Syncope in High-Risk Cardiomyopathy Patients With Implantable Defibrillators: Frequency, Risk Factors, Mechanisms, and Association With Mortality

Introduction

Syncope in heart failure patients with or without an implantable cardioverter-defibrillator (ICD) may be associated with a poor prognosis regardless of the cause of the syncope, but syncope caused by life-threatening ventricular arrhythmias is associated with an increased risk of death.^1–4 Because no large study in high-risk patients with heart failure has specifically addressed syncope as a safety end point, there is a lack of evidence on which factors are associated with all-cause syncope, arrhythmogenic syncope, and nonarrhythmogenic syncope.

Clinical Perspective on p 552

Furthermore, it is not known whether slow ventricular tachycardia (VT) (170–199 bpm) may induce hemodynamic instability and syncope in high-risk heart failure patients or whether the duration of ventricular arrhythmias until ICD treatment is initiated affects the frequency of syncope. The design of the Multicenter Automatic Defibrillator Implantation Trial–Reduce Inappropriate Therapy (MADIT-RIT)⁵ allowed us to answer these questions.

In MADIT-RIT, syncope was a prespecified safety end point, because life-threatening ventricular arrhythmias may precipitate syncope before termination/conversion of the rhythm by device therapy and thereby compromise patient safety. It was anticipated that the rate of syncope might be increased in the treatment arms that were programmed with a high-rate cutoff before therapy or programmed to delayed therapy. There was a similar incidence of all-cause syncope across the treatment arms⁵; however, the origin and cause of syncope were not analyzed separately for the safety end point.

In heart failure patients with defibrillators who experience syncope, the ICD device is interrogated to establish rhythm-symptom correlation. If no appropriate ICD therapy is given or no malignant ventricular rhythm is recorded in relation to the syncopal event, the syncope may be deemed nonarrhythmic. For both the physician and the patient, it is important to know whether nonarrhythmogenic syncope is related to an adverse outcome. Furthermore, it is unclear whether appropriate ICD therapy is associated with subsequent arrhythmogenic syncope.

In the present study, we aimed to evaluate (1) the effects of innovative ICD programming with either a high-rate cutoff VT zone or delayed therapy on risk of syncope compared with conventional programming; (2) the independent prognostic factors associated with syncope; and (3) the association between syncope, the cause of syncope, and the risk of death in patients enrolled in MADIT-RIT.

Methods

MADIT-RIT

The protocol and primary report of the MADIT-RIT study have been published previously.^5,6 The study included 1500 patients from 98 hospital centers with a primary prevention guideline indication to receive an ICD or CRT-D (cardiac resynchronization therapy defibrillator) device. The protocol was approved by the institutional review board at each participating center, and participants gave informed consent. Patients were excluded if they had experienced atrial fibrillation within 1 month before implantation; if they previously had been implanted with a pacemaker, ICD, or CRT-D; or if they had a recent myocardial infarction or revascularization procedure (within 3 months). Other exclusion criteria were described in the protocol publication. Follow-up continued until trial termination on July 10, 2012. Patients were randomly assigned to 1 of 3 ICD programming configurations for the detection and initiation of therapy for VT or ventricular fibrillation.

Device Programming and Monitor Zones

Standard transvenous implantation methods were followed and commercially available ICD and CRT-D dual-chamber devices (Boston Scientific, Natick, MA) were used in this trial. Briefly, the ICD programming configurations were set as follows. The conventional configuration (arm A) had a VT zone (first) at 170 to 199 bpm and a 2.5-second delay before antitachycardia pacing (ATP) and shock delivery, with a second zone at ≥200 bpm with a 1-second delay before ATP or shock. The high-rate programming configuration (arm B) had a first monitoring-only zone between 170 and 199 bpm and a second zone at ≥200 bpm with a 2.5-second delay before ATP or shock. The duration-delay configuration (arm C) had a first zone at ≥170 bpm with a 60-second delay and monitoring zone before therapy, a second zone at ≥200 bpm with a 12-second delay and monitoring zone before therapy, and a third zone at ≥250 bpm with a 2.5-second delay and monitoring zone before ATP or shock. For all VT zones, the first programmed therapy was ATP followed by shock therapies.

Bradycardia pacing was programmed as DDD with a lower rate of 40 bpm and hysteresis off. For CRT-D, pacing optimization was described in detail previously.⁶

Interrogation and Follow-Up

Patients were followed up every 3 months within the first year and then at 6-month intervals until trial termination. Device reprogramming was possible after the first occurrence of inappropriate therapy.

All device interrogations were independently reviewed by the interrogation adjudication committee, which was blinded to the programming arm assignments. The events were adjudicated on the basis of the committee’s interpretation of atrial and ventricular arrhythmias and classified as appropriate or inappropriate ATP and appropriate or inappropriate shock or a combination of those.

End Points

Syncopal events were a prespecified end point and were identified by the physicians at the enrolling centers and adjudicated by a 3-member independent morbidity and mortality committee using device interrogation and supporting source documents as provided by the enrolling centers. Syncope was defined as a total loss of consciousness with a rapid and spontaneous recovery with or without device therapy. Syncope with ventricular fibrillation (cardiac arrest) that was treated successfully with ICD shock was included in this definition of syncope (6 patients).

Arrhythmogenic syncope was defined as syncopal episodes with a direct rhythm-symptom correlation at adjudication and comprised a group of supraventricular arrhythmias, ventricular tachyarrhythmias, asystole, other arrhythmias, or unspecified arrhythmias. Nonarrhythmogenic syncope was classified as vasovagal, orthostatic, metabolic, or neurological transient loss of consciousness, with a separate category for unknown origin. For classification as nonarrhythmogenic syncope, rhythm-symptom correlation had been ruled out by interrogation of the devices at follow-up. The available information on the event as sent by the enrolling center, such as notes from the discharging physician, notes about symptoms, notes from emergency departments, and tests performed during an admission, if any, were taken into account in the adjudication of the event. Thus, the evaluation of syncope was not uniform or determined by an algorithmic approach, and no definitive tests were performed to allocate a patient into a category (eg, vasovagal or orthostatic hypotension syncope). MADIT-RIT was not specifically designed to address the mechanism of syncope.

All patient devices were programmed with a DDD lower rate of 40 bpm. For adjudication of syncope events, technical issues such as loss of ventricular capture or pacing inhibition caused by oversensing for pacemaker-dependent patients were taken into account.

Death was a prespecified end point and was adjudicated by the mortality committee by review of the documents and classification of death by the enrolling centers. A modified Hinkle-Thaler definition was used for the cause of cardiac death.⁷

Ventricular arrhythmias were a prespecified end point, and a 3-member independent electrogram and device-interrogation committee reviewed all interrogation events and adjudicated them into appropriate therapy (for ventricular arrhythmias) by ATP or shock and inappropriate therapy for all nonventricular arrhythmias based on tachycardia onset, morphology, and rate stability from the recorded electrogram. Furthermore, device interrogations were used to classify monitored (nontreated) ventricular and supraventricular arrhythmias based on rhythms with ≥30 beats.

Monitored events were gathered from the monitored ventricular arrhythmias in VT zones where no therapy was given in all treatment arms as recorded by the devices and as described under device programming.

Statistical Analysis

To evaluate the association of baseline characteristics with the incidence of syncope, univariable Cox proportional hazards regression models were fit.⁸ The comparison of distribution of syncope according to groups of causes was accomplished by log-rank test. The cumulative probability of syncope was displayed by the method of Kaplan–Meier using the log-rank test to compare cumulative event rates between treatment arms.

Univariable and multivariable Cox proportional hazards regression models were used to estimate the risk of all-cause syncope, arrhythmogenic syncope, and nonarrhythmogenic syncope as defined previously. In the multivariable model, we included variables found by stepwise selection from the pool of all baseline variables, setting the limits for entry into the model at 0.05. Four baseline variables had a significant association with the end point of all-cause syncope and thus were included in the multivariable Cox model, whereas only 2 variables had a significant association with the end point of arrhythmogenic syncope. To those 2 models we then added variables of time-dependent ICD therapy to examine the association between incident ICD therapy and subsequent risk of syncope. Variables including time-dependent appropriate therapy, appropriate ATP, appropriate shock, inappropriate therapy, inappropriate ATP, and inappropriate shock were tested separately to estimate whether they were significantly associated with the syncopal end points. A second model was constructed for the end point of all-cause mortality in a similar fashion and was adjusted for baseline variables that had a significant impact on the end point. To maintain a robust model, we chose only 1 variable per 10 events; thus, a maximum of 7 variables were considered for incorporation into the final model. To assess the association between all-cause syncope, arrhythmogenic syncope, and nonarrhythmogenic syncope throughout the study and the risk of death, a third model was created that used the time-dependent variables of all-cause syncope, arrhythmogenic syncope, and nonarrhythmogenic syncope, with separate evaluation of those in the multivariable model. Interaction terms were used systematically to interact with and analyze the time-dependent variables with the baseline variables and the randomization arms; no significant interactions were found, although there was limited statistical power for such an analysis. The assumption of proportional hazards was checked graphically by use of standard log(−log) survival density function plots, as well as by testing the interaction of covariates with follow-up time in the multivariable models. All assumptions were found to be valid. A 2-tailed P<0.05 was considered significant. Analyses were performed with the SAS statistical system version 9.3 (SAS Institute, Cary, NC).

Results

During a mean follow-up of 1.4±0.6 years, a total of 64 patients (4.3%) experienced at least 1 syncopal episode, with a mean time to event of 0.7±0.6 years. The overall 2.5-year cumulative probability of all-cause syncope was 8%. Syncope occurred evenly in all 3 treatment arms (in 21, 22, and 21 patients in arms A, B, and C, respectively; Figure 1). The clinical characteristics of the patients are shown in Table 1.

• Download figure
• Download PowerPoint

Cause of Syncope According to Randomization Arm

After adjudication, 21 syncopal events (33%) were classified as caused by VT or ventricular fibrillation and 4 (6%) as caused by other or unspecified arrhythmias, whereas a total of 39 events (61%) were classified as nonarrhythmogenic. In total, 11 of these nonarrhythmogenic events (17% of all events [11/64]) were classified as being of unknown cause, but only after an arrhythmogenic cause had been ruled out. Syncope caused by ventricular tachyarrhythmias did not differ between the 3 randomization arms, with 6 events in arm A, 6 in arm B, and 9 in arm C (log-rank test P=0.64; Figure 2). The adjudicated cause for first syncope is shown in Table 2. Two patients (3%) had syncope while driving (1 arrhythmogenic and 1 nonarrhythmogenic).

• Download figure
• Download PowerPoint

Recurrent Syncope

Of the 64 patients with a first episode of syncope, 16 (25%) had recurrent episodes, and a total of 85 syncopal events were registered during follow-up. The total number of recurrences was found to be 10 events in arm A, 4 in arm B, and 7 in arm C. Five of the recurrent syncopal episodes were caused by ventricular tachyarrhythmias (2 in arm A and 3 in arm C), of which 1 was a monitored VT episode with a heart rate >200 bpm and a duration of <12 seconds that terminated spontaneously, whereas the others were treated appropriately by either ATP or shock.

Frequency of Syncope According to First Appropriate Shock for Ventricular Arrhythmias

A total of 74 patients experienced first appropriate shock because of ventricular tachyarrhythmias. In 10 of these patients (14%), an appropriate ICD shock was preceded by syncope, thus rendering the patient unconscious for the subsequent shock. Of these 10 patients, 2 were randomized to arm A, 3 to arm B, and 5 to arm C. Among the remaining 64 patients who experienced first appropriate shock, 49 first shocks (77%) were given during waking hours (7 am to 11 pm). Therefore, when we eliminated the appropriate shocks given during sleeping hours, ≈17% (10/59) of the patients were unconscious when they received their first appropriate shock for either VT or ventricular fibrillation.

Distribution of Syncope According to Rate and Duration of VT and Incidence of Syncope According to Monitored and Treated Ventricular Tachyarrhythmias

Figure 3 shows the relationship between heart rate and duration of VT and syncope derived from the differences among the 3 different ICD programming settings from arms A, B, and C. Only 1 syncopal episode occurred because of VT with a rate between 170 and 199 bpm, whereas the number of syncopal events was essentially equally distributed depending on ventricular rate ≥200 bpm and ICD programming delays.

• Download figure
• Download PowerPoint

Among the treated or monitored first ventricular tachyarrhythmias in the heart rate range ≥200 bpm, 11 of 133 (8%) resulted in syncope, whereas 92% did not (Table 3). In other words, 10 patients (48%; 10/21) who experienced syncope attributable to ventricular tachyarrhythmias had already experienced a ventricular tachyarrhythmic event during follow-up.

Factors Associated With All-Cause Syncope

Independent prognostic factors for all-cause syncope included ischemic cardiomyopathy, previous ventricular arrhythmias, left ventricular ejection fraction ≤25%, and younger age, although the latter 2 were only borderline significant (Table 4). With the exception of age, the baseline variables that were associated with all-cause syncope were factors associated with an increased cardiovascular risk profile. In multivariable analysis, appropriate ICD therapy at any time before syncope was also independently associated with increased risk of subsequent all-cause syncope (hazard ratio [HR], 2.27; 95% confidence interval [CI], 1.02–5.07; P=0.045). The increased risk observed was primarily driven by appropriate shock (HR, 3.33; 95% CI, 1.28–8.68; P=0.014; data not shown in Tables).

Risk Factors Associated With Arrhythmogenic and Nonarrhythmogenic Syncope

Baseline variables associated with arrhythmogenic syncope were ischemic cardiomyopathy and previous ventricular arrhythmias (Table 5). Appropriate ICD therapy was independently associated with later occurrence of arrhythmogenic syncope (HR, 5.89; 95% CI, 2.06–16.82; P=0.001) driven by appropriate shock (HR, 8.16; 95% CI, 2.57–25.93; P<0.001; data not shown in Tables). We were unable to establish any factors associated with nonarrhythmogenic syncope from baseline patient characteristics, medications, or ICD therapy delivered throughout the study.

Association With Mortality by All-Cause, Arrhythmogenic Syncope, and Nonarrhythmogenic Syncope

Table 6 shows the association between arrhythmogenic and nonarrhythmogenic syncope, as well as all-cause syncope, and the risk of death. All-cause syncope, arrhythmogenic syncope, and nonarrhythmogenic syncope were all significantly associated with increased risk of death.

Discussion

Syncope caused by a higher threshold for ICD firing or by delayed ICD firing was a concern when MADIT-RIT was designed, and therefore, syncope was a prespecified safety end point. The present study reveals 4 major findings.

First, hemodynamic instability and subsequent syncope attributable to slow ventricular arrhythmias between 170 and 199 bpm were rare and only occurred in 1 patient during a mean follow-up of 1.4 years.

Second, a high-rate cutoff and delayed activation of ICD treatment resulted in very few syncopal events directly caused by the setting of the ICD device. Only 2 patients experienced syncope caused by fast untreated VT, monitored for 2.5 and 12 seconds, respectively. Thus, the 3 monitored ventricular arrhythmias observed terminated spontaneously before any ICD treatment was initiated.

Third, both arrhythmogenic and nonarrhythmogenic syncope were significantly associated with increased risk of death. We could not establish any risk factors associated with the risk of nonarrhythmogenic syncope, which suggests a multifactorial causality. Patients with nonarrhythmogenic syncope are at high risk despite a “negative” ICD interrogation for ventricular tachyarrhythmias, which indicates that further evaluation and close follow-up of these patients is a necessity and is recommended.

Fourth, the present study supports that syncope in heart failure patients (with a defibrillator) is primarily vasovagal, orthostatic, or otherwise nonarrhythmogenic in mechanism and underscores the fact that the presence of heart disease (in this case, ischemic or nonischemic heart failure) does not dictate that syncope has a cardiac cause, as also demonstrated by Alboni et al.⁹

Mechanistically, these results further indicate that high-risk heart failure patients with moderate to severe heart failure symptoms and reduced ejection fraction are able to tolerate rather long durations of fast VTs, whereas slow VTs for practical purposes do not result in a loss of consciousness. A small retrospective study supports this and likewise found that no patients with slow VT (<187 bpm) experienced syncope.¹⁰ Heart failure patients, who experience syncope, independent of the mechanism, have an increased cardiovascular risk profile compared with those who do not experience syncope. The association between nonarrhythmogenic syncope and increased risk of death found in the present study supports the mechanistic approach and conclusion previously formulated, that syncope in many heart failure patients represents an inability to compensate for a hemodynamic collapse rather than an arrhythmic event.¹ For both arrhythmogenic and nonarrhythmogenic syncope, we found that the symptom “syncope” was a reliable indicator of unstable hemodynamic response and subsequently significantly predicted adverse prognosis.

A practical detail for clinicians derived from the present study is the frequency of syncope when patients experience an appropriate shock. Approximately 80% of the patients were awake when they received the first appropriate shock. This should be an important message for patients about to receive an ICD device and needs to be discussed with patients more prone to anxiety with ICD device therapies, which has been shown to be a major concern.

We found that ≈8% of all sustained first ventricular tachyarrhythmias ≥200 bpm (treated or monitored) resulted in syncope, and nearly one third of the total syncopal events were attributable to ventricular tachyarrhythmias. Demonstration of the ability to produce ventricular tachyarrhythmias increases the likelihood that the patient will also be subject to a later syncope caused by a ventricular tachyarrhythmia. Despite our findings, the optimal ICD therapy cutoff, in terms of both heart rate and duration, and the balance between patient safety (syncope) and timely termination of sustained ventricular arrhythmias remain to be established. Even with expert adjudication in a randomized trial and with syncope as a specific safety end point, a large volume of syncopal events remained of unknown origin. This difficult assessment is in agreement with previous studies.^11–13

A post hoc analysis of the Sudden Cardiac Death in Heart Failure Trial (SCD-HeFT)² indicated that heart failure patients with syncope had a higher risk of death than those without syncope, and these patients did not benefit from ICD (compared with amiodarone or placebo). In SCD-HeFT and in the study by Middlekauff et al,¹ the causal link between syncope and death may be attributable to hemodynamic collapse and terminal pump failure rather than an ICD-treatable ventricular arrhythmia. In SCD-HeFT, study data on syncope were not collected systematically to define the temporal relationship between syncope and ICD therapy, and events were not adjudicated by an independent committee. Predictors of (all-cause) syncope in SCD-HeFT were New York Heart Association class III, QRS duration ≥120 ms, and lack of β-blocker use. The difference in predictors of all-cause syncope is not readily explained, but there were important differences in the study characteristics and study interventions between these 2 studies; for instance, in SCD-HeFT, 30% of patients were in New York Heart Association class III compared with 52% in MADIT-RIT. Furthermore, all patients in MADIT-RIT were implanted with an ICD or CRT-D with a backup pacing mode versus only one third of the patients in SCD-HeFT. Because 39% of all syncopal events in MADIT-RIT were arrhythmogenic compared with ≈15% in SCD-HeFT, many factors associated with nonarrhythmogenic syncope may have confounded the overall predictive model in SCD-HeFT. The incidence of first-time all-cause syncope in MADIT-RIT was much lower than in SCD-HeFT (4% versus 14%), but with follow-up of 1.4 years versus 3.8 years, respectively, with 2.5-year Kaplan–Meier rates of 8% versus 18% to 22%. It is likely that with a longer follow-up, a similar incidence of all-cause syncope would be found in MADIT-RIT. However, because all patients in MADIT-RIT and only one third of the patients in SCD-HeFT received an ICD with backup pacing mode, some of the cases classified as “unknown” or “other” in SCD-HeFT may have been bradycardic events that were essentially avoided in MADIT-RIT unless a marked vasodepressor component was present. Furthermore, the incidence may have been slightly overestimated in SCD-HeFT, because no formal adjudication was performed and some of the syncopal events may have been syncope-mimicking conditions. SCD-HeFT found that ICDs did not protect patients against all-cause syncope and only protected against the risk of sudden cardiac death. The present data support this statement when the patients from treatment arm A (conventional) are evaluated. In treatment arm A, 3 episodes of syncope attributable to ventricular fibrillation were treated successfully as ventricular fibrillation with a shock.

The MADIT-RIT syncope data represent a large volume of patients with heart failure, in which patients were randomized according to ICD programming. The difference in ICD programming between the 3 randomization arms yields insight into the mechanisms that link arrhythmogenic syncope to rate and duration of ventricular tachyarrhythmias.

Study Limitations

The small number of all-cause syncope and arrhythmogenic syncope events, as well as deaths, and the relatively short follow-up result in limited power for statistical analysis. When analyzing the risk of death, we used multivariable Cox regression models that took many confounding factors into account, but unmeasured confounding associated with patients who experienced syncope may have biased the results. For the evaluation on risk of syncope, patients were randomized at baseline to the 3 treatment arms. Despite adjudication and device interrogations, many syncopal events remained of unknown origin, and this is a limitation of the present study and other studies reported in the literature. Ventricular arrhythmias below the defined detection zones of duration and rate of the devices and adjudicated as nonarrhythmogenic syncope may have contributed to an underestimation of the frequency of arrhythmogenic syncope.

Conclusions

Syncope attributable to slow ventricular arrhythmias is a rare event, and innovative ICD programming with either high-rate cutoff or delayed therapy does not increase the risk of arrhythmogenic or nonarrhythmogenic syncope. Syncope in heart failure patients is related to an increased cardiovascular risk profile and is associated with an increased risk of death regardless of its cause.

Acknowledgments

We thank all patients who participated in MADIT-RIT and each of the study coordinators and principal investigators of the enrolling centers. We also acknowledge the great work of Bronislava Polonsky; her skills in statistical programming and data management are an invaluable asset to the University of Rochester’s Heart Research Follow-up Program.

Footnotes

CLINICAL PERSPECTIVE

Syncope is a frequent event in heart failure patients with implantable cardioverter defibrillators (ICDs). Recently it was shown that ICD programming to high-rate cutoff or prolonged monitoring did not increase the likelihood of syncope compared with conventional programming. In the present report, we further explore the specific cause and mechanism of syncope in relation to ICD programming and the impact of both arrhythmogenic syncope and nonarrhythmogenic syncope on death. Syncope caused by arrhythmias is responsible for 40%, whereas 60% of all syncopal events are caused by nonarrhythmic events, such as orthostatic hypotension syncope or vasodepressor reflex syncope. ICD programming to high-rate cutoff or prolonged monitoring algorithms did not increase the risk of syncope caused by ventricular tachyarrhythmias, and particularly slow ventricular tachyarrhythmias (in the range of 170–199 bpm) are rare causes of arrhythmogenic syncope. These results indicate that high-risk heart failure patients with moderate to severe heart failure symptoms and reduced ejection fraction are able to tolerate rather long durations of fast ventricular tachycardias, whereas slow ventricular tachycardias, for practical purposes, do not result in a loss of consciousness. In the Multicenter Automatic Defibrillator Implantation Trial–Reduce Inappropriate Therapy (MADIT-RIT), both arrhythmogenic and nonarrythmogenic syncope were significantly associated with increased risk of death. These findings suggest that syncope in heart failure patients with ICDs is a significant marker of high risk, despite the cause of the syncopal event.