ABSTRACT
Background
Cardioneuroablation (CNA) targets atrial ganglionated plexuses (GP) to treat vagally mediated bradyarrhythmias. Automated electrogram fractionation mapping has been proposed as a surrogate tool for GP localization, but optimal parameter settings remain undefined. We aimed to systematically evaluate automated fractionation mapping parameters and their association with autonomic responses during ablation.
Methods
Data from 26 patients were analyzed, including 12 patients undergoing atrial fibrillation (AF) ablation in sinus rhythm and 14 patients undergoing CNA. Using high‐density electroanatomical mapping, three predefined values were tested for each fractionation parameter—signal width, refractoriness, and amplitude threshold—yielding 27 automated configurations. A purely anatomical localization strategy was also evaluated. In the AF cohort, all configurations were systematically assessed, generating 336 fractionation maps and 870 ablation sites. Autonomic response during radiofrequency delivery was defined as an absolute RR interval change > 10%. Selected high‐performing configurations were subsequently evaluated in the CNA cohort (614 ablation sites).
Results
In the AF cohort, 25 of 27 configurations showed a significant association with autonomic response. After Bonferroni correction, 15 configurations remained significant. The configuration with signal width 15 ms, refractoriness 25 ms, and amplitude threshold 0.05 mV demonstrated the highest sensitivity, although several configurations showed comparable performance. When evaluated in the CNA cohort, this configuration yielded a sensitivity of 74.1% and specificity of 73.9%.
Conclusions
This pilot study provides a reproducible framework for evaluating automated fractionation mapping parameters associated with autonomic responses during CNA. The findings support biologically plausible parameter–performance relationships and identify parameter ranges potentially suitable for procedural use.
Keywords: atrial fibrillation, autonomic nervous system, cardioneuroablation, electrogram fractionation, pulmonary vein isolation
Conceptual overview of the study design and physiological rationale underlying programmable electrogram fractionation mapping during ablation. In the primary AF cohort, 27 predefined combinations of signal width, refractory period, and roving sensitivity were systematically evaluated across 870 ablation sites during PVI to identify electrogram parameter profiles associated with autonomic responses. The central panel illustrates how programmable fractionation parameters influence electrogram classification and mapping sensitivity. In an independent exploratory CNA cohort, the optimized high‐sensitivity configuration identified in the AF cohort was retrospectively evaluated off‐line, demonstrating preservation of the fractionation–autonomic response association. Overall, the figure summarizes a reproducible methodological framework for identifying electrophysiological signatures associated with autonomic‐responsive atrial tissue.
1. Introduction
The intrinsic cardiac autonomic nervous system comprises clusters of post‐ganglionic parasympathetic neurons, known as ganglionated plexuses (GP), located mainly within the atrial walls or nearby para‐cardiac regions [1, 2]. Recognition of these neural structures has led to the development of catheter‐based techniques to selectively ablate GP and attenuate vagal influence [3, 4]. It has been postulated that the presence of GP affects local electrogram (EGM) characteristics, such as fractionation [5]. By using Fast Fourier Transform analysis, Pachon et al. described two types of myocardium: a compact myocardium, demonstrating uniform and fast conduction properties, and a fibrillar (fractionated) atrial myocardium, characterized by fragmented and heterogeneous conduction [5]. Accordingly, fractionated potentials have been proposed as a surrogate marker that may help to localize GP during electrophysiological studies [4]. Targeting these regions with radiofrequency (RF) energy may produce observable autonomic responses during ablation [4]. In 2007, Lellouche et al., analyzed regular bipolar signals and identified key characteristics of fractionated EGM associated with parasympathetic responses during ablation, thereby supporting the concept of GP localization [6]. However, their approach relied on subjective, off‐line visual analysis and was not applicable to real‐time clinical procedures.
Recently, automated fractionation mapping software integrated into three‐dimensional (3D) electroanatomical mapping systems has enabled real‐time detection of complex EGM patterns [7, 8]. This software uses programmable parameters—signal width, amplitude threshold (referred to as “roving sensitivity” in the proprietary algorithm) and refractoriness—to define EGM fractionation. However, the optimal settings for identifying EGMs associated with autonomic responses remain undetermined [9]. As a result, parameter selection has largely relied on empirical choices rather than systematic evaluation.
In the present study, we systematically tested multiple configurations of an automated EGM fractionation mapping algorithm to identify the settings associated with autonomic responses during pulmonary vein isolation (PVI). Our aim was to establish a reproducible methodological framework for identifying EGMs associated with autonomic responses.
2. Methods
2.1. Study Design
Patients undergoing their first PVI for atrial fibrillation (AF) were selected to evaluate fractionation mapping performance across left atrial regions with high likelihood of autonomic response, as well as regions where such responses were unlikely to occur. This design enabled systematic assessment of both true‐positive and true‐negative sites. Configurations identified as highly sensitive in the AF cohort were subsequently evaluated off‐line in a separate exploratory cohort of patients undergoing cardioneuroablation (CNA), with the aim of assessing whether the most sensitive configuration in the AF cohort could identify sites associated with autonomic responses during GP ablation in patients undergoing CNA.
Because no previous studies had systematically compared multiple programmable fractionation configurations, no formal sample size calculation was performed. Statistical analysis was conducted after acquisition of 870 ablation sites from 12 patients and 336 fractionation maps, once consistent associations between several configurations and autonomic responses had already emerged.
2.2. Primary AF Cohort
This cohort consisted of prospectively enrolled patients undergoing their first RF PVI for paroxysmal AF. EGM signals from 12 consecutive patients were analyzed during sinus rhythm. In all cases, the ablation strategy consisted exclusively of electrical PVI using wide antral circumferential lesions, without delivery of any additional lesions outside the pulmonary vein antra. PVI was performed following a standardized and reproducible ablation sequence in all patients (Ablation was consistently started at the posterior‐superior aspect of the left pulmonary veins, progressing inferiorly along the posterior wall around the inferior vein and then advancing superiorly along the ridge toward the roof to complete the antral circle. Subsequently, the right pulmonary veins were isolated using the same systematic approach, beginning at the posterior‐superior region, proceeding inferiorly around the inferior vein, and completing the circle along the anterior wall).
The protocol was approved by the local ethics committee. Informed consent was obtained from all patients prior to the procedure. Exclusion criteria included moderate or severe left atrial enlargement (defined as a left atrial volume index ≥ 42 mL/m2) as well as the development of any rhythm other than sinus rhythm during the procedure, including pacing‐dependent rhythms.
2.3. Exploratory CNA Cohort
To explore whether the fractionation mapping configuration identified as the most sensitive in the primary AF cohort was associated with autonomic responses during CNA procedures, electroanatomical maps from 14 patients undergoing CNA were retrospectively analyzed. A configuration previously suggested by Aksu et al. [10] was also evaluated (configuration #16 in our protocol with refractoriness 30 ms, signal width 5 ms and roving sensitivity 0.1 mV). These patients underwent CNA ablation for functional bradycardia, functional atrioventricular block, or cardioinhibitory vasovagal syncope.
Operators were not blinded to the fractionation map. The sequence of GP ablation was determined according to primary clinical indication (in patients with cardioinhibitory syncope, ablation was systematically started at the left superior GP, followed by the left inferior, right inferior, right posteroseptal, and finally right superior GP). In patients presenting with functional atrioventricular block, ablation was preferentially started at the right posteroseptal GP. Additionally, extracardiac vagal stimulation (ECVS) before and after ablation was used in some patients to modulate and titrate the extent of ablation. Atropine testing was performed in all patients prior to procedural indication and repeated at the end of the procedure.
Importantly, because CNA procedures were performed independently from the present retrospective analysis, ablation was not guided by a single predefined fractionation configuration among those evaluated in this study. Rather, different mapping settings and procedural strategies were used according to operator preference and clinical context. This variability reflects real‐world CNA practice rather than a standardized blinded experimental protocol. Furthermore, the retrospective use of the CNA cohort was intended exclusively to explore whether the most sensitive configuration identified in the AF cohort preserved its association with autonomic responses in an independent physiological setting, rather than to evaluate procedural efficacy or compare CNA strategies.
2.4. Procedures
In both cohorts, procedures were performed under general anesthesia. Electroanatomical mapping during sinus rhythm was performed using the EnSite NavX Cardiac Mapping System (St. Jude Medical, Abbott, Sylmar, CA, USA) and a high‐density multielectrode mapping catheter (Advisor HD Grid Mapping Catheter, Sensor Enabled, Abbott, Minneapolis, MN, USA). Fractionation maps were acquired prior to any RF application, ensuring that EGM characteristics were not influenced by ablation lesions. Bipolar recordings were filtered between 30 and 300 Hz. Mapping parameters were standardized at 7 mm for internal and external projections, 7 mm for interpolation, and 0.1 mV for low‐voltage identification [10]. This filtering range and fractionation parameter framework have been used in prior EGM‐guided CNA studies [10].
In the primary AF cohort, pulmonary vein (PV) isolation was performed during sinus rhythm using an irrigated catheter (TactiCath or TactiFlex Ablation Catheter, Sensor Enabled, Abbott, Minneapolis, MN, USA). All patients underwent wide‐area circumferential ablation encircling the PVs. RF applications were delivered at a power of 35 W, reduced to 30 W in the posterior wall, with an endpoint lesion size index (LSI) of 4.5 to 5 in the posterior wall and 5.5 in other walls.
Cardiac cycle length (RR interval) and lesion locations were recorded for each RF application. In both cohorts, the baseline RR interval was measured immediately before the onset of each lesion. During RF delivery, the RR interval was defined as the mean of the three most extreme consecutive RR intervals observed during the application (depending on whether the RR interval increased or decreased).
In the exploratory CNA cohort, electroanatomical maps obtained during routine CNA procedures were reprocessed off‐line using both the most sensitive fractionation parameter configuration identified in the AF cohort and the configuration suggested by Aksu and colleagues [10]. In this analysis, fractionation maps were generated post hoc, and each ablation site was classified according to the presence or absence of fractionated electrograms based on the predefined configuration. Autonomic response at each ablation site was then assessed using the same definition applied in the AF cohort based on RR interval changes.
2.5. Off‐Line Analysis
Once the map of each patient was obtained and stored, an off‐line point‐by‐point analysis of RR response and corresponding fractionation mapping was performed. In the primary AF cohort, to determine a threshold for RR interval variation that would exceed expected physiological variability in heart rate, ten non‐consecutive RR intervals were measured in each patient during general anesthesia, with catheters in place but before any ablation. Mean RR interval and standard deviation were calculated. Because a variation exceeding two standard deviations would suggest modulation beyond expected physiological fluctuation, an absolute RR interval change > 10% was ultimately selected to define autonomic response, consistent with prior literature [11]. This definition intentionally captured both RR prolongation and shortening, reflecting bidirectional autonomic responses during RF delivery. This approach provided a pragmatic and reproducible surrogate for autonomic response during mapping procedures.
2.6. Fractionation Mapping Algorithm
The fractionation mapping software (EnSite NavX Cardiac Mapping System, St. Jude Medical, Abbott, Sylmar, CA, USA) allows adjustment of three parameters (Figure 1): width, amplitude threshold (roving sensitivity), and refractoriness. These parameters determine how individual deflections within bipolar EGM are detected and classified as fractionated signals. The tested parameter values were intentionally selected to span the full range permitted by the software, using low, intermediate, and high settings in accordance with manufacturer recommendations. These values were not intended to represent physiological thresholds but to enable systematic methodological comparison with the constraints of the mapping system.
FIGURE 1.
Definitions and schematic of fractionation mapping algorithm parameters: Maximum Width: maximum time allowed from the start of deflection until its return to isoelectric line to be counted as one deflection for the calculation of fractionation. Amplitude threshold: (Roving sensitivity) The minimum voltage amplitude required for a deflection to be counted as a fractionation. Set to just above the baseline noise to eliminate any EGM noise from being mistaken as a fractionation. Refractory: the minimum time allowed from start of one deflection until a second deflection may be counted. This parameter acts as a blanking period to prevent new detections (double counting) for a specified time after detection has occurred [7].
A combination of these parameters determines the number of deflections which, in turn, identifies fractionation based on a predetermined threshold. Fractionation areas were visualized on the map using a color scale [8]. We systematically tested three predefined values for each parameter: Width (5, 10, or 15 ms), refractoriness (20, 25, or 30 ms), and amplitude threshold (0.05, 0.1, or 0.15 mV), resulting in 27 mapping configurations (Table S1). For each combination map, each ablation point was reviewed to identify if fractionation was present and determine if a significant autonomic response occurred. For schematic purposes, each ipsilateral pulmonary vein antral region was subdivided into four anatomical quadrants (anterosuperior, anteroinferior, posterosuperior, and posteroinferior). An additional mapping configuration was created based on an anatomical approach. For this configuration, ablation zones were delineated according to anatomical schemes proposed by Sun et al. [4]. The effectiveness of this configuration was evaluated to assess the performance of a purely anatomical strategy [10].
2.7. Statistical Analysis
Statistical analyses were performed at both the point‐by‐point and patient levels. For each of the 27 automated fractionation mapping configurations and for the anatomical configuration, 2 × 2 contingency tables were constructed to evaluate the association between EGM fractionation and autonomic response. Chi‐square testing was performed for each configuration. Because 27 predefined fractionation configurations were evaluated, a Bonferroni‐corrected significance threshold of p < 0.00185 (0.05/27) was additionally applied as a conservative adjustment for multiple comparisons. The primary analysis was performed on a point‐by‐point basis, considering each ablation site as an individual observation. This strategy allowed systematic evaluation of fractionation mapping performance across atrial regions both likely and unlikely to exhibit autonomic responses.
Because multiple mapping sites originated from the same patient, patient‐level bootstrap‐adjusted sensitivity and specificity estimates were subsequently used as the primary performance metric to account for intra‐patient clustering. Mean estimates and corresponding 95% confidence intervals were derived from the bootstrap distributions.
To explore differences among high‐performing configurations, selected pairwise comparisons were additionally performed. Differences in sensitivity between selected configuration pairs were then estimated with McNemar testing and Holm–Bonferroni correction for multiple comparisons.
Continuous variables are presented as mean ± standard deviation (SD) or median with range when appropriate. Categorical variables are presented as counts and percentages. Statistical analyses were performed using Microsoft Excel for Mac (version 16.84), IBM SPSS Statistics (version 29.0.2.0), and Calcupedev Version 11.
3. Results
3.1. Primary AF Cohort
A total of 870 ablation sites from 12 patients (mean age 60.17 ± 11.14 years; 33% female) were analyzed, with an average of 72.5 sites per patient. Procedures were performed between February 2023 and February 2024. Off‐line voltage map analysis did not demonstrate significant scarring or low‐voltage areas. Patient characteristics are detailed in Table 1.
TABLE 1.
Baseline characteristics of the population.
| Primary AF cohort | Exploratory CNA cohort | |
|---|---|---|
| Patients (n) | 12 | 14 |
| Age (years) | 60.1 ± 11.1 | 30.9 ± 11.6 |
| Female (%) | 33.3 | 71.4 |
| Hypertension (%) | 66.6 | 0 |
| Sleep apnea (%) | 8.3 | 0 |
| Heart failure (%) | 8.3 | 0 |
| Ejection fraction (%) | 55 ± 8.5 | 63 ± 3.1 |
| LA volume (mL) | 37.7 ± 7.8 | 21.1 ± 5.1 |
| Mean baseline RR interval (ms) ± SD | 1166 ± 39.78 | 1115 ± 423 |
| Maps analyzed | 336 | 28 |
| Maps analyzed per patient | 28 | 2 |
| Ablation points analyzed | 870 | 614 |
| Ablation points analyzed per patient | 72.6 ± 6.2 | 43.9 ± 19.4 |
The mean baseline RR interval was 1166 ms with a mean standard deviation (SD) of 39 ms, corresponding to a coefficient of variation of 3.4%. Because a deviation greater than twice the SD would suggest variation beyond expected physiological fluctuation, RR interval changes exceeding 6.8% of baseline were considered suggestive of autonomic modulation. To provide a standardized and reproducible definition aligned with prior literature, an absolute RR interval variation greater than 10% was ultimately used to define autonomic response [11].
Twenty‐five of the 27 automated configurations demonstrated nominal statistical significance (p < 0.05), whereas the purely anatomical configuration did not.
After Bonferroni correction for 27 predefined comparisons (corrected significance threshold p < 0.00185), 15 configurations remained significantly associated with autonomic response. Exact p values and Bonferroni‐corrected results for all evaluated configurations are provided in Table S2.
Bootstrap‐adjusted sensitivity and specificity across all evaluated configurations are summarized in Figure 2 and raw point‐by‐point estimates are provided in Table S2. Bootstrap‐adjusted sensitivity ranged from 30.4% to 88.9%, whereas specificity ranged from 48.1% to 89.7%. High‐sensitivity configurations clustered within parameter sets combining intermediate‐to‐high width values (10–15 ms), intermediate refractoriness (20–25 ms), and low amplitude thresholds (0.05–0.10 mV). In contrast, highly restrictive parameter settings demonstrated progressively lower sensitivity with increased specificity. Configuration #6 (width 15 ms, refractoriness 25 ms, amplitude threshold 0.05 mV) demonstrated the numerically highest bootstrap‐adjusted sensitivity, although adjacent configurations demonstrated overlapping confidence intervals and comparable overall performance (Figure 2).
FIGURE 2.
Bootstrap‐adjusted sensitivity and specificity across all evaluated fractionation mapping configurations. Sensitivity (left) and specificity (right) are shown with 95% bootstrap confidence intervals obtained using patient‐level bootstrap resampling to account for intra‐patient clustering. Configurations are ordered by decreasing sensitivity. Dotted vertical blue lines indicate the 95% confidence interval of the most sensitive configuration (#6: Width 15 ms—Refractory period 25 ms—Roving sensitivity 0.05 mV). All other configurations have confidence intervals that lie entirely within this interval, except configurations #26 (10–30–0.15), #19 (5–20–0.15), #25 (5–30–0.15), and the empirical anatomical approach. Configuration parameters are defined as follows: Width (5, 10, or 15 ms), refractory period (20, 25, or 30 ms), and roving sensitivity (0.05, 0.10, or 0.15 mV). Raw point‐by‐point estimates are provided in Table S2.
To illustrate the impact of software parameter selection on the spatial distribution of automatically detected fractionated electrograms, Figure 3 presents representative maps obtained from the same patient using configurations #6 and #25. Markedly different patterns of fractionation were observed. Configuration #6, which demonstrated the highest sensitivity for autonomic response detection, identified broader and more continuous regions of fractionated electrograms. In contrast, configuration #25, which demonstrated the highest specificity, generated smaller and more restricted regions. Sites associated with autonomic responses were predominantly located within fractionated regions, visually illustrating the relationship between automated fractionation mapping and physiological autonomic responses.
FIGURE 3.
Representative case illustrating the effect of automated fractionation mapping settings on the spatial distribution of fractionated atrial EGMs during cardioneuroablation. Three‐dimensional electroanatomic maps of the left atrium were obtained with EnSite NavX Cardiac Mapping System (St. Jude Medical, Abbott, Sylmar, CA, USA) in the same patient and projection (antero‐posterior view). In all panels, white areas represent regions classified by the automated algorithm as fractionated according to the selected software parameters, whereas purple areas represent regions with absent or low fractionation. Multicolored borders indicate the transition across the fractionation color scale generated by the mapping system. White areas: Fractionated EGM; Purple areas: absent/low fractionation; White spheres: autonomic response sites during RF delivery; Pink/red spheres: RF lesions. Panel A: Fractionation map obtained with configuration #6, defined by a signal width of 15 ms, refractoriness of 25 ms, and amplitude threshold of 0.05 mV. This configuration showed the highest sensitivity in the primary AF cohort and identified a broader distribution of fractionated regions. Panel B: Fractionation map obtained with configuration #25, defined by a signal width of 5 ms, refractoriness of 30 ms, and amplitude threshold of 0.15 mV. This configuration showed the highest specificity and identified a more restricted distribution of fractionated regions. Panel C: Fractionation map obtained again with configuration #6 (signal width of 15 ms, refractoriness of 25 ms, and amplitude threshold of 0.05 mV) with RF ablation sites displayed as pink and red spheres. White spheres indicate sites where an autonomic response was observed during RF delivery according to the prespecified study definition. This panel illustrates the spatial relationship between areas classified as fractionated by the automated map and sites associated with functional autonomic responses during RF delivery.
Heatmap visualization of sensitivity and specificity across parameter combinations demonstrated a performance gradient according to refractoriness and amplitude threshold settings. Configurations combining low amplitude thresholds (roving sensitivity) and intermediate refractoriness consistently demonstrated higher sensitivity, whereas increasing refractoriness and amplitude thresholds shifted performance toward higher specificity at the expense of reduced sensitivity (Figure 4).
FIGURE 4.
Heatmap comparing sensitivity and specificity. Heatmaps display bootstrap‐adjusted sensitivity (left panels) and specificity (right panels) according to combinations of width, refractory period, and roving sensitivity parameters. Warmer colors indicate higher performance values. Across parameter combinations, lower amplitude thresholds and intermediate refractory periods were consistently associated with higher sensitivity, whereas increasing refractory period and amplitude threshold progressively shifted performance toward higher specificity at the expense of sensitivity.
Although configuration #6 demonstrated the numerically highest sensitivity, several adjacent configurations (#14, #5, #2, and #3) showed broadly comparable performance, supporting the concept of a cluster of high‐performing parameter combinations rather than a single universally optimal configuration. In contrast, the previous configuration used by Aksu et al. (#16) demonstrated significantly lower sensitivity in paired bootstrap‐adjusted comparisons.
Pairwise comparisons were restricted to the configurations with the highest bootstrap‐adjusted sensitivity and to the configuration previously used by Aksu et al. [10] (#16 in our classification), in order to assess whether clinically relevant differences in sensitivity existed among the most relevant parameter combinations. Detailed pairwise comparisons are summarized in Table S3.
3.2. Spatial Distribution of Autonomic Responses
The spatial distribution of autonomic responses demonstrated a non‐random anatomical pattern (Figure 5). The highest proportions of autonomic responses were observed in the left inferior‐anterior region (11.2%), left superior‐anterior region (10.4%), and right superior‐anterior region (7.4%), corresponding anatomically to regions commonly associated with the left lateral/Marshall‐related GP and the right anterior GP [4]. In contrast, posterior and inferior right‐sided regions demonstrated substantially lower response rates. Overall, 51 of 870 ablation sites (5.86%) exhibited autonomic responses (Table S4).
FIGURE 5.
Spatial distribution of sites of autonomic response. The distribution of autonomic responses observed during RF delivery is displayed according to pulmonary vein antral region. Higher response frequencies were observed predominantly within left anterior regions and the right superior‐anterior region, corresponding anatomically to areas commonly associated with the left lateral/Marshall‐related GP and the right anterior GP. In contrast, posterior and inferior right‐sided regions demonstrated substantially lower response frequencies.
3.3. Exploratory CNA Cohort
The exploratory CNA cohort included 14 patients who underwent ablation between November 2022 and March 2025 (mean age 30.9 ± 11.6 years; 71.4% female). On average, 43.9 ablation points were delivered per patient. The most common clinical indication was functional bradycardia (57%) (Table 1).
Physiological markers of autonomic modulation before and after CNA are summarized in Table 2. Mean heart rate increased following ablation, whereas atropine‐induced chronotropic response became markedly attenuated. Similarly, extracardiac vagal responsiveness was abolished in most tested patients, Wenckebach cycle length decreased, syncope and presyncope burden was reduced, and mean heart rate and variability on follow‐up were consistent with reduced parasympathetic influence (Supporting Information Clinical Observations and Supporting Information Figure).
TABLE 2.
Physiological markers of autonomic modulation before and after CNA procedures.
| Pre‐procedure | Post‐procedure | Statistical analysis | |
|---|---|---|---|
| Mean heart rate (bpm) | 63.8 ± 15.1 | 83.2 ± 10.4 |
Wilcoxon signed‐rank test, Z = −3.234, p = 0.001 |
| Atropine‐induced ΔHR (bpm) | 51 (9–74) | 4 (0–19) |
Wilcoxon signed‐rank test, Z = −3.300, p < 0.001 |
| Positive extracardiac vagal response (performed in 9 patients) | 100% (9/9) | 11.1% (1/9) |
Exact binomial test, p = 0.039 |
| Wenckebach cycle length (ms) | 456 ± 102 | 372 ± 61 |
Wilcoxon signed‐rank test, p = 0.001 |
Note: Values are presented as mean ± SD or median (range), as appropriate. ΔHR = change in heart rate after atropine administration.
Despite marked demographic and structural differences between AF and CNA cohorts, the association between fractionation and autonomic response appeared to remain preserved in the exploratory CNA cohort (Table 3).
TABLE 3.
Exploratory indirect comparison of diagnostic performance metrics between the HAFE cohort described by Lellouche et al. [6] and the present study.
| HAFE pattern (95% CI) [6] | Present study | ||
|---|---|---|---|
| Primary AF cohort (Highest‐performing configuration by metric bootstrap—adjusted) (95% CI) | Exploratory CNA cohort (95% CI) a | ||
| Sensitivity (%) |
72 (49.4–94.5) |
88.9 a (78.1–99.7) |
74.1 (63.6–82.4) |
| Specificity (%) |
91 (85.2–97.2) |
90 b (84.2–95.7) |
73.9 (70–77.5) |
Note: Different parameter configurations yielded the highest values for individual diagnostic performance metrics. Values are shown for descriptive comparison and should not be interpreted as direct statistical comparisons between studies.
Width 15 ms; refractoriness 25 ms; amplitude threshold 0.05 mV; (configuration # 6).
Width 5 ms; refractoriness 30 ms; amplitude threshold 0.15 mV; (configuration # 25).
Using the highest‐sensitivity parameter configuration identified in the primary AF cohort (configuration #6: width 15 ms, refractoriness 25 ms, amplitude threshold 0.05 mV), the association between fractionation and autonomic response remained significant in the exploratory CNA cohort (p = 0.001). Sensitivity was 74.1% (95% CI 63.6–82.4), specificity 73.9% (95% CI 70.0–77.5), positive predictive value 30.1% (95% CI 24.2–36.8), and negative predictive value 94.9% (95% CI 92.4–96.7). In contrast, the configuration commonly used in previous CNA studies (width 5 ms, refractoriness 30 ms, amplitude threshold 0.1 mV) [10] demonstrated lower sensitivity and did not retain significant association in this exploratory cohort (p = 0.39).
Restricting analysis to the first lesion per GP per patient to account for possible physiological modification after the initial RF application increased sensitivity to 82.35% and NPV to 88%, with a corresponding reduction in specificity (57.5%) and PPV (45.2%) (p = 0.006).
4. Discussion
4.1. Main Findings
This study systematically evaluated programmable parameters of automated fractionation mapping software in relation to autonomic responses during PVI in a primary AF cohort and subsequently explored the performance of selected high‐performing configurations in an independent exploratory CNA cohort. The principal findings were: (1) multiple—but not all—fractionation parameter combinations demonstrated significant association with autonomic responses, including 15 configurations that remained significant after Bonferroni correction for multiple comparisons; (2) high‐performing configurations shared a reproducible profile characterized by intermediate‐to‐high width values, intermediate refractoriness, and low‐to‐moderate amplitude thresholds; (3) a structured performance gradient was consistently observed across parameter combinations, demonstrating a progressive trade‐off between sensitivity and specificity; (4) the anatomically guided approach showed no significant association; and (5) the association between fractionation and autonomic response remained preserved in the exploratory CNA cohort.
Importantly, although 25 of the 27 configurations demonstrated nominal statistical significance, 15 configurations remained significant after Bonferroni correction for multiple comparisons. This persistence of significance across a large subset of parameter combinations supports that the observed associations are unlikely to be explained solely by chance and reinforces the biological plausibility of the fractionation–autonomic response relationship.
The observed performance gradients remained consistent after patient‐level bootstrap correction, supporting that the associations were not solely driven by clustering of multiple mapping points within individual patients. Compared with the anatomical approach, high‐performing fractionation configurations demonstrated greater sensitivity and negative predictive value, supporting their potential utility as procedural adjuncts during CNA. Bootstrap‐adjusted estimates should be interpreted as patient‐level resampling performance metrics rather than direct replacements for the pooled raw point‐by‐point estimates.
In the exploratory CNA cohort, post‐ablation attenuation of vagal responses and changes in physiological autonomic markers were consistent with effective autonomic modulation. However, because analysis of the CNA cohort was performed retrospectively, the objective was limited to determining whether the high‐sensitivity configuration identified in the AF cohort was also capable of predicting autonomic responses at ablation sites in an independent CNA population. Importantly, CNA procedures were not guided by this optimized configuration, and mapping parameters during the original procedures were not standardized across patients. Therefore, the present analysis does not allow conclusions regarding whether the high‐sensitivity configuration influenced procedural efficacy or clinical outcomes in the CNA cohort. Rather, preservation of the fractionation–response association provides exploratory support for the external applicability and physiological consistency of the mapping signal across different autonomic ablation settings.
Heatmap and parameter subset analyses demonstrated reproducible performance gradients across parameter combinations. Refractoriness and amplitude threshold exerted the greatest influence on mapping performance, whereas width demonstrated a comparatively smaller effect (Table S5).
4.2. Peculiarities of the Study Design
We intentionally selected paroxysmal AF patients with non‐severe atrial dilatation in sinus rhythm for the primary AF cohort, allowing systematic sampling of atrial regions both likely and unlikely to harbor GP and thereby balancing positive and negative sites for analysis. Voltage maps did not reveal significant low‐voltage areas in the AF cohort, suggesting that the observed EGM fractionation was unlikely to be driven by a structural substrate and may instead reflect a functional mechanism. By selecting patients without major atrial structural remodeling and using a standardized anesthesia protocol, we optimized the ability to identify software configurations associated with autonomic‐responsive regions.
Our definition of autonomic response accounted for previous observations demonstrating that both heart rate slowing and paradoxical acceleration may occur during acute autonomic modulation, particularly during ablation near the right anterior GP [12, 13]. The study was not designed to differentiate specific autonomic mechanisms, and autonomic response was therefore used as a surrogate procedural marker rather than a mechanistic endpoint.
To address potential clustering bias due to multiple observations originating from the same patient, patient‐level bootstrap correction and clustering‐adjusted analyses were incorporated into the study design. Importantly, the overall configuration gradients remained stable after clustering correction, supporting the internal consistency of the findings despite the exploratory evaluation of multiple parameter combinations.
Bootstrap and paired comparison analyses suggested that high sensitivity was observed across a cluster of adjacent configurations rather than by a single universally optimal parameter set.
Autonomic responses demonstrated a non‐random anatomical distribution, predominating in left anterior and right superior‐anterior regions commonly associated with major GP locations, further supporting the physiological plausibility of the observed associations (Figure 5). Detailed regional frequencies are provided in Table S4.
In the exploratory CNA cohort, the proportion of autonomic‐response sites increased from 5.8% to 13.2% because ablation was restricted to presumed GP regions. Despite this higher baseline prevalence, the fractionation–response association remained significant.
Amplitude threshold (formerly “roving sensitivity”) and refractoriness were the parameters most strongly associated with detection of autonomic responses. Lower amplitude thresholds (≈0.05 mV) likely facilitate detection of smaller individual deflections within fractionated electrograms, thereby increasing the sensitivity of the algorithm for complex electrogram patterns associated with autonomic responses. This does not necessarily imply that the overall electrogram is of low amplitude. Shorter refractory periods (≈20 ms) may allow counting of closely spaced deflections within fragmented electrograms. Together, these observations suggest that electrograms associated with autonomic responses may contain closely spaced and sometimes small‐amplitude component deflections that are more effectively captured using lower amplitude thresholds and shorter refractoriness settings.
Lellouche et al. [6] previously described high‐amplitude fractionation EGM (HAFE) associated with parasympathetic responses using off‐line visual analysis of bipolar EGMs. Aksu and colleagues subsequently demonstrated the feasibility of software‐based fractionation mapping during sinus rhythm, although parameter selection remained largely empirical [8]. In this context, the present findings may be interpreted as a contemporary software‐based extension of the original HAFE concept, translating visually recognized EGM characteristics into reproducible and programmable mapping parameters (Table S6).
Because extensive low‐voltage areas were not identified in the AF cohort, advanced structural remodeling was less likely to represent the primary determinant of the observed fractionation patterns. Nevertheless, EGM fractionation should not be interpreted as a specific marker of autonomic tissue.
A recent multicenter study [14] associated greater numbers of RF applications with improved clinical outcomes after CNA irrespective of localization strategy. In contrast, the present findings suggest that systematic parameter optimization may improve identification of autonomically responsive regions while potentially reducing unnecessary lesion delivery.
Although the software used in the present study could not distinguish HAFE from Low Amplitude EGM (LAFE) patterns, fractionation remained associated with autonomic responses across most configurations. The relatively low positive predictive values indicate that fractionation mapping alone remains insufficient as a standalone determinant of ablation targets and should instead be integrated with anatomical and physiological information [6].
Overall, these findings may help refine EGM‐guided strategies for GP localization during CNA procedures. By providing a systematic framework for parameter selection within automated fractionation mapping algorithms, this approach may contribute to improved reproducibility and physiological consistency of EGM‐guided autonomic modulation strategies.
5. Limitations
This pilot exploratory study has a relatively small sample size, which limits the precision of performance estimates, particularly for specificity and PPV. Although Bonferroni correction was applied to account for multiple testing, the exploratory nature of the study and the large number of evaluated parameter combinations increase the possibility of residual type I error and require confirmation in independent prospective datasets.
Findings were derived using a single mapping system and may not be directly generalizable to other software platforms employing different signal acquisition and processing algorithms. Fractionated EGMs may also reflect mechanisms other than autonomic innervation, including localized fibrosis, PV potentials, inter‐atrial signal overlap, or slow conduction (e.g., AV nodal slow pathway) [9]. In typical CNA patients treated for reflex syncope, significant fibrosis is less likely; however, these mechanisms cannot be fully excluded.
Although voltage was not analyzed as an independent substrate parameter, voltage information was intrinsically incorporated into the fractionation algorithm through the amplitude threshold (roving sensitivity) setting. Therefore, while voltage contributed to fractionation detection, the present study cannot determine whether standalone voltage mapping would provide incremental value beyond its role within the automated fractionation algorithm. Future studies integrating dedicated voltage mapping or advanced imaging may further refine discrimination between autonomic‐related and substrate‐related fractionation.
Although the definition of autonomic response intentionally captured bidirectional responses (RR prolongation and shortening), smaller autonomic effects below the prespecified threshold may not have been detected.
The primary cohort consisted of patients undergoing PVI for paroxysmal AF rather than typical candidates for CNA. However, this population was intentionally selected to enable systematic assessment of fractionation patterns in atrial regions both close to and remote from usual GP locations while minimizing major atrial structural remodeling.
Because the analysis was restricted to the left atrium in the AF cohort, the EGM features identified in this study represent functional properties of atrial tissue and may, in principle, be present in both atria. Nonetheless, the lack of right atrial evaluation of response may limit the generalizability of the findings.
As an exploratory methodological study, no formal sample size calculation was performed. Instead, the study was designed to achieve enough ablation sites to allow point‐by‐point analysis across a wide range of parameter configurations, consistent with prior studies in the field.
These findings require confirmation in larger prospective cohorts to determine whether parameter‐guided fractionation mapping translates into improved procedural or clinical outcomes after CNA procedures.
6. Conclusion
Automated fractionation mapping demonstrated reproducible associations between specific programmable EGM parameter profiles and atrial sites exhibiting autonomic responses during RF delivery. High‐performing configurations consistently clustered within parameter sets combining intermediate‐to‐high width values, intermediate refractoriness, and low amplitude thresholds, supporting the presence of a biologically plausible mapping signal.
The persistence of these associations after patient‐level clustering correction and in an independent exploratory CNA cohort supports the methodological robustness of the approach.
Given the exploratory nature of the study and the relatively small sample size—particularly within the CNA cohort—these findings should be interpreted as hypothesis‐generating and require confirmation in larger prospective studies designed to evaluate procedural and clinical outcomes (Figure 6).
FIGURE 6.
Programmable EGM parameters and autonomic responses during ablation. Conceptual overview of the study design and physiological rationale underlying programmable electrogram fractionation mapping during ablation. In the primary AF cohort, 27 predefined combinations of signal width, refractory period, and roving sensitivity were systematically evaluated across 870 ablation sites during PVI to identify electrogram parameter profiles associated with autonomic responses. The central panel illustrates how programmable fractionation parameters influence electrogram classification and mapping sensitivity. In an independent exploratory CNA cohort, the optimized high‐sensitivity configuration identified in the AF cohort was retrospectively evaluated off‐line, demonstrating preservation of the fractionation–autonomic response association. Overall, the figure summarizes a reproducible methodological framework for identifying electrophysiological signatures associated with autonomic‐responsive atrial tissue.
Author Contributions
The initial conception of the study was performed by Jesús Almendral. Final study design was developed by Juan F. Agudelo‐Uribe, Jesús Almendral, and Rafael Correa‐Velásquez. Data collection was carried out by Juan F. Agudelo‐Uribe, Juan D. Ramírez‐Barrera, and Rafael Correa‐Velásquez. Data analysis was performed by Juan F. Agudelo‐Uribe and Margarita Londoño‐Arango. The first draft of the manuscript was written by Juan F. Agudelo‐Uribe, and critical revisions were performed by Jesús Almendral, Rafael Correa‐Velásquez, and Teresa Barrio‐López. All authors reviewed and approved the final version of the manuscript.
Funding
The authors have nothing to report.
Ethics Statement
All procedures performed in this study were in accordance with the ethical standards of the institutional research committee and with the 1964 Declaration of Helsinki and its later amendments or comparable ethical standards. The study protocol was reviewed and approved by the institutional ethics committee of CardioVID Clinic.
Consent
Informed consent was obtained from all individual participants included in this study.
Conflicts of Interest
Juan F. Agudelo‐Uribe reports receiving modest honoraria from Abbott Medical Colombia for proctoring electrophysiology procedures, lectures, and scientific presentations. Juan D. Ramírez‐Barrera, Rafael Correa‐Velásquez, Teresa Barrio‐López and Jesús Almendral declares no competing interests. Margarita Londoño‐Arango is an employee of Abbott Medical Colombia.
Supporting information
Table S1: Possible combinations for fractionation mapping characteristics.
Table S2: raw point‐by‐point diagnostic performance of fractionation mapping configurations in the AF derivation and exploratory CNA validation cohorts: exact p values and Bonferroni‐corrected analysis.
Table S3: Patient‐level bootstrap sensitivity differences and Paired McNemar comparisons between selected fractionation mapping configurations.
Table S4: Spatial distribution of ablation sites exhibiting autonomic responses in primary AF cohort.
Table S5: Mean sensitivity and specificity according to programmable fractionation mapping parameter subsets in the primary AF cohort.
Table S6: Comparison between the present study and the study by Lellouche et al. [6] evaluating electrogram fractionation associated with autonomic responses.
Data Availability Statement
The datasets generated and/or analyzed during the current study are available from the corresponding author on reasonable request.
References
- 1. Armour J. A., Murphy D. A., Yuan B. X., Macdonald S., and Hopkins D. A., “Gross and Microscopic Anatomy of the Human Intrinsic Cardiac Nervous System,” Anatomical Record 247 (1997): 289–298, 10.1002/(SICI)1097-0185(199702)247:2<289::AID-AR15>3.0.CO;2-L. [DOI] [PubMed] [Google Scholar]
- 2. Randall W. C., Milosavljevic M., Wurster R. D., Geis G. S., and Ardell J. L., “Selective Vagal Innervation of the Heart,” Annals of Clinical and Laboratory Science 16 (1986): 198–208. [PubMed] [Google Scholar]
- 3. Scanavacca M., Hachul D., Pisani C., and Sosa E., “Selective Vagal Denervation of the Sinus and Atrioventricular Nodes, Guided by Vagal Reflexes Induced by High Frequency Stimulation, to Treat Refractory Neurally Mediated Syncope,” Journal of Cardiovascular Electrophysiology 20 (2009): 558–563, 10.1111/j.1540-8167.2008.01385.x. [DOI] [PubMed] [Google Scholar]
- 4. Sun W., Zheng L., Qiao Y., et al., “Catheter Ablation as a Treatment for Vasovagal Syncope: Long‐Term Outcome of Endocardial Autonomic Modification of the Left Atrium,” Journal of the American Heart Association 5 (2016): 5, 10.1161/JAHA.116.003471. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5. Pachon M. J. C., Pachon M. E. I., Pachon M. J. C., et al., ““Cardioneuroablation”—New Treatment for Neurocardiogenic Syncope, Functional AV Block and Sinus Dysfunction Using Catheter RF‐Ablation,” Europace 7 (2005): 1–13, 10.1016/j.eupc.2004.10.003. [DOI] [PubMed] [Google Scholar]
- 6. Lellouche N., Buch E., Celigoj A., et al., “Functional Characterization of Atrial Electrograms in Sinus Rhythm Delineates Sites of Parasympathetic Innervation in Patients With Paroxysmal Atrial Fibrillation,” Journal of the American College of Cardiology 50 (2007): 1324–1331, 10.1016/j.jacc.2007.03.069. [DOI] [PubMed] [Google Scholar]
- 7. Aksu T. and Guler T. E., “Electroanatomical Mapping‐Guided Ablation During Atrial Fibrillation: A Novel Usage of Fractionation Mapping in a Case With Sinus Bradycardia and Paroxysmal Atrial Fibrillation,” Journal of Interventional Cardiac Electrophysiology 57 (2020): 331–332, 10.1007/s10840-019-00633-9. [DOI] [PubMed] [Google Scholar]
- 8. Aksu T., Yalin K., and Gopinathannair R., “Fractionation Mapping Software to Map Ganglionated Plexus Sites During Sinus Rhythm,” Journal of Cardiovascular Electrophysiology 31 (2020): 3326–3329, 10.1111/jce.14753. [DOI] [PubMed] [Google Scholar]
- 9. Aksu T., Brignole M., Calo L., et al., “Cardioneuroablation for the Treatment of Reflex Syncope and Functional Bradyarrhythmias: A Scientific Statement of the European Heart Rhythm Association (EHRA) of the ESC, the Heart Rhythm Society (HRS), the Asia Pacific Heart Rhythm Society (APHRS) and the Latin American Heart Rhythm Society (LAHRS),” Europace 26 (2024): euae206, 10.1093/europace/euae206. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10. Aksu T., Gupta D., D'Avila A., and Morillo C. A., “Cardioneuroablation for Vasovagal Syncope and Atrioventricular Block: A Step‐by‐Step Guide,” Journal of Cardiovascular Electrophysiology 33 (2022): 2205–2212, 10.1111/jce.15480. [DOI] [PubMed] [Google Scholar]
- 11. Nunan D., Sandercock G. R. H., and Brodie D. A., “A Quantitative Systematic Review of Normal Values for Short‐Term Heart Rate Variability in Healthy Adults,” Pacing and Clinical Electrophysiology 33 (2010): 1407–1417, 10.1111/j.1540-8159.2010.02841.x. [DOI] [PubMed] [Google Scholar]
- 12. Chiou C. W., Eble J. N., and Zipes D. P., “Efferent Vagal Innervation of the Canine Atria and Sinus and Atrioventricular Nodes: The Third Fat Pad,” Circulation 95 (1997): 2573–2584, 10.1161/01.CIR.95.11.2573. [DOI] [PubMed] [Google Scholar]
- 13. Hu F., Zheng L., Liang E., et al., “Right Anterior Ganglionated Plexus: The Primary Target of Cardioneuroablation?,” Heart Rhythm 16 (2019): 1545–1551, 10.1016/j.hrthm.2019.07.018. [DOI] [PubMed] [Google Scholar]
- 14. Barrio‐Lopez M. T., Álvarez‐Ortega C., Minguito‐Carazo C., et al., “Predictors of Clinical Success of Cardioneuroablation in Patients With Syncope,” JACC: Clinical Electrophysiology 10 (2024): 2711–2724, 10.1016/j.jacep.2024.07.027. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Table S1: Possible combinations for fractionation mapping characteristics.
Table S2: raw point‐by‐point diagnostic performance of fractionation mapping configurations in the AF derivation and exploratory CNA validation cohorts: exact p values and Bonferroni‐corrected analysis.
Table S3: Patient‐level bootstrap sensitivity differences and Paired McNemar comparisons between selected fractionation mapping configurations.
Table S4: Spatial distribution of ablation sites exhibiting autonomic responses in primary AF cohort.
Table S5: Mean sensitivity and specificity according to programmable fractionation mapping parameter subsets in the primary AF cohort.
Table S6: Comparison between the present study and the study by Lellouche et al. [6] evaluating electrogram fractionation associated with autonomic responses.
Data Availability Statement
The datasets generated and/or analyzed during the current study are available from the corresponding author on reasonable request.