Author + information
- Received August 10, 2007
- Revision received October 29, 2007
- Accepted November 21, 2007
- Published online April 1, 2008.
- Brian Olshansky, MD, FACC⁎,⁎ (, )
- Jeanne E. Poole, MD, FACC†,
- George Johnson, BSEE∥,
- Jill Anderson, RN∥,
- Anne S. Hellkamp, MS§,
- Douglas Packer, MD, FACC‡,
- Daniel B. Mark, MD, MPH, FACC§,
- Kerry L. Lee, PhD§,
- Gust H. Bardy, MD, FACC∥,
- SCD-HeFT Investigators
- ↵⁎Reprint requests and correspondence:
Dr. Brian Olshansky, Professor of Medicine, University of Iowa Hospitals, 200 Hawkins Drive, Iowa City, Iowa 52242.
Objectives The outcome of congestive heart failure (CHF) patients with syncope is understood incompletely.
Background We analyzed data from patients enrolled in the SCD-HeFT (Sudden Cardiac Death Heart Failure Trial) to determine whether syncope predicted outcomes in patients with CHF.
Methods We compared outcomes (and associated clinical characteristics) in patients with and without syncope enrolled in SCD-HeFT.
Results In SCD-HeFT, 162 (6%) patients had syncope before randomization, 356 (14%) had syncope after randomization (similar incidence in each randomized arm), and 46 (2%) had syncope before and after randomization. A QRS duration ≥120 ms and absence of beta-blocker use predicted syncope during follow-up (hazard ratio [HR] 1.30 and 95% confidence interval [CI] 1.06 to 1.61, p = 0.014 and HR 1.25, 95% CI 1.01 to 1.56, p = 0.048, respectively). Syncope recurrence did not differ by randomization arm. However, in the implantable cardioverter-defibrillator (ICD) arm, syncope, before and after randomization, was associated with appropriate ICD discharges (HR 1.75, 95% CI 1.10 to 2.80, p = 0.019 and HR 2.91, 95% CI 1.89 to 4.47, p = 0.001, respectively). Post-randomization syncope predicted total and cardiovascular death (HR 1.41, 95% CI 1.13 to 1.76, p = 0.002 and HR 1.55, 95% CI 1.19 to 2.02, p = 0.001, respectively). The elevated relative risk of mortality for syncope versus nonsyncope patients did not vary significantly across treatment arms (ICD, HR 1.54, 95% CI 1.04 to 2.27; amiodarone, HR 1.33, 95% CI 0.91 to 1.93; and placebo, HR 1.39, 95% CI 0.96 to 2.02, test for difference p = 0.86).
Conclusions For CHF patients with ICDs, syncope was associated with appropriate ICD activations. Syncope was associated with increased mortality risk in SCD-HeFT regardless of treatment arm (placebo, amiodarone, or ICD). (SCD-HeFT Trial; NCT00000609)
Syncope is caused by conditions ranging from benign, self-limiting autonomic fluctuations to chronic, recurrent, and potentially fatal arrhythmic or cardiac mechanical causes (1). Syncope associated with cardiovascular disease may have a poor prognosis (1–3). Patients with syncope and inducible ventricular tachycardia at electrophysiology testing have a similar outcome to those patients with inducible ventricular tachycardia who have been resuscitated from cardiac arrest (4). Patients with syncope and inducible ventricular arrhythmias who undergo implantable cardioverter-defibrillator (ICD) implant use their ICDs (5–9).
The risk of death in heart failure patients with syncope is present even when the electrophysiology study is negative (10–12). Congestive heart failure appears to be an independent predictor of mortality in patients with syncope (13). Retrospective data suggest that patients with syncope and cardiomyopathy who undergo ICD implantation, whether or not they have a positive electrophysiology test, get appropriate ICD activations (9,14,15) and that this may reduce the risk of death (16–18), although the data on this are not definitive and the risk of death remains high (12,19). Therefore, ICDs have been recommended in heart failure patients with syncope (7,20). Data regarding syncope in patients with congestive heart failure are based largely on small single-site studies of highly select patients who often are not treated with standard medical therapy for heart failure.
The SCD-HeFT (Sudden Cardiac Death in Heart Failure Trial) offered an unusual opportunity to examine some of the outstanding questions about syncope in congestive heart failure.
Major trial results and patient population
The SCD-HeFT was a large, multicenter, randomized, controlled clinical trial of ischemic and nonischemic cardiomyopathy patients with New York Heart Association (NYHA) functional class II and III congestive heart failure who had left ventricular ejection fraction ≤35% who were treated with standard medical therapy for heart failure (including beta-adrenergic blocking drugs and vasodilators) and subsequently were randomized to receive amiodarone, a placebo pill, or a single-chamber, shock-only, ICD (programmed VVI-34) (21). The predetermined outcome of the trial was total mortality. The methodology and results are reported elsewhere (21).
All patients enrolled in SCD-HeFT were included in this analysis. For the purpose of this analysis, syncope was defined as a sudden loss of consciousness with loss of postural tone and with rapid complete recovery without need for cardioversion or defibrillation. Syncope was identified before enrollment and during the follow-up period (“post-randomization”).
The clinician’s judgment established the presumptive cause of syncope. No algorithm or guidance was given to manage or diagnose syncope. No standardized definitions were used for syncope diagnoses. Data were not collected systematically that could define temporal relationships between syncope and ICD shocks, heart failure exacerbations, electrocardiogram (ECG) changes, ventricular or supraventricular arrhythmias, heart block, or transient hypotension.
For a patient to be enrolled in SCD-HeFT, any episodes of syncope occurring before randomization needed to be considered by the investigator unrelated to any life-threatening ventricular arrhythmia (otherwise, the patient would have had an indication for an ICD and could not be included in the study). No registry was kept of patients excluded from SCD-HeFT for this reason. Preferably, any identified episodes of syncope were present in the distant past. Despite this, it was possible that pre-randomization syncope was misdiagnosed as a benign problem. Outcomes included “appropriate” ICD shocks (shocks for ventricular tachycardia or ventricular fibrillation as assessed by a blinded adjudicating committee) (22), cardiovascular mortality, sudden cardiac death, and all-cause mortality.
Predictors of post-randomization syncope were evaluated both univariably with single-predictor Cox proportional hazards survival models (referred to hereafter as “Cox model”) and multivariably, in a single Cox model that included candidate predictors used in the main trial (i.e., functional class, ejection fraction, heart failure etiology, atrial fibrillation, nonsustained ventricular tachycardia, QRS duration, and standard cardiovascular medications). Ventricular pacing was not included because the amount of ventricular pacing was nearly absent (but was not quantitated) in the ICD arm (programmed at 34 beats/min). Recurrent syncope, defined as more than 1 syncope occurrence during follow-up, was compared among treatment arms with a logistic regression model adjusted for heart failure etiology (ischemic or nonischemic) and NYHA functional class.
Association of post-randomization syncope with mortality, cardiovascular death, and sudden death was evaluated with Cox models. Both unadjusted (syncope as the only predictor) and adjusted (including other mortality predictors) were used to ascertain how much of syncope’s relationship with mortality risk was driven by other risk factors associated with mortality. Adjusted models included other baseline prognostic factors identified in SCD-HeFT (randomized treatment, age, gender, heart failure etiology, NYHA functional class, time since heart failure diagnosis, ejection fraction, 6-min walk distance, systolic blood pressure, diabetes, angiotensin-converting enzyme inhibitor use, digoxin use, presence of mitral regurgitation, renal insufficiency, substance abuse, baseline ECG intervals, and the Duke Activity Status Index). Other factors that appeared imbalanced between syncope and nonsyncope patients also were evaluated but were not retained in the final model if they were found to be unrelated to mortality risk. Because post-randomization syncope could occur at any time during follow-up, it was included in these models as a time-dependent covariate. In addition, an interaction test was used in the all-cause mortality model to determine whether the relationship of post-randomization syncope to mortality risk differed across randomized treatment arms, and pre-randomization syncope was added to the model as a separate variable to determine its relationship to mortality risk. Among patients who received an ICD, the associations of pre- and post-randomization syncope to appropriate shock therapy were assessed using an unadjusted Cox model.
Hazard ratios, generated by Cox models, took the entire follow-up period into account. The reported 2.5-year Kaplan Meier rates (chosen because 75% of the cohort had that much follow-up) were intended to give an indication of the absolute rates of syncope, which hazard ratios do not provide.
Syncope before or after randomization in SCD-HeFT occurred in 19% (472) of enrolled subjects during a median follow-up of 45.5 months. Presumptive causes for syncope are listed in Table 1.
Before randomization, 162 patients (6%) had syncope. After randomization, 356 patients (14%) had at least 1 episode of syncope. Forty-six of these patients (2%) had syncope before and after randomization. The median (25th, 75th percentiles) time between the episode of syncope and randomization into SCD-HeFT was 11 (4, 39) months. The median (25th, 75th percentile) time from randomization to the first episode of syncope was 15 (5, 28) months.
Predictors of syncope post-randomization
In univariable tests, NYHA functional class III, QRS duration ≥120 ms, and absence of beta-blocker use were predictors of post-randomization syncope (Table 2). Variables unrelated to syncope risk were randomization arm, heart failure etiology, left ventricular ejection fraction, atrial fibrillation or flutter, nonsustained ventricular tachycardia, and baseline angiotensin-converting enzyme inhibitor, warfarin, or statin use. In the multivariable model, the only significant predictors were QRS duration ≥120 ms (hazard ratio [HR] 1.30, 95% confidence interval [CI] 1.06 to 1.61, p = 0.014) and the absence of beta-blockers (HR 1.25, 95% CI 1.01 to 1.56, p = 0.048).
Patients with syncope before randomization had median (25th, 75th percentile) systolic blood pressure at baseline of 118 (108, 130) mm Hg compared with patients without syncope before randomization, who had a systolic blood pressure of 118 (105, 131) mm Hg (p = 0.90). When this was added to the mortality model, blood pressure had no impact on the estimates for either pre- or post-randomization syncope.
Seventy-one patients had more than 1 episode of syncope (3% of all enrolled patients). Twenty-three patients in the ICD arm, 25 patients in the placebo arm, and 23 patients in the amiodarone arm had syncope recur. There were no differences in recurrence rates between arms (p = 0.95).
Association of syncope with appropriate ICD shocks
In SCD-HeFT, 811 patients received an ICD. Of these patients, 6% (52) had syncope before randomization and 20 (38%) of those with syncope before randomization had an appropriate shock. In contrast, 19% (157) with no syncope had an appropriate shock. Syncope before randomization was associated with a greater likelihood of an appropriate ICD shock (HR 1.75, 95% CI 1.10 to 2.80, p = 0.019).
In the ICD arm, 16% (128) had syncope after randomization and 52 of those (41%) with syncope had an appropriate shock, although the syncope occurred before the shock in only 27 patients (21% of the total and 52% of those having an appropriate shock). Of these 27 patients, 18 had syncope and a shock in the same month. In the remaining 9, the shock was 6 to 41 months after syncope. In contrast, of those who had no syncope after randomization, 125 (12%) had an appropriate shock. Syncope after randomization was associated with increased risk of an appropriate ICD shock (HR 2.91, 95% CI 1.89 to 4.47, p = 0.001) (23,24).
Syncope and mortality
Syncope occurring before randomization was not associated with death (HR 0.98, 95% CI 0.73 to 1.33; p = 0.91). However, syncope occurring after randomization was associated with increased risk (Table 3). Patients with syncope, compared with those without syncope, had a higher risk of all-cause mortality (HR 1.41, 95% CI 1.13 to 1.76, p = 0.002) and a greater risk of cardiovascular death (HR 1.55, 95% CI 1.19 to 2.02, p = 0.001). Risk of sudden cardiac death was not significantly greater (HR 1.41, 95% CI 0.90 to 2.21, p = 0.13). Unadjusted HRs for post-randomization syncope were 1.61, 1.80, and 1.54 for death, cardiovascular death, and sudden cardiac death, respectively. These are only slightly greater than the adjusted HRs, indicating that the relationship between syncope and mortality risk was largely independent of other correlated factors.
Importantly, syncope’s association with mortality risk was independent of randomization arm (interaction p = 0.86). Syncope HRs were similar for patients randomized to amiodarone (HR 1.33, 95% CI 0.91 to 1.93), to placebo (HR 1.39, 95% CI 0.96 to 2.02), and to ICD (HR 1.54, 95% CI 1.04 to 2.27). A mechanistic relationship between syncope and mortality was not established.
Syncope was common in the SCD-HeFT population, a population well-treated with standard medical therapy for heart failure. Post-randomization syncope was associated with increased risk of all-cause mortality, cardiovascular mortality, and sudden cardiac death (despite randomization to an ICD). Those patients randomized to an ICD, who had syncope, were more likely to receive appropriate ICD shocks than those without syncope; yet, a single-chamber, shock-only ICD with backup ventricular pacing did not protect patients against syncope and did not protect against the risk of death.
Syncope has been associated with adverse outcomes in patients with structural heart disease, especially when cardiomyopathy and congestive heart failure are present (1,25–27) and even in some patients without heart failure (28). This is true whether or not ventricular arrhythmias are inducible at electrophysiology testing (1,5,8,17,25–27,29).
The results of our analyses provide several important and novel observations. First, syncope after randomization in SCD-HeFT was a risk factor for mortality, and those subjects randomized to an ICD had similar outcomes compared with those randomized to amiodarone or placebo. Second, syncope in heart failure patients is a risk factor for appropriate ICD discharge.
In SCD-HeFT, excess mortality for syncope patients compared with the nonsyncope patients was not ameliorated by a single-chamber, shock-only ICD. Syncope, a marker for mortality in this population, may be due to hemodynamic collapse and may represent evidence for a poorer substrate that was not determined by other means. Although the mechanism for syncope was uncertain, it remains an important predictor and the excess risk is not altered by amiodarone or ICD intervention.
Several reports implicate a greater risk for arrhythmic episodes in patients with heart failure, cardiomyopathy, and syncope. Sudden death due to a life-threatening arrhythmia is suspected to be the reason that syncope patients who have structural heart disease and congestive heart failure die. The tacit assumption is that arrhythmias responsible for syncope may contribute to the increased mortality of these syncope patients. Several small, single-center, nonrandomized, retrospective analyses supporting this thesis have had substantial influence on therapies and have affected guideline development (12,30,31), despite the fact that these reports are very small, not comparative, and not randomized.
The benefits of ICD therapy and of a standard medical regimen for heart failure in the syncope patient remain uncertain. One recent nonrandomized retrospective report evaluated 51 patients with unexplained syncope, cardiomyopathy, and a negative electrophysiology test. Those who underwent an ICD implant had a lower probability of death or cardiac arrest compared with those who did not (17). It is likely that syncope and death are due to hemodynamic collapse rather than an arrhythmia in many heart failure patients (10).
The mechanisms responsible for and the relationships between syncope and death in heart failure patients are generally obscure. Syncope may be an indicator of an unstable, end-stage, cardiomyopathic process in this patient population and, therefore, predicts a poor prognosis. Death in patients with syncope may be due to sudden hemodynamic collapse rather than an arrhythmia. In SCD-HeFT, patients received what were considered optimal doses of angiotensin-converting enzyme inhibitors, beta-adrenergic blockers, spironolactone, and other standard medications for congestive heart failure. The doses were titrated, as tolerated, for the patients, generally by heart failure physicians. Abrupt, unexpected hypotension due to, or in lieu of, medical therapy is still possible.
Perhaps, syncope was not due to malignant ventricular arrhythmias for many SCD-HeFT patients. Rather, a malignant ventricular arrhythmia, like syncope, may be another marker for a more advanced cardiovascular condition and just associated with syncope. Patients in the SCD-HeFT with syncope had a greater risk of appropriate ICD shocks, yet ICDs (with back-up bradycardia pacing) did not protect these patients from dying.
It is even possible that a malignant ventricular arrhythmia was exacerbated by the ICD (32) or that ICD shocks facilitated death. On the other hand, even for patients without an ICD, the mortality was high. Perhaps ICD shocks are a marker for a poor prognosis (24).
The SCD-HeFT syncope data represent the largest collection of patients with syncope, cardiomyopathy, and heart failure treated with or without an ICD. The SCD-HeFT offered the opportunity to assess the incidence of syncope and subsequent outcomes based on randomization to amiodarone, to a placebo, or to an ICD. The high incidence of syncope in this population likely testifies to the fact that heart failure patients, such as those enrolled in SCD-HeFT, can be at immanent risk of hemodynamic collapse. No specific therapy has been shown to improve outcomes in these syncope patients. In fact, ICD shocks may indicate a poor prognosis as they correlate with poorer heart function (unpublished data, Poole JE, 2008).
The causes of syncope in this analysis of SCD-HeFT were presumptive and not determined definitively. The diagnosis was not determined by an algorithmic approach. The evaluation of syncope was not uniform or mandated. The SCD-HeFT was not designed specifically to address issues related to syncope or its mechanism.
This paper does not address all issues regarding syncope in heart failure such as why syncope is associated with a particularly bad prognosis. Further studies may be needed, but this report is the first to show that patients with syncope, cardiomyopathy, and heart failure die at the same rate whether they receive an ICD or not.
Syncope is common in a well-treated heart failure population. After randomization, syncope in SCD-HeFT patients was associated with an increased mortality risk. In the ICD arm, syncope predicted appropriate ICD shocks, but ICDs did not protect against and may have predicted death.
Dr. Olshansky is a consultant, an investigator, on the Speaker’s Bureau, and has participated in scientific studies for Medtronic; is a consultant, an advisor, an investigator, on the Speaker’s Bureau, and has participated in scientific studies for Boston Scientific and Guidant; is a consultant and advisor for CV Therapeutics; is an advisor for BioControl; is on the Speakers’ Bureau for Reliant/Baxter; and is a consultant for Blackwell. Dr. Poole is on the Speakers’ Bureau of Medtronic, Guidant, and St. Jude Medical. Dr. Packer is a consultant, an investigator, on the Speakers’ Bureau, and has participated in scientific studies for Medtronic; is an investigator for and has participated in scientific studies for Biosense Webster, Boston Scientific, and Endocardial Solutions; is a consultant, an investigator, on the Speakers’ Bureau, and has participated in scientific studies for Cryocath; is an investigator, on the Speakers’ Bureau, and has participated in scientific studies for Guidant; is a consultant for Cryocor; and is on the Speakers’ Bureau for Berlex. Dr. Mark is a consultant, an investigator, and has participated in scientific studies for Medtronic. Dr. Lee is a consultant/advisor and has participated in scientific studies for Medtronic. Dr. Bardy has received research grants from Medtronic and the National Institutes of Health; is on the Board and has equity and intellectual property in Cameron Health; and has done consulting for Philips.
- Abbreviations and Acronyms
- confidence interval
- hazard ratio
- implantable cardioverter-defibrillator
- New York Heart Association
- Received August 10, 2007.
- Revision received October 29, 2007.
- Accepted November 21, 2007.
- American College of Cardiology Foundation
- Steinberg J.S.,
- Beckman K.,
- Greene H.L.,
- et al.
- Menon V.,
- Steinberg J.S.,
- Akiyama T.,
- Beckman K.,
- Carillo L.,
- Kutalek S.
- Andrews N.P.,
- Fogel R.I.,
- Pelargonio G.,
- Evans J.J.,
- Prystowsky E.N.
- Link M.S.,
- Costeas X.F.,
- Griffith J.L.,
- Colburn C.D.,
- Estes N.A. 3rd.,
- Wang P.J.
- Middlekauff H.R.,
- Stevenson W.G.,
- Stevenson L.W.,
- Saxon L.A.
- Knight B.P.,
- Goyal R.,
- Pelosi F.,
- et al.
- LeLorier P.,
- Krahn A.D.,
- Klein G.J.,
- Yee R.,
- Skanes A.C.
- Packer D.L.,
- Bernstein R.,
- Wood F.,
- et al.
- Poole J.E.,
- Johnson G.,
- Callans D.J.,
- et al.
- Poole J.E.,
- Johnson G.W.,
- Hellkamp A.S.,
- et al.
- Shen W.K.,
- Decker W.W.,
- Smars P.A.,
- et al.
- Olshansky B.,
- Mazuz M.,
- Martins J.B.
- Pires L.A.,
- May L.M.,
- Ravi S.,
- Parry J.T.,
- Lal V.R.,
- Nino C.L.
- Strickberger S.A.,
- Benson D.W.,
- Biaggioni I.,
- et al.