Real-Time Closed-Loop Suppression of Repolarization Alternans Reduces Arrhythmia Susceptibility in Vivo

Abstract

Background –

Repolarization alternans (RA) has been implicated in the pathogenesis of ventricular arrhythmias and sudden cardiac death.

Methods –

We have developed a real-time, closed-loop system to record and analyze RA from multiple intra-cardiac leads, and deliver dynamically R-wave triggered pacing stimuli during the absolute refractory period. We have evaluated the ability of this system to control RA and reduce arrhythmia susceptibility, in vivo.

Results –

R-wave triggered pacing can induce RA, the magnitude of which can be modulated by varying the amplitude, pulse width and size of the pacing vector. Using a swine model (n = 9), we demonstrate that to induce a one μV change in the alternans voltage on the body surface, coronary sinus (CS) and left ventricle (LV) leads, requires a delivered charge of 0.04 ± 0.02, 0.05 ± 0.025 and 0.06 ± 0.033 μC, respectively, while to induce a one unit change of the K_score, requires a delivered charge of 0.93 ± 0.73, 0.32 ± 0.29 and 0.33 ± 0.37 μC, respectively. For all body surface and intra-cardiac leads, both Δ(Alternans Voltage) and ΔK_score between baseline and R-wave triggered paced beats increases consistently with an increase in the pacing pulse amplitude, pulse width and vector spacing. Additionally, we show that the proposed method can be used to suppress spontaneously occurring alternans (n = 7), in the presence of myocardial ischemia. Suppression of RA by pacing during the absolute refractory period results in a significant reduction in arrhythmia susceptibility, evidenced by a lower S_rank score during programmed ventricular stimulation compared to baseline prior to ischemia.

Conclusions –

We have developed and evaluated a novel closed-loop method to dynamically modulate RA in a swine model. Our data suggest that suppression of RA directly reduces arrhythmia susceptibility and reinforces the concept that RA plays a critical role in the pathophysiology of arrhythmogenesis.

Introduction

Repolarization alternans (RA), a pattern of electrocardiographic (ECG) alternation during the repolarization phase of the cardiac action potential, is associated with heightened ventricular arrhythmia susceptibility across a wide range of pathophysiologic conditions ^1–4.

Surges in RA have been demonstrated on body-surface ECGs prior to spontaneous ventricular tachycardia/ventricular fibrillation (VT/VF) in patients with coronary artery disease ⁵ and in those hospitalized for acute heart failure (HF) ⁶. Our group ⁷ and others ⁸ have shown close correlation between RA measured simultaneously from body-surface and intra-cardiac electrograms (EGMs), suggesting that these measures are detecting the same electrical phenomenon. Additionally, analysis from implantable cardioverter-defibrillator (ICD) EGMs has shown a surge in RA amplitude immediately prior to spontaneous VT/VF events ^9–11, ¹².

There is substantial experimental ^13–16 and computational ^17–19 evidence demonstrating that discordant action potential duration (APD) alternans increases spatio-temporal heterogeneity and dispersion in ventricular repolarization ¹³, ¹⁴. Increased dispersion of refractoriness can give rise to functional lines of conduction block and wavefront fractionation, which can provide the substrate for reentry and arrhythmogenesis ^{13, 20–25}. Thus, RA and arrhythmogenesis are closely linked mechanistically and the same pathophysiologic processes that lead to RA may also lead to VT/VF ²⁵. This suggests that the heart either passes through a period of heightened RA during the transition to VT/VF, or surges in RA occur concurrently with the development of VT/VF. In this paradigm, states of heightened RA may promote the development of an unstable electrical substrate such that an appropriately timed trigger, such as a premature ventricular impulse, can lead to the onset of arrhythmia. This paradigm also suggests that elevated levels of RA may have clinical importance as a predictor of impending VT/VF ³.

The association between a heightened state of RA, a vulnerable electrophysiologic substrate and the observation of impending arrhythmias suggests that dynamic, real-time modulation of RA may prevent the onset of VT/VF by preventing the establishment of a vulnerable substrate. In this context, we have recently developed an intra-cardiac recording system capable of real-time quantification of RA from multiple intra-cardiac sites in vivo ⁷. We have also developed a closed-loop system capable of delivering R-wave triggered pacing pulses during the absolute refractory period (ARP) to suppress RA. In a previous proof of concept study in a swine HF model, we have demonstrated that this system can successfully suppress spontaneously occurring RA²⁶. However, whether RA suppression reduces arrhythmia susceptibility, which would provide a more definitive causal link between RA and arrhythmogenesis, remains to be proven in vivo.

In this study, we assess the impact of suppressing RA using R-wave triggered pacing during the ARP on arrhythmia susceptibility.

Methods

The data that support the findings of this study are available from the corresponding author upon reasonable request.

Animal Preparation

Animal studies were approved by the institutional review board and the subcommittee on research animal care at Massachusetts General Hospital, and all methods were performed in accordance with the relevant guidelines and regulations.

Yorkshire swine (40–45 kg), underwent anesthesia induction and maintenance with Telazol (4.4 mg/kg) IM and Isoflurane (1.5–5%), respectively. Animals were intubated and mechanically ventilated. An arterial line was used to measure invasive blood pressure.

Percutaneous vascular access was obtained in the femoral veins and arteries and jugular veins using the Seldinger technique^{27, 28}. Multipolar catheters ⁷ were placed under fluoroscopic guidance in the coronary sinus (CS), right ventricle (RV), left ventricle (LV), and right atrium. A catheter was also inserted in the inferior vena cava as a reference electrode for unipolar signals. Standard ECG electrodes were placed on the chest and limbs ⁷.

Coronary Artery Occlusion

Coronary ischemia was induced in a closed-chest model via percutaneous techniques, as previously reported ^{7, 29–32}. Briefly, an angioplasty balloon was placed in the mid left circumflex or mid left anterior descending coronary arteries. Intravenous unfractionated heparin was administered (4000 units prior to engaging the coronary artery followed by 1000 units/hour during balloon inflation). Hand injections of contrast were used to confirm balloon location and occlusion.

Equipment and Data Collection

Body surface (leads II and V4) and intra-cardiac ECG signals, as well as arterial blood pressure, were recorded through a Prucka Cardiolab (GE Healthcare, Buckinghamshire, UK) electrophysiology system and inputted into a real-time, multi-channel, signal acquisition, display and analysis system consisting of custom software written in MATLAB 7.6 (Mathworks, Natick, MA) and LabVIEW 8.5 (National Instruments, Austin, TX) ⁷, ²⁶.

Intra-cardiac and Body Surface ECG Data Analysis

The algorithm for measuring RA has been described previously ^{10, 33, 34, 7}. Briefly, the algorithm obtains preliminary R-wave annotations which are subsequently refined. Abnormal beats (i.e. premature ventricular complexes [PVCs]) are identified by using a QRS template-matching algorithm and replaced with a median odd or even template beat (estimated from the odd or even ‘normal’ beats in the 128 beat sequence), depending on whether the abnormal beat is odd or even ^34,7. Subsequently, repolarization interval boundaries for RA analysis are independently determined using a wavelet-transform (WT)-based technique ³⁵ for each body surface and intra-cardiac lead individually, to account for variability in the morphology and timing of the T-wave between leads ³⁵.

RA is estimated using the spectral method for each 128-beat data sequence, using a 512-point power spectrum to improve frequency-domain resolution, as previously described ^{10, 33, 34, 7}. For each lead, spectral analysis is performed independently to account for the spatial variability of RA. RA indices are estimated as:

$$\text{alternans}\text{voltage}\text{(μV)=}\sqrt{{\text{alternans peak}-\text{μ}}_{\text{noise}}}$$

$${\text{K}}_{\text{score}}=\frac{\text{alternans}\text{peak}-{\text{μ}}_{\text{noise}}}{{\text{σ}}_{\text{noise}}}$$

The alternans peak is the peak of the power spectrum at 0.5 cycles/beat and the mean (μ_noise) and standard deviation (σ_noise) of spectral noise are measured in a predefined noise window (0.43–0.46 cycles/beat). The alternans voltage is a direct measurement of alternans, while the K_score reflects the statistical significance of alternans voltage, relative to background noise ^{10, 33, 34, 7}.

We employ an algorithm to compute a phase index (PI) which differentiates between the two phases of RA ³². The algorithm computes the integral of the T-wave on a beat-by-beat basis employing T-wave onset and offset annotations determined by the WT ³⁶. To overcome subtle changes in beat-to-beat T-wave morphology, which may affect phase estimation, we use the signed derivative of the integral of the T-wave to classify each beat to a binary phase index,−1 or +1, corresponding to small or large integral values, respectively ³². Finally, we employ an artificial alternating reference phase index sequence (−1,+1,−1,+1,…) against which the phase time series of each lead is compared ³² to minimize the sensitivity of the algorithm to an arbitrary ECG reference channel,.

The R-wave Triggered Programmable Stimulator

We then developed a closed-loop programmable stimulator (Figure 1A) that is capable of both estimating RA and delivering pacing pulses on a beat-to-beat basis in real time based on detection of the ECG R-wave. Briefly, the stimulator consists of the following components (for more details, please see the Supplement):

Figure 1.

(A) Closed-loop triggered pacing system. The triggered pacing system is a test platform that provides 16-channel electrogram data collection, real-time analysis, and real-time deterministic triggered pacing delivery on two independent channels. The triggered pacing system consists of two computers (the ‘host’ computer, and the ‘target’ computer which runs the LabVIEW Real-Time operating system), a multifunction data acquisition card installed in the target computer and a stimulus generator. The triggered pacing system interfaces with the patient through the analog outputs of a GE Cardiolab electrophysiology system. (B) Representative electrograms obtained by pacing during the ARP on an every other beat basis (pulse amplitude = −5 mA, width = 30 ms and coupling = 30 ms to the R-wave). (C) Schematic depicting the timeline for the experimental protocol. For each animal, first, 10 min of ECG was recorded as baseline, followed by programmed ventricular stimulation (PVS) to gauge arrhythmia susceptibility during baseline. Repolarization alternans (RA) was then induced in the healthy heart by pacing during the ARP from RV12 using two stimulation protocols: (1) pacing on every beat (RV12 Ev) and (2) pacing on every other beat (RV12 EvO). This was followed by a 10 min wait period while the animal was prepared for myocardial infarction (MI) induction. Post balloon occlusion, a new 10 min baseline was recorded during which spontaneous occurrence of RA was observed as a result of successful MI induction. Pacing during the ARP was then applied to suppress alternans. Next, PVS was applied during suppression pacing to gauge the efficacy of pacing during the ARP in reducing arrhythmia susceptibility. Finally, suppression pacing was turned OFF, which caused RA to reappear, following which PVS was reapplied.

Host computer

the host computer allows the user to interact with the triggered pacing program. All alternans analysis occurs on the host computer, which is connected to the target computer via an ethernet crossover cable.

Target computer

the target computer runs the LabVIEW Real-Time 8.5 operating system to control all data acquisition and trigger-pacing timing. It includes a multifunction data acquisition card (National Instruments PCI-6255) which is used to digitally sample sixteen analog inputs every millisecond from the Prucka Cardiolab electrophysiology system. The real-time software performs peak detection to determine the R-wave annotations and can deliver two independent digital trigger signals at two independently programmable coupling intervals with 1 ms resolution.

Closed-loop RA Suppression

Real-time R-wave triggered pacing is controlled by the user. The user can select the alternans estimation channel from among 2 body surface and 12 intra-cardiac electrodes. Having estimated the RA, the system calculates the charge required to deliver to suppress RA to levels less than established thresholds ⁷ by calibrating the amount of charge to the real-time estimations of alternans voltage and K_score, as described below (section “Induction of RA in a Structurally Normal Heart“).

Stimulus generator

The stimulus generator STG2008 (Multi Channel Systems, Reutlingen, Germany) is an eight-channel stimulator, from which two channels are used by the triggered-pacing system. It can output constant current pulses of up to 16 mA.

Two customizable digital trigger output channels are independently configurable by the user. Each trigger output channel delivers a timed digital pulse to the multi-channel stimulus generator with 1 ms resolution and each of the two independent stimulus generator channels receives each trigger pulse and delivers a customizable pacing output to any intra-cardiac electrode pair. Each trigger output channel is dynamically coupled to the QRS complex (R-wave triggered). Triggered pacing can be delivered from a single trigger output channel or from both channels simultaneously. For each trigger output, the user can select R-wave coupling interval (ms), pacing amplitude (mA), pulse width (ms) and whether to trigger on every beat, every even beat, or every odd beat.

Finally, to determine the appropriate pacing pulse polarity, the algorithm uses the estimated RA phase. Specifically, the product of: (phase polarity) x (pacing polarity) is estimated ³². Phase polarity is defined as positive/negative when the R-wave triggered stimulus is delivered on beats with the opposite/same phase to the phase of alternans (out of versus in- phase pacing) ³². Using this approach, we have observed that out-of-phase pacing with positive pulse polarity and in-phase pacing with negative pulse polarity increase RA whereas out-of-phase pacing with negative pulse polarity and in-phase pacing with positive pulse polarity decrease RA. Accordingly, we propose that when the RA phase and the pacing pulse have opposite polarity, R-wave triggered pacing will decrease RA. This information is used to customize the pacing pulses to suppress RA. On detection of significant RA, the system displays proposed pacing pulse parameters (amplitude, pulse width, coupling to the R wave and polarity) on the control console.

All system components including hardware and software communicate with each other in real-time. Following significant RA detection, the proposed pacing pulse parameters define are delivered to intra-cardiac pacing sites in either the CS, LV or RV to deliver dynamic R-wave triggered pacing to suppress RA.

Assessment of Arrhythmia Susceptibility

Arrhythmia susceptibility, under varying states of RA, was assessed using programmed ventricular stimulation (PVS) ³⁷. A positive PVS was defined as sustained VT or VF lasting > 30 secs or requiring external defibrillation.

Pacing pulses during PVS were delivered from a widely spaced bipole (between electrodes 1 and 5 of the decapolar LV catheter) with an amplitude of twice the diastolic threshold and pulse width of 2 ms. PVS was initiated with a drive train of 8 beats (S1) at a cycle length of 400 ms with an extra-stimulus (S2) delivered at a coupling interval of approximately 300 ms. The drivetrain (i.e. S1) for all PVS sequences was triggered to the preceding R-wave to avoid pacing on the refractory period from prior beats. The coupling interval for S2 was reduced in 10 ms steps until ventricular refractoriness was reached, at which point S2 was fixed at 20 ms above the effective refractory period (ERP) and an S3 was added beginning at a coupling interval 10 ms less than S2. This process was repeated until sustained VT/VF was induced or ventricular refractoriness was reached on S6, in which case PVS was deemed non-inducible under those conditions.

PVS was performed in the absence and presence of triggered pacing to suppress RA. If sustained VT/VF was induced, biphasic external defibrillation was performed using 150 joules with paddles placed on the chest of the animal. A rest period of ~10 min was allowed after each positive PVS and defibrillation.

To quantify the outcome of PVS across different RA states, we developed a single “score” rank parameter (S_rank) which assigned the highest score (highest arrhythmia susceptibility) to the intervention that required (i) the smallest number of extra-stimuli during PVS to induce an arrhythmia, or (ii) if the number of extra-stimuli was the same, when the coupling interval between S1 and S_last was the smallest. Both of these criteria suggest less aggressive stimulation was necessary to induce sustained VT/VF, reflecting a more vulnerable arrhythmic substrate. In clinical practice, the results of PVS are generally reported in binary fashion (positive or negative) and there is no validated method for reporting arrhythmia susceptibility in a quantitative or semi-quantitative fashion. We propose the S_rank score not as a surrogate for inducible VT/VF (as a binary outcome), but rather, as a method to describe a semi-quantitative relationship between the level of RA and the likelihood of inducing VT/VF.

Statistical Analysis

Variables are expressed as mean ± standard deviation. Box-plots that include the median, 90–10% and 95–5% percentiles are used to present the statistical properties of the estimated RA sequences. Values of p < 0.05 were considered statistically significant (‘*’ denotes statistical significance of p < 0.05, while ‘+’ denotes 0.05 < p < 0.1). Statistical analysis was performed using MATLAB (MathWorks Inc, Natick, MA). Comparison of K_score with and without R-wave triggered pacing was performed using the paired T-test (Figure 2). Two-way ANOVA was used to gauge the effect of RV triggered pacing vector spacing on alternans voltage and K_score (Figures 3A–3B), while the paired T-test was used to compare slopes, alternans voltage and K_score between baseline and triggered pacing (Figures 3C–3D and 4D-4E). The Kruskal-Wallis test was used to elucidate the significant reduction in alternans voltage and K_score with triggered pacing turned ON, as compared to baseline and with pacing turned OFF (Figures 4A–4B). The Wilcoxon signed rank test was used to gauge the effects of PVS during R-wave triggered pacing (Figures 5 and 6).

Figure 2.

Effects of the current amplitude (A), width (B) and coupling interval (C) on the K_score with and without triggered pacing (ΔK_score) for different pacing interventions (x-axis) measured on the body surface and intra-cardiac leads from the coronary sinus (CS1U, CS2U, CS7U), left ventricle (LV1U, LV2U, LV9U) and right ventricle (RV3U RV4U, RV7U, RV8U). Annotation with an “*” indicates p<0.05 for the comparison of K_score with and without triggered pacing (paired T-test). Data are presented as median (horizontal solid line), 75–25% percentiles (box) and 90–10% percentiles (error bars).

Figure 3.

Effects of RV triggered pacing vector spacing on alternans voltage and K_score in non-RV leads. R-wave triggered pacing pulses are delivered from electrodes on the RV catheter on an every other beat basis (n=8) with −5 mA amplitude, 10 ms pulse width, and 10 ms R-wave coupling interval. (A) Δ(Alternans Voltage) between baseline and triggered data segments as a function of RV pacing vector spacing, from 3 to 30 mm. Both pacing vector spacing and lead type independently affect alternans voltage (two-way ANOVA, p < 0.000001 for each variable). Δ(Alternans Voltage) increases as the pacing vector spacing increases, as estimated from CS, LV and body surface leads. For each lead, all pair wise comparisons of 3 mm spacing versus 18 mm or greater spacing were statistically significant (all p ≤ 0.016). (B) ΔK_score between baseline and triggered data segments as a function of RV pacing vector spacing, from 3 to 30 mm. Both pacing vector spacing and lead type independently affect K_score (two-way ANOVA, p < 0.000001 for each variable). ΔK_score increases as the pacing vector spacing increases, as estimated from CS, LV and body surface leads. For each lead, all pair wise comparisons of 3 mm spacing versus 9 mm or greater spacing were statistically significant (all p ≤ 0.016). (C) Δ(Alternans Voltage)/mm spacing for CS, LV and body surface leads, expressed as mean ± standard error of the mean (n=8). For each lead, all triggered slope comparisons which are statistically greater than the baseline slope comparisons (paired T-test) are indicated on the plot by an “*” (all p < 0.01). (D) ΔK_score/mm spacing for CS, LV and body surface leads, expressed as mean ± standard error of the mean (n=8). For each lead, all triggered slope comparisons (paired T-test) which are statistically greater than the baseline slope comparisons are indicated on the plot by an “*” (all p < 0.01).

Figure 4.

(A) Representative example on the use of R-wave triggered pacing to suppress spontaneously occurring alternans during acute myocardial ischemia, (Kruskal-Wallis test). Alternans voltage (A) and K_score (B) are plotted for the triangular intra-cardiac bipolar lead configuration CS2CS7, CS2LV3 and CS2LV10 in a coronary balloon occlusion model which demonstrates spontaneous alternans. Suppression pacing intervention is delivered from the right ventricle apex (RV12; amplitude: 4 mA, width: 10 ms, coupling to R-wave: 10 ms, as determined by our algorithm). a: spontaneously occurring alternans is visible at baseline, b: triggered pacing is delivered from RV12 on every even beat with a positive polarity pulse, c: triggered pacing is discontinued and both the Alternans voltage and K_score increase to the baseline level during sinus rhythm. Transitions a to c occur correspondingly at times marked by solid vertical black lines, while the colored horizontal lines during each intervention indicate the mean value of the alternans voltage/K_score during that intervention. (C) ECG morphology changes during the corresponding recordings a to c, above; panels show the median odd(red)/even(blue) beats in a 128-beat sequence of the intra-cardiac lead configuration CS2CS7, CS2LV3 and CS2LV10 during each intervention. (D-E) Summary results on the use of R-wave triggered pacing to suppress spontaneously occurring alternans measured from RVCS leads. At baseline, during balloon occlusion in the presence of acute ischemia, markedly elevated levels of RA are observed. When triggered pacing is initiated, using the customized pulse parameters (amplitude, width, coupling interval and phase), a significant decrease in alternans voltage and K_score is observed (paired T-test) (E). With cessation of pacing, alternans magnitude again returns to the elevated levels seen during baseline recordings in the absence of pacing (paired T-test).

Figure 5.

Summary results of programmed ventricular stimulation (S_rank), and pacing during the ARP (using RV12 every pacing denoted by RV12 Ev or RV12 every other pacing denoted by RV12 EvO) at baseline (n=5), using the Wilcoxon signed rank test. (A) S_rank, (B) ST segment alternans voltage and K_score, (C) T-wave alternans voltage and K_score.

Figure 6.

Summary results of programmed ventricular stimulation (S_rank), and pacing during the ARP to suppress repolarization alternans during acute MI. (A) S_rank, (B) ST segment alternans voltage and K_score, (C) T-wave alternans voltage and K_score. (n=11) *: p<0.05; +: 0.05<p<0.1, using the Wilcoxon signed rank test.

Results

Figure 1B displays representative examples of low-noise, high-fidelity, artifact-free body surface and unipolar intra-cardiac electrograms obtained while pacing on every other beat during the ARP. Figure 1C, presents a schematic depicting the timeline for the experimental protocol. Briefly, for each animal, we first obtained a 10 min ECG baseline recording, followed by PVS to assess arrhythmia susceptibility during baseline. Then, RA was induced in the healthy heart by pacing during the ARP from RV12 for about 3 minutes, during which the onset of RA was immediate, and could be quantified by the algorithm after the first 128-beat sequence had been recorded. RA persisted while pacing was ON and estimates of RA were obtained both during pacing and after pacing was turned OFF. RA was induced using two stimulation protocols: (1) pacing on every beat (RV12 Ev) and (2) pacing on every other beat (RV12 EvO). The order of the RA induction protocols was varied randomly across experiments to eliminate bias. PVS was applied after successful induction of RA. This was followed by a 10 min wait period while the animal was prepared for MI induction. Following coronary artery occlusion, a new 10 min baseline was recorded during which spontaneous occurrence of RA was observed as a result of successful MI induction. Pacing during the ARP was then applied to suppress alternans. Next, PVS was applied during suppression pacing to assess the efficacy of pacing during the ARP in reducing arrhythmia susceptibility. Finally, suppression pacing was turned OFF, which caused RA to reappear, following which PVS was reapplied.

RA Induction in the Structurally Normal Heart

We evaluated the ability of pacing during the ARP to induce RA in the structurally normal heart (n=9). To begin, the right atrium was paced at a rate slightly above the sinus rate to minimize spontaneous fluctuations in heart rate variability. Simultaneously, bipolar current pulses were delivered from the two most distal electrodes of a decapolar catheter in the RV apex with the following sets of pulse parameters: amplitude (−5, −1, 1 and 5 mA), width (10 and 30 ms) and coupling interval to the R-wave (10 and 30 ms). Coupling intervals of 10 and 30 ms after the R-wave ensure that the pulses fall during the ARP of the ventricular myocardium. By visual inspection of electrograms obtained from leads nearby or remote to the pacing-electrodes, we have not observed any captured ventricular beats.

In Figure 2 we show results of RA induction using R-wave triggered pacing on an every other beat basis. The effect of the current amplitude on K_score (ΔK_score) is plotted for body surface (ECGII and V4) and intra-cardiac RV (RV3U, RV4U, RV7U, RV8U), CS (CS1U, CS2U, CS7U), and LV (LV1U, LV2U, LV9U) electrodes. In Figure 2A, we assess the effect of the pacing pulse amplitude by holding both the coupling interval to the R-wave (10 ms) and the current pulse width (30 ms) constant. Regardless of the polarity of the current pulse (−1 vs. +1 mA and –5 vs. +5 mA), stimulation during the ARP results in a significant increase in RA which is more pronounced in electrodes close to the pacing site (i.e. RV leads). These data suggest that the amplitude of RA can be locally modulated by R-wave triggered pacing during the ARP. Additionally, higher amplitude pacing pulses (i.e. −5/+5 mA) result in RA which is detectable at more distant intra-cardiac electrodes (i.e. LV and CS) and even on the body surface. This data demonstrate that amplitude of triggered pacing pulses during the ARP can modulate the spatial impact of RA induced by this method.

Figure 2B demonstrates the effect of changing the pulse width (10 ms vs. 30 ms) while holding the current amplitude constant (−5 mA). These data demonstrate that similar to the pulse amplitude, the pulse width is an important parameter in controlling RA at proximal (i.e. RV leads) and remote (i.e. LV, CS and body surface) sites from the site of triggered pacing (p < 0.05). Figure 2C shows results evaluating the effect of the coupling interval with respect to the R-wave (10 ms and 30 ms) on the K_score during triggered pacing, while holding the current amplitude (−5 or 5 mA) and pulse width (30 ms) constant. We observe that, at least in this range of values, the coupling interval with respect to the R-wave does not significantly affect the induced RA magnitude.

In aggregate, we have found that to induce a one μV change in the alternans voltage on CS, LV and the body surface leads requires a delivered charge of 0.05 ± 0.025 μC, 0.06 ± 0.033 μC and 0.04 ± 0.02 μC, respectively. Similarly, to induce a one unit change in the K_score requires a delivered charge of 0.32 ± 0.29 μC, 0.33 ± 0.37 μC and 0.93 ± 0.73 μC, respectively. We subsequently used these values to calibrate the charge required to suppress RA.

Of note, analysis of the recordings presented in Figure 2 did not demonstrate any change in the number of PVCs per total number of beats (PVC %) between baseline (i.e. without triggered pacing) and during triggered pacing ( 1.71 ± 0.33% vs. 2.10 ± 0.37%p = 0.203). This finding further corroborates that stimulation using these parameters during the ARP does not induce global ventricular myocardial depolarization.

Effect of the Pacing Pulse Electric Field Size on RA

To further evaluate the effect of pacing during the ARP on RA, we performed a series of experiments (n=8) in which the animal was again paced from the atrium at a rate slightly above the sinus rate. To examine the effect of increasing the stimulus field size from 3 mm (pacing from a bipolar configuration between electrodes RV1 and RV2) to 30 mm (pacing between electrodes RV1 and RV8), current pulses were delivered (amplitude: −5 mA, width: 10 ms, coupling to the R wave: 10 ms) on every other beat from the RV apex. We evaluated the effect on RA of increasing the distance between the RV bipolar pacing electrodes (i.e. increasing the size of the stimulation electric field) on the RV (Supplement), CS and LV EGMs, as well as on body surface leads (II and V4). In Figures 3A–3B, we show the Δ(Alternans Voltage) and ΔK_score between baseline and pacing during the ARP as a function of the RV pacing vector spacing from 3 to 30 mm. Δ(Alternans Voltage) and ΔK_score increase consistently as the pacing vector spacing increases for all CS, LV and body surface leads. For each lead, all pair wise comparisons of 3 mm spacing versus 18 mm or greater spacing were statistically significant (Figures 3A–3B). For both Δ(Alternans Voltage) and ΔK_score, both the pacing vector spacing and lead type (CS, LV and body surface) independently affected the RA estimate (two-way ANOVA, p < 0.001 for each variable).

To better define the relationship between stimulation field size and global magnitude of RA induction, we estimated the Δ(Alternans Voltage)/mm (Figure 3C) and ΔK_score/mm (Figure 3D) for CS, LV and body surface leads where, for each animal and for each lead, this value was calculated as the slope of the best linear fit constrained to pass through (0,0). For each lead, all slope comparisons were statistically greater during pacing than the baseline (p < 0.01). We observed that for both alternans voltage and K_score, the biggest effect per mm is observed in leads CS7 and CS8, which are closest to the site of stimulation (i.e. proximal CS closest to the right ventricle). These findings again demonstrate a strong spatial relationship between field size and magnitude of RA induction. The K_score of the body surface leads is also significantly affected by field size, likely due to inherently smaller noise in these leads. We have found that to induce a one μV change in alternans voltage on CS, LV and body surface leads for a fixed charge (amplitude: −5 mA, width: 10 ms, coupling: 10 ms), a field size of 0.62 ± 0.12 mm and 0.66 ± 0.46 mm and 0.27 ± 0.05 mm, respectively, is required. In addition, a one unit change of K_score requires a field size of 5.50 ± 1.26 mm, 3.47 ± 0.07 mm and 13.11 ± 0.56 mm, respectively.

Spontaneous RA Suppression During Acute Ischemia

Following establishment of parameters for RA induction via R-wave triggered pacing during the ARP, we sought to evaluate the ability of this method to suppress spontaneously occurring RA during acute ischemia, in vivo.

Figure 4, shows data from a single animal where the alternans voltage and K_score are plotted using a previously described, triangular, intra-cardiac bipolar lead system optimized for intra-cardiac alternans detection ⁷. This lead system is comprised of bipoles from electrodes CS2CS7, CS2LV3 and CS2LV10. During coronary artery occlusion (Figures 4A-B, baseline, beats 0–600), significant spontaneous RA is present, as evidenced by elevated levels of alternans voltage and K_score. Upon detection of significant RA, the system estimates in real-time the RA-suppression pacing pulse parameters and instructs the user to pace on every even beat with the following pacing pulse parameters: amplitude: 4 mA; width: 10 ms; coupling: 10 ms. Pacing from catheter electrodes in the RV apex (beats 600–1100) results in a significant reduction (to levels below the spontaneously occurring RA at baseline during ischemia) in both the alternans voltage and K_score (Figures 4A–4B). Finally, pacing is discontinued (beats 1100–1600) and alternans voltage and K_score increase again to levels similar to those observed prior to initiation of pacing. Figure 4C shows ECG morphology changes during each intervention corresponding to recordings in panels a to c, above. The panels show the median odd(red)/even(blue) beats in a 128-beat sequence of leads CS2CS7, CS2LV3 and CS2LV10. In this example, R-wave triggered pacing during the ARP on every even beat results in a significant decrease of spontaneous RA during acute ischemia (alternans voltage: ~4-fold average decrease compared to baseline in panel a, p < 0.0001, Kruskal-Wallis test; K_score: ~14-fold average decrease compared to baseline in panel a, p < 0.0001, Kruskal-Wallis test). In panel c, RV12 pacing is discontinued, leading to an increase of alternans voltage and K_score (alternans voltage: ~3-fold average increase compared to alternans suppression in panel b, p = 0.0001, Kruskal-Wallis test; K_score: ~9-fold average increase compared to alternans suppression in panel b, p < 0.0001, Kruskal-Wallis test).

Figures 4D and 4E present summary results across all experiments (N= 7 animals; n=11 records) where R-wave triggered pacing during the ARP was used to suppress spontaneously occurring RA. In an attempt to evaluate the broad efficacy of this form of pacing in suppressing RA and reducing arrhythmia susceptibility, we have attempted pacing from different sites: RV apex (recorded from catheters in the LV and CS); basal LV (recorded from catheters in the RV and CS) and distal CS ( recorded from catheters in the RV and proximal CS). We have observed a similar effect in suppressing RA regardless of the pacing or recording site. The figure demonstrates the alternans voltage (top row) and K_score (bottom row) measured from RVCS leads depicting the suppression of RA when pacing is delivered. During balloon occlusion (baseline), in the presence of acute ischemia, markedly elevated levels of RA are observed. When triggered pacing is initiated, using the customized parameters (amplitude, pulse width, coupling interval and phase), a significant decrease in alternans voltage and K_score is observed. With cessation of pacing, the magnitude of RA again returns to the heightened levels observed during baseline recordings, in the absence of pacing.

These findings highlight the ability of customized pacing protocols during the ARP to suppress spontaneously occurring RA during acute ischemia. They further provide evidence that R-wave triggered pacing suppresses spontaneously occurring RA and provide the first demonstration of a closed-loop approach which is capable of delivering dynamic and adaptive therapy to suppress RA.

Impact of RA Suppression on Arrhythmia Inducibility in the Normal Heart

In order to assess whether the presence of significant RA is pro-arrhythmic, we performed a series of studies in the normal swine heart in which we used pacing during the ARP on every other beat to artificially induce alternans and assess the susceptibility to arrhythmias.

In Figure 5, we present results of PVS and RA without pacing (Baseline), pacing during the ARP from RV12 on every beat (RV12 Ev) or every other beat (RV12 EvO) in the structurally normal heart (n=5). In Figure 5A, at baseline the S_rank score is low, suggesting low arrhythmogenicity, which is also supported by low levels of alternans voltage and K_score in the ST segment (B) and T-wave (C). Pacing on every other beat significantly increases S_rank, as expected, and a rise in RA is evident (B-C). In contrast, pacing on every beat does not significantly alter alternans voltage or K_score and arrhythmia is not easily inducible, as demonstrated by the low S_rank score. We used the Wilcoxon signed rank test (two-sided) to demonstrate that the alternans voltage and K_score expectedly increase while pacing on an every other beat basis, compared to either baseline or pacing on every beat (0.05 < p <0.1), and these changes were accompanied by a trend toward increased arrhythmia susceptibility when pacing every other beat (0.05 < p < 0.1).

Impact of RA Suppression on Arrhythmia Inducibility during Acute Ischemia

To further assess the impact of RA suppression on arrhythmia susceptibility, we performed PVS during RA occurring in the setting of coronary artery occlusion.

In Figure 6A, at baseline (prior to coronary balloon occlusion), the S_rank score is low, suggesting low inducibility of arrhythmias during PVS, which is also supported by low levels of alternans voltage and K_score in the ST segment (B) and T-wave (C). Following balloon occlusion, as expected, a significant rise in RA is seen (B-C). PVS in the setting of acute ischemia results in a significant increase in the S_rank score (panel A), suggesting greater ease of VT/VF induction and a more pro-arrhythmic substrate. R-wave triggered ARP pacing was then applied from electrode RV12 with customized parameters as dictated by the algorithm. Despite the presence of ongoing ischemia (the balloon in the coronary artery remained inflated), the use of ARP pacing results in a reduction in the magnitude of RA (B-C), although RA levels remain elevated compared to baseline given the ongoing presence of myocardial ischemia. The reduction in RA magnitude also results in a trend toward reduced arrhythmia susceptibility when PVS is performed during pacing in the setting of on-going myocardial ischemia (A). The S_rank score while using ARP pacing was lower than without pacing and was not significantly different than the baseline, prior to the onset of coronary ischemia. This data provides evidence that ARP pacing can suppress RA, have a stabilizing effect on the underlying eletrophysiologic substrate and in turn reduce the propensity for arrhythmogenesis. It also supports the concept that RA plays a direct role in the pathogenesis of ventricular arrhythmias.

Discussion

Several lines of clinical evidence lend support to the role of RA in the genesis of cardiac arrhythmias. Therefore, the ability to suppress RA may hold promise in preventing the development of ventricular tachy-arrhythmias by homogenizing the degree of heterogeneity of repolarization, a fundamental trigger of arrhythmias under diverse pathophysiological conditions.

This study provides a systematic and comprehensive approach into exploring the ability to suppress RA in vivo using a programmable, proprietary, closed-loop stimulator capable of simultaneous multi-channel electrogram acquisition from multiple intra-cardiac and body surface sites, real-time RA estimation, and using these estimates to dynamically determine the parameters of calibrated R-wave triggered pacing pulses and delivery electrical therapy to suppress RA. Using this experimental set-up, we have evaluated the feasibility of suppressing RA and arrhythmia susceptibility in vivo, by pacing during the ARP.

Using a novel experimental set-up, we demonstrated for the first time in a closed-loop system the feasibility of using ARP stimulation to suppress RA in-vivo. Our results lead to the following conclusions. First, RA can be induced in a structurally normal heart by pacing on an every other beat basis, and the amplitude and width of the current pulse can modulate the magnitude of RA. Second, the spacing of the pacing vector is a significant determinant of the magnitude of induced RA. Third, the characteristics (phase, pulse amplitude and duration) of a pacing protocol can be estimated in real-time for dynamic and efficient suppression of RA. Lastly, this approach can be used to suppress spontaneously occurring RA during acute MI, in-vivo, and reduce the risk for ventricular tachyarrhythmias.

From a biophysical point of view, stimulation during the ARP may serve as a pseudo I_to1, which results in reactivation of the L-type channels and an increased depolarizing current which results in APD prolongation. This hypothesis is in agreement with pre-clinical studies investigating the mechanisms of cardiac contractility modulation (CCM) which have shown that non-excitatory ARP stimulation can control the APD, in part by modulating cellular calcium transients ³⁸. Winter et al. ³⁹ have demonstrated that the impact of CCM stimuli on APD₉₀ is critically dependent on the amplitude of the stimulus, the pulse width and location of the stimulus, but not on the coupling interval. Our in vivo findings, using similar sized pacing pulses ³⁹, suggest that a pacing stimulus can be reliably delivered during the ARP without triggering another action potential.

Further support for the rationale behind the use of pacing during the ARP to suppress RA comes from the work of Li and Otani ⁴⁰ who have shown in a canine ventricular cell model that the optimal time during an AP to apply a stimulus to control alternans is during the early plateau phase. Furthermore, the delivered charge of the control stimulus during repolarization is two orders of magnitude smaller than one delivered later during the AP.

In addition to modulation of the APD, it is also conceivable that stimulation during the ARP may impact conduction velocity (CV) and local activation time as a potential mechanism for modulating RA. We evaluated local activation times with and without R-wave triggered pacing in the RV, LV and CS in our experimental set-up and we were not able to detect a significant difference in CV during ARP stimulation. It’s possible that a higher density of electrodes in the RV, adjacent to the site of triggered pacing, may be necessary to detect a local impact on CV. Additional studies will be necessary to determine whether modulation of CV plays a role in the mechanism of RA suppression by triggered pacing.

Our data provides novel insights on the spatial effects contributing to the ability of R-wave triggered pacing during the ARP to control RA. In particular, we have shown that the effectiveness of controlling RA attenuates over distance. Moreover, analysis of the effect of the pacing pulse electric field size on RA demonstrates a direct spatial relationship where an increase in the stimulus field size results in augmentation of the magnitude of local RA induction.

We have also shown that spontaneously occurring RA originating in the LV, during myocardial ischemia, can be suppressed from an RV site by applying appropriate R-wave triggered pacing stimuli. An important consideration in evaluating the use of triggered pacing to control RA is the potential for pro-arrhythmic effects of stimulation during the ARP. Similar to medium-term human studies (3–12 months) of CCM delivered for 5–7 hours daily ⁴¹, we did not identify any increase in ventricular ectopy or adverse clinical events during triggered pacing in our studies, suggesting that intermittent sub-threshold stimulation may be safe, even in the presence of myocardial ischemia.

The findings of this study have important clinical implications. Thousands of patients who are at increased risk for VT/VF are implanted with ICDs annually. These individuals may be subjected to painful defibrillator shocks if VT/VF occurs. The ability to deliver pacing during the ARP through the ICD to suppress RA, before VT/VF occurs, has the important potential to deliver anti-arrhythmic therapy without needing painful ICD shocks. Additionally, our prior work has shown that a triangular electrode configuration comprised of leads in the RV and CS yielded optimal RA detection⁷ and therefore, our results could be easily translatable to a recording lead system readily available in commercial cardiac resynchronization defibrillators (CRT-D). However, significant technological advances will need to be made to currently available ICDs in order for them to be capable of executing the closed-loop RA suppression algorithm.

In conclusion, our findings support the idea that a closed-loop system can be used to deliver R-wave triggered pacing based on real-time RA estimations to suppress RA at the whole heart level. These findings open the door to novel implantable device-delivered pacing therapies to suppress RA and prevent ventricular arrhythmias.

What is known:

• Repolarization alternans (RA) is closely associated with the substrate which gives rise to ventricular arrhythmias

• Whether suppression of RA can prevent the onset of arrhythmias has not been demonstrated

What the Study Adds:

• We have developed a closed-loop system capable of detecting RA in real-time and delivering pacing stimuli during the absolute refractory period which are capable of dynamic RA suppression

• In a swine model of acute coronary ischemia, real-time RA suppression resulted in reduced arrhythmia susceptibility

• This data demonstrates that RA plays a causative role in ventricular arrhythmogenesis and therapies which suppress RA may be effective at preventing the onset of arrhythmias

Nonstandard Abbreviations and Acronyms

RA

Repolarization Alternans

ECG

Electrocardiograph

VT/VF

Ventricular Tachycardia/Ventricular Fibrillation

HF

Heart Failure

EGM

Electrograms

ICD

Implantable Cardioverter-Defibrillator

APD

Action Potential Duration

ARP

Absolute Refractory Period

CS

Coronary Sinus

RV

Right Ventricle

LV

Left Ventricle

PVC

Premature Ventricular Complexes

WT

Wavelet Transform

PI

Phase Index

PVS

Programmed Ventricular Stimulation

ERP

Effective Refractory Period

CCM

Cardiac Contractility Modulation

CV

Conduction Velocity

CRT-D

Cardiac Resynchronization Therapy-Defibrillator