Article
Reference
Sustained and self‐terminating atrial fibrillation induced immediately after pulmonary vein isolation exhibit differences in coronary sinus
electrical activity from onset
JOHNER, Nicolas, NAMDAR, Mehdi, SHAH, Dipen
Abstract
Little data exists on the electrophysiological differences between sustained atrial fibrillation (sAF; >5 minutes) vs self-terminating nonsustained AF (nsAF;
JOHNER, Nicolas, NAMDAR, Mehdi, SHAH, Dipen. Sustained and self‐terminating atrial fibrillation induced immediately after pulmonary vein isolation exhibit differences in coronary sinus electrical activity from onset. Journal of Cardiovascular Electrophysiology, 2020, vol.
31, no. 1, p. 150-159
DOI : 10.1111/jce.14296 PMID : 31778260
Available at:
http://archive-ouverte.unige.ch/unige:135846
Disclaimer: layout of this document may differ from the published version.
1 / 1
1
Sustained and Self-terminating Atrial Fibrillation Induced
Immediately after Pulmonary Vein Isolation Exhibit Differences in Coronary Sinus Electrical Activity from Onset
Short title: Coronary sinus electrograms predict sustained AF
Authors: Nicolas Johner, MD, Mehdi Namdar, MD, PhD and Dipen C. Shah, MD, FHRS.
Affiliations: Cardiology Division, Geneva University Hospitals, Geneva, Switzerland Email: Nicolas Johner: nicolas.johner.ge@gmail.com; Mehdi Namdar:
mehdi.namdar@hcuge.ch; Dipen C. Shah: dipen.shah@hcuge.ch.
Correspondence: Dipen C. Shah, Cardiology Division, Geneva University Hospitals, Rue Gabrielle-Perret-Gentil 4, 1205 Geneva, Switzerland. Tel: +41-223727202/Fax: +41-223727229.
Word count: 5’111
Funding sources: Dipen C. Shah is supported by the Swiss National Science Foundation (Grant #173007).
Disclosures: Nicolas Johner: received educational grants from Boston Scientific, Abbott and Cardinal Health. Mehdi Namdar: received consultant fees and travel grants from Boston- Scientific, Biotronik and Biosense-Webster. Dipen C. Shah received consultant fees from Biosense Webster, Biotronik, St. Jude Medical and Boston Scientific. He has benefited from research grants from Biosense Webster, St. Jude Medical and Boston Scientific via the Cardiology Division, Geneva University Hospitals. He received Speaker Board member fees from Biosense Webster, St. Jude Medical and Boston Scientific.
2
Abstract
1
INTRODUCTION: Little data exists on the electrophysiological differences between sustained 2
(sAF, >5min) versus self-terminating non-sustained AF (nsAF, <5min). We sought to investigate 3
the electrophysiological characteristics of coronary sinus (CS) activity during post-pulmonary 4
vein isolation (PVI) sAF versus nsAF.
5
METHODS AND RESULTS: We studied 142 patients post-PVI for paroxysmal AF. In a 50- 6
patient subset, CS electrograms in the first 30 seconds of induced AF were analyzed manually.
7
A custom-made algorithm for automated electrogram annotation was derived for validation on 8
the whole patient set. In patients with sAF post-PVI, CS fractionated potentials were ablated.
9
Manual analysis showed that patients with sAF exhibited higher activation pattern variability (2.1 10
vs 0.5 changes/sec, p<0.001); fewer proximal-to-distal wavefronts (25 vs 61%, p<0.001); fewer 11
unidirectional wavefronts (60 vs 86%, p<0.001); more pivot locations (4.3 vs 2.1, p<0.001);
12
shorter cycle lengths (190 vs 220ms, p<0.001); and shorter cumulative isoelectric segments (35 13
vs 44%, p=0.045) compared to nsAF. These observations were confirmed on the whole study 14
population by automated electrogram annotation and sample entropy computation (SampEn:
15
0.29±0.15 in sAF vs 0.15±0.05 in nsAF, p<0.0001). The derived model predicted sAF with 78%
16
sensitivity, 88% specificity; agreement with manual model: 88% (Cohen’s kappa=0.76). CS 17
defragmentation resulted in AF termination or non-inducibility in 49% of sAF.
18
CONCLUSION: In paroxysmal AF patients post-PVI, induced sAF shows greater activation 19
sequence variability, shorter cycle length and higher SampEn in the CS compared to nsAF.
20
Automated electrogram annotation confirmed these results and accurately distinguished self- 21
terminating nsAF episodes from sAF based on 30-second recordings at AF onset.
22 23
3 1
Keywords: Atrial fibrillation, catheter ablation, coronary sinus, electrograms, prediction model, 2
pulmonary vein isolation, signal processing, inducibility, substrate modification.
3
Abbreviations: AF, atrial fibrillation; AFCL, atrial fibrillation cycle length; AUC, area under the 4
curve; CI, confidence interval; CS coronary sinus; LA, left atrium; nsAF, non-sustained atrial 5
fibrillation; OAT; organized atrial tachycardia; PAF, paroxysmal atrial fibrillation; PV, pulmonary 6
vein; PVI, pulmonary vein isolation; RA, right atrium; ROC, receiver operating characteristic;
7
sAF, sustained atrial fibrillation; SampEn, sample entropy.
8 9 10
Introduction
11
While pulmonary vein isolation (PVI) is the cornerstone of atrial fibrillation (AF) ablation, it is 12
considered likely that additional extra-PV substrate modification can further improve outcomes.
13
The lack of an established ablation endpoint beyond PVI limits patient selection for extra-PV 14
ablation. AF non-inducibility by atrial pacing has been proposed as an endpoint to test the 15
neutralization of the arrhythmogenic substrate(1). We sought to characterize the 16
electrophysiological differences between induced sustained AF (sAF) and self-terminating non- 17
sustained (nsAF) AF after PVI.
18
Very scarce data exists analyzing the electrophysiological properties of sAF versus nsAF in 19
humans(2–5). We hypothesized that post-PVI sAF is maintained longer than nsAF due to more 20
extensive extra-PV – including LA and CS – substrate. Based on this hypothesis, we selected 21
for measurement electrogram characteristics that have been previously associated with AF 22
4 substrate and/or AF persistence(6–13). A 5-minute cut-off duration was selected empirically to 1
define sAF and the sensitivity of the results to the cut-off value was tested subsequently.
2
The goal of this study was: 1) to investigate the electrophysiological characteristics in the CS of 3
induced sAF (>5 minutes) versus self-terminating nsAF (<5 minutes) after PVI, and 2) to confirm 4
the observations using an automated objective electrogram annotation method.
5 6
Methods
7
Design and study population
8A retrospective analysis was performed on consecutive patients who underwent a first-ever 9
paroxysmal AF (PAF) ablation between 2009 and 2015 at our institution. Written informed 10
consent was obtained and the study protocol was approved by the institutional review board.
11
Exclusion criteria were the absence of post-PVI induced AF (defined as AF lasting >30 12
seconds), and transition from induced AF to an organized atrial tachyarrhythmia (OAT) within 13
the first 5 minutes of the episode. OATs were excluded by analyzing FF interval regularity and 14
consistency of activation patterns.
15
First, preliminary analysis of candidate electrogram characteristics was performed on a 16
subgroup of 50 patients randomly selected from the study population with a 1:1 ratio of sAF and 17
nsAF (Derivation cohort). Preliminary analysis included manual electrogram annotation as well 18
as simple algorithmic quantification of isoelectric baseline. After identification of the relevant 19
electrogram features which best discriminate sAF and nsAF, a custom-made electrogram 20
analysis algorithm was developed for automated and reproducible electrogram annotation and 21
analysis. Automated electrogram analysis was then tested on the whole study population.
22
5 Figure 1 summarizes the study flow.
1 2
Electrophysiology study
3All antiarrhythmic drugs were discontinued 5 half-lives before the procedures, except 4
amiodarone, which was discontinued for ≥72h. A 10-electrode catheter (2-5-2 mm electrode 5
spacing, Abbott, Lake Bluff, IL) was placed with its tip in the distal CS or great cardiac vein and 6
all electrodes within the CS so that the tip of the catheter was placed at 2 o’clock around the 7
mitral annulus on the left anterior oblique view. PVI was performed using a 20-pole circular 8
catheter (Biosense Webster, Diamond Bar, CA) and an irrigated-tip ablation catheter 9
(Thermocool, Biosense Webster)(14). Following PVI, sAF inducibility was tested using a 10
standardized decremental burst pacing protocol. Eight-paced beat sequences were delivered 11
from a CS bipole or the left atrial appendage with 1-2 ms stimulus duration and decremental S1- 12
S1 intervals from 350 ms down to effective atrial refractory period in steps of 10 ms. The test 13
was considered positive when sAF (>5 minutes) was induced, hypothetically indicative of non- 14
PV substrate. In case of non-inducibility, this protocol was repeated to confirm the phenotype.
15
Patients with induced sAF post-PVI underwent selective ablation of continuous or rapid 16
fractionated potentials in the CS with a delivered maximum power of 25 W. The burst-pacing 17
induction protocol was then repeated to evaluate the effect of CS defragmentation on AF 18
inducibility.
19
Endocardial bipolar electrograms were amplified, filtered (band-pass 30-500 Hz) and digitalized 20
at 1’000 Hz (LabSystem Pro, Boston Scientific).
21 22
Manual electrogram analysis: Derivation cohort
236 For each patient, the initial 30 seconds of post-PVI induced AF were analyzed. Near-field 1
electrograms were selected and local activation times were annotated as shown in Figure 2 2
(details in Supplemental Material). Over the 30-second recording, 5 markers of complexity were 3
analyzed: 1) the beat-to-beat CS activation pattern, defined as the pair of electrodes showing 4
the earliest and latest annotated electrogram (Figure 3, Panel A). 2) Local activation pivots(15), 5
defined as wavefronts making a 180° change of direction of activation sequence (Figure 3, 6
Panel B) were characterized based on location and activation sequence. 3) The predominant 7
grade of electrogram organization in the CS was categorized using a modified Wells 8
classification(16) (Figure 4). 4) Mean AFCL was measured separately in the proximal and distal 9
CS bipole during the initial 5 seconds and during seconds 10-25. For all the above-mentioned 10
analyses, activation sequences obscured by far-field ventricular deflections were excluded from 11
analysis. 5) The cumulative duration of isoelectric signal defined as amplitudes <0.04 mV 12
uninterruptedly for at least 40 ms was measured in the proximal and distal CS bipoles using a 13
custom-made algorithm (MATLAB R2014a, The MathWorks, Natick, MA).
14 15
Automated electrogram annotation and analysis
16Based on the electrogram features of interest identified at preliminary analysis, a custom-made 17
algorithm was designed for automated CS electrogram analysis, including beat-to-beat local 18
activation time annotation, CS activation sequences identification, AF cycle length (AFCL), 19
electrogram organization quantification, and Sample entropy (SampEn) measurement.
20
Automated electrogram analysis and SampEn measurement was performed on the whole study 21
population and also separately on the derivation cohort. SampEn is defined as the negative 22
logarithm of the probability that two segments of the signal which are similar for length m remain 23
similar for length m+1. Therefore, SampEn is a measure of signal predictability, with zero 24
7 indicating regular periodic signal and progressively greater values indicating random irregular 1
signal. Algorithm details are described in the Supplemental Material.
2 3
Follow-up 4
Patients underwent follow-up visits at 1, 3 and 6 months (accompanied by 24-48h Holter 5
monitoring) and every 6-12 months thereafter, when rhythm monitoring was performed only in 6
case of suggestive symptoms. Recurrent AF and organized atrial arrhythmias were defined in 7
accordance with current guidelines(17). No arrhythmias within the initial 3 months were counted.
8
Patients without any induced AF post-PVI (noAF; N=107, see Study overview, Figure 1) were 9
included in the outcome analysis for comparison as well as to achieve a representative sample 10
of paroxysmal AF ablation patients.
11 12 13
Statistical analysis and model design
14Continuous variables are expressed as mean (±1 standard deviation) if normally distributed and 15
as median (interquartile range) if not. Categorical variables are expressed as number 16
(percentage of patients). Gaussian distribution was tested using the Kolmogoroff-Smirnov test.
17
Differences between two groups were tested using Student’s t-test for normally distributed 18
continuous variables and Mann-Whitney U test in case of non-normal distribution. The Fischer 19
exact test was used to compare categorical variables. Univariable analyses and multivariable 20
logistic regression were used to identify demographic parameters associated with sAF.
21
Significance was defined at bilateral α<0.05. For prediction models, univariable analyses were 22
8 performed to preselect covariates. Stronger associations (p<0.2) were included to build the 1
multivariable model, which was designed using backward elimination to maximize area under 2
the receiver-operating characteristic (ROC) curve. Ten-fold cross-validation was performed to 3
assess the unbiased predictive accuracy. Agreement between automated and manual model 4
prediction was measured by Cohen’s kappa coefficient. Analyses were performed using 5
Statistica 13 (Dell, Round Rock, TX) and STATA 15 (StataCorp, College Station, TX).
6 7
Results
8
Patient demography
9A total of 263 patients with PAF were assessed for eligibility, 142 of which were included in the 10
study: 76 sAF, 66 nsAF, and 121 were excluded (Figure 1).
11
Patient demographics are summarized in Table 1. Compared to patients with post-PVI nsAF, 12
patients with post-PVI sAF were older, had a longer history of PAF, had a larger LA volume 13
index (LAVI) at echocardiography and more frequently had a history of stroke/TIA. Independent 14
predictors of sAF after adjustment were age, LAVI and a longer history of PAF.
15
Demographics of the derivation subgroup are summarized in Supplemental table 1.
16 17
AF induction and reproducibility
18The 142 selected AF episodes were induced with an average 226±32 ms S1-S1 interval in sAF 19
(range: 160-350 ms) and 214±22 ms in nsAF (range: 180-300 ms, p=0.008). Of the 142 20
patients, 69 had nsAF or no arrhythmia during the first induction protocol run. When the 21
9 induction protocol was repeated for reproducibility, 66 (96%) patients remained non-inducible;
1
73 (96%) of 76 sAF episodes were induced on first attempt.
2
The median duration of induced AF was 114 seconds (range: 30-293 seconds) in nsAF and 24 3
minutes 31 seconds (range: 309 seconds – 80 minutes 48 seconds) in sAF. All 66 nsAF 4
episodes terminated spontaneously to sinus rhythm. Seven sAF episodes terminated 5
spontaneously (duration range: 320 – 503 seconds), 48 were terminated by internal electrical 6
cardioversion and 21 terminated during CS defragmentation.
7 8
Effect of CS ablation
9Of the 76 sAF patients, 65 underwent CS defragmentation, with an average 12.6±7.3 10
radiofrequency pulses corresponding to 8.4±4.9 minutes of radiofrequency time. CS 11
defragmentation resulted in termination to sinus rhythm or non-inducibility in 32 (49%), of which 12
21 (32%) were terminations to sinus rhythm and an overlapping 24 (37%) were non-inducible 13
after defragmentation. Radiofrequency time was shorter in patients with termination or non- 14
inducibility post-CS ablation compared to patients with no effect (7.2±4.7 versus 9.5±4.5 min, 15
p=0.02).
16 17
Manual analysis of coronary sinus electrograms
18A derivation subgroup of 50 patients was selected randomly for preliminary manual electrogram 19
analysis. A total of 34’866 deflections, accounting for 7’174 AF CS activation sequences, were 20
analyzed on a total of 1500 seconds of recordings. Compared to induced nsAF (N=25), patients 21
with induced sAF (N=25) showed higher activation pattern variability (2.1±1.02 vs 0.5±0.4 22
pattern changes per second, p<0.0001, Figure 5). Analysis of CS activation patterns (Figure 6) 23
10 showed, in sAF compared to nsAF patients, fewer proximal-distal activation sequences
1
(25%±18% vs 61%±36% of wavefronts, p<0.001), fewer unidirectional activations (60%±16% vs 2
86%±15% of wavefronts, p<0.0001) and a shorter longest sequence of consecutive identical 3
wavefronts (19±15.5 vs 74±40.5 consecutive beats, p<0.0001).
4
There were more numerous pivot morphologies (differing in location and/or activation sequence) 5
in sAF compared to nsAF (4.3±1.8 versus 2.1±1.6 different morphologies, p<0.0001), and more 6
frequent pivoting wavefronts with a distal-proximal-distal activation sequence (0.1 (0.03-0.27) 7
versus 0 (0-0.07) per second, p=0.003), whereas proximal-distal-proximal pivots were evenly 8
distributed (0.2 (0.13-0.77) versus 0.33 (0-0.67) per second, p=0.86).
9
Disorganized activity was more prevalent in sAF compared to nsAF: 15/25 (60%) sAF episodes 10
versus 2/25 (8%) nsAF episodes showed predominantly type II activity, p<0.001. All other 11
episodes showed predominantly type I activity and none showed dominant type III. Segments 12
without organized wavefronts (Wells type III) were absent from nsAF and were present in 13/25 13
(52%) sAF episodes, p<0.0001. In sAF, the duration of uninterruptedly disorganized segments 14
represented a median 1.7% (0-1.7%) of the signal.
15
AFCLs were shorter in sAF compared to nsAF for all combinations of site and time of 16
measurement (p<0.05). AFCL averaged 190±33 ms (range 121-254 ms) in sAF and 220±23 ms 17
(range 187-303) in nsAF, p<0.001 (Figure 5). AFCL did not change significantly from seconds 0- 18
5 to seconds 10-25 for all combinations of site and group (p>0.4).
19
The percentage of cumulative isoelectric periods in the proximal CS (bipole furthest from 20
catheter tip) was smaller in sAF compared to nsAF (35%±18% vs 44%±15%, p=0.045). There 21
was no difference between both groups in the distal CS (bipole closest to catheter tip;
22
56%±15% vs 59%±14%, p=0.392). Intra-patient comparison showed shorter isoelectric signal 23
11 duration in the proximal CS compared to the distal CS in both sAF (34.6%±17.5% vs
1
55.7%±14.6%, p<0.0001) and nsAF (44.3%±15.1% vs 59.3%±14.4%, p=0.002).
2 3
Preliminary prediction model
4The derived multivariable logistic regression model’s parameters and coefficients are described 5
in Supplemental tables 2-3.
6
The model predicted sAF with 96% sensitivity and 100% specificity, area under the ROC curve 7
(AUC)=0.98. Ten-fold cross-validation resulted in a mean AUC of 0.93±0.16 (95% confidence 8
interval 0.82-1.03). The model was robust to alternative definitions of sAF ranging from 1 to 15- 9
minute cut-off (see Supplemental Material).
10
For representative examples of nsAF and sAF with corresponding model predictions, see 11
Supplemental Material.
12 13
Automated electrogram analysis
14Results from manual electrogram analysis were reproduced by automated analysis on the 15
whole sample (N=142) and are summarized in Figure 7: patients with post-PVI induced sAF 16
(N=76) showed higher activation pattern variability (4.4±0.9 vs 3.1±1.2 pattern changes/sec, 17
p<0.0001), fewer proximal-distal activations (11±14% vs 31±30%, p<0.0001), more numerous 18
non-uniform wavefronts (81±16% vs 59±28%, p<0.0001), shorter AFCL (208±19 vs 226±20 ms, 19
p<0.0001) and higher SampEn (0.29±0.15 vs 0.15±0.05, p<0.0001) compared to patients with 20
self-terminating nsAF (N=66). The derived multivariable logistic regression model predicted sAF 21
with 88% specificity, 78% sensitivity, area under the ROC curve = 0.86 (Figure 8, Table 2;
22
12 Intercept refers to the constant in the regression model and has no clinical correlate). The 1
parsimonious model is reported in Supplemental results (independent predictors of sAF:
2
SampEn and number of proximal-distal CS activations/30 seconds, AUC = 0.85). Ten-fold 3
cross-validation resulted in a mean AUC of 0.85±0.03 (95% confidence interval 0.78-0.91).
4
Agreement between manual and automated model classification on the derivation subgroup 5
(N=50) was 88%, Cohen’s kappa = 0.76 (p<0.0001). Detailed correlations between manual and 6
automated electrogram annotation are reported in Supplemental figures 10-14. The evolution of 7
electrogram characteristics later during sAF episodes is reported in the Supplemental results.
8
Comparison of CS electrogram characteristics between patients in whom AF terminated during 9
CS defragmentation versus those in whom AF did not terminate is also reported in the 10
Supplemental results.
11 12
Long-term outcome
13Six patients were lost to follow-up prior to the end of the blanking period and were excluded 14
from survival analysis.
15
Among the remaining 243 patients, AF-free survival at 24 months was 82% (200/243 patients).
16
AF recurrence occurred in 20/74 (27%) patients with sustained induced AF post-PVI (sAF), 17
11/66 (17%) patients with self-terminating non-sustained AF post-PVI (nsAF) and 12/103 (12%) 18
patients without induced AF post-PVI (noAF; p for trend = 0.014 by the log-rank test). The rate 19
of AF recurrence increased significantly with post-PVI inducibility phenotype (sAF > nsAF >
20
noAF: HR = 1.56, 95% CI: 1.09-2.23, p=0.016 by Cox regression). AF-free survival did not differ 21
significantly between sAF and nsAF patients (73% vs 83%, respectively, p=0.15 by the log-rank 22
test; HR = 1.70, 95% CI: 0.81-3.54, p=0.16 by Cox regression). AF-free survival was lower in 23
13 sAF patients compared to nsAF + noAF patients (73% vs 86%, respectively, p=0.016 by the log- 1
rank test; HR = 2.06, 95% CI: 1.13-3.75, p=0.018 by Cox regression).
2 3
Discussion
4
This study demonstrates that sAF induced after PVI for PAF exhibits distinct 5
electrophysiological properties in the CS compared to induced nsAF during the initial 30 6
seconds of the induced arrhythmia. The main findings were that sAF showed: increased CS 7
activation sequence variability, fewer proximal-distal CS activations, fewer organized 8
(unidirectional) activation patterns, more spatially distinct pivot morphologies within the CS, 9
lower grade of organization, shorter AFCL and a lesser isoelectric signal duration in the 10
proximal CS. These observations were first evaluated manually on a patient subgroup and 11
subsequently confirmed by automated electrogram analysis on the whole study population and 12
separately on the derivation subgroup as well. Additionally, follow-up data showed an 13
association between post-PVI AF inducibility and AF recurrence at 24 months.
14
Only four previous studies have compared the electrophysiological properties of sAF and nsAF 15
using heterogenous definitions of sAF, ranging from 5-15 minutes. Karch et al.(3), mapped the 16
right atrium (RA) and found longer AFCL and more organized AF electrograms in nsAF versus 17
sAF, predominantly at the septal RA. Three studies focused on FF intervals and reported 18
shorter AFCL in sAF versus nsAF on surface ECG(2) and endocardial RA and CS(4,5). To the 19
best of our knowledge, the present data uniquely describes beat-to-beat CS activation and 20
electrogram characteristics in post-PVI sAF and nsAF.
21 22
Electrogram characteristics indicative of AF substrate
2314 While some electrogram characteristics reported here have not been used previously to predict 1
AF duration or inducibility, their selection was based on the following rationale.
2
Complexity parameters that have been correlated with AF perpetuation or substrate include 3
electrogram fractionation(8,9,11), Wells’ AF types(11), activation pattern variability(9–11), 4
complex activation patterns during AF (including inside the CS)(9–11), as well as automated 5
measures of disorganization such as SampEn(6) and dispersion of dominant frequency 6
distribution(6). Moreover, Kalifa et al.(7) showed that maximal electrogram fractionation and 7
activation pattern variability are found in the immediate vicinity of areas exhibiting maximal 8
dominant frequency, compatible with high-frequency excitation from AF drivers. A shorter AFCL 9
has been associated with electrical remodeling(18) and AF persistence(9,11). Furthermore, 10
Haïssaguerre et al.(19) showed that PVI resulted in AFCL prolongation in the CS, the 11
magnitude of which correlated with AF termination and non-inducibility. Local CS pivots have 12
been attributed to electrically active partial CS-LA connections(15). We hypothesized a 13
correlation between pivot dispersion and CS activation sequence variability, and pivots were 14
analyzed separately to account for locally complex activations. Correlation between the number 15
of distinct pivot morphologies and the number of activation pattern changes was confirmed by 16
linear regression (R=0.63, p<10^-6).
17 18
Anatomic basis of CS substrate
19The CS represents a convenient and fixed, spatially limited structure with extensive connections 20
to the LA and allowing a view of the electrical activity in the adjacent LA. The CS is surrounded 21
by a myocardial sleeve continuous with the ostial RA myocardium and extending 25-50 mm 22
from the ostium(20). Myocardial connections between the CS and LA have been observed in 23
89-100% of human hearts, with the most common location being the CS midportion(20,21). In 24
15 an anatomic study of 65 patients using computed tomography(21), 4 morphologic variants of 1
CS-LA connections were identified: 40% of patients had a single connection in the midportion of 2
the CS, 23% had two connections, 15% had complete CS-LA attachment and 11% had no 3
visible connections.
4
CS-LA connections likely play a significant role in AF perpetuation, as evidenced by reports of 5
local reentry mechanisms(12) and CS to LA conduction of local arrhythmias triggering and 6
perpetuating AF(13). Moreover, partial or total CS-LA disconnection results in AF termination 7
and/or decreased AF inducibility(13,22,23), which is consistent with our results.
8
Haïssaguerre et al.(22) reported an increase in AFCL and a decrease in CS activation pattern 9
variability following partial CS-LA disconnection, suggesting a causal relationship between these 10
electrogram characteristics and the extent of CS-LA connections. Therefore, the results of the 11
present study are compatible with more extensive CS-LA connections in sAF compared to 12
nsAF. Of note, in the above-mentioned study, CS ablation resulted in AF termination in 35% of 13
patients, which is consistent with the present report (32%).
14
Activation pattern variability in the CS during rapid pacing has been shown in vitro to be related 15
to intermittent functional block at CS-LA connections resulting in unstable reentrant circuits 16
involving the CS musculature and randomly using variable CS-LA connections(24). In addition, 17
this type of activity was associated with AF-like ECG activity as well as AF induction. Finally, the 18
presence of functional conduction block at CS-LA connections along with a variable CS 19
activation sequence has been shown in humans to be associated with AF inducibility(25) and 20
with a history of AF(26). These studies, as well as the present data (particularly the variability in 21
CS activation patterns and pivot locations), are compatible with a substantial role of non-fixed 22
CS-dependent reentry through varying CS-LA connections in the induction and maintenance of 23
24 AF.
16 1
Clinical implications
2The present data suggests that post-PVI sAF inducibility relates to distinct electrophysiological 3
properties of extra-PV substrate. Because continuous and rapid fractionated potential ablation in 4
the CS resulted in AF termination/non-inducibility in 49% of sAF patients, a contribution of the 5
CS to AF maintenance is likely in these patients.
6
The prediction model may serve as a multidimensional measure of AF complexity. The reliable 7
prediction of AF self-termination may have clinical applications particularly in the context of the 8
possible relationship of AF burden with AF-related complications. Finally, one may speculate 9
that catheter ablation of AF sustaining mechanisms or sources should result in similar changes 10
of diminishing electrogram complexity which may therefore provide a definable electrogram 11
based ablation endpoint instead of actual arrhythmia termination.
12
The association between post-PVI AF inducibility and long-term AF-free survival indicates that 13
sAF is related to clinically relevant extra-PV AF substrate. In addition, it supports the use of AF 14
inducibility as a prognostic indicator of rhythm outcomes.
15 16
Study limitations
17There is an overlap between the characteristics of near-field and far-field electrograms(15,27);
18
therefore, complete discrimination during AF may not be achievable based on bipolar CS 19
electrograms alone. It is likely that a proportion of far-field potentials from the adjacent LA were 20
inappropriately annotated. In addition, various criteria have been used(28) to define local 21
activation time during AF. Therefore, the reported activation sequences should be interpreted in 22
17 the context of the annotation method used. Of note, alternative definitions of a change of
1
activation pattern were tested and yielded similar results (Supplemental Material).
2
Because this cohort included PAF patients after PVI only, the reproducibility of these results 3
needs to be verified with respect to the broader AF population.
4
Because of the non-randomized stepwise ablation protocol used in the present cohort, the 5
relative contribution of the CS in sAF maintenance cannot be determined from the present data.
6
Further studies investigating these electrogram characteristics at other atrial sites along with the 7
clinical impact of ablation at those sites are needed.
8
The mapping catheter used in this study covers a limited area of tissue with limited resolution.
9
Accordingly, the present electrophysiological findings should be interpreted with caution. The 10
present method, however, identified substantial differences in electrical activity between 11
sustained and self-terminating AF, using readily accessible tools and routine laboratory setup.
12
Implementation of high-density could conceivably add valuable insight but high density mapping 13
during an irregular arrhythmia such as AF is difficult to interpret. Panoramic mapping (body 14
surface ECG mapping) on the other hand has limitations in resolving complex low voltage 15
activation but may merit further study.
16
While CS defragmentation resulted in AF termination or non-inducibility in a substantial 17
proportion of patients, the present study was not designed to assess the utility of such an 18
ablation strategy, which remains to be evaluated prospectively. We acknowledge that the 19
clinical relevance of AF non-inducibility as a procedure outcome in PAF is controversial. While 20
the majority of studies, including the present one, showed an association with arrhythmia-free 21
survival(1), the prognostic accuracy reported in the literature was low and heterogenous 22
definitions of sAF (ranging 1-10 minutes) were used. Therefore, when interpreting the present 23
18 results, one should keep in mind that, although indicative of worse clinical outcome, post-PVI 1
sAF does not equate clinical AF recurrence.
2
Finally, it should be noted that the comparison of outcomes of sAF versus other patients is 3
confounded by the fact that sAF patients received additional electrogram-guided substrate 4
modification (targeting complex fractionated electrograms in the CS as well as other LA sites in 5
case of persistently inducible sAF) whereas nsAF and noAF patients did not. Under the 6
hypothesis that adjunctive extra-PV ablations are useful for long-term sinus rhythm 7
maintenance, this confounding factor is likely to decrease the difference in AF-free survival 8
between sAF and nsAF + noAF patients. Therefore, this observational analysis should be 9
interpreted with caution and may not be representative of the prognostic utility of post-PVI 10
inducibility testing in general.
11 12
Conclusion
13
Induced sAF shows decreased electrogram organization, more variable activation sequences, 14
shorter AFCL and higher SampEn in the CS compared to self-terminating nsAF in PAF patients 15
after PVI. Automated analysis of CS electrograms within the initial 30 seconds of AF episodes 16
accurately distinguished self-terminating from sustained AF.
17 18 19
Acknowledgments
20
We thank Dr Nicolas Vachicouras for his help in writing Matlab scripts and figure editing.
21
19 1
2 3 4
References
5
1. Baker Md M, Kumar Mbbs P, P Hummel Md J, K Gehi Md Fhrs A. Non-Inducibility Or 6
Termination As Endpoints Of Atrial Fibrillation Ablation: What Is The Role? J Atr 7
Fibrillation. 2014;7(3):1125.
8
2. Bollmann A, Sonne K, Esperer H-D, Toepffer I, Langberg JJ, Klein HU. Non-invasive 9
assessment of fibrillatory activity in patients with paroxysmal and persistent atrial fibrillation 10 using the Holter ECG. Cardiovasc Res. 1999;44(1):60–6.
11
3. Karch MR, Ndrepepa G, Zrenner B, Saur C, Schneider MAE, Schomig A, Schmitt CS.
12
Simultaneous multisite endocardial mapping of sustained and non-sustained atrial 13 fibrillation in humans. J Invasive Cardiol. 2003;15(5):257–62.
14
4. Capucci A, Biffi M, Boriani G, Ravelli F, Nollo G, Sabbatani P, Orsi C, Magnani B. Dynamic 15
Electrophysiological Behavior of Human Atria During Paroxysmal Atrial Fibrillation.
16 Circulation. 1995;92(5):1193–202.
17
5. Kumagai K, Yasuda T, Tojo H, Noguchi H, Matsumoto N, Nakashima H, Gondo N, Saku K.
18
Role of Rapid Focal Activation in the Maintenance of Atrial Fibrillation Originating from the 19 Pulmonary Veins. Pacing Clin Electrophysiol. 2000;23(11P2):1823–7.
20
6. Akar JG, Everett TH, Kok L-C, Moorman JR, Haines DE. Effect of electrical and structural 21
remodeling on spatiotemporal organization in acute and persistent atrial fibrillation. J 22 Cardiovasc Electrophysiol. 2002;13(10):1027–34.
23
7. Kalifa J, Tanaka K, Zaitsev AV, Warren M, Vaidyanathan R, Auerbach D, Pandit S, 24
Vikstrom KL, Ploutz-Snyder R, Talkachou A, Atienza F, Guiraudon G, Jalife J, Berenfeld O.
25
Mechanisms of Wave Fractionation at Boundaries of High-Frequency Excitation in the 26
Posterior Left Atrium of the Isolated Sheep Heart During Atrial Fibrillation. Circulation.
27 2006;113(5):626–33.
28
8. Konings KT, Smeets JL, Penn OC, Wellens HJ, Allessie MA. Configuration of unipolar 29
atrial electrograms during electrically induced atrial fibrillation in humans. Circulation.
30
1997;95(5):1231–41.
31
9. Allessie MA, Groot NMS de, Houben RPM, Schotten U, Boersma E, Smeets JL, Crijns HJ.
32
Electropathological Substrate of Long-Standing Persistent Atrial Fibrillation in Patients With 33
20 Structural Heart Disease Longitudinal Dissociation. Circ Arrhythm Electrophysiol.
1 2010;3(6):606–15.
2
10. Groot NMS de, Houben RPM, Smeets JL, Boersma E, Schotten U, Schalij MJ, Crijns H, 3
Allessie MA. Electropathological Substrate of Longstanding Persistent Atrial Fibrillation in 4
Patients With Structural Heart Disease Epicardial Breakthrough. Circulation.
5 2010;122(17):1674–82.
6
11. Gaita F, Calò L, Riccardi R, Garberoglio L, Scaglione M, Licciardello G, Coda L, Di Donna 7
P, Bocchiardo M, Caponi D, Antolini R, Orzan F, Trevi GP. Different patterns of atrial 8
activation in idiopathic atrial fibrillation: simultaneous multisite atrial mapping in patients 9 with paroxysmal and chronic atrial fibrillation. J Am Coll Cardiol. 2001;37(2):534–41.
10
12. Chugh A, Oral H, Good E, Han J, Tamirisa K, Lemola K, Elmouchi D, Tschopp D, Reich S, 11
Igic P, Bogun F, Pelosi F, Morady F. Catheter ablation of atypical atrial flutter and atrial 12
tachycardia within the coronary sinus after left atrial ablation for atrial fibrillation. J Am Coll 13 Cardiol. 2005;46(1):83–91.
14
13. Oral H, Ozaydin M, Chugh A, Scharf C, Tada H, Hall B, Cheung P, Pelosi F, Knight BP, 15
Morady F. Role of the Coronary Sinus in Maintenance of Atrial Fibrillation. J Cardiovasc 16 Electrophysiol. 2003;14(12):1329–36.
17
14. Macle L, Jaïs P, Weerasooriya R, Hocini M, Shah DC, Choi K-J, Scavée C, Raybaud F, 18
Clémenty J, Haïssaguerre M. Irrigated-Tip Catheter Ablation of Pulmonary Veins for 19 Treatment of Atrial Fibrillation. J Cardiovasc Electrophysiol. 2002;13(11):1067–73.
20
15. Pascale P, Shah AJ, Roten L, Whinnett Z, Wilton SB, Jadidi AS, Scherr D, Miyazaki S, 21
Pedersen M, Derval N, Knecht S, Sacher F, Hocini M, Haïssaguerre M, Jaïs P. Disparate 22
Activation of the Coronary Sinus and Inferior Left Atrium During Atrial Tachycardia After 23
Persistent Atrial Fibrillation Ablation: Prevalence, Pitfalls, and Impact on Mapping. J 24
Cardiovasc Electrophysiol. 2012;23(7):697–707.
25
16. Wells JL, Karp RB, Kouchoukos NT, MacLean WA, James TN, Waldo AL. Characterization 26
of atrial fibrillation in man: studies following open heart surgery. Pacing Clin Electrophysiol 27 PACE. 1978;1(4):426–38.
28
17. Calkins H, Hindricks G, Cappato R, Kim Y-H, Saad EB, Aguinaga L, Akar JG, Badhwar V, 29
Brugada J, Camm J, Chen P-S, Chen S-A, Chung MK, Cosedis Nielsen J, Curtis AB, 30
Davies DW, Day JD, d’Avila A, Natasja de Groot NMS, Di Biase L, Duytschaever M, 31
Edgerton JR, Ellenbogen KA, Ellinor PT, Ernst S, Fenelon G, Gerstenfeld EP, Haines DE, 32
Haissaguerre M, Helm RH, Hylek E, Jackman WM, Jalife J, Kalman JM, Kautzner J, 33
Kottkamp H, Kuck KH, Kumagai K, Lee R, Lewalter T, Lindsay BD, Macle L, Mansour M, 34
Marchlinski FE, Michaud GF, Nakagawa H, Natale A, Nattel S, Okumura K, Packer D, 35
Pokushalov E, Reynolds MR, Sanders P, Scanavacca M, Schilling R, Tondo C, Tsao H-M, 36
Verma A, Wilber DJ, Yamane T, Document Reviewers: 2017 37
HRS/EHRA/ECAS/APHRS/SOLAECE expert consensus statement on catheter and 38
surgical ablation of atrial fibrillation. Eur Eur Pacing Arrhythm Card Electrophysiol J Work 39
Groups Card Pacing Arrhythm Card Cell Electrophysiol Eur Soc Cardiol. 2018;20(1):e1–
40
160.
41
21 18. Hobbs WJC, Fynn S, Todd DM, Wolfson P, Galloway M, Garratt CJ. Reversal of Atrial 1
Electrical Remodeling After Cardioversion of Persistent Atrial Fibrillation in Humans.
2
Circulation. 2000;101(10):1145–51.
3
19. Haïssaguerre M, Sanders P, Hocini M, Hsu L-F, Shah DC, Scavée C, Takahashi Y, Rotter 4
M, Pasquié J-L, Garrigue S, Clémenty J, Jaïs P. Changes in Atrial Fibrillation Cycle Length 5
and Inducibility During Catheter Ablation and Their Relation to Outcome. Circulation.
6 2004;109(24):3007–13.
7
20. Chauvin M, Shah DC, Haïssaguerre M, Marcellin L, Brechenmacher C. The Anatomic 8
Basis of Connections Between the Coronary Sinus Musculature and the Left Atrium in 9 Humans. Circulation. 2000;101(6):647–52.
10
21. Saremi F, Thonar B, Sarlaty T, Shmayevich I, Malik S, Smith CW, Krishnan S, Sánchez- 11
Quintana D, Narula N. Posterior Interatrial Muscular Connection between the Coronary 12
Sinus and Left Atrium: Anatomic and Functional Study of the Coronary Sinus with 13 Multidetector CT. Radiology. 2011;260(3):671–9.
14
22. Haïssaguerre M, Hocini M, Takahashi Y, O’neill MD, Pernat A, Sanders P, Jonsson A, 15
Rotter M, Sacher F, Rostock T, Matsuo S, Arantés L, Teng Lim K, Knecht S, Bordachar P, 16
Laborderie J, Jaïs P, Klein G, Clémenty J. Impact of Catheter Ablation of the Coronary 17
Sinus on Paroxysmal or Persistent Atrial Fibrillation. J Cardiovasc Electrophysiol.
18 2007;18(4):378–86.
19
23. Morita H, Zipes DP, Morita ST, Wu J. Isolation of canine coronary sinus musculature from 20
the atria by radiofrequency catheter ablation prevents induction of atrial fibrillation. Circ 21 Arrhythm Electrophysiol. 2014;7(6):1181–8.
22
24. Morita H, Zipes DP, Morita ST, Wu J. The role of coronary sinus musculature in the 23 induction of atrial fibrillation. Heart Rhythm. 2012;9(4):581–9.
24
25. Huang D, Marine JE, Li J-B, Zghaib T, Ipek EG, Sinha S, Spragg DD, Ashikaga H, Berger 25
RD, Calkins H, Nazarian S. Association of Rate-Dependent Conduction Block Between 26
Eccentric Coronary Sinus to Left Atrial Connections With Inducible Atrial Fibrillation and 27
Flutter. Circ Arrhythm Electrophysiol. 2017;10(1).
28
26. Kasai A, Anselme F, Saoudi N. Myocardial connections between left atrial myocardium and 29
coronary sinus musculature in man. J Cardiovasc Electrophysiol. 2001;12(9):981–5.
30
27. Shah D, Haissaguerre M, Jais P, Hocini M, Yamane T, Macle L, Choi KJ, Clementy J. Left 31
Atrial Appendage Activity Masquerading as Pulmonary Vein Potentials. Circulation.
32
2002;105(24):2821–5.
33
28. Cantwell CD, Roney CH, Ng FS, Siggers JH, Sherwin SJ, Peters NS. Techniques for 34
automated local activation time annotation and conduction velocity estimation in cardiac 35
mapping. Comput Biol Med. 2015;65:229–42.
36 37 38
22
Figure Legends
1
Figure 1. Study overview. AF, atrial fibrillation; nsAF, non-sustained AF; OAT, organized atrial 2
tachyarrhythmia; PVI, pulmonary vein isolation; sAF, sustained AF.
3
Figure 2. Manual electrogram annotation. Electrograms not included for analysis were not near 4
field, within refractory period or inconsistent with the activation sequences. Inset: example of 5
manual approximation of dV/dt.
6
Figure 3. (A) Defining activation patterns. First two activations are “CS5 to CS3”, followed by 7
“CS1 to CS4” patterns, based on the earliest and latest electrograms. (B) Local pivots.
8
Wavefronts bifurcate at CS3, with one arm continuing to CS1 and one pivoting back to CS5 9
(proximal-distal-proximal).
10
Figure 4. Modified Wells AF classification. (1) Type 1: discrete electrograms separated by 11
isoelectric signal. (2) Type 2: discrete electrograms with perturbations of the baseline. (3) Type 12
3: continuous electrical activity.
13
Figure 5. Total number of activation pattern changes in sAF and nsAF. Dots represent 14
individual patients; random horizontal scatter for visibility (A). Mean FF interval in sAF and nsAF 15
at seconds 0-5 and 10-25 (B). Brackets show 95% confidence intervals. AF, atrial fibrillation.
16
Figure 6. Total number of wavefronts (percentage) sorted by CS activation pattern in nsAF and 17
sAF. P value and 95% confidence intervals are shown relative to the number of unidirectional 18
activations. AF, atrial fibrillation.
19
Figure 7. CS electrogram characteristics as measured by the automated electrogram 20
annotation algorithm. AF, atrial fibrillation; CS, coronary sinus; nsAF, non-sustained atrial 21
fibrillation; sAF, sustained atrial fibrillation.
22
23 Figure 8. Prediction model accuracy. Receiver-operating characteristic (ROC) curve of the 1
automated prediction model for predicting sAF (whole cohort, N=142).
2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19
24
Figures
1
Figure 1
23 4 5
6 7
25
Figure 2
1 2
3 4 5 6 7 8 9 10
26
Figure 3
1 2
3
27
Figure 4
1
2
28
Figure 5
1
2 3 4 5 6 7 8 9 10 11 12 13
29
Figure 6
1
2 3 4 5 6 7 8 9 10 11
30
Figure 7
1
2 3 4 5
6 7
8 9 10
11 12
13 14
31
Figure 8
1
2 3 4
5 6 7
8 9
32
Tables
1
Table 1. Patient characteristics
All Patients (N=142)
nsAF Group (N=66)
sAF Group (N=76)
P Value
Uni- variable
P Value adjusted
Age (y) 60.2 (±11.3) 57.3 (±10.8) 62.7 (±11.1) 0.004* 0.023*
Male gender 101 (71%) 48 (73%) 53 (70%) 0.69 0.97 Body mass index (kg/m2) 26.4 (±4.7) 26.6 (±5.3) 26.3 (±4.1) 0.75 0.19 History of PAF (mo) 29.5
(12,64.5)
15 (10,48) 45
(18.25,72)
0.035* 0.024*
Medication
Number of failed AADs 1 (1,2) 1 (1,2) 1 (1,1) 0.10 0.081 Concomitant medication†
Amiodarone 21 (15%) 10 (15%) 11 (14%) 0.96 0.22 Beta-blockers 44 (31%) 26 (39%) 18 (24%) 0.06 0.11 ACEI and ARB 28 (20%) 14 (21%) 14 (18%) 0.71 0.56 Comorbidity
Arterial hypertension 45 (32%) 21 (31%) 24 (32%) 0.98 0.81
Diabetes mellitus 6 (4%) 3 (5%) 3 (4%) 1.0 0.39
Hypercholesterolemia 35 (25%) 17 (26%) 18 (24%) 0.81 0.46
33 Structural heart disease 10 (7%) 6 (9%) 4 (5%) 0.51 0.36 Prior stroke/TIA 9 (6%) 1 (2%) 8 (11%) 0.037* 0.29 CHADS2-VASc score ≥ 2 50 (35%) 18 (27%) 32 (42%) 0.07 0.99 Left atrial volume index
(ml/m2)
35.3 (±11.5) 32.2 (±9.3) 37.8 (±12.5) 0.003* 0.016*
Left ventricular function
Normal 138 (97%) 64 (97%) 74 (97%)
1.0 0.81 Slightly reduced 4 (3%) 2 (3%) 2 (3%)
AAD, antiarrhythmic drug; ACEI, angiotensin-converting-enzyme inhibitor; ARB, angiotensin receptor blocker; nsAF, non-sustained AF; PAF, paroxysmal AF; sAF, sustained AF; TIA, transient ischemic attack.
*p<0.05
†Other AADs not shown (discontinued).
1 2 3 4 5 6 7
34
Table 2. Parameters of the final automated model
Parameter Odds ratio Regression
coefficient
P value Number of activation pattern changes (n/30
seconds)
1.016 0.016 0.217
Number of cycles in the longest segment of consecutive activations without change of activation pattern (n)
1.008 0.008 0.866
Number of activations with CS5 to CS1 pattern (n/30 seconds)
0.986 -0.014 0.102
Sample entropy (dimensionless) 1.45*105 11.883 0.001
(Intercept) 0.025 -3.68 0.03
1
