Neuromodulation Therapy for Atrial Fibrillation

Abstract

Atrial fibrillation has a multifactorial pathophysiology influenced by cardiac autonomic innervation. Both sympathetic and parasympathetic influences are pro-fibrillatory. Innovative therapies targeting the neurocardiac axis include catheter ablation or pharmacologic suppression of ganglionated plexi, renal sympathetic denervation, low-level vagal stimulation, and stellate ganglion blockade. To date, these therapies have variable efficacy. As our understanding of atrial fibrillation and the cardiac nervous system expands, our approach to therapeutic neuromodulation will continue evolving for the benefit of those with AF.

INTRODUCTION

Atrial fibrillation (AF) is a common arrhythmia that will ultimately affect 1 in 5 individuals¹. The frequency of AF is matched by its complexity. AF has a multifactorial pathophysiology influenced by cardiac autonomic innervation^{2, 3}. There are even distinct autonomic-driven phenotypes of AF with distinct management strategies (Figure 1). For example, in patients with vagal AF, defined as occurring postprandially or nocturnally, beta-blocker therapy may aggravate or precipitate persistent or permanent AF⁴. Both AHA and ESC guidelines suggest that disopyramide may be useful for vagal AF due to its anticholinergic activity, although the evidence is limited.^{5, 6}

Figure 1.

Vagal versus Adrenergic Atrial Fibrillation, adapted from Coumel, et al.⁵⁰ and de Vos, et al. ⁴

A comprehensive understanding of the neurocardiac axis is essential in treating AF. The present review will elucidate the anatomy and physiology of the neurocardiac axis as well as the rationale and evidence for therapeutic neuromodulation in atrial fibrillation.

ANATOMY & PHYSIOLOGY OF CARDIONEURAL AXIS

The autonomic nervous system (ANS) influences heart rate (chronotropy), conduction speed (dromotropy), contractility (inotropy), and relaxation (lusitropy), with opposing parasympathetic (PNS) and sympathetic (SNS) effects⁷. ANS innervation is categorized as central, arising from the brain cortex, brainstem, and spinal cord; intrathoracic extra-cardiac, characterized by sympathetic chain ganglia; and intrinsic cardiac, comprised of neurons organized into ganglionated plexi with interconnecting neurons within epicardial fat pads⁸.

Sympathetic and Parasympathetic Pathways to the Heart

Central autonomic signaling arises from the cingulate cortex, amygdala, thalamus, and cerebellum, with parasympathetic and sympathetic anatomical variation⁹. SNS output descends to the cervical and thoracic spinal cord where preganglionic neurons in the intermediolateral columns project to superior cervical, middle cervical, cervicothoracic (stellate), and mediastinal ganglia¹⁰. Long, postganglionic axons project to epicardial ganglia and cardiac myocytes directly, stimulating the release of norepinephrine to beta-1 and beta-2 receptors of the sinoatrial (SA) and atrioventricular (AV) nodes, as well as atrial and ventricular myocytes¹¹.

In contrast, long preganglionic parasympathetic neurons originate in the brainstem, at the nucleus accumbens and dorsal motor nucleus of the medulla oblongata. These project through the vagus nerve to short postganglionic neurons within intrinsic cardiac ganglia, releasing acetylcholine at muscarinic (M2) receptors of the SA and AV nodes as well as atrial myocytes.

Sympathetic innervation at adrenergic receptors classically stimulates G-protein coupled receptor (GPCR) G_s which activates adenylyl cyclase (AC), increasing intracellular cAMP¹². cAMP plays a complex role in mediating chronotropy, predominantly through binding hyperpolarization-activated cyclic nucleotide gated (HCN) channels to enhance channel opening probability by positively shifting voltage thresholds^{13, 14}. cAMP additionally activates protein kinase A (PKA), which phosphorylates molecules implicated in automaticity pathways regulated by calcium cycling¹⁵. Many nonclassical pathways are associated with beta-2 activation and influence chronotropy through cAMP-independent mechanisms¹². Sympathetic innervation additionally strengthens contractility through cAMP/PKA-mediated augmentation of intracellular calcium^{11, 12} SNS-driven, inward calcium via L-type channels, in addition to promoting automaticity and contractility, has been implicated in early afterdepolarizations that trigger AF through promoting forward Na/Ca exchange¹⁶.

Parasympathetic release of acetylcholine at M2 receptors slows depolarization and heart rate by both negatively influencing GPCR activation as well as binding to muscarinic-gated potassium channels (I_KAch) for an inward, hyperpolarizing current^{17, 18}. Acetylcholine impacts heart rate greater than contractility due to higher M2 receptor density in nodal and atrial tissue¹¹. Parasympathetic stimulation ultimately shortens the atrial effective refractory period (AERP)¹⁹, which synergistically with sympathetic input promotes AF initiation¹⁶. Local circuit neurons facilitate SNS/PNS communication in the cardiac conduction system^{11, 20}.

Anatomy and Physiology of Epicardial Ganglionated Plexi

Extrinsic autonomic nerves enter the intrinsic cardiac nervous system (ICNS) through the cardiac hilum and synapse onto epicardial ganglionated plexi (GPs)^{10, 21, 22}. One GP contains 200–1000 local, interconnecting neurons embedded in epicardial adipose tissue²³. Some have colloquially referred to the ICNS as the heart’s “little brain”²⁴.

Atrial and Ventricular Ganglionated Plexi

GP topography is variably described but consistent between hearts; 75% of epicardial ganglia lie in the cardiac dorsum, with the highest density at the left atrial (LA) - pulmonary vein (PV) junction²⁵. An early study identified 5 atrial and 5 ventricular ganglia, in which 4 major atrial GP were localized near the PVs. These 4 atrial GP were further explored as ablation targets and renamed: (1) anterior right (ARGP), concentrated immediately anterior to the right PVs, (2) superior left (SLGP), residing in the LA roof medial to the left superior PV, (3) inferior right (IRGP) and (4) inferior left (ILGP), both located inferior to the right and left PVs on the LA posterior wall^{23, 25–27}.

Atrial GP integrate signals from bilateral vagosympathetic trunks to the SA and AV nodes^{28, 29}. Input to the SA node is mediated by the ARGP, which receives innervation either directly from the right vagosympathetic trunk (predominant pathway) or indirectly from bilateral vagosympathetic trunks after traversing the SLGP²⁸. The SLGP and ARGP converge on the IRGP before modulating the AV node²⁹. The ILGP also influences the AV node and contributes to ventricular slowing^{23, 28, 29}. Pauza, et al. apply a different nomenclature to their findings in human hearts, which largely coincides with the aforementioned GPs but adds emphasis to ventral GPs residing along coronary arteries and atrial surfaces^{21, 30} (Figure 2).

Figure 2.

The anatomy of the cardiac autonomic ganglionated plexi, as discovered through anatomic dissection of canines (Armour et al., black ²⁶), anatomic dissection of human hearts (Pauza et al., blue ^{21, 30}), and electroanatomic mapping of human hearts (Po et al., green ²⁷). (A) Posterior, (B) Cross-sectional, (C) Right, (D) Left. Abbreviations: Armour et al., PL-LAGP=posterolateral left atrial, OMGP=obtuse marginal, SLAGP=superior left atrial, PDGP=posterior descending, PM-LAGP=posteromedial left atrial, PRAGP=posterior right atrial, SRAGP=superior right atrial; Po et al., SLGP=superior left, ILGP=inferior left, ARGP=anterior right, IRGP=inferior right; Pauza et al., MDsGP=middle dorsal ganglionated subplexus, LDsGP=left dorsal ganglionated subplexus, VLAsGP=ventral left atrial ganglionated subplexus, DRAsGP=dorsal right atrial ganglionated subplexus, VRAsGP=ventral right atrial ganglionated subplexus, LCsGP=left coronary ganglionated subplexus, RCsGP=right coronary ganglionated subplexus. Anatomy: SVC=superior vena cava, PA=pulmonary artery, LAA=left atrial appendage, VOM=vein of Marshall, LSPV=left superior pulmonary vein, LIPV=left inferior pulmonary vein, RSPV=right superior pulmonary vein, RIPV=right inferior pulmonary vein, IVC=inferior vena cava, SA=sinoatrial, AV=atrioventricular.

Pulmonary Veins and Vein of Marshall

PV ectopy plays an established role in initiating and maintaining AF^{31, 32}. The PVs become encased in atrial tissue as they return from the lungs and enter the LA, forming a myocardial sheath composed of circumferentially oriented cardiomyocytes resembling a “mesh-like” network^{23, 32, 33}. The irregular orientation of these fibers may contribute to anisotropy and re-entry. Studies have also identified both cholinergic and adrenergic activity neighboring the PV–LA junctions^{23, 34}. Strategies for pulmonary vein isolation (PVI) have evolved to circumferential ablation for increased maintenance of sinus rhythm^{32, 35}, which further improves when neighboring autonomic innervation is eliminated³⁶. PV innervation by both adrenergic and vagal fibers is highly co-localized³⁴, engendering difficulty in selectively ablating either pathway.

The Vein of Marshall (VOM) runs within the vestigial ligament of Marshall (LOM) and is another landmark implicated in AF³⁷. The intracardiac VOM extends from the coronary sinus to the left superior PV base and pulmonary artery, attaching superiorly to the pericardium^{23, 38}. Not only is VOM ectopy seen in patients with sustained AF³⁹, but also the VOM is richly innervated by cholinergic and adrenergic fibers implicated in atrial tachyarrhythmias^{40, 41}. Recent clinical trial data has demonstrated that adjunctive alcohol ablation of the VOM improves maintenance of sinus rhythm in persons with persistent AF treated with PVI⁴².

Inaccessible Ganglionated Plexi

Several GP are difficult to access with an endocardial approach due to their association with the great vessels, including: (1) transverse sinus GP (TSGP), between the pulmonary artery and base of the aorta, and (2) superior vena caval-aortic GP (SVC-Ao GP), in the posteromedial wall of the SVC, the anterolateral aortic wall, and superior to the right pulmonary artery^{23, 26, 43}. The SVC-Ao GP has been identified as the “head station” of parasympathetic input to the heart, receiving most vagal fibers that ultimately synapse on the atria, SA, and AV nodes^{23, 44}. The SVC-Ao GP may also be implicated in apnea-related AF, with reduced AF inducibility following ablation in canines⁴⁵.

Renal Influences on Sympathetic Nervous System Activity

The kidney contains mechanical and chemical receptors which ascend along afferent renal fibers and travel through renal nerves and dorsal roots (T9-L4) that ultimately project centrally to enact widespread sympathetic activation⁴⁶. This pathway is implicated in hypertension and its denervation has been targeted for refractory hypertension with varying efficacy^{47, 48}. Recent data suggests that renal denervation in conjunction with PVI reduces the risk of recurrent atrial arrhythmias (hazard ratio=0.57; 95% CI, 0.38 to 0.85; P=0.006)⁴⁹.

RATIONALE FOR TARGETING THE AUTONOMIC NERVOUS SYSTEM

Role of the ANS in Atrial Fibrillation

While early clinical studies formed the basis of our understanding of “vagal AF” versus “adrenergic AF”^{4, 50}, this distinction is not always clinically apparent. Biomarkers may be useful in delineating autonomic profiles that contribute to AF. Insulin-like growth factor 1 (IGF1) and N-terminal pro-B-type natriuretic peptide (NT-proBNP) are associated with the risk of incident AF^51–53. Decreased IGF1 corresponds with increased risk of AF⁵¹, possibly due to IGF1’s inverse relationship with age⁵⁴, as well as its deactivating role on the SNS that may be protective against arrhythmia. No molecular biomarkers have been identified that distinguish vagal versus adrenergic AF.

Heart rate variability (HRV) is a physiologic biomarker that measures autonomic imbalance with both time and frequency domain measurements⁵⁵. High frequency (HF) represents vagal predominance while low frequency (LF), when expressed in normalized units, reflects sympathetic tone^{55, 56}. In 782 patients with hypertension, higher HRV and HF were associated with AF occurrence⁵⁷, while prior studies noted lower HRV and LF to confer AF risk^{58, 59}. This inconsistency may be influenced by varied autonomic phenotypes. Interestingly, patients undergoing PVI without recurrent AF have lower HF and elevated LF/HF in the months after ablation compared to those with recurrence³⁶.

No study has distinguished HRV patterns in patients with clinical vagal or adrenergic AF. Future studies should evaluate the utility of physiologic or molecular biomarkers in confirming phenotypic variants in those with symptoms, as well as distinguishing phenotypes when clinical indicators are lacking. This approach may clarify conflicting evidence on these classifications and the varying efficacy of AF treatments.

Ultimately, defining “vagal” versus “adrenergic” AF may oversimplify the interdependent relationship of sympathetic and parasympathetic activation. In the first large observational study of vagal versus adrenergic AF, an autonomic pattern was only identifiable in 33% of patients with paroxysmal AF (PAF)⁴. Another later study could not distinguish autonomic phenotypes by history alone; however, there was evidence of increased adrenergic tone in the 20 minutes preceding AF followed by a predominance of vagal tone in the last 5 minutes in both patients with lone AF and those with structural heart disease⁶⁰, suggesting a synergistic role of both systems.

Sharifov, et al. demonstrated that isoproterenol and adrenaline infusions induced AF in 21% and 17% of canines, respectively, while acetylcholine induced AF in 100%. The effects of acetylcholine were potentiated with simultaneous isoproterenol infusion, decreasing the threshold acetylcholine concentration required for AF induction⁶¹. This study, and others, have demonstrated that the parasympathetic and sympathetic systems exert independent and synergistic effects on AF^{62, 63}.

Electrophysiologic Effects of Parasympathetic and Sympathetic Stimulation

Proposed autonomic-driven mechanisms underlying AF include those that facilitate reentry or provoke calcium mishandling causing early or delayed afterdepolarizations^{64, 65}.

Parasympathetic input shortens the atrial effective refractory period (AERP)^{19, 66, 67}, which is a known trigger for AF and has been implicated in PV re-entry, as suggested by optical mapping⁶⁸. Medications like verapamil and class III antiarrhythmic drugs attenuate AERP shortening^{69, 70}. Patients with PAF exhibit greater baseline AERP dispersion compared to those without AF, and the degree of AERP dispersion increases during episodes of supraventricular arrhythmia⁷¹. PVs also have, on average, shorter AERPs when compared with the LA and when compared with PVs of patients without AF⁷².

Sympathetic stimulation, in contrast, predominantly influences AF by predisposing myocytes to afterdepolarizations. Delayed afterdepolarizations (DADs) occur after full repolarization and typically arise from abnormal diastolic increases in cytoplasmic calcium (usually from abnormal ryanodine receptors) causing inward depolarizing sodium through Na/Ca exchange⁷³, a mechanism for AF common in heart failure⁷⁴. In contrast, early afterdepolarizations (EADs) have been associated with sympathetic stimulation and are caused by excitation during phase 2 or 3 of myocyte repolarization⁶⁵. EADs can arise from prolonged repolarization allowing L-type calcium channel recovery⁷⁵, and more recently have been associated with brief refractory periods coupled with increased vulnerability to calcium cycling^{16, 72}.

Sympathetic stimulation can provoke AF via the latter mechanism by transiently augmenting calcium flow through L-type calcium channels, often termed calcium transient currents^{16, 19, 76}. In a canine model, Patterson, et al. demonstrated that within PV myocytes, EAD formation is favored by enhanced calcium transient currents and subsequent increased forward Na/Ca exchange¹⁶. In these same myocytes, norepinephrine and isoproterenol infusions significantly increased calcium transient currents and EAD formation, both independently and with the addition of pacing interventions.

Both early (I_{Na, early}) and late (I_Na,late) sodium currents are also implicated in arrhythmogenesis. I_{Na, early} drives the initial upstroke in the non-pacemaker action potential, while I_{Na, late} courses through recovered sodium channels that have reopened during the plateau phase⁷⁷. Increased I_Na,late can prolong action potential duration^{77, 78} and provokes EADs through the former mechanism of increased calcium current through recovered L-type channels⁷⁷. Sympathetic stimulation can enhance I_Na,late and contribute to AF through PKA mediated posttranslational modifications^{79, 80}.

Autonomic modulation of the SA node may also promote reentry. In computer models, cholinergic stimulation caudally shifts the leading pacemaker site which, when coupled with an atrial premature beat and/or unidirectional block within the SA node, promotes reentry and atrial tachycardia⁸¹. This is attributed to an intrinsic heterogeneity in response to autonomic stimulation within the SA node, and is influenced by enhanced L-type calcium currents that increase the probability of atrial premature beats. Another study describes the onset of AF as “flickering” between sinus rhythm and AF, and that simultaneous sympathetic and parasympathetic modulation of the sinus node contribute to this “tug-of-war.”⁸²

Finally, while sympathetic and parasympathetic stimulation have distinct electrophysiologic effects, their ability to provoke AF is synergistic. In the study by Patterson, et al.,EADs provoked by sympathetic stimulation only led to sustained arrhythmias when acetylcholine was administered simultaneously¹⁶. Burashnikov, et al. similarly found that EADs formed after both rapid pacing and after AF termination in the presence of acetylcholine, but not in its absence, and that acetylcholine was an additional prerequisite for AF initiation⁸³. Interestingly, simultaneous PNS and SNS activation shortens AERPs as an algebraic sum of their individual effects⁶².

These findings unify theories of parasympathetic and sympathetic synergy; parasympathetic activation shortens effective refractory periods and action potential durations, which allows calcium mishandling induced by sympathetic stimulation to favor EADs that ultimately incite AF. Furthermore, this pattern of rapid atrial pacing followed by termination bears some similarity to patterns of HRV (increased adrenergic followed by vagal predominance) that occur prior to the onset of AF⁶⁰, underscoring that surges in parasympathetic and sympathetic tone, taken together, play a collaborative role in precipitating and sustaining AF.

Overall Autonomic Tone and AF

The heart has physiologic, circadian variations in rhythm⁸⁴ implicated in both atrial⁸⁵ and ventricular⁸⁶ arrhythmias, as well as sudden cardiac death^{87, 88}. Patients with frequent tachyarrhythmias often have increased atrial tachycardia and AF at night⁸⁹, possibly mediated by nocturnal increases in vagal activity⁹⁰.

Cardiac autonomic neuropathy, often observed in diabetes, is associated with AF⁹¹. Diabetes can cause both sympathetic and parasympathetic cardiac denervation⁹². Changes in HRV have been associated with asymptomatic episodes of AF in patients with type 2 diabetes and no prior history of AF⁹³. Additionally, increased epicardial adipose tissue (where most autonomic GP reside) is an independent predictor of AF recurrence, possibly due to underlying inflammatory responses in adipose tissue or electrical remodeling⁹⁴.

Finally, manipulation of autonomic outflow from GPs induces AF. ARGP stimulation can convert isolated premature depolarizations (PD) into AF-inducing PDs⁹⁵. Stimulating the base of the right superior PV progressively slows the canine heart rate and initiates AF after fewer PDs⁹⁶. This effect is attenuated by lidocaine injection at the site of GP stimulation⁹⁶. Ablation of epicardial and coronary sinus atrial tissue attenuates vagally induced AERP shortening in canines and reduces AF inducibility from cervical vagal stimulation⁹⁷. Moreover, high frequency stimulation of atrial GP, when synchronized with AERPs, can cause PV ectopy and subsequent AF in humans⁹⁸. Finally, response to GP stimulation independently predicts the recurrence of AF following PVI in patients with PAF, further suggesting that GP ablation may reduce AF⁹⁹.

THERAPEUTIC INTERVENTIONS

Clinical Trials Evaluating Therapeutic Modulation of the Neurocardiac Axis

Modulation of the neurocardiac axis has emerged as a promising therapy for treating AF. Techniques include GP ablation, pharmacological treatments, renal denervation, low-level stimulation of the vagus nerve, and blockade of the sympathetic ganglia, including the stellate ganglion (Figure 3).

Figure 3.

Therapeutic Modulation of the Neurocardiac Axis.

Therapeutic Interventions Targeting the Ganglionated Plexi

GP identification can be guided by (1) empiric/anatomical targeting and/or (2) high frequency stimulation (HFS). The latter method involves identifying GP regions that, when stimulated, induce sinus rate slowing (R-R lengthening) and/or blood pressure changes consistent with “vagal responses”¹⁰⁰. Additional methods for GP identification include imaging-based approaches (SPECT/CT guidance)¹⁰¹ or ectopy-triggered approaches¹⁰². The best method for identifying GPs has not yet been established as each method has its own limitations.

Ganglionated Plexi Ablation

Approaches to GP ablation (GPA) include transcatheter, endocardial techniques and thoracoscopic or open epicardial ablation (Supplementary Table 1). The cornerstone of AF ablation is PVI^103–105, and GPA has been investigated both alone and adjunct to PVI with varying results.

Endocardial Approaches to Ganglionated Plexi Ablation

Early studies established the feasibility and safety of endocardial GPA both when guided by HFS^{27, 36, 106} or anatomic mapping¹⁰⁷. The preferred order of GPA is (1) SLGP, (2) ILGP, (3) ARGP, (4) IRGP, because the IRGP acts as a gateway for the other GP to access the AV node. Early IRGP ablation may prematurely eliminate parasympathetic responses to stimulation of remaining GPs, which can make the endpoint of ablation ill-defined²⁷.

The efficacy of GPA when performed alone is variable, which is partly influenced by ablation methods. In small observational studies, Lemery, et al. found that HFS-guided GPA alone led to 50% freedom from AF at 8 months (N=14 patients)¹⁰⁶, while Katritsis, et al. employed anatomic-guided GPA with less promising results (GPA alone: 26% freedom from AF, PVI: 63%)¹⁰⁷. However, in a larger study of 80 patients randomized to HFS- or anatomic-guided GPA, Pokushalov, et al. demonstrated that anatomic-guidance was superior to HFS (77.5% freedom from AF vs 42.5%, p=0.02)¹⁰⁸. They postulated that their “cloud-like” lesions were superior to the “linear” lesions employed by Katritsis, et al., conferring greater efficacy with a higher area of ablation. Furthermore, GP sites have both sympathetic and parasympathetic innervation; in those with significant sympathetic contribution, HFS may not induce adequate vagal responses, ultimately underestimating GP regions requiring ablation¹⁰⁹. Interestingly, even in a study that only included patients with “vagal AF,” endocardial HFS-guided GPA alone was unsuccessful in eliminating AF inducibility; 17 of 18 patients required additional PVI to achieve freedom from AF at 15 months¹¹⁰.

Several studies suggest that the efficacy of endocardial GPA combined with PVI is greater than the efficacy of either technique alone. In an observational study of 297 patients with PAF, Pappone, et al. ablated GP targets with vagal reflexes (sinus bradycardia<40 bpm) in addition to PVI and found greater 1-year freedom from symptomatic AF compared to PVI alone (99% vs 85%, p=0.0002)³⁶. Katritsis, et al. randomized 242 patients with symptomatic PAF to (1) PVI alone, (2) anatomic GPA alone, and (3) PVI plus anatomic GPA. Two-year freedom from recurrent arrhythmia was most optimal in the PVI + GP ablation group (PVI only: 56%, GPA only: 48%, PVI+GPA: 74%, p=0.004)¹⁰⁹, with no differences in safety events, including post-procedural atrial flutter. In this trial, Katritsis et al. employed “cloud-like” lesions instead of the linear lesions they used previously¹⁰⁷. In another randomized trial of 264 patients with persistent forms of AF, PVI + GPA was superior to PVI + LA linear ablation with increased maintenance of sinus rhythm at three years (49% vs 34%, p=0.035) and less atrial flutter (6% vs 18%, p=0.002)¹¹¹. Taken together, these studies demonstrate that endocardial GPA improves maintenance of sinus rhythm when performed with PVI.

Epicardial Approaches to Ganglionated Plexi Ablation

Epicardial approaches to GPA have more variable efficacy. The Ganglion Plexus Ablation in Advanced Atrial Fibrillation Trial (AFACT) randomized 240 patients with paroxysmal or persistent AF undergoing surgical ablation (PVI+/−Dallas lesion set) to additional epicardial ablation of the 4 major GPs and LOM versus no additional ablation. At 1-year there was no added benefit of GP and LOM ablation¹¹². The GPA group also required more frequent pacemaker implantation due to sinus node dysfunction and AV block. Similarly, an observational study found no added benefit of epicardial GPA added to LA cryoablation during cardiac surgery in maintaining sinus rhythm with or without antiarrhythmic medications¹¹³.

The lack of benefit observed in the AFACT trial may be influenced by inadvertent GPA during PVI in the control group. While HFS-vagal responses were present in at least 1 GP in 87% of control patients compared to 0% of the GPA group, the differences between patients who only had 1 active GP site after PVI alone and patients who received full GPA may be too small for a clinical impact. Moreover, if the IRGP was ablated first, absent vagal responses to HFS may have prematurely indicated successful ablation. Ultimately, the epicardial approach to GPA does not appear to offer significant safety or efficacy advantages.

Hybrid approaches that combine epicardial ablation with endocardial ablation may show promise. While not designed to test GPA, the Converge Trial is the largest and only randomized clinical trial comparing hybrid and endocardial ablation in patients with persistent forms of AF¹¹⁴. The hybrid ablation arm (epicardial PVI followed by endocardial ablation to close gaps) was superior to endocardial ablation alone (67.7% freedom from AF vs 50%, p=0.036). Further studies are needed to clarify the impact of hybrid approaches on GP elimination, autonomic modulation, and improved outcomes.

Right Sided versus Left Sided Left Atrial GP Ablation

Ablation of either right-sided or left-sided left atrial GP (LAGP) alone can reduce AF inducibility thresholds independently in preclinical studies¹¹⁵. AF inducibility is virtually eliminated after ablation of all 4 GP plus the LOM¹¹⁵, suggesting that global GPA is most optimal.

As discussed previously, the recommended order of GPA ends with ARGP and IRGP to avoid falsely indicating inactivity in the SLGP and ILGP (and thereby successful ablation). In a retrospective analysis of 67 patients undergoing mini-maze surgery, patients were divided into: (1) right sided LAGP ablation first, and (2) left sided LAGP ablation first¹¹⁶. In those who underwent anatomic ablation, the order did not affect rates of AF recurrence. However, in those who underwent HFS-guided GPA, patients who received right sided ablation first had significantly higher AF recurrence than those with left sided ablation first (p=0.016), suggesting that HFS-guided ablation may cause incomplete ablation when approached from the right.

Ultimately, the efficacy of neuromodulation treatments is influenced by many factors. Which cells are responsible for ablation outcomes is not known. Not only are both sympathetic and parasympathetic cells highly co-localized³⁴, but also ablation damages both neurons and glial cells¹¹⁷. In an immunohistochemical study of S100B-expressing cells in the ICNS, S100B+ glial cells were found in high density surrounding neuronal cell bodies, which in turn expressed the S100B receptor¹¹⁸. In the same study, of 112 patients with AF receiving PVI, reduced AF recurrence was associated with increased S100B following ablation¹¹⁸, suggesting that the extent of glial cell injury may influence ablation efficacy.

Furthermore, in their study of inputs from porcine right atrial GP on the SA node, Hanna, et al. identified a diversity of gene and protein receptor expression, giving rise to highly heterogeneous neuronal subpopulations by transcriptome analysis¹¹⁷. The authors postulated that the heterogeneity in neuronal subtypes within GP may contribute to the heterogeneity in GPA results. Interestingly, selectively ablating cardiac neuronal subpopulations has become an avenue of research for improving ablation outcomes, and has been the focus of some preclinical studies in attenuating apnea-induced AF¹¹⁹ or ventricular tachycardia/fibrillation subsequent to cardiac remodeling¹²⁰, but has not yet been explored in humans.

Pharmacological Interventions

Studies have investigated botulinum toxin or calcium chloride injections into epicardial GP for preventing post-operative AF (POAF) (Supplementary Table 2). Botulinum toxin (BTX), the neurotoxin produced by Clostridium botulinum, interferes with parasympathetic signaling by blocking the release of acetylcholine from presynaptic vesicles¹²¹, while calcium chloride injections cause calcium induced neurotoxicity¹²².

Botulinum Toxin

In canines, BTX injection at sinus and AV nodal epicardial fat pads can temporarily suppress vagal AF for 1 week¹²³. Similarly, BTX injection into the ARGP, IRGP, SLGP and ILGP temporarily prolongs AERPs and eliminates bradycardic responses, while producing a more prolonged effect on diminishing AF burden at 3 months in canines¹²⁴.

In a pilot study, Pokushalov, et al. randomized 60 patients with PAF undergoing coronary artery bypass grafting (CABG) to either LA epicardial BTX injection or placebo. The BTX group not only had significantly less POAF (p=0.024), but also had 0% recurrent AF at 1 year compared to 27% of the placebo group (p=0.02)^{125, 126}. A subsequent randomized clinical trial (N=130) investigated peri-operative BTX injection in both LA and anterior epicardial fat pads during cardiac surgery¹²⁷. There was an 11% lower absolute risk of POAF in the BTX group compared with placebo that did not meet statistical significance, potentially due to lower than anticipated power. The patient population in this study was higher-risk (compared to the Pokushalov study), included patients undergoing valve surgery, had higher median age, and high prevalence of COPD; consequently, included patients may have had more arrhythmogenic substrate. Additionally, few patients had prior PAF, and those with a history of persistent AF were excluded. Because patients with AF are known to have GP hyperactivity when compared to healthy controls¹²⁸, it is possible that in patients without prior AF, the pathophysiology of POAF is less related to autonomic remodeling and GP hyperactivity and rather due to inflammation or sympathetic hyperactivity¹²⁹. A multicenter, international Phase 2 randomized clinical trial evaluating BTX for POAF after cardiac surgery is currently under way (NCT03779841)¹³⁰.

Calcium Chloride Injection

Calcium chloride injection induces apoptosis within cardiac GP and prevents POAF in animal studies¹³¹. In a randomized clinical trial, 200 patients without AF were randomized to calcium chloride injection of the major GP or placebo during CABG¹³². At 7 days, only 15% of patients receiving calcium chloride injection developed POAF compared to 36% of placebo (p=0.001); those who developed AF had significantly reduced AF burden. Studies of HRV demonstrated a balanced reduction in parasympathetic and sympathetic tone.

Renal Denervation

Renal denervation (RDN) involves ablating renal afferent and efferent sympathetic nerves with the purpose of reducing sympathetic tone¹³³. In animal models, RDN reduces blood pressure, progression of kidney injury, cardiac fibrosis and diastolic dysfunction¹³⁴. While RDN was less promising in treating resistant hypertension in the SIMPLICITY-HTN-3 trial⁴⁸, some suggest these results may have been influenced by incomplete denervation¹³⁵. Nevertheless, RDN may abate AF through reducing overall sympathetic tone.

In the multicenter, single-blind ERADICATE-AF trial, 302 patients with PAF were randomized to either (1) PVI alone (N=148) or (2) PVI+RDN (N=154)⁴⁹. After 12 months, the PVI+RDN group had significantly higher freedom from AF, atrial flutter, or tachycardia (72.1% vs 56.5%, p=0.006), with no significant difference in procedural complications. The benefit of RDN in patients with persistent AF has been demonstrated in smaller studies¹³⁶, and will be further evaluated in an ongoing clinical trial (NCT04055285).

In a meta-analysis that included 6 trials (N=725 patients), adjunctive RDN significantly reduced recurrent AF compared to PVI alone (Risk Ratio 0.68, 95% CI 0.55–0.83, p=0.0002)¹³⁷. Another trial in patients with paroxysmal or persistent AF and refractory hypertension trended towards increased efficacy in the ablation plus RDN group without reaching statistical significance at one year (p=0.34)¹³⁸, but is still undergoing extended follow-up and patient recruitment. Additional randomized clinical trials are ongoing (NCT01990911, NCT02115100), including a trial in ultrasound-based RDN (NCT04182620).

Low-Level Vagal Nerve Stimulation

In a small study of patients with PAF, 1 hour of low-level tragus stimulation during AF ablation significantly suppressed pacing-induced AF¹³⁹. In a sham-controlled, double-blind, randomized clinical trial in the ambulatory setting, patients with PAF received 1 hour of low-level tragus stimulation per day over 6 months via an ear clip (N=26) or sham control (N=27)¹⁴⁰. Median AF burden in those receiving stimulation was 85% lower than in the control group, with no device related side effects. In the stimulation arm, device compliance decreased over time (from 86% at 1 month to 74% at 6 months), and AF burden was higher at baseline in the treatment group, possibly amplifying the intervention’s effects¹⁴¹. In another randomized study, low-level vagal stimulation delivered via a bipolar wire sutured to vagal pre-ganglionic fibers alongside the superior vena cava decreased the incidence of POAF¹⁴².

Stimulation parameters and patient selection optimization are crucial in adequately understanding the efficacy of neuromodulation treatments. The heterogeneity of prior studies in nerve stimulation for heart failure has been attributed to varying site, strength, and cycles of stimulation^{143, 144}. Currently, there are few studies investigating nerve stimulation for arrhythmia suppression. Future studies should determine the comparative safety and efficacy of implantable or wearable devices, as well as define optimal stimulation parameters. Finally, while not directly related to vagus nerve stimulation, a pilot randomized study recently found that temporary spinal cord stimulation decreased POAF incidence from 31% to 4%¹⁴⁵.

Stellate Ganglion Block

In a placebo-controlled study, 36 patients with PAF were randomized to temporary, transcutaneous stellate ganglion block (SGB) unilaterally with lidocaine, or placebo¹⁴⁶. Prior to SGB, AF was inducible in all patients by atrial pacing; after SGB, AF was inducible in only 54% of patients (p < 0.01) and had significantly shorter duration (p < 0.01). The same results were achieved regardless of right-sided versus left-sided SGB. A pilot study of 25 patients undergoing cardiac surgery found that left-sided SGB had a lower POAF rate than the institution’s average (18.2% versus 27%)¹⁴⁷. A larger, randomized clinical trial of 200 patients undergoing lung lobectomy similarly found that patients randomized to right-sided SGB had significantly lower POAF (3% versus 10%, p=0.045). Interestingly, the SGB group also had lower WBC counts, suggesting a potential interaction between autonomics and inflammation. The long-term effects of SGB on AF have yet to be studied.

Conclusions

Atrial fibrillation has a multifaceted, complex pathophysiology, of which cardiac autonomic innervation plays a central role. Modulation of the neurocardiac axis has emerged as an innovative therapy with varying efficacy. As our understanding of atrial fibrillation and the ICNS expands, our approach to therapeutic neuromodulation will continue to evolve for the benefit of those with AF.