Drivers in AF colocate to sites of electrogram organization and rapidity: Potential synergy between spectral analysis and STAR mapping approaches in prioritizing drivers for ablation

Stochastic trajectory analysis of ranked signals (STAR) mapping has recently been used to ablate persistent atrial fibrillation (AF) with high rates of AF termination and long‐term freedom from AF in small, single‐arm studies. We hypothesized that rapidity and organization markers would correlate with early sites of activation (ESA).


| INTRODUCTION
Evidence is conflicting as to whether electrogram characteristics are useful as surrogate markers for localized drivers in persistent atrial fibrillation (AF) in humans. [1][2][3][4][5][6] Whilst markers of organization have better identified sites that play a mechanistic role in AF, 1,7 markers of rapidity have been less reliable. 1,2 Optical mapping animal studies have shown that AF is maintained by sites demonstrating fastest cycle lengths (CL) and highest dominant frequency (DF) [8][9][10] ; however, in humans, this has proven to be a poor predictor of localized driver sites. 1 The poor correlation may be because of the lack of spatiotemporal stability of drivers in AF, which may explain the apparent inconsistency of sites of rapidity. 2,6 In addition, optical mapping animal studies have shown intra-atrial high-to low-frequency gradients at rotor sites. 9 A novel mapping strategy called stochastic trajectory analysis of ranked signals (STAR) mapping has recently been used to guide AF ablation, resulting in high rates of AF termination and subsequent freedom from AF. 11 We hypothesized that using a novel methodology for CL analysis, sites demonstrating greater rapidity and organization would colocate to AF driver sites identified using the STAR mapping method. Furthermore, we hypothesized that the driver sites identified using this approach that did demonstrate greater rapidity and organization would be more mechanistically important in AF maintenance, as evidenced by a greater likelihood of AF termination with ablation.

| METHODS
Patients undergoing catheter ablation for persistent AF (<24 months and no previous AF ablation) and that were prospectively included in the STAR mapping study (clinicaltrials.gov NCT02950844) were included (Supporting Information Material). Patients provided informed consent for their study involvement which was approved by the UK National Research Ethics Service (16/LO/1379).

| STAR mapping method
The STAR mapping method has been described in detail previously. 12,13 In brief, the STAR mapping method principle is to use data from multiple individual wavefront trajectories to identify atrial regions that most often precede the activation of neighboring areas.
By gathering data from many thousands of activations, a statistical model can be formed. This permits regions of the atrium to be ranked according to the amount of time those activations precede those of adjacent regions (Supporting Information Material). 12,13 A STAR map consists of color-coded electrode positions projected on to a replica of the patient's atrial geometry created in CARTO. Each color represents the proportion of time the electrode spends leading in relation to the other paired electrodes as highlighted by the color scale on the right-hand side of the STAR maps. An early site of activation (ESA) is defined as an electrode leading ≥75% of the time compared with paired electrodes.

| STAR mapping study
All patients underwent at least two 5-minute unipolar recordings in the left atrium (LA) pre-and post-pulmonary vein isolation (PVI) utilizing a whole-chamber basket catheter (Constellation; Boston Scientific Ltd, Marlborough, MA or FIRMap, Abbott, CA). These recordings were used to create STAR maps. All patients underwent PVI followed by STAR mapping-guided ablation of ESA (Supporting Information Material). 13 A study-defined ESA ablation response included AF termination or CL slowing ≥30 ms.

| Bipolar voltage and ESA (Supporting Information Material)
ESA identified were categorized as existing in either non-low voltage zone (LVZ) or LVZ. The relationship of ESA mapped to LVZs and achieving AF termination on ablation was assessed. The relationship between the proportion of LVZs and the number of ESA identified was also evaluated.

| Spectral analysis and ESA
Utilizing the unipolar signals recorded from the basket catheter, the CL was determined at each electrode pole in contact over each of the 5-minute recordings in each patient. The CL was measured as the time difference between two consecutive atrial signals using a custom-written Matlab script (Matlab 2017b; MathWorks, Natick, MA). Predefined refractory periods of 70 ms [12][13][14] were used to avoid double-counting of fractionated electrograms. Ventricular far-field signals were filtered using a method previously published 11 (Supporting Information Material). Unipolar activation timing was taken as the maximum negative deflection (peak negative dv/dt). A minimum voltage cutoff of 0.1 mV was used. All the CL measurements over the 5-minute recording for each electrode was used to create a histogram of the CLs with all CLs identified on the x-axis (rounded to the nearest whole millisecond) and the percentage of recording made up by each CL on the y-axis were plotted for each individual electrode for each 5-minute recording for each patient.
The histograms were used to determine the dominant CL and minimum-CL (Min-CL) (Supporting Information Material and Figure   S1). The parameters for CL determination was uniform for all the cases. There was no attempt to compensate for spectral leakage or

| Regional CL and frequency gradients at ESA
The Min-CL at ESA was compared with the Min-CL obtained at adjacent electrodes to elicit whether there was a CL gradient from ESA to surrounding areas. This was repeated for all ESA identified on each STAR map in each patient. Neighboring electrodes were defined as electrodes that were within 3 cm of the ESA. This was also repeated using DF measurements to assess for frequency gradients from ESA to surrounding areas (Supporting Information Material). The relationship between LA site(s) with the overall highest DF, as per those values in the top decile, and ESA was also assessed.

| Impact of PVI on spectral analysis data
Offline analysis of STAR maps pre-PVI was performed to determine whether ESA identified post-PVI were present pre-PVI. Using the method described above, the Min-CL and lowest CLV was determined at the ESA using the pre-PVI unipolar recordings. Any impact PVI had on these values were evaluated at ESA that were identified pre-and post-PVI.

| Statistical analysis
Statistical analyses were performed using SPSS (IBM SPSS Statistics, Version 24; IBM Corp, New York, NY). Continuous variables are displayed as mean ± SD or median (IQR). Categorical variables are presented as a number and percentage. The Student t test, or its nonparametric equivalent, the Mann-Whitney U test when appropriate was used for comparison of continuous variables. Fisher's exact test was used for the comparison of nominal variables. Sensitivity and specificity were used to determine the predictive power of the proportion of LVZs, rapidity, and organization markers in identifying ESA and particularly those associated with an ablation response. The odds ratio was also calculated to determine the probability of an ESA being associated with an ablation response depending on whether it colocated or not to spectral analysis markers and LVZs. One-way analysis of variance was used to compare the Min-CL and highest DF changes from an ESA to neighboring poles. Results with P < .05 were regarded as significant.

| RESULTS
Thirty-two patients underwent STAR mapping-guided ablation. 13 Baseline characteristics are demonstrated in Table S1. The mean AF duration was 15.4 ± 4.3 months (n = 4 AF duration <12 months and n = 28 AF duration >12 months) and 21 of the 32 (65.6%) patients were on an antiarrhythmic drug preablation. The total procedural radiofrequency (RF) time whilst in AF was 52.1 ± 7.7 minutes. The total RF time to achieve sinus rhythm with ablation (including PV isolation, ESA ablation + AT ablation) was 64.5 ± 9.7 minutes.

| Ablation response at ESA
In brief, 92 ESA were identified in the 32 patients (2.8 ± 0.8 per patient) on the post-PVI STAR maps of which 83 (90.2%, 2.6 ± 0.7 per patient) were targeted with ablation ( Figure 1 and Table S2).
The nine ESA that were not ablated were in patients in whom ablation at a previous site had resulted in AF termination. An ablation response was achieved with 73 ESA (2.3 ± 0.6 per patient) F I G U R E 1 A flow chart of the study that shows the ESA identified and those associated with an ablation response and how many of these colocated to sites of Min-CL, lowest CLV, regional DF gradients, and LVZs. CLV, cycle length variability; DF, dominant frequency; ESA, early sites of activation; LVZ, low-voltage zone which included at least one response in all 32 patients. On a per ESA basis, AF termination was achieved with ablation at 24 sites (18 organization to AT and 6 termination to sinus rhythm) and CL slowing of ≥30 ms was achieved with 49 sites. On a per patient basis, AF termination was achieved in 24 patients (6 sinus rhythm and 18 AT) and CL slowing of ≥30 ms in the remaining 8 patients.

| Bipolar voltage and ESA
On a per patient basis, an average of 2.5 ± 0.6 well-defined LVZs were identified. A majority of the ESA identified were mapped to LVZs (62/92, 67.4%). Out of the 83 ESA that were ablated, 60 were mapped to LVZs (72.3%) of which 59 (98.3%) were associated with an ablation response.
Patients with more than 50% of the LA comprised of LVZs (59.2 ± 6.5 mV) were more likely to have more than two ESA identified. There was a strong positive correlation between the proportion of LVZs present and the number of ESA identified (r s = 0.91; P < .001).
However, LVZ alone was of little predictive value in predicting ESA (Table S2).
The association of ESA with an LVZ was highly predictive of an ablation response (Table 1). ESA mapped to LVZs were also more frequently associated with AF termination on ablation vs ESA mapped to non-LVZs (odds ratio = 20.0, 95% confidence interval [CI], 1.1-351.5; P = .04). and 56 (60.9%) colocated to one of these sites, respectively. These markers were of little predictive value in predicting ESA (Table S3).  3.5.2 | Min-CL sites within the top decile and 5% of the "Absolute Min-CL"

| Spectral analysis and ESA
Of the 73 ESA with an ablation response, 54 (74.0%) colocated to one of the fastest Min-CL sites by either definition (Figures 1 and 2A-F).
The predictive power of these parameters in identifying ESA with an ablation response is demonstrated in Table 1. with an ablation response is demonstrated in Table 1.

| DISCUSSION
This study has demonstrated that sites of rapidity (identified using Min-CL) and organization (identified using CLV) are associated with ESA in terms of colocating. Although this association was unlikely to were also more frequently associated with AF termination on ablation in contrast to those that were not.

| Spectral analysis and ESA
This study has shown that approximately 75% of ESA with an ablation response colocated to sites of lowest CLV or to sites of group has previously shown that AF drivers frequently colocate to sites of lowest CLV. 1,7 Other studies have also shown that AF termination is more common in patients with higher baseline organization index. 15,16 The data are, however, conflicting in regard to markers of rapidity. Some studies have shown that targeting sites of highest DF is associated with electrophysiological endpoints on ablation 5  Defining the fastest Min-CL sites as Min-CLs within 5% of the Absolute Min-CL was associated with greater specificity in identifying a driver site than when using the top decile.
Even though spectral analysis data showed an association with ESA, the strength of that association is unlikely to enable identification of driver sites in itself. However, ESA that colocated with these novel markers of rapidity and organization were much more likely to terminate AF on ablation (odds ratio, 29 for lowest CLV and 34 for fastest Min-CL). This suggests that these markers could potentially be of use to aid in differentiating localized drivers that are critical for the maintenance of AF from those that are less mechanistically important. These markers could, therefore, potentially be useful for stratifying the importance of ESA, potentially allowing a hierarchal approach to ablation of localized drivers.

| Regional CL and DF gradients and ESA
Studies have shown the presence of a left to right atrial gradient in the context of AF 20,21 and that PVI results in a loss of this gradient. 20 However, others have demonstrated an absence of left to right atrial gradients in persistent AF 22 or in the context of drivers outside the PVs. 23 In animal studies, regional DF gradient have been reported at sites of rotor activity; however, the presence of intra-atrial gradients has not been assessed in human studies. This is the first study that has demonstrated the presence of regional CL and DF gradients at sites of localized drivers in AF. Due to the conflicting evidence seen with regard to DF ablation and the poor correlation between sites of highest DF in the atrium and localized drivers, this finding highlights that even though localized drivers do not reliably colocate to the sites of highest DF, as defined using the current methodology to calculate DF, drivers sites do represent sites of highest regional DF. This suggests that rather than comparing DF across the whole atrium to identify sites of highest DF, it may instead be feasible to use DF to look for a regional frequency gradient. Applying this strategy in guiding AF ablation needs to be prospectively evaluated to determine whether it is additive in clinical practice.

| Impact of PVI
The lack of PVI impact on spectral characteristics at driver sites highlights their independence from the PVs. However, as we have shown that drivers are more likely to be identified post-PVI 1,13,14,24 and with studies demonstrating that PVI eliminates wavebreaks and organizes wavefronts, 25 this does clearly still highlight the importance of PVI in the ablation strategy for persistent AF. However, these data suggest that they may be less important for the maintenance of AF than the drivers identified as ESA.

| Bipolar voltage and ESA
This study has shown that ESA were predominantly mapped to LVZs and that the proportion of LVZs strongly correlated to the number of ESA identified. These findings are supported by sites of structural remodeling resulting in alteration in CV dynamics that promote reentrant mechanisms 26,27 and that reentrant activity anchors to sites of structural heterogeneity. 28 The bipolar voltage map can thereby be used by the operator to plan the extent of their ablation.

| Limitations
In this study, ESA was used as a surrogate for localized drivers in AF and the focus was on EP endpoints to determine the mechanistic significance of these sites since there is arguably no other way to verify that a driver has been mapped. Others have used AF termination or CL prolongation as a surrogate for the interruption of mechanisms important for AF maintenance. 16,24,29 Our previous work has shown that targeting these sites prospectively is associated with a high freedom from AF/AT supporting their mechanistic role in AF. 13 A 5-minute recording duration was used for the initial study using STAR mapping and for the spectral analysis. However, our group has subsequently shown that drivers can be identified on STAR maps HONARBAKHSH ET AL.