disease and 11 (5.5%) control subjects were found to have LVSD. Twenty-six (28%) stroke patients, 22 (26%) patients with TIA and 24 (31%) patients with PVD had LVSD. Left ventricular systolic dysfunction was symptomatic in 44% of patients and in 35% of control subjects.
Conclusions Left ventricular systolic dysfunction is five times more common among patients with stroke, TIA and PVD than among age- and gender-matched control subjects. Asymptomatic LVSD is more common than symptomatic LVSD in these patients. These findings suggest that routine screening of all patients with noncardiac vascular episodes for LVSD should now be considered. Future studies should investigate whether identifying and treating LVSD in these patients would reduce their known high rate of cardiac death.
Symptomatic heart failure is present in 0.4% to 2% of the population and continues to rise as the population ages (1). In order to substantially reduce and/or delay the incidence of heart failure and its consequences, we need to detect and treat presymptomatic left ventricular (LV) dysfunction.
In the Western world, heart failure is largely due to LV systolic dysfunction (LVSD). It is clear that angiotensin-converting enzyme (ACE) inhibitors and beta-blockers can prevent the progression of LVSD to heart failure and reduce morbidity and mortality (2–4). In addition, spironolactone, warfarin and implantable cardioverter-defibrillators (ICDs) are possible therapies in selected cases. Indeed, ICD placement is a new therapy that particularly benefits those with LVSD (5,6).
Left ventricular systolic dysfunction is found in 2% to 12% of the general population, with >60% of patients being asymptomatic (7). Previous screening studies have focused on the general population, but screening whole populations is unlikely to be cost-effective. More selective screening for LVSD is already routinely undertaken in post-myocardial infarction (MI) patients while they are in the coronary care unit. Patients who present with their first noncardiac vascular episode represent another high-risk group that might be worth screening for treatable LVSD. Currently, such patients in Europe will only get a routine echocardiogram if they have atrial fibrillation. Echocardiography resources are inadequate to image all such patients routinely, certainly in nonteaching hospitals. Previous surgical and angiographic studies of patients with noncardiac vascular episodes have found LVSD in 16% to 68% of patients with vascular disease who are awaiting carotid endarterectomy or femoral revascularization (8,9). All of these previous results have come from small surgical series of patients with vascular disease; therefore, they are heavily biased, because LVSD is likely to be a major factor causing vascular patients to be excluded from surgical series.
Thus, the first reason for wanting to know the extent of LVSD among patients with noncardiac vascular episodes is that such patients might have a great amount of LVSD, and it may be very cost-effective to screen for treatable LVSD in such patients. The second reason is that patients who have had noncardiac vascular episodes are known to be at exceptionally high risk of cardiac death, and because of that, we commonly optimize the treatment of conventional risk factors, such as blood pressure and cholesterol, in such patients. However, in clinical practice, we do not routinely seek a much bigger risk factor (i.e., LVSD) in these patients, despite the fact that specific treatments exist for LVSD that are known to reduce cardiac deaths (for example, a patient with a transient ischemic attack [TIA] could well have an arrhythmic death due to asymptomatic LVSD, despite optimal control of blood pressure and cholesterol). Better knowledge about the incidence of LVSD in such patients would provide information on these two issues. First, are such patients an enriched source of LVSD? This might raise the possibility of screening such patients for LVSD. Second, do these patients have enough LVSD to be an independent contributor to the high rate of cardiac death that they are known to experience? If so, screening for and vigorously treating asymptomatic LVSD in such patients may be a cost-effective way of reducing their currently high rate of cardiac death.
We report an epidemiologic survey, using two-dimensional echocardiography, to establish the prevalence and clinical characteristics of LVSD in patients presenting to the hospital with their first noncardiac vascular episode (i.e., stroke, TIA or overt peripheral vascular disease [PVD]), and we compare these patients with age- and gender-matched control subjects recruited from general practice.
A total of 700 men and women, aged 45 to 87 years, were invited to participate. We identified 400 patients with stroke, TIA or PVD at first presentation to Ninewells Hospital (Dundee, Scotland, United Kingdom), and we matched them with 300 age- and gender-matched control subjects recruited from the general practice population. Of the 700 subjects who were initially approached, 522 subjects agreed to participate (75% response rate). Those patients who declined to participate were generally older and male and had prolonged hospital stays for their vascular episode. Stroke patients were identified from the hospital’s stroke admission database, and this diagnosis was confirmed on computed tomographic scan. Patients with TIA included only those patients with classic symptoms. These patients were recruited from referrals to the vascular laboratory for ultrasound of the carotid artery. Their clinical history, not their ultrasound findings, determined inclusion in the study cohort. Patients with vague symptoms or a referral for investigation of dizzy spells/funny turns were excluded. Patients with PVD were also recruited from the vascular laboratory, based on a history of new-onset, intermittent claudication in conjunction with an ankle-arm index <0.8 at rest. Patients with PVD with acute leg ischemia were excluded. Any vascular patient with a previous stroke, TIA or PVD was excluded, as this would constitute their second noncardiac vascular event. A previous MI or history of coronary artery disease did not preclude any patient from being recruited. Age- and gender-matched control subjects were recruited from the general practice population. These subjects differed from patients solelyby the absence of a previous stroke or TIA or history of PVD. Control subjects were matched for age and gender according to our own hospitals’ MEMO database, which represents the general population (10). Patients with a cerebrovascular accident, TIA or PVD were removed from the control group.
Two-dimensional echocardiography (Sonos 2000, Hewlett-Packard, Palo Alto, California) was performed with each patient reclining at 40° in the left lateral position. Images were acquired and stored on videotape and analyzed on-line. The LV ejection fraction (LVEF) was measured by using the biplane summation method of discs (modified Simpson’s rule) (11). Each ejection fraction was taken as a mean value of three cardiac cycles. Echocardiograms were deemed acceptable if ≥75% of the endocardium was visible. Echocardiography was performed by a single observer (R. K.). A random sample of 50 echocardiograms was repeated, with the first results concealed by the same observer, and 25 of these were re-analyzed by a second observer. The interobserver variability was 8%, and the intraobserver variability was 8.9%.
Each patient underwent a full history and physical examination. Symptoms of breathlessness were recorded in accordance with New York Heart Association functional class. Each patient in this study gave written, informed consent. The local Tayside Ethics Committee approved the study.
Blood pressure was recorded with an automated sphygmomanometer. Each reading was taken as an average of three readings from both arms, with the patient seated after a 5-min rest.
Standard 12-lead electrocardiograms (ECGs) were recorded and classified for the presence of atrial fibrillation, atrial flutter, left bundle branch block, LV hypertrophy, pathologic Q waves and ischemia. Ischemia included ST-segment depression, any T-wave inversion and Q waves (12).
The clinical definitions that we used in this study are as follows. Ischemic heart disease (IHD)was defined as a history of angina, MI or taking anti-anginal medications. Hypertensionincluded a known history of such or a history of taking antihypertensive medications, or both, or blood pressure >160/95 mm Hg at presentation. Hypercholesterolemiareferred to a patient with known total serum cholesterol >5.2 mmol/l or a patient on a cholesterol-lowering diet, medication or both. Smokingreferred to patients with a history of cigarette smoking at the time of the noncardiac event or to current smokers. It did not include patients who had stopped smoking before their initial presentation to hospital. Alcohol excessincluded patients who admitted to alcohol consumption above the recommended weekly allowances. Left ventricular systolic dysfunctionwas defined as an ejection fraction ≤40%. Symptomatic LV dysfunctionwas defined by the presence of symptoms of cardiac dyspnea, ankle swelling, fatigue, orthopnea and paroxysmal nocturnal dyspnea. Asymptomatic LV dysfunction was defined by the absence of symptoms of cardiac dyspnea.
The data were analyzed using the STATA (X) statistics program (version 5, 1997, Timberlake Consultants Ltd., London, UK). The results were prepared separately for patients with vascular disease and control subjects. The results presented are univariate odds ratios. Proportions were compared between groups by using the chi-square test. Multivariate analysis was done using logistic regression after adjusting for age and gender. Model testing was done using graphs of residuals, leverages and areas under the receiver-operating characteristic curves.
We were able to measure the ejection fraction in 255 patients with stroke, TIA or PVD and in 202 control subjects. There were 245 (54%) men and 212 (46%) women. The epidemiologic characteristics of the patients with stroke, TIA and PVD and control subjects are shown in Table 1. Of patients with stroke, TIA or PVD, 27% had IHD, 48% had hypertension and 12% with diabetes, as compared with 16% of control subjects with IHD, 35% with hypertension and 5% with diabetes.
The mean LVEF in patients with stroke, TIA or PVD was 45 ± 8.9% (range 17% to 68%, 95% confidence interval [CI] 44% to 46%). In control subjects, the mean LVEF in all patients was 50.6 ± 6.9% (range 20% to 63%, 95% CI 49.7% to 51.6%). In those patients with LVSD, the mean LVEF was 34.5 ± 5.2% (range 16% to 40%, 95% CI 33.3% to 35.7%) in the stroke, TIA and PVD group, as compared with 33 ± 6.5% (range 20% to 40%, 95% CI 29.3% to 36.7%) in the control group.
Left ventricular systolic dysfunction was found in 72 patients with stroke, TIA and PVD (28%), as compared with 11 age- and gender-matched control subjects (5.5%). Twenty-six (28%) stroke patients, 22 (26%) patients with TIA and 24 (31%) patients with PVD had LVSD. In the stroke, TIA and PVD group, LVSD was more common in men than in women (20% vs. 9%), and in the control group, 4.6% of men versus 1% of women had LVSD. As age increased up to 75 years, so did the prevalence of LVSD (Figs. 1 and 2). ⇓⇓
Symptomatic LVSD was found in 44% of the stroke, TIA and PVD group and in 35% of the control group. Asymptomatic LVSD was found in 56% of the stroke, TIA and PVD group and in 65% of the control group.
Table 2outlines the risk factors for LVSD in both patient groups. Of the patients with stroke, TIA or PVD with LVSD, 54% had nohistory of IHD, whereas 91% of control subjects with LVSD had a history of IHD. The presence of diabetes mellitus and IHD, together with noncardiac vascular episodes, significantly increased the likelihood of LVSD. In the control group, IHD, diabetes and atrial fibrillation significantly increased the likelihood of LVSD. Atrial fibrillation alone did not predict LVSD in patients with stroke, TIA or PVD. Sixty-two patients with stroke, TIA or PVD (86%) and 9 control subjects (82%) with LVSD were in sinus rhythm (Table 2).
In both groups of patients, an abnormal ECG was much more likely to be associated with LVSD than with normal LV function. Of all patientswith stroke, TIA or PVD, 60% had abnormal ECGs, as compared with 29% of control subjects. In the patients with stroke, TIA or PVD with ECG evidence of IHD, 64% had LVSD, as compared with 36% with normal LV function (p < 0.001). In control subjects with an ischemic ECG, 82% had LVSD and 19% had normal LV function (Table 3).
In a more detailed logistic regression analysis, the independent, useful predictors of LVSD were IHD and an abnormal ECG, particularly an ischemic ECG, in the presence of a noncardiac vascular event (Table 4). In patients with more severe LVSD (ejection fraction <35%), a history of IHD and an abnormal ECG remained significant risk factors for development of LVSD in both patients and control subjects, as the severity of LV dysfunction progressed.
Hypercholesterolemia was observed in more patients with stroke, TIA or PVD than in control subjects, but it did not predict LVSD. Excessive alcohol consumption was more common among patients with normal LV function than among those with LVSD (7% vs. 4%). Alcohol excess was less frequent among patients with stroke, TIA or PVD than among control subjects (6% vs. 12%).
Of the patients with stroke, TIA or PVD, 60% were taking aspirin. Only 24% of control subjects with LVSD were taking an ACE inhibitor, a diuretic, a beta-blocker or combination of these medications. Of the control subjects with LVSD, 52% were taking a diuretic, an ACE inhibitor, a beta-blocker or a combination of these.
Left ventricular systolic dysfunction is extremely common in patients who present to the hospital with their first stroke or TIA or their first presentation of PVD. We have found that 28% of these patients have LVSD, as compared with 5.5% of age- and gender-matched control subjects. This amounts to five times more LVSD in patients with noncardiac vascular disease than in the general population. As with studies in Glasgow and Southampton, asymptomatic LVSD was found in almost two-thirds of all patients with LVSD (7,13). As for patients with stroke, TIA or PVD, there are no previous studies available to make a direct comparison with our study. It should be remembered that many patients with stroke or PVD will often be unable to exercise to become symptomatic, and this may increase the prevalence of “hidden” asymptomatic LVSD in these patients.
If we classify LVSD by its severity, the prevalence of mild LVSD was much greater than that of severe LVSD in both study groups. In comparison to the population studies of McDonagh et al. (7)and Morgan et al. (13), we have certainly identified much more LVSD in our high-risk patients than they found in their population screening.
Prevalence of LVSD as defined by echocardiography
We chose to define LVSD as LVEF ≤40%. We prespecified this value as our cutoff, because it represents the value used in many ACE inhibitor studies of patients with LV dysfunction, and this is the level of function below which ACE inhibitor therapy is commonly prescribed. Unfortunately, most studies of LVSD using echocardiography have selected different cutoff values to represent LVSD (7,13). We chose 40% as a compromise between some studies that used LVEF <30% to 35% and some that used LVEF <45%. The prevalence of LVSD would be higher (48%) if the cutoff value were increased to LVEF <45%, as in the Southampton study (13). It will be interesting to learn the follow-up analysis of this intermediate group (LVEF 40% to 45%) of patients to see whether they also progress to heart failure, and its resultant mortality and morbidity. Previous studies have shown that echocardiographic measurements of LV function correlate extremely well with both radionuclide and contrast ventriculographic measurements (14). Interestingly, the prevalence of LVSD in our patients with stroke, TIA or PVD was even greater than that detected by open-access echocardiography clinics in the United Kingdom, where patients suspected of having LVSD are referred by their family doctor. Most open-access echocardiography clinics report a 20% to 32% prevalence of LVSD in selected populations (15). In patients taking diuretics, the prevalence of LVSD is usually around 21% (16).
Clinical significance of LVSD in noncardiac vascular disease
A key point concerning patients with noncardiac vascular episodes is that they are well known to have a high incidence of subsequent cardiac death (17). The assumption has generally been made that their cardiac death is due to fresh ischemia/infarction, but clearly it is also distinctly possible, in view of these results, that some cardiac deaths are “arrhythmic,” due to LVSD, in which case, beta-blockade or even ICD therapy might reduce cardiac deaths in those patients with asymptomatic LVSD. In addition, ACE inhibitors might slow down and/or prevent the development of overt heart failure and hospital admissions. Furthermore, the judicious use of warfarin in patients with LVSD with previous vascular episodes might prevent further embolic events, even in those in sinus rhythm, as it appeared to do in the Survival And Ventricular Enlargement (SAVE) trial (2). Indeed, it is ironic that after a noncardiac vascular episode, it is standard practice to optimize the treatment of distant risk factors, such as blood pressure and cholesterol, and yet we ignore whether a bigger risk factor, like LVSD, already exists. This is even stranger when one considers that LVSD is easily detected by a noninvasive test and easily treated by beta-blockers, ACE inhibitors, spironolactone, warfarin, ICDs or a combination of these.
The recent Heart Outcomes Prevention Evaluation (HOPE) study has shown that treating all high-risk patients with ramipril will reduce cardiac events (18). However, we must remember that, so far, the HOPE study is the only study to show such benefits. Indeed, in the QUinapril in Ischemic Events Trial (QUIET), quinapril did not prevent MI or reduce ischemic events in patients with coronary artery disease with normal LV function (19). Most commentators have suggested that we await the results of the European trial on Reduction of Cardiac Events with Perindopril in Stable Coronary Artery Disease (EUROPA) and the Prevention of Events with Angiotensin-Converting Enzyme inhibition (PEACE) study before we consider putting all vascular patients on ACE inhibitors, independent of their LV function (20,21). However, even if the HOPE strategy were implemented in full, this does not invalidate our suggestion that LVSD should be identified in these patients, because a diagnosis of LVSD will influence whether other non-ACE inhibitor therapies are initiated in these patients (i.e., the possibility of using beta-blockade, ICDs, spironolactone and warfarin) and increases the importance of differentiating between LVSD and normal LV systolic function in these patients (4,22–24).
Should we screen for LVSD?
The large number of patients with LVSD that we have identified by echocardiography suggests that screening of patients with stroke, TIA or PVD for LVSD should be considered. If cost constraints make this difficult, screening could be selective by focusing on those with abnormal ECGs and/or a history of MI, but we have shown that echocardiographic screening should to be done in more patients than just those with atrial fibrillation, which is the current practice. A significant amount of cardiac deaths might be preventable by identifying and aggressively treating LVSD in patients who present with their first noncardiac vascular episode.
☆ This study was supported by a grant from the Scottish Health Office, Scotland, United Kingdom.
- angiotensin-converting enzyme
- confidence interval
- Heart Outcomes Prevention Evaluation study
- implantable cardioverter-defibrillator
- ischemic heart disease
- left ventricular
- left ventricular ejection fraction
- left ventricular systolic dysfunction
- myocardial infarction
- peripheral vascular disease
- transient ischemic attack
- Received May 8, 2001.
- Revision received September 21, 2001.
- Accepted October 19, 2001.
- American College of Cardiology
- Kelly R,
- Struthers A.D
- the SAVE Investigators,
- Pfeffer M.A,
- Braunwald E,
- Moye L.A,
- et al.
- the MERIT-HF Study Group,
- Hjalmarson A,
- Goldstein S,
- Fagerber B,
- et al.
- Moss A.J
- National Institute for Clinical Excellence
- Evans J.M.M,
- McDevitt D.G,
- MacDonald T.M
- Davie A.P,
- Love M.P,
- McMurray J.J
- Morgan S,
- Smith H,
- Simpson I,
- et al.
- Hampton J
- Francis C.M,
- Caruana L,
- Kearney P,
- et al.
- Dennis M.S,
- Burn J.P,
- Sandercock P.A,
- et al.