Comparison of electrogram characteristics in persistent atrial fibrillation

Abstract

Introduction:

Multiple analysis techniques evaluate electrograms during atrial fibrillation (AF), but none have been established to guide catheter ablation. This study compares electrogram properties recorded from multiple right (RA) and left atrial (LA) sites.

Methods:

Multisite LA/RA mapping (281 ± 176/239 ± 166 sites/patient) was performed in 42 patients (30 males, age 63 ± 9 years) undergoing first (n = 32) or redo-AF ablation (n = 10). All electrogram recordings were visually reviewed and artifactual signals were excluded leaving a total of 21 846 for analysis. Electrogram characteristics evaluated were cycle length (CL), amplitude, Shannon’s entropy (ShEn), fractionation interval, dominant frequency, organizational index, and cycle length of most recurrent morphology (CL_R) from morphology recurrence plot analysis.

Results:

Electrogram characteristics were correlated to each other. All pairwise comparisons were significant (p < .001) except for dominant frequency and CL_R (p = .59), and amplitude and dominant frequency (p = .38). Only ShEn and fractionation interval demonstrated a strong negative correlation (r = −.94). All other pairwise comparisons were poor to moderately correlated. The relationships are highly conserved among patients, in the RA versus LA, and in those undergoing initial versus redo ablations. Antiarrhythmic drug therapy did not have a significant effect on electrogram characteristics, except minimum ShEn. Electrogram characteristics associated with ablation outcome were shorter minimum CL_R, lower minimum ShEn, and longer mimimum CL. There was minimal overlap between the top 10 sites identified by one electrogram characteristic and the top 10 sites identified by the other 10 characteristics.

Conclusion:

Multiple techniques can be employed for electrogram analysis in AF. In this analysis of eight different electrogram characteristics, seven were poorly to moderately correlated and do not identify similar locations. Only some characteristics were predictive of ablation outcome. Further studies to consider electrogram properties, perhaps in combination, for categorizing and/or mapping AF are warranted.

CENTRAL ILLUSTRATION 1

A summary of the study workflow, correlation analysis, and outcomes. This figure illustrates the workflow of the study and the main findings. In this study, the left and right atrium were mapped in 32 patients undergoing first catheter ablation procedure for persistent atrial fibrillation (AF) and 10 patients undergoing a second catheter ablation procedure for persistent AF. Mapping was performed with multielectrode catheters and 21 846 recordings were obtained. Eight electrogram characteristics were analyzed: (1) CL = cycle length; (2) CL_R = cycle length of the most recurrent morphology; (3) Rec% = frequency of the most recurrent morphology; (4) DF = dominant frequency; (5) OI = organizational index; (6) FI = fractionation interval; (7) ShEn = Shannon’s entropy; (8) AMP = median amplitude. Correlation analyses were done stratified by atrium and ablation group. Of the eight parameters studied, only Shannon’s entropy and fractionation interval demonstrated a strong negative correlation, while the rest of the pairwise comparisons demonstrated poor to moderate correlation, suggesting they provide independent information about the electrogram in AF. The effect of antiarrhythmic use was explored, and 1-year follow-up outcomes are presented.

1 |. INTRODUCTION

Catheter ablation of cardiac arrhythmias has the greatest success rates when intracardiac mapping can be employed to identify the substrate for the arrhythmia. In addition to using electrograms for mapping of activation time, characteristic morphologic signatures such as accessory pathway potentials, slow pathway potentials, and mid-diastolic potentials are often described and useful for the identification of the arrhythmia substrate. The inability to map atrial fibrillation (AF) in the clinical electrophysiology laboratory using standard activation mapping has generated multiple alternative approaches to mapping and ablation. With regard to ablation, the most widely used approach—pulmonary vein isolation (PVI)—is an empiric, anatomic approach not guided by electrogram mapping which has only a moderate success rate of approximately 50% for persistent AF. Other anatomic approaches, such as linear ablation,¹ targeting MRI fibrosis,² and posterior wall isolation³ have not improved outcomes compared to PVI. A variety of electrogram-guided approaches have been tested including targeting of complex fractionated atrial electrograms (CFAE) and dominant frequency (DF) of the electrogram signal without uniform or reproducible success. Finally, there are other electrogram analysis approaches that provide different ways to categorize the electrogram signal in AF, including Shannon’s entropy (ShEn) and a technique we described for AF electrograms—morphology recurrence plots.^4,5 In this latter technique, the predominant electrogram morphology in the recording is identified and its frequency and cycle length (CL) are measured, with a goal of identifying electrograms with similar/repetitive morphology at a rapid rate that could represent drivers of AF.^4,6–8 It is unknown which of these techniques, in isolation or combined, could provide a useful approach to mapping AF. But given that the electrogram reflects both underlying tissue properties and electrophysiologic characteristics, it seems likely that an electrogram-based approach for mapping and ablation of AF will emerge. The importance of identifying electrogram approaches to improve upon current outcomes of ablation for persistent AF is highlighted by the negative results of the recent DECAAF II² and CAPLA³ studies. Notably, in DECAAF II,² ablation of areas of fibrosis identified by cardiac MRI did not improve outcomes but was also associated with harm.

Electrogram analysis in AF is generally focused on different properties such as amplitude, rate, complexity, and repeatability. There is no a priori reason for these electrogram properties to be correlated unless they are linked by the underlying substrate, for example, if areas of fibrosis characterized by low amplitude serve as an anchor for rotors that are drivers for AF. To date, there has been no formal assessment testing the hypothesis that these electrogram properties are related to each other. Given the multiplicity of electrogram analysis techniques in use for mapping AF, usually using a limited set of characteristics, it is important to assess whether these methods are correlated and whether key features identify similar regions. In this report, we compare electrogram properties recorded from over 20 000 right (RA) and left atrial (LA) sites in patients presenting for catheter ablation for AF. We explore the effect of antiarrhythmic therapy on electrogram properties and assess which electrogram properties are related to ablation outcome. Finally, we also test whether the 10 sites identified by the individual characteristics as possible areas of interest are common among the different electrogram characteristics.

2 |. METHODS

2.1 |. Study population

A total of 42 patients with persistent AF at the time of planned electrophysiologic studies and catheter ablation were prospectively enrolled at the University of Miami, between December 2016 and September 2020, of which 32 (Group 1) underwent first-time catheter ablation and 10 (Group 2) had undergone prior catheter ablation. Persistent AF was defined as AF lasting more than 7 days.⁹ Patients who had undergone more than 1 prior ablation procedure were excluded. Ten patients were on active antiarrhythmic drug therapy at the time of the procedure. All patients provided written informed consent. This study was approved by the Institutional Review Board of the University of Miami.

2.2 |. Mapping and electrogram recordings

Electrophysiologic studies were performed using standard techniques for transvenous access for RA mapping and transseptal access for LA mapping. Multielectrode catheters (Reflexion [2 mm interelectrode spacing] n = 25, High Density grid [3 mm interelectrode spacing] n = 5—Abbott; LASSO [2 mm interelectrode spacing] n = 1, PentaRay [4 mm interelectrode spacing] n =11—Biosense Webster) were used for mapping and sequentially recording bipolar intracardiac electrograms from multiple sites in the LA (281 ± 176 sites per patient) and RA (239 ± 166 sites per patient) before catheter ablation. In Group 2, the PentaRay was used in all 10 patients, while in Group 1, the catheter used was based on operator preference. Bipolar electrogram recordings were obtained for 15 s at each site at a sample rate of 1200 Hz using commercially available amplifiers (g.tec). Additionally, a 30 Hz highpass filter and notch filter were applied before recordings were stored and analyzed offline using customized software.^4,10 Custom software written in C# (Microsoft) was used for all aspects of signal processing and off-line analysis of electrogram recordings. All electrogram recordings were visually reviewed and signals with substantial artifacts, predominantly ventricular activations, or no electrograms (electrodes not in contact with myocardium or in areas of scar) were excluded leaving a total of 21 846 for analysis.

2.3 |. Electrogram analyses

Figure 1 shows sample electrogram recordings. The following parameters were calculated for each 15 s recording:

FIGURE 1.

Sample (15 s) electrogram recordings from two sites from a single patient. The pulmonary vein (PV) recording demonstrates a highly repetitive electrogram morphology—the adjacent morphology recurrence plot is a color-coded cross-correlation matrix of all activations with the red indicating cross-correlation values near 1 (see color scale). In contrast, the posterior wall recording shows more variable electrogram morphology, quantified by the lower recurrence percentage (Rec%). The results of all the other electrogram characteristics are shown.

2.3.1 |. Median amplitude (AMP)

For each 15 s recording, a previously validated algorithm was used to identify each unique electrogram.¹⁰ The peak-to-peak amplitude was measured for each 100 ms window centered at the detected electrogram. The AMP was defined as the median of all the peak-to-peak amplitudes of the 15 s recording.

2.3.2 |. ShEn

ShEn is a statistical measure of the complexity of a signal¹¹ and has been previously applied to characterize AF electrograms.^12,13 It is calculated by

$$H(X)=\frac{-{\sum }_{i=1}^{n}p(\text{xi})\text{log}10(p(\text{xi}))}{\text{log}10(n)},$$

where X is the signal, n is the number of possible amplitude values, and p(x_i) is the probability of any sample of the signal to have a particular amplitude value. Entropy increases as the signal spends more time away from the baseline (thus, a discrete atrial electrogram in sinus rhythm would have very low ShEn as the signal is mostly at baseline).

2.3.3 |. Fractionation interval

Fractionation interval was defined as the mean interval between all detected peaks of the 15 s recording. Peaks below a predetermined threshold were excluded. In this study, one-half of the standard deviation of the 15 s voltage readings was used as the threshold. Peaks that were within 30 ms of a peak of higher amplitude were also excluded. A fractionation interval <100 ms has been previously used to define CFAE.

2.3.4 |. DF and organizational index

To perform power spectral analysis, each 15 s recording was divided into 15 1-s segments. Each segment was bandpass filtered with cutoff frequencies of 40 and 250 Hz using a third-order Butterworth filter. The resulting signals were then rectified and then lowpass filtered at 20 Hz with a third-order Butterworth filter. A Hanning window was applied before performing a Fast Fourier Transform to obtain a power spectrum. The DF was defined at the frequency containing the greatest power in the power spectrum. DF is thought to provide an estimation of activation rate in Hz or activations per second. To calculate organizational index (OI), the total power contained within 1 Hz bands centered at the DF and the next four harmonics of the DF was calculated and divided by the total power of the spectrum above 3 Hz. OI is a measure of temporal regularity with a value closer to 1 suggesting a signal with high regularity. The mean DF and OI of the 15 segments was calculated for each signal.

2.3.5 |. Recurrence percent and recurrence CL

These were calculated as previously described.⁴ Briefly, for each 15 s recording, a previously validated algorithm was used to identify each unique electrogram.¹⁰ A 100 ms window for each detected activation was cross-correlated with every other activation in the recording, allowing for the identification of the most recurrent electrogram morphology. The frequency (Rec%) of the most recurrent morphology was calculated. The CL of just the most recurrent morphology (CL_R = CL/Rec%) which is the mean interactivation interval for only the most recurrent morphology was calculated. These parameters define how repeatable the electrogram morphology is and how fast the most repeatable morphology is being activated.

2.4 |. Ablation and follow-up

As a uniform ablation strategy (PVI) was applied only in Group 1 patients, we report follow-up for these patients and the associated electrogram properties recorded from mapping in the pulmonary veins. There were four patients who also underwent posterior wall isolation. As PVI was the predominant treatment modality, we evaluated the predictive value of the electrogram analysis from the pulmonary veins to predict freedom from recurrence in “PVI responders” and “PVI nonresponders.” Patients were followed for recurrent AF for at least 1 year. All patients had at least one ambulatory monitor of at least 2 weeks duration. Recurrence of persistent AF, paroxysmal AF, and atrial flutter was documented.

2.5 |. Data analysis

Patient characteristics were tabulated as mean ± standard deviation for continuous variables and counts with percentages for categorical variables. As each patient had several hundred mapped data points, correlation between pairs of electrogram characteristics (CL, CL_R, Rec %, DF, OI, FI, ShEn, and AMP) were calculated using Pearson correlation coefficient for each patient individually. These individual patient correlations were summarized by mean, standard deviation, and coefficient of variation. Similar methods were used to compare electrogram properties in the RA versus LA, but correlations were subset by the LA and RA for each patient and then summarized by atrium. To compare the correlations among those patients receiving a first versus second ablation procedure, individual patient correlations were summarized by group (first or second ablation). p Values were combined using Fisher’s method. Fisher’s Z-Test was used as a significance test for the difference between two correlation coefficients. As there were 28 electrogram characteristic comparisons, a Bonferroni-adjusted p < .0018 was used to determine significance.

To further examine the correlations among pairs of electrogram characteristics in the whole cohort, partial Pearson correlations were calculated and adjusted for the subjects, RA versus LA, and first versus second ablation. Pearson correlations were defined as low (|<0.5|), moderate (|0.5–0.8|), and high (|>0.8|).

To explore the differences in electrogram characteristics between patients on versus off antiarrhythmic therapy as well between patients who responded versus those who did not respond to PVI (Group 1 only), Student’s t-test was used for variables with normal distribution while Wilcox test was utilized for variables not normally distributed. Shapiro–Wilk test was used to check for normal distribution. Receiver operating characteristic (ROC) curves were constructed for the shortest CL_R to find the optimal cutpoint by evaluating Youden Index. A logistic regression model utilized the binary classification of the shortest CL_R to predict AF recurrence within 1 year after catheter ablation. ROC curves and Area Under the Curve (AUC) were calculated to assess the model.

3 |. RESULTS

The Central Illustration 1 provides a snapshot of the workflow of the study, and the main findings of correlation analysis and outcomes.

3.1 |. Patient characteristics

As previously reported in this cohort,⁸ there were 30 males (71%) with mean age 63 ± 9 years. CHA2DS2-VASc score was 2.4 ± 1.5. Left ventricular ejection fraction was 48 ± 12% and hypertension was present in 67%, diabetes mellitus in 17%, prior stroke in 7%, and coronary artery disease in 19%. Body mass index was 28.7 ± 8.0 kg/m². Antiarrhythmic drugs actively taken at the time of the procedure in Group 1 patients were Amiodarone (n = 3), Sotalol (n = 3), Flecainide (n = 1), and Dronedarone (n = 1). In Group 2, two patients were taking Amiodarone.

3.2 |. RA versus LA

Figure 2 shows the correlation coefficients for all the comparisons among the electrogram characteristics in the RA and in the LA. Qualitatively, the correlation coefficients were similar for the LA versus RA, though there were statistically significant (p < .001) differences between LA and RA for all comparisons except for CL_R and CL, CL_R and DF, CL_R and OI, CL_R and FI, CL_R and AMP, Rec% and DF, OI and FI, ShEn and DF, AMP and Rec%, AMP and OI, AMP and FI, as well as AMP and ShEn. In only one case was there a directional change in the correlation coefficient—for AMP versus DF, where the strength of correlation was very weak (−0.05 in the LA and 0.01 in the RA). The color coding for the strength of the correlations in the RA versus LA was qualitatively similar for all comparisons.

FIGURE 2.

Correlation coefficients for all the comparisons among electrogram characteristics stratified by the left versus right atrium. Summary of individual patient correlation coefficients for all the comparisons among electrogram characteristics stratified by the left versus right atrium. Values are average Pearson correlation coefficient (SD). All correlations are significant (p < .001). AMP, median amplitude; CL, cycle length; CL_R, cycle length of the most recurrent morphology; DF, dominant frequency; FI, fractionation interval; OI, organizational index; Rec%, frequency of the most recurrent morphology; ShEn, Shannon’s entropy.

3.3 |. First versus second ablation procedure

Figure 3 shows the correlation coefficients for all the comparisons among the electrogram characteristics for patients undergoing an initial versus repeat ablation procedure. Qualitatively, the correlation coefficients were similar for initial versus repeat ablation procedure, though there were statistically significant (p < .001) differences for CL and OI, CL ShEn, CL_R and Rec %, DF and OI, OI and AMP, and FI and ShEn. The color coding for the strength of the correlations in those undergoing first versus second ablation was qualitatively similar for all comparisons.

FIGURE 3.

Correlation coefficients for all the comparisons among electrogram characteristics stratified by first versus second ablation. Summary of individual patient correlation coefficients for all the comparisons among electrogram characteristics stratified by first versus second ablation. Values are average Pearson correlation coefficient (SD). All correlations are significant (p < .001). AMP, median amplitude; CL, cycle length; CL_R, cycle length of the most recurrent morphology; DF, dominant frequency; FI, fractionation interval; OI, organizational index; Rec%, frequency of the most recurrent morphology; ShEn, Shannon’s entropy.

3.4 |. Patient correlation summary accounting for individual patients, LA versus RA, and first versus second procedure

Figure 4A shows the mean ± standard deviation for the individual correlation coefficients for all the comparisons among electrogram characteristics for all the recordings from each individual patient. FI and ShEn had a very high negative correlation (−0.94). Moderate negative correlations were noted among CL and DF (−0.63), ShEn and Rec% (−0.62), CL_R and Rec% (−0.58), ShEn and AMP (−0.52), and DF and OI (−0.50), while moderate positive correlations were noted among FI and Rec% (0.62) and FI and AMP (0.52). The other 20 comparisons, while statistically significant (p < .001), demonstrated poor correlation (absolute value of correlation coefficient 0.05–0.44).

FIGURE 4.

Correlation coefficients and fully adjusted partial correlation for all the comparisons among electrogram characteristics. Comparison of individual patient correlation coefficients and fully adjusted partial correlation for all the comparisons among electrogram characteristics. (A) Values are average individual Pearson correlation coefficient (SD). All correlations are significant (p < .001). (B) Partial Pearson correlation for the whole cohort adjusted by patient, left versus right atrium, and first versus second ablation. Values are partial Pearson correlation coefficient (p value). All correlations are significant (p < .001) except for DF and CL_R (p = .59) as well as AMP and DF (p = .38). AMP, median amplitude; CL, cycle length; CL_R, cycle length of the most recurrent morphology; DF, dominant frequency; FI, fractionation interval; OI, organizational index; Rec%, frequency of the most recurrent morphology; ShEn, Shannon’s entropy.

Figure 4B shows the fully adjusted partial correlation coefficients for all the comparisons when combined into a single analysis. All correlations were significant (p < .001) except for DF and CL_R (p = .59) as well as AMP and DF (p = .38). The single high correlation between FI and ShEn was observed in both analyses. While the strength of correlation between Rec% and CL_R, AMP and FI, as well as AMP and ShEn varied between moderate and poor in the two analyses, the correlations were close to 0.5 in both analyses. AMP and DF showed very low correlation in both analyses that were directionally different (−0.05 and 0.01). All other comparisons had concordant poor or moderate correlation in both analyses.

The coefficients of variation for the correlation coefficients among all patients are shown in Table 1. For the strong negative relationship between FI and ShEn, this was minimal at −1.71%. For the moderate correlations, the coefficients of variation varied between −31.26% and 20.94%. For the poor correlations, the coefficients of variation were much larger and varied between −240.57% and 138.50%. This demonstrates a high degree of consistency in the relationship between FI and ShEn only.

TABLE 1.

For each pair of comparisons (variable 1 vs. variable 2), the correlation coefficient and coefficient of variation is shown.

Variable 1	Variable 2	Correlation coefficient	Coefficient of variation (%)
ShEn	FI	−0.94	−1.71
DF	CL	−0.63	−16.87
FI	Rec%	0.62	11.69
ShEn	Rec%	−0.62	−10.76
Rec%	CLR	−0.58	−12.49
AMP	FI	0.53	20.94
AMP	ShEn	−0.52	−17.71
OI	DF	−0.50	−31.26
ShEn	CLR	0.44	14.46
OI	Rec%	0.43	26.82
OI	CL	0.41	33.68
FI	OI	0.37	31.44
FI	CLR	−0.37	−17.27
FI	CL	0.36	32.39
ShEn	OI	−0.36	−31.87
AMP	Rec%	0.35	30.47
FI	DF	−0.35	−28.54
ShEn	CL	−0.35	−31.98
Rec%	CL	0.34	37.44
ShEn	DF	0.34	28.56
AMP	CLR	−0.23	−25.66
DF	Rec%	−0.22	−51.70
AMP	OI	0.21	63.36
OI	CLR	−0.16	−70.78
AMP	CL	0.14	102.10
CLR	CL	−0.11	−85.65
DF	CLR	0.07	138.5
AMP	DF	−0.05	−240.57

Note: The data are ranked ordered by absolute value of the correlation coefficient in descending order (strongest correlations at the top and weakest at the bottom).

Abbreviations: AMP, median amplitude; CL, cycle length; CL_R, cycle length of most recurrent morphology; DF, dominant frequency; FI, fractionation interval; OI, organizational index; Rec%, frequency of most recurrent morphology; ShEn, Shannon’s entropy.

3.5 |. Comparison of electrogram properties in patients on and off antiarrhythmic therapy

Table 2 shows the results based on antiarrhythmic drug use at the time of the procedure. Most electrogram properties did not differ. Only the minimum ShEn was significantly higher (p = .024) in patients who were on antiarrhythmic therapy. Moreover, patients on antiarrhythmic therapy did trend to have lower maximum DF (p = .09). Further analysis focused only on pulmonary vein electrograms in Group 1 showed no statistically significant differences in electrogram properties for those on antiarrhythmic drugs, though there was a trend for minimum CL_R and maximum ShEn (Table 2).

TABLE 2.

Comparison of all electrograms and only pulmonary vein electrograms in patients on and off antiarrhythmic therapy for total cohort and for Group 1, respectively, and all electrograms and only pulmonary vein electrograms in PVI responders and nonresponders.

	N	Min CLR (ms)	Highest Rec(%)	Min CL (ms)	Shortest FI (ms)	Highest DF (Hz)	Max OI	Min OI	Max AMP (mV)	Min ShEn	Max ShEn
Overall N = 42
Off AAD	32	171.5 ± 23.1	98.3 ± 2.9	109.3 ± 7.4	42.7 ± 0.9	9.7 ± 1.2	0.58 ± 0.05	0.28 ± 0.02	4.57 ± 2.04	0.535 ± 0.047	0.951 ± 0.010
On AAD	10	190.6 ± 33.6	95.7 ± 9.8	106.4 ± 3.2	42.4 ± 0.7	8.9 ± 1.2	0.57 ± 0.04	0.28 ± 0.02	4.71 ± 3.29	0.572 ± 0.039	0.958 ± 0.013
p Value		.12	.59	.50	.41	.09	.83	.75	.67	.02	.12
Pulmonary veins N = 32
Off AAD	24	175.1 ± 28.5	95.5 ± 6.6	125.9 ± 13.4	43.4 ± 1.1	9.0 ± 1.2	0.56 ± 0.04	0.32 ± 0.03	1.83 ± 0.847	0.595 ± 0.063	0.939 ± 0.016
On AAD	8	227.1 ± 81.3	87.3 ± 19.6	126.4 ± 20.7	43.3 ± 1.0	8.7 ± 1.5	0.55 ± 0.05	0.32 ± 0.02	1.51 ± 0.547	0.626 ± 0.064	0.951 ± 0.016
p Value		.08	.89	.68	.73	.29	.53	.96	.31	.26	.10
Overall N = 26
PVI responders	18	158.7 ± 12.8	98.7 ± 2.2	111.2 ± 8.4	42.8 ± 0.9	9.9 ± 1.2	0.58 ± 0.03	0.27 ± 0.02	5.12 ± 2.37	0.535 ± 0.038	0.949 ± 0.009
PVI nonresponders	8	191.5 ± 33.5	94.8 ± 11.1	106.8 ± 3.2	42.6 ± 0.7	9.5 ± 1.2	0.59 ± 0.04	0.29 0.02	4.74 ± 2.68	0.568 ± 0.037	0.952 ± 0.013
p Value		.01	.78	.37	.92	.36	.49	.07	.81	.06	.68
Pulmonary veins N = 26
PVI responders	18	162.7 ± 11.9	97.5 ± 3.5	125.6 ± 9.1	43.3 ± 1.2	9.2 ± 1.2	0.56 ± 0.03	0.32 ± 0.03	1.98 ± 0.908	0.586 ± 0.05	0.942 ± 0.01
PVI nonresponders	8	230.5 ± 80.9	87.5 ± 19.4	117.5 ± 15.2	43.6 ± 0.9	9.0 ± 1.5	0.57 ± 0.04	0.31 ± 0.03	1.52 ± 0.530	0.656 ± 0.05	0.939 ± 0.021
p Value		.01	.31	.048	.43	.73	.47	.67	.16	.01	.70

Abbreviations: AAD, antiarrhythmic drugs; AMP, median amplitude; CL, cycle length; CL_R, cycle length of most recurrent morphology; DF, dominant frequency; FI, fractionation interval; OI, organizational index; PVI, pulmonary vein isolation; Rec%, frequency of most recurrent morphology; ShEn, Shannon’s entropy.

3.6 |. Comparison of electrogram properties in PVI responders and nonresponders

In Group 1, one patient was lost to follow-up and one was censored at the time of cardiac surgery during the 3-month postablation blanking period. Follow-up data were available in the remaining 30 patients. Four patients who did not have recurrent AF or flutter, but were maintained on antiarrhythmic drugs were excluded from the outcome analysis. Of the 18 patients who were free of AF or flutter and off antiarrhythmic drugs (PVI responders), one patient had undergone additional posterior wall isolation. In one patient who had multiple monitors, there was a single 1-min episode of AF in one monitor. This patient was classified as a PVI responder. Only one PVI responder was on antiarrhythmic drugs at the time of the ablation (on Amiodarone). Of the eight patients who were PVI nonresponders, only two had undergone additional posterior wall isolation. Of these eight patients, four had recurrent paroxysmal AF, two had recurrent persistent AF, and two had atrial flutter recurrence. Five patients who were PVI nonresponders were on antiarrhythmic drugs at the time of the ablation (two on Amiodarone, one on Sotalol, one on Flecainide, and one on Dronedarone). Table 2 shows the electrogram results for the PVI responders (n = 18) versus the PVI nonresponders (n =8). When comparing the overall electrograms characteristics, minimum CL_R was significantly lower among PVI Responders (158.7 ± 12.8 vs. 191.5 ± 33.5 ms, p = .01) compared to PVI nonresponders. When comparing pulmonary vein electrogram characteristics, PVI Responders had shorter minimum CL_R (p = .01), lower minimum ShEN (p = .01), and longer minimum CL (p = .048) compared to PVI nonresponders (Table 2). Multivariable analysis was not performed due to the small sample size. The optimal cutpoint for the shortest CL_R ≥ 192 ms predicted AF recurrence 1-year postablation with sensitivity of 62.5% and specificity of 100% (AUC = 0.81, 95% confidence interval [CI]: [0.53–1.00]) (Figure 5).

FIGURE 5.

Shortest cycle length of most recurrent morphology (CL_R) shows high performance predicting responders versus nonresponders to catheter ablation at 1 year in patients with persistent atrial fibrillation (AF) undergoing initial catheter ablation. In a sample size of 26 (group 1) patients with persistent AF undergoing initial catheter ablation, the AF recurrence was 30.8%. The shortest CL_R optimal cutpoint to predict AF recurrence at 1-year postablation was calculated with a receiver operating characteristic curve in R. Grouping the shortest CL_R by ≥192 and <192 ms resulted in area under the curve = 0.81 [95% confidence interval: 0.53–1.00), Youden Index = 0.625, sensitivity = 62.5%, specificity = 100%.

3.7 |. Correlation among top 10 sites identified by each of the electrogram characteristics

For each electrogram parameter, the top 10 values and associated sites corresponding to these values were identified. Results are shown in Table 3. It is clear that there is minimal overlap between the top 10 sites identified by one electrogram characteristic and the top 10 sites identified by the other electrogram characteristics.

TABLE 3.

For each parameter and for each patient, the top 10 values (smallest or largest depending on the parameter), were identified.

	Average ± SD of the top 10 values [min, max]	Total no. of patients with a top 10 electrogram parameter per location						Average ±SD % of overlap in the top 10 with other electrogram characteristics
		Group 1			Group 2
Parameter		PV	LA (non-PV)	RA	PV	LA (non-PV)	RA	Shortest CL	Highest DF	Highest OI	Highest ShEN	Shortest FI	Shortest CLR	Highest Rec	Lowest AMP
Shortest CL (ms)	128.6 ± 11.7 [108.0, 158.0]	19	29	27	6	10	9	100	19.2 ± 14.8	0.2 ± 1.5	4.3 ± 6.5	3.7 ± 5.3	1.0 ± 3.0	0.0 ± 0.0	9.4 ± 12.7
Highest DF (Hz)	8.0 ± 0.8 [6.8, 10.9]	28	32	25	4	10	9	19.9 ± 15.0	100	0.2 ± 1.5	5.9 ± 7.0	4.5 ± 7.6	3.6 ± 6.9	1.4 ± 3.5	3.1 ± 6.8
Highest OI	0.5 ± 0.0 [0.4, 0.6]	27	26	25	8	10	7	0.2 ± 1.5	0.2 ± 1.5	100	0.2 ± 1.5	4.3 ± 11.1	23.3 ± 19.6	23.1 ± 18.5	4.3 ± 7.4
Highest ShEN	0.9 ± 0.0 [0.9, 1.0]	30	30	29	10	10	6	4.5 ± 6.7	6.0 ±7.0	0.2 ±1.5	100	58.3 ± 17.7	0.5 ± 2.2	0.0 ±0.0	20.2 ± 12.8
Shortest FI (ms)	44.5 ± 1.3 [43.0, 48.0]	32	30	32	10	10	10	2.6 ± 3.8	2.9 ± 5.0	2.7 ±5.8	43.0 ±16.2	100	0.2 ± 1.2	0.2 ± 1.2	30.7 ±14.9
Shortest CLR (ms)	222.8 ± 48.8 [156.0, 377.0]	29	27	24	6	8	10	0.9 ± 2.9	3.5 ± 6.9	22.4 ± 18.0	0.5 ± 2.2	0.2 ± 1.4	100	67.9 ± 20.6	1.6 ± 3.7
Highest Rec (%)	83.0 ± 13.3 [46.0, 100.0]	30	26	29	7	8	8	0.0 ± 0.0	1.4 ±3.4	21.0 ± 16.3	0.0 ± 0.0	0.2 ± 1.5	65.0 ± 22.4	100	1.9 ± 4.5
Lowest AMP	124.1 ± 43.3 [57.3, 227.1]	28	21	30	10	7	6	9.8 ± 13.0	3.1 ± 6.8	4.3 ± 7.4	20.2 ± 12.8	42.4 ± 17.9	1.7 ± 3.8	1.9 ± 4.5	100

Note: The table shows the average and range of these top 10 values and their location (PV, non-PV, LA, and RA), as well as the overlap of sites with the top 10 values for each pair of electrogram characteristics.

Abbreviations: AMP, median amplitude; CL, cycle length; CLR, cycle length of most recurrent morphology; DF, dominant frequency; FI, fractionation interval; LA, left atrial; OI, organizational index; PV, pulmonary vein; RA, right atrial; Rec, frequency of most recurrent morphology; SD, standard deviation; ShEN, Shannon’s entropy.

3.8 |. Comparison of HD-grid to reflexion electrogram characteristics

Table 4 shows the comparison of electrogram characteristics recorded in five patients by the high density (HD)-grid and in 25 patients by the Reflexion catheter, overall, by chamber, and by region. Only rare and small differences were observed, suggesting that these catheter types do not significantly impact the electrogram characteristics.

TABLE 4.

Comparison of electrogram characteristics recorded in five patients by the HD-grid and in 25 patients by the Reflexion catheter.

Atrium/region	Catheter type	N	Min CLR (ms)	Highest Rec (%)	Min CL (ms)	Shortest FI (ms)	Highest DF (Hz)	Max OI	Min OI	Max AMP (mV)	Min ShEn	Max ShEn
Overall	HD-grid	5	166.8 ± 33.1	99.6 ± 0.9	107.2 ± 7.3	42.0 ± 0.7	10.3 ± 1.0	0.61 ± 0.02	0.28 ± 0.02	13.2 ± 3.8	0.509 ± 0.033	0.953 ± 0.010
	Reflexion	25	174.2 ± 25.9	96.7 ± 6.7	108.76 ± 6.0	42.9 ± 0.8	9.5 ± 1.2	0.57 ± 0.04	0.27 ± 0.02	10.4 ± 4.1	0.550 ± 0.041	0.951 ± 0.011
	p Value		.6568	.0468	.6706	.0460	.1252	.0094	.5259	.1854	.0429	.7847
RA	HD-grid	5	181.4 ± 30.1	98.0 ± 2.6	113.8 ± 8.7	42.2 ± 0.5	8.7 ± 1.0	0.60 ± 0.04	0.29 ± 0.02	12.0 ± 4.1	0.509 ± 0.033	0.942 ± 0.013
	Reflexion	25	209.0 ± 34.3	88.4 ± 11.7	116.4 ± 14.5	43.5 ± 1.6	8.3 ± 0.8	0.54 ± 0.04	0.29 ± 0.03	9.6 ± 4.0	0.568 ± 0.044	0.935 ± 0.013
	p Value		.1157	.0010	.5987	.0005	.4920	.0158	.9275	.2852	.0095	.3312
LA	HD-grid	5	169.2 ± 32.2	99.6 ± 0.9	107.2 ± 7.3	42.2 ± 0.8	10.3 ± 1.0	0.58 ± 0.02	0.28 ± 0.02	9.6 ± 3.7	0.576 ± 0.029	0.952 ± 0.010
	Reflexion	25	180.6 ± 40.2	94.2 ± 9.6	112.3 ± 8.2	43.1 ± 0.9	9.5 ± 1.3	0.56 ± 0.40	0.28 ± 0.02	7.1 ± 3.4	0.591 ± 0.065	0.950 ± 0.011
	p Value		.5137	.0102	.2072	.0684	.1209	.2705	.7729	.2101	.4330	.7074
SVC	HD-grid	5	239.0 ± 47.5	88.2 ± 20.4	141.0 ± 18.5	49.4 ± 9.6	7.1 ± 0.8	0.52 ± 0.07	0.35 ± 0.05	4.5 ± 1.6	0.582 ± 0.088	0.908 ± 0.037
	Reflexion	25	288.0 ± 72.4	76.0 ± 19.7	140.2 ± 19.4	47.5 ± 7.0	7.2 ± 0.6	0.51 ± 0.04	0.33 ± 0.03	4.3 ± 2.6	0.613 ± 0.079	0.907 ± 0.035
	p Value		.0921	.2663	.9363	.6875	.8547	.7373	.3949	.7892	.4974	.9473
RAA	HD-grid	5	207.2 ± 37.7	90.4 ± 11.1	169.0 ± 43.6	100.0 ± 55.2	6.6 ± 1.8	0.54 ± 0.08	0.41 ± 0.10	9.6 ± 2.6	0.613 ± 0.038	0.823 ± 0.081
	Reflexion	25	252.6 ± 63.6	69.6 ± 15.4	147.5 ± 16.9	70.0 ± 19.1	7.2 ± 0.8	0.49 ± 0.05	0.34 ± 0.04	6.5 ± 2.5	0.662 ± 0.039	0.850 ± 0.035
	p Value		.0591	.0082	.3347	.2938	.5186	.2176	.1816	.0538	.0409	.5037
Septal RA	HD-grid	5	218.6 ± 62.9	88.2 ± 15.2	126.8 ± 15.3	49.2 ± 7.7	8.1 ± 1.4	0.55 ± 0.03	0.33 ± 0.02	9.5 ± 3.9	0.550 ± 0.045	0.902 ± 0.027
	Reflexion	25	258.8 ± 61.8	72.4 ± 16.6	133.0 ± 16.3	51.2 ± 11.4	7.97 ± 0.7	0.51 ± 0.05	0.32 ± 0.03	5.4 ± 3.0	0.634 ± 0.048	0.895 ± 0.031
	p Value		.2408	.0799	.4417	.6417	.7685	.0548	.3167	.0727	.0096	.6654
Other RA	HD-grid	5	193.6 ± 21.7	96.8 ± 4.0	116.0 ± 8.2	43.2 ± 0.8	8.3 ± 0.6	0.57 ± 0.03	0.29 ± 0.02	10.9 ± 3.7	0.533 ± 0.041	0.933 ± 0.008
	Reflexion	25	239.6 ± 43.6	75.7 ± 12.6	119.9 ± 15.6	44.8 ± 3.4	8.0 ± 0.8	0.52 ± 0.05	0.30 ± 0.03	8.5 ± 3.4	0.611 ± 0.045	0.925 ± 0.018
	p Value		.0043	<.0001	.4382	.0427	.4016	.0232	.4557	.2341	.0088	.1380
Left PVs	HD-grid	5	188.4 ± 41.1	96.0 ± 4.3	135.6 ± 23.8	43.8 ± 1.1	8.9 ± 2.1	0.55 ± 0.04	0.34 ± 0.05	4.6 ± 1.8	0.596 ± 0.046	0.935 ± 0.024
	Reflexion	25	207.6 ± 66.8	85.8 ± 17.7	138.0 ± 16.5	46.8 ± 4.8	8.6 ± 1.2	0.54 ± 0.05	0.33 ± 0.03	4.2 ± 2.7	0.644 ± 0.072	0.922 ± 0.020
	p Value		.4201	.0181	.8358	.0105	.7542	.6834	.8148	.6447	.0844	.3183
Right PVs	HD-grid	5	193.0 ±41.5	94.0 ± 7.6	129.0 ±12.6	42.2 ±0.8	8.4 ± 1.3	0.56 ± 0.02	0.32 ±0.03	5.3 ± 1.4	0.630 ±0.046	0.945 ± 0.022
	Reflexion	25	232.7 ± 104.0	82.3 ± 19.6	133.6 ±18.8	44.1 ±2.5	8.0 ±0.9	0.53 ± 0.04	0.33 ± 0.04	3.7 ±2.4	0.636 ±0.089	0.939 ±0.018
	p Value		.1736	.0372	.5173	.0058	.4999	.0422	.6397	.0743	.8539	.5899
Post LA	HD-grid	5	202.8 ± 33.2	81.8 ± 13.5	117.8 ± 8.8	44.8 ± 2.5	8.7 ± 1.0	0.55 ± 0.04	0.32 ± 0.02	4.6 ± 1.5	0.664 ± 0.070	0.927 ± 0.024
	Reflexion	25	289.6 ± 134.7	65.8 ± 25.6	122.4 ± 10.4	45.5 ± 2.8	8.6 ± 1.3	0.53 ± 0.05	0.31 ± 0.03	4.0 ± 2.0	0.712 ± 0.075	0.931 ± 0.017
	p Value		.0090	.0674	.3392	.6053	.8939	.4213	.3926	.4983	.2217	.7011
Septal LA	HD-grid	5	241.8 ± 50.1	82.0 ± 16.1	129.6 ± 12.7	47.4 ± 2.1	8.1 ± 0.9	0.52 ± 0.04	0.31 ± 0.02	4.0 ± 2.1	0.662 ± 0.081	0.924 ± 0.005
	Reflexion	25	332.1 ± 263.6	67.3 ± 23.0	132.0 ± 19.7	48.2 ± 6.9	7.8 ± 0.8	0.50 ± 0.05	0.32 ± 0.04	3.9 ± 1.7	0.696 ± 0.078	0.921 ± 0.022
	p Value		.1261	.1252	.7334	.6183	.4571	.3043	.2363	.8808	.4265	.5517
Other LA	HD-grid	5	234.0 ± 42.0	75.2 ± 16.0	107.4 ± 7.7	43.8 ± 1.3	9.3 ± 1.0	0.54 ± 0.04	0.30 ± 0.03	5.0 ± 1.0	0.676 ± 0.057	0.941 ± 0.011
	Reflexion	25	252.0 ± 65.3	70.7 ± 18.6	123.7 ± 17.3	44.4 ± 1.5	8.6 ± 1.0	0.52 ± 0.06	0.30 ± 0.03	5.2 ± 2.3	0.664 ± 0.069	0.934 ± 0.015
	p Value		.4539	.5967	.0049	.4241	.1948	.4823	1.000	.7962	.6896	.2572
LAA	HD-grid	5	225.6 ± 99.8	83.6 ± 21.9	137.8 ± 19.8	70.8 ± 28.3	7.6 ± 1.3	0.56 ± 0.03	0.39 ± 0.08	8.7 ± 4.4	0.636 ± 0.071	0.869 ± 0.047
	Reflexion	25	294.6 ± 231.8	69.4 ± 21.5	149.0 ± 23.8	62.2 ± 27.2	7.2 ± 1.0	0.52 ± 0.05	0.38 ± 0.07	6.0 ± 3.1	0.672 ± 0.081	0.882 ± 0.055
	p Value		.3012	.2353	.3041	.5590	.5713	.0247	.7531	.2508	.3553	.5964

Note: Those with Bonferroni-corrected significant p values are highlighted in red.

Abbreviations: AMP, median amplitude; CL, cycle length; CLR, cycle length of most recurrent morphology; DF, dominant frequency; FI, fractionation interval; HD, high density; LA, left atrium; LAA, left atrial appendage; OI, organizational index; PV, pulmonary veins; PVI, pulmonary vein isolation; RA, right atrium; RAA, right atrial appendage; Rec%, frequency of most recurrent morphology; ShEn, Shannon’s entropy; SVC, superior vena cava.

4 |. DISCUSSION

In this study, we evaluated the relationship among eight electrogram characteristics that have been studied in the setting of AF. Notably, these variables were all correlated but the vast majority of comparisons were poor to moderately correlated. Only ShEn and FI demonstrated a strong negative correlation, as expected by their definition. The relationships are highly conserved among patients and in the RA versus LA as well as between those undergoing initial versus second ablation procedures. This study, therefore, identifies seven unique electrogram characteristics that are not highly correlated to each other that can be used as an initial library to map and/or classify electrograms in AF. Antiarrhythmic drug therapy did not appear to have strong effects on these electrogram properties, though the data are cross-sectional and not measured pre- and post-therapy. Importantly, there were some electrogram features that were identified to be potential predictors for ablation outcome that should be evaluated in larger cohorts. As the sites identified by the individual characteristics as possible areas of interest do not overlap among the different electrogram characteristics, it is critical to identify which of these characteristics may be helpful for mapping AF. As electrogram analysis is the foundation for successful ablation of many cardiac arrhythmias, these data provide the background data and rationale to support further studies to determine whether appropriate electrogram analysis techniques in AF—either in isolation or in combination—can provide the requisite mapping for the purpose of identifying sites or regions to target for ablation.

4.1 |. Electrogram characteristics in AF

Several approaches have been assessed for characterizing electrograms during AF. The principal goal is to identify electrogram characteristics that can help guide ablation as standard activation mapping in the electrophysiology laboratory in humans is not feasible and empiric approaches have limited success rates. The basic approaches are to evaluate signal complexity/reproducibility, rate, and amplitude.

Nademanee et al.¹⁴ reported mapping and ablation of AF guided by CFAEs, defined by (1) fractionated electrograms composed of two or more deflections, and/or perturbation of the baseline with continuous deflection of a prolonged activation complex over a 10-second recording period; (2) electrograms with a very short CL (≤120 ms) averaged over a 10-s recording period. These criteria are also captured quantitatively in the FI and ShEn. While this report demonstrated an impressive 91% freedom from AF at 1 year, subsequent evaluation of CFAE-guided ablation in randomized studies^1,15 did not demonstrate any benefit to this approach. One way to reconcile these differences is that there is some other unknown/unmeasured factor that might be consistently applied in the setting of a single-site study that could not be generalized to a multicenter clinical trial. Thus, rather than abandoning this criterion, it might be prudent to explore other features that could enhance its utility, that is, by combining FI or ShEn with other electrogram characteristics that are not highly correlated to these measures.

DF analysis of atrial electrograms is an interesting theoretical approach to assess the periodicity of AF. Sites with higher frequency presumably are the drivers of AF in the rest of the atrium (fast regions have to drive slower regions and not vice versa). Several studies have explored the utility of mapping AF with DF analysis.^16–18 Sanders et al.¹⁶ found that AF CL prolonged by 10% when ablation was performed at the high DF sites with no prolongation at other sites. Atienza et al.¹⁷ showed that ablation of maximal DF sites was associated with a higher probability of maintenance of sinus rhythm. However, prospective comparison of DF-guided ablation plus PVI versus PVI alone did not show any benefit.¹⁸ There are many technical factors that could impact the utility and reproducibility of DF analysis related to the varying amplitudes and intervals, as well as the sharp morphology of the electrograms.^19–21 As the DF is an index of AF frequency, it is moderately correlated with AF CL—they are actually inversely related. The OI quantifies the regularity of the arrhythmia. It has been suggested that OI maps may better localize AF sources than DF maps.²²

We first reported ShEn as a measure of electrogram complexity in AF.¹² This has subsequently been evaluated in human and animal mapping studies showing that high ShEn was consistently colocated at the pivot zone of rotational activity.¹³ This has not yet been applied as a target for ablation of AF.

Low electrogram amplitude is a key feature associated with atrial fibrosis.²³ As atrial fibrosis is a key pathophysiologic substrate for AF, this has been targeted for ablation. A meta-analysis of six studies demonstrated benefit to this approach.²⁴ Yagishita et al.²⁵ compared ablation outcomes in those with no low voltage areas outside the pulmonary veins and those with low voltage areas that were targeted for ablation and showed similar outcomes, suggesting a positive effect of targeting the low voltage areas. While seemingly simple, there are many factors that may influence atrial electrogram amplitude in AF,²⁶ including patient characteristics, catheter characteristics, recording characteristics, and the variability of the signal. DECAAF II² recently demonstrated that simply targeting fibrosis detected by cardiac MRI, in addition to PVI, does not improve outcomes.

Morphology recurrence plots provide an assessment of the reproducibility of the electrograms in a recording.⁵ Bipolar recordings near a source or driver of AF would be expected to have highly reproducible and rapid electrograms. This can be quantified in the Rec% and CL_R. A retrospective evaluation of these parameters demonstrated their potential utility.⁴ A similar concept of evaluating electrogram similarity and rate has also been proposed to identify potential drivers of persistent AF.^6,7 Neither approach has yet been applied for ablation of AF.

Small studies of combined mapping approaches have been reported with some success—DF + CFE,^27,28 DF + low AMP,²⁹ CFE + low AMP.³⁰ The goal of these approaches is to identify the anatomic and electrophysiologic substrates for AF. The tools to provide comprehensive evaluation of electrogram complexity/reproducibility, rate, and amplitude during AF have been developed. Some potential electrogram characteristics were identified in this study that are associated with outcome—OI, ShEn, CL_R, AMP, and CL. These parameters need to be comprehensively evaluated in a large cohort to develop an optimal algorithm to identify drivers of AF that can be targeted for ablation. If this can be achieved, this should provide better outcomes than current empiric approaches.

5 |. LIMITATIONS

While this study included only 42 patients from a single site, the electrogram analysis incorporated 21 846 electrograms from which each of the electrogram characteristics were derived and used for comparison. The present study did not assess the temporal stability of the electrogram characteristics. As extensive LA and RA mapping was performed, this could not be incorporated into this protocol. The focus of this report was the comparison of the electrogram characteristics to each other, so all electrogram characteristic data are simultaneous. As all patients had clinical persistent AF, it is also unknown whether these relationships are preserved for patients with induced AF. It should be noted that the effect of antiarrhythmic drugs on electrogram characteristics were not performed on the same patient without and with treatment. Though this would be ideal, it is impractical as it would require two procedures and the exact positioning of the catheters at both studies. Thus, the cross-sectional approach undertaken in this study is the only practical approach. Finally, there was no concerted effort to target particular electrogram characterstics for ablation in this study and the outcomes analysis is underpowered with only 26 patients. Nevertheless, some potential candidate electrogram characteristics displayed differences between PVI responders and PVI nonresponders supporting further evaluation.

6 |. CONCLUSIONS

There is an urgent need for improving precision of AF ablation. Multiple approaches have been tried without success, including some limited attempts at electrogram-based approaches. As electrograms in AF are determined, at least in part, by underlying tissue properties, electrophysiologic characteristics, and autonomic input, it seems likely that incorporating electrogram analysis in the ablation strategy could be helpful, but this requires further study. While the alternative of performing extensive atrial ablation to “debulk” the LA may achieve success for rhythm control, more extensive ablation and fibrosis may lead to loss of atrial function^31–33—the stiff LA^34,35—detracting from the benefits of maintaining sinus rhythm. Developing novel approaches for activation mapping of AF could provide such an approach, but the most recent data with FIRM mapping (REAFFIRM, NCT02274857) has not been encouraging. While individual electrogram-based approaches have also not performed well in multicenter, clinical trials, comprehensive evaluation of all the tools available, particularly in combination, to analyze electrogram complexity/reproducibility, rate, and amplitude has not been performed. The multiple techniques for electrogram analysis evaluated in this report can provide largely independent information and, therefore, represent a tool chest that is worth developing for better mapping of AF and potential identification of ablation targets.

Abbreviations:

AMP

median amplitude

CL

cycle length

CL_R

cycle length of most recurrent morphology

DF

dominant frequency

FI

fractionation interval

OI

organizational index

PVI

pulmonary vein isolation

Rec%

frequency of most recurrent morphology

ShEn

Shannon’s entropy

DATA AVAILABILITY STATEMENT

Any inquiries regarding the data can be submitted to the corresponding author and it will be addressed accordingly.