A total of 82 patients were enrolled in the study between April 2017 and August 2022. The NPAF group included 40 patients (24 with persistent AF and 16 with long-standing persistent AF) who underwent initial radiofrequency catheter ablation (RFCA). The control group consisted of 42 patients with structurally normal hearts and no previous history of AF who underwent left-sided catheter ablation procedures [N = 35 left-sided accessory pathway, N = 7 idiopathic ventricular premature contractions (VPC)]. NPAF was defined as AF episodes lasting ≥7 days. If sinus rhythm (SR) could not be restored by cardioversion prior to electroanatomic mapping, those NPAF patients were excluded from the study. All antiarrhythmic medications were discontinued for a minimum of 5 half-lives, and amiodarone was ceased for a duration of 2 months prior to catheter ablation. Patient demographics, clinical characteristics, and medications details were obtained from electronic medical records.

This study was approved by the ethics committee of the Third People's Hospital of Chengdu and conducted in compliance with the ethical principles of the Declaration of Helsinki. All patients provided written informed consent.

The aim of the LGE-MRI study was to quantify the extent of fibrosis in the LA and LAA. Patients with NPAF underwent LGE-MRI of the LA and LAA using a 3-Tesla scanner (Siemens Skyra, Germany) and a 32-channel cardiac coil within 1 week before the electrophysiological assessment and/or ablation procedure. All patients were in atrial fibrillation (AF) during scanning.

The specific scanning protocol employed a 3D inversion recovery gradient-echo sequence, combined with respiratory navigation and electrocardiogram gating, and was performed 20 min after intravenous administration of gadolinium contrast agent (Gadobutrol, 0.2 mmol/kg). Scanning parameters were as follows: field of view, 350–390 mm; echo time (TE), 1.3–1.6 ms; repetition time (TR), 700–870 ms; inversion time, 310–350 ms; in-plane resolution, 1.37 × 1.37 mm; slice thickness, 1.5 mm; and voxel size, 1.25 × 1.25 × 2.5 mm (reconstructed dimensions: 0.625 × 0.625 × 1.25 mm). Additional typical acquisition parameters included: TR/TE, 5.5/3.0 ms; flip angle, 25°; in-plane resolution, 1.25 × 1.25 mm; and slice thickness, 2.5 mm (reconstructed dimensions: 0.625 × 0.625 × 1.25 mm).

For image post-processing, LGE CMR images were analyzed and segmented using MASS imaging software (Medis Medical Imaging Systems, Leiden, Netherlands). First, raw LGE CMR data were imported into the MASS software. The endocardial and epicardial contours of the LA were then manually delineated on each of the 96 cardiac MRI slices. Based on these segmentation results, the software automatically generated a 3D model of the LA. Myocardial tissue was classified by signal intensity ratio (SIR) as follows (15): healthy atrial myocardium (SIR < 1.2), interstitial fibrosis (SIR: 1.2–1.32), and dense scarring (SIR > 1.32).

In patients with NPAF, pre-procedure computerized tomography angiography images of the LA were obtained, and transesophageal echocardiography was performed within 48 h prior to the study to rule out left atrial thrombus. Catheter ablation procedures were performed under general anesthesia and muscular paralysis with hemodynamic and electrocardiographic monitoring. A decapolar coronary sinus catheter (2-5-2 mm interelectrode spacing, P-curve, Biosense Webster) was positioned via femoral venous approach under fluoroscopic guidance. Intravenous heparin was administered to maintain an activated clotting time of 300–350 s. Subsequently, LA access was obtained by transseptal puncture, through which a 20-polar mapping catheter (2 mm center-to-center spacing, Pentaray, Biosense Webster) was advanced to create the electroanatomic map. After mapping, a 3.5 mm tip irrigated ablation catheter (1-6-2 mm interelectrode spacing, ThermoCool SmartTouch, Biosense Webster) was advanced to perform ablation in the NPAF study group.

Intracardiac electrograms (EGMs) were recorded using an electrophysiological recording system (WorkMate Claris System, One St. Jude Medical, St. Paul, MN, USA) coupled with a three-dimensional electrophysiological navigating system (CARTO 3 system, Biosense Webster, Diamond Bar, CA, USA). EGM recording was filtered at 30–500 Hz. A detailed fast anatomic map (FAM) was performed with a distribution of points using a fill threshold of 15 mm and an internal point filter was set to limit electric potential data acquisition to within 5 mm from the geometrical surface. The respiration gating method, which ensured the acquisition of anatomic points limited to the end period of expiration during mechanical ventilation, was adopted to minimize the breath related mapping shifts. The CONFIDENSE module (Biosense Webster) was used for tissue proximity detection and to assess adequate tissue-catheter contact of each acquired point during FAM. All mappings were performed by a single, highly experienced electrophysiologist, who implemented rigorous quality control both during real-time data acquisition and offline post-processing.

The following steps were performed to assess bipolar voltage in AF and SR in the NPAF group: 1.

In patients presenting in AF, cardioversion was performed to obtain the first electroanatomic map in SR.

In the control group, ablation of accessory pathway or VPC was performed after electroanatomic mapping of the LA and LAA in SR using the same technique as above for NPAF in SR.

The bipolar voltage of dense scar was defined using the cut off value of 0.1 mV. Conventionally, any point with a bipolar voltage ≤ 0.5 mV during SR was classified as a low voltage point. LVA were defined as sites of ≥3 adjacent low voltage points, which were <5 mm apart. For the individualized LVA definition: sites with bipolar voltage ≤ the normalized LVA detection threshold were designated as abnormal substrate. The borders of LVAs were automatically reconstructed in 3D using the CARTO® VOLTAGE VISITAG™ module.

Categorical data were presented as counts and percentages (%) and the comparison between two groups was assessed using the appropriate chi-square test. Continuous data were depicted as either the mean with standard deviation or the median with interquartile range (IQR), and the comparison between two groups was evaluated using the t-test or Mann–Whitney U-test, respectively. Linear regression analysis was used to evaluate the relationships in LA and LAA voltages. The relationships between LA and LAA voltages with SR and AF were assessed using Spearman correlation coefficients. A two-tailed p-value less than 0.05 was considered statistically significant. All statistical analyses were performed with either SPSS statistics version 26.0 (IBM Corporation, Chicago, IL, USA) or GraphPad Prism software version 9.0 (GraphPad Software, CA, USA).

The baseline characteristics of all study participants are presented in Table 1. Compared to patients without AF, those with NPAF had larger LA, were older, male, and exhibited a higher prevalence of hypertension and coronary heart disease.

During SR, a mean of 422 ± 185 and 498 ± 372 bipolar voltage points per patient were analyzed in the LA for the control and NPAF group, respectively; An average of 276 ± 139 and 167 ± 67 bipolar voltage points per patient were analyzed in the LAA for the two groups, respectively.

Compared to patients without NPAF, the 5th, 25th, 50th, 75th, and 95th percentiles of bipolar voltage in the LA were significantly lower in patients with NPAF (Figure 1A). The bipolar voltage in the LAA was significantly higher than the LA in patients with NPAF, but not in the control group (Figure 1A). The median bipolar voltage in the LA was significantly lower in patients with NPAF when compared to the control group [2.970 (1.599) vs. 0.870 (0.560), P < 0.001; Figure 1B]. However, there was no significant difference in LAA bipolar voltage between the two groups [3.303 (1.796) vs. 3.100 (1.045), P > 0.05; Figure 1B]. The 5th and 95th percentiles of bipolar voltage in the LA were significantly lower in patients with NPAF compared to the control group [1.103 (0.529) vs. 0.085 (0.096), P < 0.001, Figure 1C; 7.937 (3.282) vs. 3.232 (2.087), P < 0.001, Figure 1D]. No significant difference was observed in the 95th percentile of bipolar voltage in the LAA (V_LAA95%) between the two groups [8.089 (3.571) vs. 7.604 (3.404), P > 0.05; Figure 1D]. The V_LAA95% in patients without NPAF conformed to a normal distribution. We used the V_LAA95% as a benchmark given the relative similarity in patients with and without AF.

In previous studies, in control patients without AF, the 5th percentile of all LA mapping points (V_LA5%) was designated as a bipolar voltage threshold indicating abnormal atrial substrate (17, 18). Thus, a point with a voltage lower than V_LA5% was defined as low voltage. In the present study, a strong linear correlation between V_LA5% and V_LAA95% (P < 0.0001, V_LA5% = 0.1324 × V_LAA95%, Figure 2A) was observed in the control group. The ratio of each point's voltage to V_LAA95% in each patient was termed the relative voltage index (RVI). When adjusting each point's voltage in the LAA and LA by V_LAaA95% and expressing them as RVI, the scattergrams revealed a notably stronger linear correlation of the electrogram amplitudes between SR and AF (Figures 2B,D). However, when directly using the absolute voltage of each point (AVI) in the LAA and LA, the scattergrams displayed a significant but weaker linear correlation of the bipolar voltage between SR and AF (Figures 2C,E).

Paired t-tests revealed that the absolute voltage index (AVI) values of the LA and LAA exhibited statistically significant differences in the same patient between SR and atrial AF states. In contrast, the RVI showed no statistically significant differences between SR and AF states in either the LA or LAA regions (Figures 3A,B). These results suggest that the correlation between V_LAA95% and V_LA5% established in control patients could applicability to patients with NPAF. These results suggest that the correlation between V_LAA95% and V_LA5% established in control patients could be applicable to patients with NPAF.

In the control group, the RVI of V_LA5% to V_LAA95% approaches a constant as indicated by the regression equation mentioned above. This constant was identified as the standard relative voltage index (SRVI) in this study, serving as the threshold value for the RVI criterion in defining LVA. The V_LA5% of normal LA is defined as the low voltage threshold value, and any point with a voltage lower than “normal” V_LA5% should be classified as a low voltage point for that patient. However, the “normal” V_LA5% in a patient with NPAF is unavailable due to diseased myocardium. Herein, the “normal” V_LA5% for each NPAF patient can be calculated with the equation: V_LA5% = 0.1324 × V_LAA95%.

Following electroanatomic mapping of the LA and LAA, the V_LAA95% is determined. Utilizing the formula V_LA5% = 0.1324 × V_LAA95%, the patient-specific V_LA5% is calculated, which functions as the normalized threshold for identifying LVA within the LA in patients with NPAF. In the cohort of 40 NPAF patients, the SRVI-guided substrate mapping can be visually appreciated as a better match to the fibrotic regions of LGE in both rhythms as compared to the conventional voltage threshold for LVA (Table 2). Two retrospective cases of NPAF were presented to compare the distributions of LVA defined by SRVI and conventional threshold of 0.5 mV with fibrosis regions of LGE-MRI in both AF and SR (Figure 4). The V_LA5% for Case 1 was 0.85 mV in AF and 1.0 mV in SR (Figures 4A,B). For Case 2, the V_LA5% was 0.28 mV in AF and 0.32 mV in SR (Figures 4C,D). In comparison to the conventional LVA threshold of 0.5 mV, LVA in both the anterior and posterior walls of the LA in NPAF patients during AF, as defined by the normalized criteria-guided substrate mapping, exhibited a stronger correlation with the fibrotic regions of LGE.

A low voltage index was employed to define the heterogeneity of the bipolar voltage amplitude by computing the coefficient of variance (standard deviation/mean value) of the voltage of all points (18). The regional and overall differences in the low voltage index based on SRVI and conventional LVA detection threshold of 0.5 mV in both rhythms are summarized in Table 3. During both rhythms, the low voltage index was significantly higher in AF than SR across all LA segments, regardless of the threshold settings. The correlation between the low voltage index in both rhythms, as defined by the normalized RVI-guided substrate mapping, was superior to the conventional LVA detection threshold value of 0.5 mV.

The origin of low amplitude atrial electrograms can be multifactorial but atrial fibrosis is considered a primary structural cause (19). Bipolar voltage mapping is used as a surrogate for identification of histopathological LA fibrosis despite inherent limitations (8–10, 13, 20). Studies have demonstrated a strong correlation between bipolar voltage mapping and LGE-MRI, which is considered the gold standard for identification of LA fibrosis (5, 6). However, there is no unified criteria for detecting LVA in patients with NPAF across different rhythms and with different recording electrodes. Voltage is influenced by the extent of LA fibrosis which relates to clinical factors such as age, LA size, comorbidities, and duration of AF (18, 21, 22). Voltage is also affected by the rhythm mapped, regional wall thickness, and the duration of acquisition (21–25). Using a universal absolute voltage cut-off to define LVA in NPAF disregards the variable factors that may influence bipolar voltage in each individual case. Our study refines the definition of LVA in NPAF to account for a population with heterogeneous fibrotic burden and distribution.

Voltage thresholds to identify low voltage have been classified based on the 95th percentile of voltage in patients with and without AF (17, 18, 26). Multiple thresholds have been reported for identifying LVA across different phenotypes of AF (PAF and NPAF) during both AF and SR. A universal constant, such as 0.5 mV and 0.4 mV, has been established as the absolute threshold for identifying LVA in AF patients (18, 27–29). Fibrotic burden by MRI remains the gold standard for identification of substrate. Low voltage thresholds have limited specificity and sensitivity for detection of fibrosis by MRI but appear to be more predictive when mapping in AF as compared to SR (25). Using the LAA voltage as a reference, the equation for determining a normalized LVA threshold value in the LA for NPAF is highlighted in the current study (V_LA5% = 0.1324 × V_LAA95%). We have validated this voltage index with gold standard MRI in our study and provided a more accurate representation of histopathological fibrosis during electroanatomic mapping.

The observed voltage discrepancy between the LAA as compared to the LA in NPAF is consistent with the study by Yagishita et al. noting that the highest voltages were observed in the LAA in patients with PAF and NPAF in SR and AF (24). Our study expands the analysis to patients with normal LA (control group—no PAF or NPAF). We found no significant difference in the distribution of LAA voltage between the control group and NPAF group during SR, especially for the V_LAA95%. Notably, we extended this observation by including a control cohort (patients with normal LA, no PAF/NPAF) and demonstrating selective preservation of LAA voltage in NPAF: while all LA voltage percentiles (5th, 25th, 50th, 75th, 95th) were significantly lower in NPAF vs. controls, LAA voltage—including V_LAA95%—has no significant difference between groups (median LAA voltage: 3.303 [1.796] vs. 3.100 [1.045], P > 0.05; V_LAA95%: 8.089 [3.571] vs. 7.604 [3.404], P > 0.05) during SR. This stability indicates the electrophysiological characteristics of LAA is spared from the structural remodeling that drives LA voltage attenuation in NPAF.

Consistent with prior studies in NPAF, which have demonstrated a lower LGE burden in the LAA compared with the LA, this reduced fibrotic involvement may explain the LAA's electrophysiological stability (18, 25). Factors that may make the LAA less susceptible to fibrosis include a lower proportion of interstitial cells, robust arterial supply, and enhanced wall thickness (30, 31). Intuitively, the LAA voltage (V_LAA95%) would be the most appropriate benchmark to establish a standard RVI for detection of LVA in each case of NPAF.

Studies exploring the relationship between bipolar voltage in SR and AF have had mixed results. Linear correlations have been shown but the influence of other electrophysiological factors such as electrogram fractionation may play a significant role in voltage differences between SR and AF (22, 24). Moisés et al. suggested that bipolar voltage thresholds for LVA should be adjusted depending on the rhythm and found that employing a modified lower detection threshold for LVA during mapping might produce valid and reliable results for identifying regions with a voltage of <0.5 mV in SR (23). There is a paucity of data on the ideal voltage threshold for identifying LVA in AF, however. Differences due to wavefront collision, spiral waves, and critical repolarization/depolarization mass may result in significant voltage threshold changes in SR vs. AF. In one study, the LVA threshold during AF was half of the standard threshold used in SR (23). Variability in AF cycle length, location of recording, and substrate heterogeneity may limit the reproducibility of LVA thresholds when used during AF. The personalized RVI that we have proposed in this study simplifies the assessment of LVA during electroanatomic mapping. Our findings suggest that the RVI can be used for LVA assessment in a rhythm-independent fashion.

A simplified and tailored approach to LVA threshold assessment with the RVI enhances the identification of atrial substrate in each individual patient without the influence of the type of rhythm. This may lead to improved efficiency of mapping and a better understanding of the underlying pathology for prognostication, especially if advanced cardiac imaging (LGE-MRI) is unavailable or undesired. Whether the use of RVI can more accurately identify and improve outcomes after voltage-based extra-pulmonary vein substrate modification during AF catheter ablation is unknown and deserves further investigation.

There were several limitations in our study. Patients with AF at the beginning of the case were cardioverted to SR for the first mapping study. This may not be achievable in all cases of NPAF. Although our LA mapping density meets current high-density standards, we acknowledge it may be relatively low depending on the contemporary mapping system used; future studies employing ultra-high-density mapping could potentially achieve finer substrate characterization. Voltage measurements in SR may be underestimated due stunned myocardium after cardioversion. Apart from comparing bipolar voltage, we did not analyze other electrogram characteristics, such as unipolar voltage, cycle length, fractionated electrograms, and rotational activity across varied voltage settings and rhythms. We also did not explore the influence of additional wavefronts during sinus rhythm on voltage amplitudes. Additionally, prior studies have shown that patients with left-sided accessory pathways [which constitute the majority of Wolff-Parkinson-White (WPW) syndrome cases] exhibit a significantly higher prevalence of AF compared to the general population, with paroxysmal AF occurring in up to one-third of WPW patients (32, 33). This association indicates that left-sided accessory pathways are linked to an atrial substrate with heightened AF susceptibility and the selection of such patients as the control group may introduce potential confounding.