Cardiac autonomic control is generated by a hierarchy extending from the central nervous system to extracardiac thoracic plexuses and an intrinsic cardiac nervous system embedded mainly in epicardial fat. This hierarchy does not operate as a simple opposition between a sympathetic accelerator and a parasympathetic brake. Afferent information, efferent commands, and local circuit activity are integrated at several levels, while sympathetic and parasympathetic fibres converge within the same cardiac neural territories. This distributed organization is fundamental for understanding physiological control, autonomic contributions to arrhythmia, and the anatomical rationale—and limitations—of cardioneuroablation. [1–4]

1  Extrinsic autonomic pathways

Preganglionic sympathetic neurons supplying the heart arise in the intermediolateral spinal cord, predominantly in upper thoracic segments. Their axons reach cervical and upper thoracic sympathetic ganglia, where they synapse with postganglionic neurons. Cardiac sympathetic nerves then approach the heart through the cervical and thoracic cardiac branches and the superficial and deep cardiac plexuses around the great arteries and tracheal bifurcation. They enter the cardiac hilum along the aorta, pulmonary trunk, caval veins, and pulmonary veins, before continuing within epicardial nerves. Sympathetic activation increases sinus rate, atrioventricular conduction, myocardial contractility, and ventricular excitability, but regional effects depend on the distribution of fibres and receptors. [1,2]

Parasympathetic preganglionic neurons arise in the medulla and travel through the vagus nerves. Cervical and thoracic vagal cardiac branches join the extracardiac cardiac plexuses and reach the heart along the great vessels. Unlike the classic diagram in which vagal fibres terminate in one discrete cardiac ganglion, preganglionic axons synapse with postganglionic neurons distributed among numerous intrinsic ganglia. The right and left vagal pathways show functional tendencies rather than absolute territorial exclusivity: effects on the sinus and atrioventricular nodes overlap and are modified by interconnections within the intrinsic network. Visceral afferents travel through both vagal and spinal pathways and provide the information required for cardiovascular reflexes. [2–4]

Extrinsic nerves become progressively intermingled before they reach their myocardial effectors. For the interventionist, therefore, a nerve seen macroscopically near a great vessel cannot be assumed to contain a single functional fibre type. Nor can an atrial region be labelled purely parasympathetic. The target of cardioneuroablation is not the cervical or thoracic vagus itself, but atrial tissue anatomically related to intrinsic ganglia and their postganglionic projections.

2  The intrinsic cardiac nervous system

The intrinsic cardiac nervous system comprises ganglia, interconnecting nerves, and a heterogeneous population of neurons situated within the heart. Most human ganglia lie in epicardial adipose and connective tissue, especially on the atrial surfaces and around the cardiac hilum. They contain parasympathetic efferent neurons, afferent neurons, and local circuit neurons; sympathetic postganglionic fibres pass through and interact with this network. Immunohistochemical studies also demonstrate neurochemical diversity that cannot be reduced to a binary sympathetic–parasympathetic classification. The intrinsic system can integrate local mechanical and chemical information and modify cardiac behaviour even while remaining under central and intrathoracic control. This functional capacity explains the useful, though simplified, expression “little brain on the heart.” [3–6]

Ganglia are linked by epicardial and subepicardial nerves that form plexuses rather than isolated islands. Neural tissue is most abundant around the venous pole of the heart: the posterior and superior atrial walls, pulmonary venous junctions, interatrial groove, superior cavoatrial region, and inferior paraseptal region. Smaller nerves continue towards the sinus node, atrioventricular junctions, atrial myocardium, coronary vessels, and ventricles. Epicardial fat provides a pathway and microenvironment for these elements, but the visible boundaries of a fat pad do not define the microscopic limits or function of a plexus. [1,4,5]

3  Human ganglionated subplexuses: the anatomical map of Pauza

Pauza and colleagues examined whole human hearts using acetylcholinesterase histochemistry and reconstructed the epicardiac neural network as seven ganglionated subplexuses. Each subplexus was defined by the course of its principal nerves and the epicardial territory supplied, rather than by a single focal ganglion. The largest concentrations occurred close to the cardiac hilum and on the dorsal and dorsolateral left atrial regions; in their material, these territories contained approximately half of the identified ganglia. Individual hearts contained hundreds of ganglia, with substantial variation in number, size, distribution, and interconnection. Their estimates of total intrinsic neuronal number were method-dependent and varied markedly with age, so they should be understood as evidence of the scale and variability of the network, not as fixed normal values. [5]

This work changed the anatomical interpretation of cardiac ganglia. The clinically familiar scheme of four, five, or six named ganglionated plexuses around pulmonary veins and interatrial landmarks is a useful procedural abstraction, but it is not a one-to-one translation of Pauza’s seven subplexuses. A clinical “plexus” may correspond to part of a larger anatomical subplexus, to a zone in which several neural routes converge, or simply to an endocardial site from which an epicardial neural population can be influenced. Terminology should therefore be stated explicitly whenever anatomical and electrophysiological maps are compared. [3,5]

The network also varies between individuals and changes with age. A lesion set based only on schematic dots cannot guarantee contact with the same neural components in every patient. Conversely, neural effects at a distance do not prove that the catheter is directly opposite a ganglion: stimulation or ablation can engage interconnecting nerves, myocardium, or adjacent plexuses. This distinction is particularly important when procedural endpoints are used to infer anatomy.

4  Relationships with the sinus and atrioventricular nodes

Autonomic influence on the sinus node reaches the superior right atrium through neural routes concentrated near the superior caval vein, the right superior pulmonary venous region, and the superior interatrial groove. Clinically, this broad territory is often represented by a superior right atrial or superior paraseptal ganglionated plexus. Modification can produce sinus acceleration, but the response reflects withdrawal of vagal influence, direct sympathetic effects, procedural stress, and baseline autonomic state. Sinus acceleration alone is therefore neither a precise anatomical locator nor proof of complete denervation. [3,7]

Neural pathways influencing atrioventricular conduction are concentrated inferiorly around the atrial septal region, coronary sinus, and inferior atrioventricular junction. The inferior paraseptal region is closely related to the coronary sinus ostium, the inferior part of the triangle of Koch, the right and left atrial vestibules, and the epicardial tissues of the inferior pyramidal space. The compact atrioventricular node and penetrating conduction axis lie nearby, but they are not themselves the intended neural target. From the right or left atrial endocardium, the operator is approaching an epicardial or intramural neural network across atrial myocardium and fibrofatty tissue. The thickness and three-dimensional arrangement of these layers help explain why responses differ between sites and why excessive energy close to the conduction axis carries a distinct risk. [1,3,7]

Regional specialization is real but not exclusive. Superior plexuses may influence atrioventricular conduction, and inferior plexuses may affect sinus rate through interconnections. Experimental descriptions of “integration centres” should not be interpreted as sharply bounded human organs. The safest anatomical model is a distributed network with preferential inputs and outputs.

5  Anatomical basis of cardioneuroablation

Cardioneuroablation uses catheter energy to attenuate excessive cardiac parasympathetic influence in selected patients with reflex syncope or functional bradyarrhythmias. Proposed indications include a cardioinhibitory or mixed reflex syncope phenotype, functional sinus node dysfunction, and functional atrioventricular block after careful exclusion of intrinsic conduction disease. The conceptual target is the postganglionic parasympathetic component of the intrinsic cardiac nervous system. Because the ganglia are predominantly epicardial, conventional endocardial ablation reaches them indirectly through the atrial wall; lesion depth, tissue composition, and the course of intervening nerves all matter. [3,7,8]

Anatomical strategies target predetermined atrial regions associated with ganglionated plexuses. Functional strategies use high-frequency stimulation to provoke transient atrioventricular block, sinus slowing, or another vagal response. Electrogram-guided approaches seek fractionated or high-frequency atrial signals thought to mark neural interfaces, while spectral methods analyse signal components. Many procedures combine anatomical and functional information. None of these methods directly images every ganglion, and agreement between them is incomplete. A fractionated electrogram may arise from complex myocardial architecture; failure to provoke a vagal response may reflect stimulation parameters or anaesthesia rather than absence of neural tissue. [3,7,8]

The relationship between target and organ must remain explicit. A left atrial approach can access posterior and pulmonary venous neural territories, whereas a right atrial approach can access superior and inferior paraseptal territories without transseptal puncture. Biatrial ablation may increase the number of neural pathways affected, but also increases tissue injury and procedural complexity. The anatomy does not establish that one approach is universally superior. Choice depends on the clinical phenotype, mapping method, operator experience, and the endpoints being sought. [7,8]

Procedural endpoints have included abolition of evoked vagal responses, sustained increases in sinus rate, shortening of atrioventricular nodal conduction indices, changes in heart-rate variability, and loss of bradycardic responses during extracardiac vagal stimulation. Extracardiac vagal stimulation can test the final cardiac response more directly than local high-frequency stimulation, but it is not universally available and methodology varies. No single endpoint has yet become a universally validated measure of selective and durable parasympathetic denervation. Moreover, autonomic measurements are sensitive to sedation, respiration, medications, volume state, and time after ablation. [7,8]

6  Clinical importance and unresolved questions

The clinical attraction of cardioneuroablation is clear: in an appropriately selected younger patient with recurrent cardioinhibitory reflex syncope or functional bradycardia, autonomic modification might reduce symptoms while avoiding permanent pacing. Published observational series and smaller comparative studies report encouraging reductions in syncope and improvements in conduction. These results have driven rapid adoption and increasingly systematic techniques. Nevertheless, the 2024 international scientific statement led by Aksu and endorsed by major rhythm societies emphasized the limitations of the evidence: heterogeneous patient selection, variable lesion sets and endpoints, relatively small populations, limited randomized evidence, and incomplete long-term follow-up. Cardioneuroablation should therefore be presented as an evolving therapy, not as a mature anatomical cure for all forms of syncope or bradycardia. [8]

Patient selection is inseparable from anatomy and physiology. Syncope may include vasodepressor and cardioinhibitory components; eliminating a bradycardic response cannot necessarily correct peripheral vasodilatation. Likewise, degenerative disease of the sinus node, atrioventricular node, or His–Purkinje system is not expected to respond reliably to neural ablation. Documentation of a functional vagal mechanism and exclusion of structural conduction disease are therefore essential before an anatomical target is considered. Age, spontaneous rhythm, response to atropine or exercise, monitoring findings, and reflex testing may contribute, but no isolated test substitutes for integrated clinical assessment. [7,8]

Selectivity is another unresolved issue. Intrinsic ganglia contain heterogeneous neurons and receive both sympathetic and parasympathetic inputs. Radiofrequency lesions also affect myocardium, small vessels, connective tissue, and axons passing through the target region. Describing the procedure as pure “vagal denervation” overstates what current methods can demonstrate. The desired clinical effect may result from partial network modification rather than elimination of a discrete parasympathetic centre. This interpretation is more consistent with Pauza’s anatomical findings. [3,5]

Durability is influenced by lesion maturation, neural plasticity, collateral sprouting, and possible reinnervation. Early changes in heart rate or variability may diminish without necessarily predicting recurrence, while persistent denervation may have consequences that require longer observation. Potential adverse effects include inappropriate sinus tachycardia, atrial injury or arrhythmia, coronary or extracardiac damage depending on the target, and inadvertent injury to the normal conduction axis. The balance between adequate autonomic modification and unnecessary lesion burden remains a central research question. [7,8]

Future progress requires a shared anatomical language, reproducible physiological endpoints, direct or surrogate imaging of neural structures, and trials comparing cardioneuroablation with established management—including pacing when indicated. Correlating clinical maps with human subplexus anatomy will be more informative than forcing both into an identical nomenclature. In the atlas, the intrinsic nervous system should therefore be shown as a three-dimensional, interconnected epicardial network related to endocardial landmarks, not as a series of isolated coloured circles.