Impact of Denervated Myocardium on Improving Risk Stratification for Sudden Cardiac Death

Abstract

Between 184,000 and 462,000 Americans die suddenly each year. Fifty percent to 70% of these deaths are due to ventricular tachycardia/fibrillation (VT/VF). We tested whether hibernating myocardium or myocardial sympathetic denervation identifies patients at high-risk for developing VT/VF independently of ejection fraction (EF). Positron emission tomography (PET) was used to quantify myocardial sympathetic denervation (¹¹C-meta-hydroxyephedrine [¹¹C-HED]), perfusion (¹³N-ammonia), and viability (insulin-stimulated ¹⁸F-2-deoxyglucose [¹⁸FDG]) in patients with ischemic cardiomyopathy (EF < 35%) eligible for a primary prevention implantable cardioverter defibrillator (ICD). The primary end-point was sudden cardiac arrest (SCA) defined as arrhythmic death or ICD discharge for VT/VF > 240 bpm. Volumes of total denervated (P = .001) and viable denervated myocardium (¹¹C-HED-¹⁸FDG mismatch, P = .03) predicted SCA, whereas hibernating and infarcted myocardium did not. Multivariate analysis identified four independent predictors of SCA: denervated myocardium > 37.6% of left ventricule (LV), LV end-diastolic volume > 98 mL/m², creatinine level > 1.49 mg/dL, and no angiotensin- inhibition therapy. Denervated myocardium had a hazard ratio of 3.5 for SCA (10.3%/year vs. 3.0%/year, p=0.001). Absence of all four factors predicted low risk (44% of cohort; SCA <1%/y) whereas two or more factors identified subjects at high-risk (20% of cohort; SCA 12%/y). Denervated myocardium quantified using PET strongly predicts risk of SCA, and is independent of EF, infarct volume, and other clinical variables.

INTRODUCTION

The annual incidence of sudden cardiac death in the United States is between 184,000 and 462,000, with estimates that 50% to 70% of the deaths are due to ventricular tachycardia (VT) or ventricular fibrillation (VF). Availability of therapies that have been shown to reduce sudden death in various at-risk groups, including beta-blockers, angiotensin-inhibition therapy, statins, aldosterone blockers, and the implantable cardioverter defibrillator (ICD), emphasize the need to accurately identify patients who will develop VT/VF within some specified period and exclude those who will not (1).

Multiple noninvasive and invasive approaches have been developed to detect the arrhythmogenic factors that initiate and maintain VT/VF in patients with ischemic heart disease. The conditions that lead to VT/VF may occur transiently or develop during the course of healing from injury to ventricular myocardium and persist. Factors known to trigger or modulate VT/VF include changes in autonomic nervous system activity, metabolic disturbances, myocardial ischemia, electrolyte abnormalities, acute volume and/or pressure overload of the ventricles, ion channel abnormalities, and proarrhythmic actions of cardiac and non-cardiac drugs. Death of myocardial cells from ischemia, toxins, infectious agents, or chronic pressure/volume overload leads to scar formation, alterations in chamber geometry, and electrical and anatomical remodeling. The electrophysiological alterations induced by these conditions initiate and maintain VT/VF in humans most likely via a reentrant mechanism, though abnormal automaticity, triggered activity, or combinations of these mechanisms may be operative (2).

Specific techniques developed to detect the presence of factors known to serve as a substrate or trigger of VT/VF and/or abnormalities in ventricular conduction and repolarization that are critical to reentry include: 1) slowed conduction (QRS duration, signal-averaged electrocardiogram); 2) heterogeneities in ventricular repolarization (QT interval, QT dispersion, T-wave alternans); 3) imbalance in autonomic tone heart rate variability, heart rate turbulence, heart rate recovery after exercise, and baroreceptor sensitivity); 4) extent of myocardial damage and scar formation (left ventricular ejection fraction [LVEF]); 5) ventricular ectopy (long-term ambulatory monitoring); and 6) electrophysiological testing (inducible VT/VF) (1).

Despite the availability of these predictive approaches, there is currently no optimal strategy for risk stratification. The most widely used strategy is based on LVEF and falls far short of the optimal goal. Current approaches dichotomize patients into low- and high-risk groups. Risk, however, is a continuum; and many risk functions are likely dynamic. Moreover, the majority of episodes of arrhythmic death occur in patients with low to intermediate risk factors. Accordingly, the risk stratification field requires further development (3).

Basic and clinical studies have shown an important role for sympathetic activation in the development of VT/VF, and inhomogeneity in myocardial sympathetic innervation may create a myocardial substrate particularly vulnerable to arrhythmic death (4). This inhomogeneity can reflect sympathetic denervation from infarction as well as reversible ischemia (5). In chronic coronary disease, reversible ischemia also creates inhomogeneity in myocardial sympathetic innervation that is independent of infarction, occurring in both stunned as well as hibernating myocardium (6). Preclinical models of hibernating myocardium have a high rate of death from VT/VF that develops in the absence of infarction and heart failure (7). In these patients, ¹¹C-meta-hydroxyephedrine (¹¹C-HED) positron-emission tomography (PET) shows extensive sympathetic denervation (8).

Based on these observations, we tested the hypothesis that inhomogeneity in myocardial sympathetic innervation and/or hibernating myocardium, quantified using PET, increased the risk of arrhythmic death independently of left ventricular function. The PAREPET study (Prediction of Arrhythmic Events with Positron Emission Tomography) was designed as an initial step towards this goal evaluating primary prevention implantable cardioverter defibrillator (ICD) candidates with coronary artery disease (9).

METHODS

The NIH-sponsored PAREPET trial (HL-76252, NCT01400334) was a prospective observational cohort study designed to determine whether imaging hibernating or denervated myocardium can predict arrhythmic death in ischemic cardiomyopathy. The full manuscript will be published in the Journal of the American College of Cardiology (10).

Study Design

The study population (n = 204) included patients with ischemic cardiomyopathy who were eligible to receive a primary prevention ICD (pre-enrollment LVEF ≤ 35% for class ≥ II and ≤ 30% for class I). They had stable ischemic heart disease and heart failure on optimal medical therapy and were not considered candidates for coronary revascularization. Exclusion criteria included a prior cardiac arrest or ICD discharge, recent infarction (<30 days) or revascularization (percutaneous coronary intervention < 3 months; bypass grafting < 1 year) (9, 10).

Echocardiography and PET

Two-dimensional echocardiography was performed on the day of PET imaging as previously described (9, 11). An echocardiographer blinded to events quantified cardiac volumes, LVEF, and mitral regurgitation as recommended by the American Society of Echocardiography.

PET imaging was performed on a ECAT EXACT HR+ (CTI, Knoxville, Tennessee) PET scanner (15.5 cm axial field-of-view; resolution ∼5.4 mm³ full-width-at-half-maximum) (9, 11). Sympathetic innervation was assessed with ¹¹C-HED [740 MBq], resting perfusion with ¹³N-ammonia (¹³NH3 [740 MBq]), and viability with ¹⁸F-2-deoxyglucose (¹⁸FDG [241 MBq]) during a hyperinsulinemic-euglycemic clamp (11). Attenuation correction was performed using a ⁶⁸Ge rod source (9, 11). Of the planned 585 PET images, 96% were completed and quantifiable with 176 subjects having complete data.

Quantitative PET Analysis

Blinded analysis used FlowQuant (Ottawa Heart Institute, Canada) (11, 12) and decay-corrected reconstructed images (zoom 2; Hann filter cutoff 0.3 cycles/pixel). Late uptake defined the LV with bottle-brush sampling (13). Late myocardial uptake was averaged from four frames of each imaging set; 15 to 60 minutes after ¹¹C-HED; 3 to 19 minutes after ¹³NH₃; 15 to 40 minutes after ¹⁸FDG. Myocardial activity was normalized to the highest 5% of sectors (496 sector model) (13). Normal ¹³NH₃ and ¹⁸FDG uptake were ≥ 80% of peak (9, 12). Normal ¹¹C-HED uptake was considered ≥ 75% of peak, based on the estimated ratio of reduced versus normal ¹¹C-HED retention fraction among zones with normal resting flow (14). All PET parameters were expressed as %LV. Infarcted and hibernating myocardium were quantified from a mismatch analysis between ¹³NH₃ and ¹⁸FDG (12). To evaluate the potential influence of global downregulation in ¹¹C-HED uptake, ¹¹C-HED retention was determined for the segment with maximal ¹¹C-HED uptake in each subject. ¹¹C-HED retention (min⁻¹) was calculated by dividing mean segment activity from the late uptake image by the integrated arterial activity (from 0- to 60 minutes) (8).

Classification of Events

Subjects were contacted at 3-month intervals to review interval ICD therapy, hospitalizations, and symptoms. The primary end-point was sudden cardiac arrest (SCA). This included arrhythmic death or ICD discharge for VF or VT > 240 bpm (9). The frequency of ICD discharge for these arrhythmias approximates the reduction in mortality with an ICD (9, 15, 16).

Deaths were classified using modified Hinkle-Thaler criteria (17, 18). Available details (medical records, witnesses, family, and death certificates), activity levels, and symptoms before death were reviewed by two cardiologists. In the event of disagreement, a consensus was reached with a third reviewer. Cardiac transplantation and arrhythmic events with end-stage heart failure or hospice care were classified as cardiac non-sudden deaths (17). Revascularization procedures were performed on 26 subjects after enrollment.

Statistical Analyses

Data are presented as mean ± SD. Subjects with and without SCA were compared with unpaired t tests (continuous data) and chi-square analysis (categorical data). The time to SCA was analyzed graphically using Kaplan-Meier plots and tested using Cox proportional hazard models and the log-rank statistic. Optimal cut-points for continuous variables were determined retrospectively by values that maximized the log-rank statistic. Stepwise selection was used to generate the optimal multivariate model to predict time to SCA from PET, demographics, medications, echocardiographic, hemodynamic, and laboratory parameters. All statistical analyses were performed with commercial software (Microsoft Excel and SAS version 9.1).

RESULTS

Subject Population

The average LVEF was 27 ± 9%, age 67 ± 11 years, New York Heart Association (NYHA) functional class 2.1 ± 0.8 and Canadian Cardiovascular Society angina class 1.8 ± 0.7. There were 21 women (10%). Diabetes was present in 47%; 78% had prior coronary revascularization; and 81% had an ICD before an event. Almost all subjects were treated with anti-platelet therapy or warfarin (99%), β-blockers (96%), and angiotensin inhibition therapy (90%) (10).

PET Quantification of Denervated, Infarcted, and Viable Myocardium

Quantitative image analysis determined that the average infarct volume was 20 ± 9% of the LV. In comparison, the average volume of denervated myocardium was considerably larger, averaging 27 ± 11% of the LV (P < .001 vs infarcted), with 8 ± 6% of the LV denervated but viable. Hibernating myocardium (viable myocardium with reduced resting perfusion) was uncommon, reflecting the high prevalence of prior revascularization, and averaged 3 ± 3% of the LV (10).

Clinical and Imaging Correlates of SCA

During a median follow-up period of 4.1 years (range, 2.5 to 7.2 years) there were 33 adjudicated SCAs. There were no differences between the 22 subjects with arrhythmic death versus the 11 with ICD discharges for VF or VT > 240 bpm. All SCA events occurred in men (P = .06). Subjects developing SCA had significantly larger volumes of denervated and viable denervated myocardium, whereas infarct volume and hibernating myocardium were not different. Although the LVEF and LV mass were not significantly different, those who experienced SCA had larger LV volume indices and more mitral regurgitation. Among laboratory parameters, serum creatinine and B-type natriuretic peptide (BNP) levels were higher in patients developing SCA (10).

Quantitative Imaging and SCA

The primary analysis of PAREPET included evaluation of PET imaging parameters as continuous variables to predict time to SCA. The volume of denervated myocardium had the strongest correlation with SCA (P = .001). Patients in the highest tertile of sympathetic denervation had an SCA event rate of ∼6.7%/y whereas the intermediate and lowest tertiles had event rates of 2.2%/y and 1.2%/y, respectively. Each 1% increase in the volume of denervated myocardium was associated with a 5.7% increase in the risk of SCA. Similarly, the volume of viable denervated myocardium was significantly associated with the time to SCA (P = .025), with risk of SCA increasing by 6.7% for every 1% increase in viable, denervated myocardium. The ability of denervated myocardium to predict SCA was not influenced by any global downregulation in ¹¹C-HED uptake. ¹¹C-HED retention in segments with maximal ¹¹C-HED uptake was 0.136 ± 0.036 min⁻¹, with no difference among those with and without subsequent SCA (0.134 ± 0.027 vs 0.137 ± 0.037 min⁻¹, P = .69). There was no significant association between SCA and the volume of infarcted or hibernating myocardium (10).

Other parameters significantly associated with SCA as continuous variables included larger LV end-diastolic and end-systolic volume indices, elevated BNP level, elevated creatinine, larger left atrial volume, lower LVEF, and elevated LV mass index (P < .05 for each). In addition, SCA was associated with no angiotensin inhibition therapy (23% vs 9%, P = .02). Multivariate analysis of these continuous and categorical variables revealed that denervated myocardium remained an independent predictor of time to SCA, in addition to LV end-diastolic volume index, creatinine, and no angiotensin inhibition therapy (10).

Predictors of Risk Using Optimized Cut-Points

A post-hoc analysis was performed to determine the optimal cut-point for each of the continuous predictors. Time to SCA was predicted by greater denervated myocardium (>37.6% of the LV, 19% of subjects), larger LV end-diastolic volume index (cut-point = 99 mL/m², 34% of subjects), elevated creatinine (cut-point = 1.49 mg/dL, 26% of subjects), and no angiotensin inhibition therapy (10% of subjects). For example, less denervated myocardium (<37.6% LV) identified a large subgroup (81% of the cohort) at lower risk of SCA (SCA rate of 3.0%/y [95% CI 1.9–4.7] vs 10.3%/y [95% CI 6.2–16.1], where CI is the confidence interval; P = .001). We then tested how well these dichotomized factors could predict risk of SCA. Subjects having no high-risk predictor represented 44% of the cohort and had a very low risk of SCA (0.9%/y). In contrast, those with one risk factor (36% of the cohort) had a SCA rate of 3.9%/y, and those with two or more high-risk factors (20% of the cohort) had an annual risk of SCA of 11.7%/y (10).

DISCUSSION

Results show that myocardial sympathetic denervation quantified using ¹¹C-HED PET imaging identifies patients with ischemic cardiomyopathy who are at high risk of SCA. SCA in this primary prevention population was dependent on the total volume of denervated myocardium and independent of other imaging parameters including LVEF, infarct volume, and hibernating myocardium. Thus, quantifying the extent of myocardial sympathetic denervation may provide a novel approach to identify a subgroup of patients with ischemic cardiomyopathy who would derive the most benefit from a primary prevention ICD.

ICD Therapy as A Surrogate for Aborted Arrhythmic Death

There currently is no consensus regarding which ICD therapies provide an appropriate surrogate for arrhythmic death in observational studies. The device therapies included in the primary end-point were considerably more restrictive than the prevailing approach in contemporary observational imaging studies, which typically include all appropriate ICD therapies (anti-tachycardia pacing and ICD discharge) (19–21). Appropriate ICD therapies occur at least three times as frequently as fatal arrhythmic events (16, 22), and self-terminating ventricular arrhythmias may have substrates and triggers that differ from those resulting in arrhythmic death (23). Our prospectively defined primary end-point was SCA, which included adjudicated arrhythmic death and an ICD equivalent end-point based on device therapies most clearly associated with lethal ventricular arrhythmias (VF and VT > 240 bpm). In the MADIT II trial, the frequency of these device therapies plus death in patients randomized to an ICD (18.3% at 2 years) was similar to the rate of death in the medically treated group (18.0% at 2 years) (15). Although this analysis supports using these events as surrogates of arrhythmic death in patients with an ICD (15, 16), we cannot exclude the possibility that this conservative definition may have underestimated the rate at which arrhythmic death would have occurred in our population.

Myocardial Sympathetic Innervation and SCA

Recent studies have imaged global cardiac sympathetic innervation in heart failure using ¹²³I-meta-iodobenzylguanidine (MIBG). They have consistently shown an association between reductions in MIBG uptake and end-points reflecting a variety of composite cardiovascular outcomes in patients with heart failure (19, 24–26). However, there is disagreement regarding the preferred imaging parameter (reduced heart-to-mediastinum [H/M] ratio versus accelerated tracer washout rate) and the added benefit from regional tracer quantification. There is also a paucity of outcome data for cause-specific mortality from SCA.

In a small study of heart failure patients (LVEF < 40%; n = 106, 52% ischemic cardiomyopathy), MIBG imaging was shown to predict arrhythmic death, as well as pump failure death and total cardiac mortality (25). Both H/M ratio and washout rate predicted arrhythmic death, but only washout rate and LVEF remained independent after multivariate analysis. In contrast to our study, subjects taking beta-blockers, insulin, or those with significant renal dysfunction were excluded. Thus, more than 97% of our study population would have been excluded, raising questions about generalizing these findings to risk stratification for SCA among contemporary heart failure patients.

A larger observational study (ADMIRE-HF) imaged 961 subjects with an LVEF ≤ 35% (66% with ischemic cardiomyopathy) and NYHA class II–III heart failure symptoms (26). The composite primary end-point was heart failure progression (hospitalization or clinical assessment), potentially life-threatening arrhythmias (sustained VT, resuscitated cardiac arrest, or any appropriate ICD therapy including anti-tachycardia pacing) and cardiac death. A late H/M ratio < 1.6 (reflecting both denervated myocardium and accelerated washout from increased sympathetic nerve activity) predicted the composite end-point (hazard ratio [HR] 2.5). Evaluating regional sympathetic innervation and perfusion did not improve this over global H/M ratio (26). In contrast, another study evaluating heart failure patients referred for an ICD (n =116, 74% with ischemic cardiomyopathy), found neither MIBG H/M ratio or washout rate to be predictors of appropriate ICD therapy (19). A semiquantitative defect score predicted ICD therapy, which is consistent with our findings and supports the importance of quantifying regional sympathetic neuronal function for risk stratification of arrhythmic death (19).

Imaging of Myocardial Ischemia and Infarct Volume to Predict SCA

Acute myocardial ischemia, fibrosis, and infarction have been shown to be arrhythmogenic substrates. For example, in a large cohort with coronary disease referred for perfusion imaging, the extent of infarction plus ischemia (summed stress perfusion score) was the only independent imaging parameter to predict arrhythmic death (27). This is consistent with our study in that the volume of denervated myocardium approximates the area at risk for ischemia (5, 8). Nevertheless, due to our study design and the stable nature of symptoms in our subjects, coronary anatomy was not evaluated and we did not directly assess the presence or extent of stress-induced ischemia at enrollment. Therefore, the potential roles of inducible ischemia and coronary disease progression in SCA in this study are not defined.

Cardiac magnetic resonance imaging (MRI) has facilitated the quantification of infarct volume as well as patchy fibrosis (“gray zone”) surrounding the infarct borders. Although cause-specific mortality from arrhythmic death has not been assessed, several studies using MRI have correlated scar volume with cardiac end-points, including appropriate ICD therapy (20, 21, 28). Some (20), but not all studies (21, 28), have shown that quantification of the gray zone improves risk prediction. The MRI gray zone is similar in size to viable, denervated myocardium assessed by PET (the difference between denervated and infarcted myocardium). Thus, MRIs may ultimately prove to be assessing a similar substrate with increased risk of VT/VF. Further studies comparing infarct volume, gray zone, and denervated volume will be required to determine whether the two approaches provide similar risk stratification for SCA.

Predictive Model to Assess the Risk of SCA

The post-hoc multivariate analysis showed that quantifying the volume of denervated myocardium along with three other variables (LV end-diastolic volume index, and creatinine and angiotensin inhibition therapy) independently predicted arrhythmic risk. Quantifying infarct size, LVEF, and BNP (and other variables) did not improve the predictive model. Whereas other parameters have been associated with cardiovascular mortality, they either preferentially predict progressive heart failure (as opposed to SCA), or their prognostic information was superseded by the four independent factors. Importantly, more than 40% of our study population had none of the risk factors and were at very low risk with an annual SCA rate of <1%. This is actually lower than the rate of arrhythmic death in patients with an LVEF between 35% and 50% and not currently candidates for an ICD. At the other extreme, 20% of our subjects had two or more risk factors and high annual SCA rates of ∼12%. A limitation of this analysis relates to its post hoc nature, inclusion of multiple parameters that could result in over-fitting and the optimization of cut points on a modest sample size. Nevertheless, identification of four independent predictors is reasonable in relation to the number of observed events. Further studies will be needed to prospectively validate the model, but because it is independent of EF among those with moderate to severe LV dysfunction, it could ultimately be useful to improve the risk stratification of a much larger number of patients at risk of SCA (29).

CONCLUSIONS

Results indicate that quantifying regional sympathetic denervation with ¹¹C-HED PET may provide a novel approach to identify patients at high risk of arrhythmic death that is independent of LVEF. Although PET/computed tomographic scanning is widely available, the use of ¹¹C-HED would be limited to centers in close proximity to a cyclotron. Fortunately, ¹⁸Fluorine-labeled norepinephrine analogs have a longer half-life that could facilitate broader application (30). With these, molecular imaging of myocardial sympathetic denervation in those patients who have ischemic cardiomyopathy may afford a new approach to improve the targeting of ICDs to those at greatest risk of SCA.

DISCUSSION

Hochberg, Baltimore: So I am not a cardiologist and this is very, very interesting to me. But in the area that I work in, we often do observational studies with, obviously, outcomes which are not as important as sudden cardiac death. A lot of the statisticians have developed sort of new methods of, let's say, mining the data in order to come up with true algorithms as opposed to multiple variable modeling. Do I assume correctly that the statisticians that were involved in your group are going to do that in terms of techniques like recursive partitioning or looking at net reclassification indices for how well different variables will improve the classification? The second question is do you currently have plans to validate these results in another observational cohort?

Cain, Buffalo: The short answer to both is yes!

Calkins, Baltimore: Mike, I enjoyed your talk. What about a nonischemic cardiomyopathy? Have you begun to look at denervated myocardium in that setting and to see what its role is? I'd be interested in where you are going to go with this.

Cain, Buffalo: The group has not yet done that, Hugh. It's a good idea, but we have not yet done that using this particular approach.

Abboud, Iowa City: Tell me more about your thinking about what kind of denervation is happening. Is this a sympathetic denervation — efferent sympathetic — because there are other important afferents, like vagal afferents, that can determine the outcome by parasympathetic activation? So the denervation you are talking about is a sympathetic?

Cain, Buffalo: Correct.

Abboud, Iowa City: Adrenergic denervation. … efferent?

Cain, Buffalo: Correct. So the hypothesis is that this border zone is denervated and adjacent to normal myocardium that, during this process, may be hypersensitive to catecholamines. You have this border zone then that has the two extremes: denervation and hypersensitivity.

Abboud, Iowa City: I'm just suggesting that there are many other afferent nerves — particularly vagal afferents — that can determine the efferent output in a significant way.

Gotto, New York: Very nice presentation, Michael. I recall when the debate was going on within Medicare about whether to reimburse — and selecting the criteria for reimbursement — for implantable defibrillators after an acute coronary syndrome. One of the key criteria that you considered was a QT interval? I didn't see that anywhere on your variables.

Cain, Buffalo: To the best of my knowledge, there were no differences between those two groups. But I don't have in my armamentarium today the data that actually shows that slide.

Weisfeldt, Baltimore: Any comments on use of beta blockers in the patients that you studied?

Cain, Buffalo: So in this study, patients did have good post-MI management treatment that did include beta blockers and ACE inhibitors. So there were no differences between the two groups, and it was good background medical therapy in both groups.