Atrial fibrillation septal pacing: translation of modelling results

Introduction

Although atrial fibrillation (AF) is the most common sustained cardiac arrhythmia, some of the underlying mechanisms still remain unclear due to their high complexity, and no single therapy has been shown to be effective in all individual patients.^1,2 Existing therapies range from antiarrhythmic drugs to surgical/catheter ablation or electrical cardioversion delivering an intra-cardiac or external electrical shock. In addition, some recent pacemakers and defibrillators have incorporated features to prevent or terminate atrial tachycardia via pacing.^3–5 The polymorphic dynamic nature and the complexity of AF hinder the development of a therapy effective for all AF types. Furthermore, AF usually also involves a clinical progression from paroxysmal to permanent AF due to structural and electrophysiological remodelling. This polymorphic nature of AF is difficult to study experimentally because of the complex interactions across multiple temporal and spatial scales in the atria that often lead to unpredictable outcomes.⁶ Computer modelling is therefore considered today as a potential new tool in the clinical environment to better understand AF mechanisms and to develop associated therapeutic options.^1,7,8

Current pacemakers and implantable defibrillators often include pacing algorithms developed for the prevention or termination of atrial arrhythmia.^9–11 Compared to electrical cardioversion, pacing has the advantage of being painless, safe and energy efficient in implantable devices. However, while the possibility to obtain local capture of AF by rapid pacing has been demonstrated in animals and humans,^12,13 no clinical study has yet proven the long term clinical benefit in terms of AF termination. A model-based study by Uldry et al.¹⁴ showed that the septum area was the only location allowing local capture of both atria during rapid pacing of AF from a single site. Atrial fibrillation septal pacing represents a simple and attractive approach since it produces simultaneous biatrial stimulation using a single lead, without the need of complex methods for synchronizing multiple pacing sites. Rapid pacing from a ring of electrodes around the septum showed promising results for AF termination when tested in a model of homogenous substrate (up to 29% AF termination in computer simulations).¹⁴ However, clinical AF substrates involve various types and levels of heterogeneities and these changes in atrial substrate have a significant impact on rapid pacing outcomes.¹⁵ It was also shown that AF cycle length (AFCL) is not predictive of the ability to capture AF,¹⁵ which might explain why patients with similar AFCL can have different capture results in the presence of rapid pacing.¹⁶ This suggests that a more detailed assessment of each patient’s atrial substrate would be needed to assess the potential response to a pacing therapy.

The ability to reproduce specific AF dynamics corresponding to anatomy and substrate of each AF patient makes the modelling framework attractive for this task. One essential phase for every model-based approach remains the translation of research results into clinically available therapeutic options.¹⁷ This translation is often not straightforward because experimental testing in animal models differs from simulations in human computer models in many anatomic and electrophysiological ways, while testing on human subjects in clinical studies is often limited by cost and ethical factors. In our case, translation of the AF septal pacing concept implies the development of an electrode-device system that could be tested experimentally and clinically. Previous model-based studies on AF septal pacing delivered pacing to the whole septal area, which can be seen as a theoretical limit representing the best performance achievable.^14,15 Translation of this modelling concept requires the practical realization of an electrode ring that will be placed intravenously in contact with the septal wall and where all electrodes will be stimulated simultaneously. The present study evaluated the number of stimulation electrodes needed in this septal ring in order to achieve AF capture for different AF dynamics.

Methods

Modelling atrial fibrillation

The atrial anatomy in our model was reconstructed from computed tomography (CT) scans of a patient in permanent AF referred for an ablation procedure. The patient was in chronic AF and the atria were slightly dilated. A 3D model was reconstructed from the CT scans.¹⁸ Among the several connections existing between the right atrium and left atrium, the most significant two were represented: the Bachmann’s Bundle (BB) and the rim of the septum. Rule-based fibre orientation was included.^18,19 The atrial anatomy was meshed using cubic elements of 330 µm, corresponding to 1.04 million computational nodes in the whole atria. Since the testing of different pacing protocols with different atrial substrates involved a high number of simulations, computational load was a major issue. Therefore, a simplified surface model (triangular mesh with 117 000 nodes and 450 µm resolution) representing the atrial epicardium was derived from the 3D model (Figure 1), resulting in a reduction of number of nodes and computational time.

Figure 1

Modelling AF. The atrial geometry is showed in the left upper corner (anterior and posterior view) with the major vessels and valves: tricuspid valve (TV), mitral valve (MV), pulmonary veins (PVs), superior and inferior vena cava (SVC and IVC) and coronary sinus (CS). Two different atrial substrates were implemented on the atrial geometry, corresponding to different types of sustained AF dynamics: (i) AF maintained by stable rotors (Cholinergic AF: AF1 and AF2), (ii) AF maintained by one to three waves meandering through the atrial surface (Meandering AF: AF3 and AF4). For each atrial substrate, two different AF episodes were induced by varying the AF initiation protocol. Transmembrane potential maps, atrial cellular action potential, AFLC and average number of wavelets (#W) are separately indicated for each AF episode.

Open in new tabDownload slide

The atrial substrate was formed by a grid of atrial units coupled with their neighbours via electrical resistors representing gap junctions. Conductivity values of these resistors were selected to reproduce conduction velocities measured in humans. An atrial cell membrane kinetics model using the Courtemanche-Ramirez-Nattel human atrial model (CRN model) was assigned to each atrial unit.²⁰ The CRN model allows the simulation of vagal stimulation through acetylcholine or the inclusion of the effect of regional heterogeneities in repolarization. Electrical propagation on the mesh of connected atrial units was simulated using the monodomain equations.²¹ Models of pathological tissue are needed to create conditions for sustained AF. Various pathologies or atrial substrates can be the cause of AF in human heart, which might explain inter-patient variability to AF therapies. In this study, we modelled two different types of AF substrates found in individual patients:¹⁸

1. Cholinergic AF: Atrial substrate based on a homogeneous atrial tissue with a 4:1 anisotropy and the CRN model in which an acetylcholine-dependent K+ current was added and I_CaL was reduced by 70% and the currents I_to and I_Kur were altered by a factor 0.5. Cholinergic AF dynamics is maintained by stable rotors (AF1 and AF2 in Figure 1).

2. Meandering Wavelets: Atrial substrate based was based on a homogeneous atrial tissue with a 4:1 anisotropy and the CRN model in which the currents I_CaL, I_to, I_Kur, and I_Kr were altered by a factor 0.7, 0.2, 0.1, and 1.5, respectively. AF is maintained by one to three wavelets meandering through the atrial surface (AF3 and AF4 in Figure 1).

AF was initiated with a pacing protocol involving multiple pacing sequences of maximum 5 s located at random locations on the atria. Most resulting AF episodes were sustained after pacing was stopped. For each of the two AF substrates, we selected two sustained AF episodes with different dynamics as shown in Figure 1: AF1 and AF2 are episodes induced in the Cholinergic AF substrate and AF3 and AF4 are episodes induced in the Meandering Wavelets substrate. They differ in terms of AFCL and average number of wavelets present in the atrial tissue (#W). AFCL was computed by averaging the cycle length (CL) values measured over 10 s at 24 sensing electrode pairs uniformly distributed on the atria. AFCL and #W were reported as mean ± SD. Cholinergic AF had faster dynamics and higher number of wavelets (AFCL 87 ± 9 ms and #W 5.0 ± 1.6) compared to Meandering Wavelets (AFCL 165 ± 18 ms and #W 1.7 ± 0.8).

Modelling atrial fibrillation septal pacing

In order to perform computer simulations equivalent to animal or human experiments, results were repeated and averaged on several simulations of AF rapid pacing. For each AF episode (AF1 to AF4), 10 instantaneous transmembrane potential maps were randomly selected as initial conditions for the subsequent simulation of the AF septal pacing algorithm. Rapid pacing was applied by injecting intracellular current inside the atrial cells located within the septal pacing electrodes area. The total duration of each rapid pacing sequence was 20 s and the pacing cycle length (PCL) was selected in the range 60–100% AFCL with 4% increments.

Preliminary model-based studies on AF septal pacing delivered pacing to the whole septal area (indicated in red in the ‘full septum’ drawing in Figure 2).^14,15 The translation of this concept towards a pacing electrode-device application requires the design of an electrode configuration reproducing the depolarization pattern offered by the ideal full septum pacing scheme. This implies developing a suitable stimulation electrode design based on a ring of multiple electrodes around the septal area. It is therefore needed to determine how many electrodes are needed to reproduce the results of the full septum scheme. The different stimulation electrode configurations tested in this article are presented in Figure 2: a ring of 1, 2, 3, 4, 6, 8, or 12 electrodes located around the fossa ovalis. Rapid pacing was applied simultaneously to all electrodes on the septal ring.

Figure 2

AF septal pacing electrode configurations. Electrode location on the septum is indicated in red for the different configurations (1, 2, 3, 4, 6, 8, 12 electrodes). The reference is the full septum. A cut-out of the left atrium is represented with tricuspid valve (TV), mitral valve (MV), pulmonary veins (PV), BB and fossa ovalis (FO).

Open in new tabDownload slide

Assessment of atrial fibrillation capture

'Local capture of AF was defined as the ability of the pacing algorithm to take control over certain areas of the atria. Capture was automatically assessed using 24 pairs of sensing electrodes evenly distributed on the atrial surface (Figure 3). For each pair of electrodes, the CL and the direction of propagating waves were assessed using the local transmembrane potentials. A pair of electrodes was considered to be within the capture zone if the propagation direction was from the septum area towards the appendages and if the measured CL was within ±5% of the PCL. The percentage of capture was computed as the number of pairs within the capture zone divided by the total 24 pairs of electrodes. The optimal PCL was defined as the PCL with the highest percentage of capture and the capture window was assessed as the range of PCL for which the percentage of capture was above 50% (only if applicable). Capture results were averaged across the 10 initial AF conditions.

Figure 3

Automatic assessment of AF capture during septal pacing. Each of the 24 pairs of electrodes evenly distributed on the atrial surface (in red) was considered as captured if the propagation direction was from the septum area towards the appendages and if the measured CL was within ±5% of the PCL.

Open in new tabDownload slide

Results

Influence of AF substrate on AF capture

Figure 4 provides a quantitative summary of AF capture results. Overall better capture results were observed for the Cholinergic AF substrate even if this type of AF has a lower AFCL (faster AF) and a higher number of wavelets compared to Meandering Wavelets substrate. The optimal PCL leading to the best AF capture (in green in Figure 4) was comparable for both AF episodes induced in the same substrate but greatly differed when the substrate was changed: 84 and 88% AFCL for Cholinergic AF and 96 and 100% AFCL for Meandering Wavelets. However significant differences in capture windows (in orange in Figure 4) were observed for the two AF episodes induced in the same atrial substrate (AF1 and AF2 for the Cholinergic AF substrate and AF3 and AF4 for the Meandering Wavelets substrate). The AF3 episode had limited capture compared to the other three AF episodes. It should be noted however, that Figure 4 represents an average on 10 different AF conditions, meaning that for some of these initial conditions AF capture results were higher.

Figure 4

Summary of AF capture results. Percentage of atrial tissue captured during AF was simulated for PCLs in the range 60–100% AFCL and the different septal electrode configurations of Figure 2. Percentages higher than 50% are highlighted in orange. The highest capture for each AF episode is highlighted in green.

Open in new tabDownload slide

Figure 5 shows examples of transmembrane potential maps and spatial extent of atrial capture obtained during rapid septal pacing at optimal PCL and optimal electrode configuration. It can be observed that for episodes with good capture results (such as AF2 and AF4), the septal pacing could capture up to around 90% of the atrial surface with a few remaining wavelets in the area of the right appendage and around the mitral and tricuspid valves.

Figure 5

Detailed capture results for the optimal AF septal pacing configuration for each AF episode (indicated in green in Figure 4). The first column shows the electrode configuration and PCL leading to the highest percentage of captured atrial tissue. The middle column shows the corresponding transmembrane potential maps observed during AF capture. The last column shows the spatial extent (in red) of the captured atrial tissue.

Open in new tabDownload slide

Influence of the number of stimulation electrodes on AF capture

For all AF episodes it was possible to find a PCL for which capture was higher than 50% even when pacing only from one electrode. For the Cholinergic AF substrate, adding more electrodes did not significantly impact capture results when pacing at optimal PCL, even when comparing results to full septum pacing. On the other hand, for the Meandering Wavelets substrate, the number of electrodes on the septal ring needed to achieve AF capture was different for the two AF episodes (Figure 4). For the AF4 episode with a slower AF and lower number of wavelets compared to AF3, good capture was achieved independently of the number of electrodes in the septum ring. For the AF3 episode with a more complex dynamics, a higher number of electrodes led to a better AF capture.

Discussion

This model-based study confirms that atrial substrate greatly influences the ability to capture AF with rapid pacing from the septum area.¹⁵ In the Cholinergic AF substrate, AF was faster (lower AFCL) and with a higher number of wavelets than Meandering Wavelets and yet overall AF capture results were better. This confirms the observation that AFCL alone is not predictive of the ability to capture AF¹⁵ and it might explain why patients with similar AFCL can have different capture results in the presence of rapid pacing.¹⁶ The PCL leading to optimal capture was also dependent on atrial substrate: 84–88% AFCL for Cholinergic AF and 96–100% AFCL for Meandering Wavelets. This suggests that a detailed assessment of each patient’s atrial substrate is needed to estimate the potential response to a pacing therapy and to determine the parameters of the optimal septal pacing sequence.

Even for a given AF substrate, great variations in capture results were observed between the two simulated AF episodes. For example in the Meandering Wavelets substrate, the AF4 episode showed only 1–2 meandering waves while in AF3 the main meandering waves were additionally accompanied by wavebreaks around anatomical obstacles, leading to lower capture results. However, the optimal PCL did not vary between these two AF episodes for a given substrate. Our simulations also show that the outcome of septal pacing depends on the dynamic of the re-entrant waves observed in the atrial tissue at the beginning of rapid pacing, meaning that the same pacing sequence will have a different impact depending on timing. Compared to other therapies for AF, pacing has the advantage that it is well tolerated and each AF septal pacing sequence is short (20–30 s).¹⁵ The pacing sequence could therefore be repeated several times until termination occurs. This ability to repeat the pacing sequence multiple times allows a successful therapy outcome even if the termination rate of a single sequence is low. Multiple attempt tolerance also allows the pacing parameters to be varied and adapted automatically.

The number of stimulation electrodes used in septal pacing ring had a lesser impact on AF capture that AF dynamic. In the Cholinergic AF substrate, it was always possible to find a PCL leading to local AF capture even when using only one stimulation electrode with results comparable to pacing the full septum area. In the Meandering Wavelets substrate however, AF episodes with a complex AF dynamics required a higher number of electrodes for local AF capture. This is in agreement with the clinical observation that while single site rapid pacing algorithms existing in pacemakers can be successful at terminating organized tachyarrhythmia, no evidence can be found about treating AF with such algorithms.¹⁴ The key feature of the AF septal pacing concept is the stimulation of multiple electrodes around the septum.¹⁴ Based on our simulation results, we can conclude that a ring of 3–4 stimulation electrodes around the septum would be sufficient for most AF dynamics. This could be practically implemented using a decapolar electrophysiology catheter placed intravenously in contact against the left septal wall around the fossa ovalis using a transseptal approach. Rapid stimulation will be performed simultaneously on 3–4 electrodes from the decapolar catheter at 80–100% AFCL. The optimal PCL will have to be determined individually for each AF patient.

Moving to clinical validation is an essential phase with every model-based result. The present study investigated the translation of the AF septal pacing modelling concept. The results of this study will serve as a basis for experimental studies to determine whether AF septal pacing induces similar AF capture mechanisms in AF patients as in computer simulations, and for which types of AF the procedure might be the most effective. AF septal pacing could represent an alternative to drugs and catheter-based ablation for some AF patients, but the therapy should be specific to each patient’s atrial substrate.

Funding

This work was supported by the Theo-Rossi di Montelera Foundation (Lausanne, Switzerland) and Medtronic Europe (Tolochenaz, Switzerland).

Conflict of interest: Nathalie Virag and Todd Kallmyer are full time employees of Medtronic.

References