Left atrial deceleration outperforms regional conduction velocity in predicting arrhythmia recurrence following atrial fibrillation ablation

Abstract

Background

The slowest regional conduction velocity (CV_min) is associated with atrial arrhythmia (AA) recurrence following atrial fibrillation (AF) ablation; however, the role of conduction deceleration has not been investigated.

Objective

The study sought to assess whether true deceleration (TD) is a better marker than CV_min in identifying abnormal left atrial (LA) substrate and AA recurrence in patients undergoing de novo pulmonary vein isolation (PVI).

Methods

Eighty AF patients and 6 control subjects underwent LA electroanatomic mapping during atrial pacing. The LA was divided into 6 anatomical regions and the regional low-voltage area (LVA), CV_min, and maximum true deceleration (TD_max) were quantified. TD was calculated as the largest continuous decline in CV along the propagating wavefront divided by the change in activation time. AF patients underwent PVI and AA recurrence was assessed during 12-month follow-up.

Results

A median of 1 to 2 TDs were found in each LA region of AF patients, and the TD_max only weakly correlated with the regional CV_min. AF patients with AA recurrence had a significantly larger LVA, lower CV_min, and greater TD_max on the anterior wall. Multivariate modeling demonstrated that the TD_max (when >110 m/s²) and not the CV_min (when <0.2 m/s) predicted AA recurrence (C-statistic = 0.74). Clinical TD sites (defined as TD_max >110 m/s²) only colocalized with LVA sites in a minority of LA regions (range 13%–44%) and were absent from control subjects.

Conclusion

TD is a novel metric for quantifying LA remodeling and predicting AA recurrence post-PVI that outperforms CV_min. This may guide future trials focusing on improving success from substrate-based AF ablation.

Graphical abstract

Key Findings.

• ▪True deceleration (the change in conduction velocity with respect to time) is a novel method of assessing atrial electrophysiology.
• ▪Patients undergoing pulmonary vein isolation had at least 1 left atrial deceleration zone, which correlates weakly to either slowest regional conduction velocity in the region or low-volage areas.
• ▪Multivariate modeling identified deceleration sites >110 m/s² as the main predictor for atrial arrhythmia recurrence postablation.

Introduction

Pulmonary vein isolation (PVI) has evolved into a safe and effective treatment for atrial fibrillation (AF); however, arrhythmia recurrence afterward is not infrequent, and the optimal invasive strategy in the setting of isolated veins continues to pose a significant challenge.¹ One major source of recurrence is abnormal atrial structural remodeling characterized by progressive atrial fibrosis, in which nonuniform conduction predisposes to localized block and re-entry.² This atrial myopathy can be recognized during an electrophysiology study by a combination of low voltage areas (LVAs), a shorter atrial effective refractory period, and a slowed conduction velocity (CV). While voltage is the simplest variable to rapidly acquire using conventional mapping systems, its interpretation using a cutoff in bipolar voltage of 0.5 mV is recognized to have a number of pitfalls in accurately identifying fibrosis, particularly in the early stages of structural remodeling.³ To improve substrate characterization, regional CVs have been evaluated with the premise that marked conduction slowing identifies functionally abnormal arrhythmogenic substrate. However, there is a wide range in the methodology used to measure CV,⁴ and as a result, the reported cutpoints vary widely from <0.27 to <0.89 m/s for predicting AF recurrence.^5,6 More recently, isochronal late activation mapping (ILAM) has been employed to visually represent CV, in which “deceleration zones” (DZs) (identified as >3 isochrones across a narrow 1-cm radius) are the threshold for significant CV slowing. However, this does not capture true deceleration (TD), as it is evaluated over a fixed time period (in m/s) rather than time-dependent changes in CV (in m/s²).

We hypothesized that TD may be more sensitive than slow local CV in identifying abnormal atrial substrate, thereby prognosticating freedom from arrhythmia after PVI. Our objective was to evaluate TD in the left atrium (LA), including its relationship to LVAs and determine whether TD was more predictive of atrial arrhythmia (AA) recurrence compared with slow regional CV.

Methods

Study population

Consecutive patients with drug-refractory paroxysmal or persistent AF undergoing first-time PVI were prospectively enrolled. Patients were excluded if they had prior catheter ablation, iatrogenic lines of conduction block (atrial or valvular surgery), atrial septal defect, or severe valvular abnormalities. In addition, a control group with no history of AF was prospectively enrolled. These patients had an indication for LA mapping for supraventricular tachycardia. Control subjects were excluded if they had structural heart disease evidenced by echocardiography. The study was approved by our institutional research ethics boards, and all patients provided written, informed consent.

LA mapping

Among AF patients and control subjects, antiarrhythmic drugs were held for 5 half-lives with the exception of amiodarone, which was discontinued 1 month before ablation. In the postabsorptive state, those patients in AF were electrically cardioverted to sinus rhythm. Access to the LA was obtained via the transseptal approach, and systemic anticoagulation was achieved with intravenous heparin to maintain an activated clotting time over 300 seconds. High-resolution electroanatomic mapping (fill threshold 4 mm) was performed in the LA using a 20-electrode, 5-spline catheter (Pentaray; Biosense Webster) and electroanatomic mapping system (CARTO 3 V7; Biosense Webster) during high right atrial (RA) pacing at 750 ms. Mapping points >4 mm from the endocardium, ectopic beats, and noisy beats were excluded from analysis. Mapping points arising from the appendage, mitral annulus, and pulmonary veins were also excluded. Bipolar electrograms (EGMs), recorded with a filter setting of 30–500 Hz and a sampling rate 1000 Hz, were used to construct bipolar voltage and activation maps of the LA from the remaining points.

Analysis protocol for slowest CV, TV and LVAs

There are multiple methods described for the calculation of atrial CV (see Coveney and colleagues).⁴ Fundamentally, these are a balance of a lengthy analysis of complex anisotropic propagation over a 3-dimensional structure with multiple wavefronts against a measurement that can be acquired in real-time but oversimplify conduction into a 2-dimensional linear construct. Given that our aim was clinical in nature (whether TD could predict AA recurrence), a real-time assessment was favored if it could be translated clinically into guiding an ablation strategy. As such, we prespecified the analysis protocol outlined subsequently given its simplicity and reproducibility.

The LA was divided into 6 anatomically distinct regions: anterior wall, posterior wall, inferior wall, roof, septum, and lateral wall. The septum and lateral wall were not considered because their activation is often complex, with multiple wave breakouts from the RA and wave collision, respectively.⁷ For each remaining LA region, 2 different CVs were computed along a propagating vector traveling perpendicular to the greatest number of 10-ms isochrones over the shortest distance across the respective region. This represented the slowest regional propagating vector and was independently verified by a second observer. The first CV computed was that of the propagating vector from the beginning to end of the respective LA region (the regional CV). Local activation times (LATs) of the propagating vector were defined by the steepest negative slope (–dV/dt_max) of the local unipolar EGM within the bounds of the local bipolar EGM (Figure 1A). The second CV computed was the minimum CV (CV_min) within the propagating vector based on local CVs of the constituent 10-ms isochrones (Figure 1B).

Figure 1.

Defining regional slowest conduction velocity (CV_min) and true deceleration (TD). A: An activation map of the left atrium (LA) for 1 atrial fibrillation (AF) patient is shown during high right atrial (RA) pacing at 750 ms. The vector (white dashed line) of the propagating wavefront is first determined for each LA region by dividing the map into 5-ms isochrones. Next, the regional CV is calculated traveling parallel with the vector and perpendicular to the greatest number of isochrones over the shortest distance. Examples of the conduction velocities of the anterior wall (left, vector AB) and posterior wall (right, vector CD) are highlighted. The regional CVs are 0.78 and 1.1 m/s, respectively. B: Schematic activation maps of a single LA region of varying length (in millimeters), divided into 6 colored 10-ms isochrones parallel to the propagating wavefront. The difference between the regional CV, CV_min, and TD are illustrated. The region where TD occurs is highlighted by the black arrows. i: CV is slow and constant, so there is no TD. ii: Conduction slowing is evident at the third isochrone and occurs over 1 isochrone. CV_min is 0.4 m/s, while TD is 60 m/s². iii: Conduction slowing is evident at the third isochrone and occurs over 2 isochrones. CV_min is 0.2 m/s, while TD is 40 m/s². (iv) Conduction slowing is evident at the third isochrone and occurs over 3 isochrones. CV_min is 0.1 m/s, while TD is 30 m/s². EGM = electrogram.

TD was evaluated only along the propagating vector used to define the CV_min because this region was most likely to demonstrate deceleration. TD occurred only when there was a decrease in CV from one 10-ms isochrone to the next. The magnitude of TD was defined as the change in CV across consecutive isochrones demonstrating decreasing CV over the time interval represented by these isochrones as shown in Figure 1B. If CV increased transiently, but then slowed down again, a second TD was computed for the same region in a similar manner as the first TD, and so on. The number of TDs and the maximum TD (TD_max) were computed for each LA region.

LVA was quantified based on the total surface area of all manually planimetered regions with at least 3 closely spaced points having peak-to-peak bipolar voltage <0.5 mV, expressed as a percentage of the respective LA region’s surface area.

Catheter ablation and clinical follow-up

Following electroanatomic mapping, circumferential PVI ablation was performed with an irrigated, contact force–sensing ablation catheter (SMARTTOUCH ST-SF; Biosense Webster) using radiofrequency energy of 25 W (ablation index 400) on the posterior wall and 35 W (ablation index 500) on the anterior wall. Additional ablation in the LA or RA was not performed. The procedural endpoint was entrance and exit conduction block into all pulmonary veins. Antiarrhythmic drugs were discontinued after ablation. Patients were followed for 1 year with a quarterly 48-hour ambulatory rhythm recorder, and additional recordings were arranged for symptoms suggestive of AA recurrence. The clinical endpoint was defined as AA recurrence >30 seconds, including AF, atrial flutter (AFL), and atrial tachycardia (AT), by 1-year postablation on or off antiarrhythmic drugs, excluding a 3-month postablation blanking period.

In silico modeling to demonstrate conduction block at TD sites

To identify how regions of TD resulted in propagation block, a single cable was simulated with unidirectional activation. In brief, the cable was divided into 4 regions: normal conduction, fast conduction, slow conduction, and ending in normal conduction, in which the transition from fast to slow simulated TD. Fast and slow regions were created by altering the relative sodium conductance and relative cable conductivity as discrete variables, in which 1 was normal, values <1 resulted in reduced conductance producing conduction slowing, and values >1 resulted in enhanced conductance producing faster conduction. Propagation failure was determined by calculating the safety factor (the amount of charge delivered to a cell that exceeds the minimum necessary to sustain action potential propagation)⁸ for each region, as well as the interface between regions, in which values <1 represented propagation failure. A detailed overview of the model can be found in Supplemental Methods.

Statistical analysis

Continuous data are presented as mean ± SD or the median (interquartile range [IQR]) for normal and non-normal datasets, respectively. Differences between groups (paroxysmal vs persistent AF; recurrence vs no AA recurrence) were assessed using Wilcoxon signed rank test, while differences between LA regions were assessed using Friedman’s test. Categorical variables are presented as frequency or percentage and were compared by chi-square or Fisher's exact test where appropriate. Correlation was performed using the Pearson or Spearman test, as appropriate. Univariable and multivariable Cox regression analysis was used to assess the predictive value of CV_min, TD_max, and other established covariates for AA recurrence. Regression results are presented as the hazard ratio (HR) and 95% confidence interval (CI). Multivariable models included covariates that predicted AA recurrence with a univariable significance level of P < .1 and were adjusted for age and sex. Model discrimination was assessed using Harrell’s C-statistic. All assumptions of the Cox proportional hazards regression model were verified, and multicollinearity was not observed between any of the potential predictor variables (variance inflation factor <3). Reproducibility was assessed with intraclass correlation coefficient. A 2-tailed P < .05 was considered statistically significant. All statistical analyses were performed using MATLAB (version 8.0; The MathWorks) and SPSS (version 20.0; IBM).

Results

Patient characteristics

Eighty consecutive patients (age 59 ± 11 years, 79% male) with AF (71% paroxysmal and 29% persistent) were prospectively enrolled. Their baseline clinical characteristics are summarized in Table 1. In addition, 6 control subjects were prospectively enrolled (age 34 ± 12, 67% male) who required LA mapping and ablation for supraventricular tachycardia (5 left-sided accessory pathways, 1 focal atrial tachycardia).

Table 1.

Patient characteristics (N = 80)

Age, y	59 ± 11
Male	63 (79)
BMI, kg/m2	30 ± 5
AF type
Paroxysmal	57 (71)
Persistent	23 (29)
AF duration, y	2.5 (1.0–5.0)
CHA2DS2-VASc score	1.0 (1.0–2.0)
LA volume, mL/m2	40 ± 14
RA size, cm2	20 ± 5
LVEF, %	57 ± 9
Cardiomyopathy
Ischemic	1 (1.3)
Nonischemic	9 (11.3)
Hypertrophic	4 (5.0)
MI/CAD	5 (6.3)
Comorbidities
CVA/TIA	4 (5.0)
Hypertension	21 (26)
Diabetes	5 (6.3)
Sleep apnea	36 (45)
Obesity	39 (49)
Renal dysfunction	3 (3.8)
Antiarrhythmic drugs
Flecainide/propafenone	28 (35)
Sotalol	4 (5.0)
Amiodarone	22 (28)

Values are mean ± SD, n (%), or median (interquartile range). Obesity is classified as BMI >30 kg/m²; renal dysfunction is classified as creatinine clearance <50 mL/min/m².

AF = atrial fibrillation; BMI = body mass index; CAD = coronary artery disease; CHA₂DS₂-VASc = congestive heart failure, hypertension, age ≥75 years, diabetes mellitus, prior stroke or transient ischemic attack or thromboembolism, vascular disease, age 65–74 years, sex category; CVA = cerebrovascular accident; LA = left atrial; LVEF = left ventricular ejection fraction; MI = myocardial infarction; TIA = transient ischemic attack.

CV_min and TD_max in control patients and AF patients

Among AF patients and control patients, the total number of LA points after editing was 1261 ± 403 (13 ± 4 points/cm²) and 786 ± 215 (12 ± 3 points/cm²), respectively. We illustrate our definition of the CV_min and TD_max in a representative LA region for 1 control patient and 3 AF patients in Figure 2. For each patient, a propagating vector is identified with the slowest CV across the entire wall. In the control patient, the anterior wall had no LVA and the CV_min in the propagating vector is 0.76 m/s. There is no CV slowing across the 10-ms isochrones and therefore no TD. In AF patient 1, the roof has a LVA of 4% and the CV_min in the propagating vector is 0.1 m/s. As with the control patient, there is no CV slowing across the 10-ms isochrone and therefore no TD. By contrast, AF patient 2 has 0.88% LVA on the inferior wall and the CV_min in the propagating vector is 0.17 m/s. This CV is the result of CV slowing across 2 isochrones, which represents a TD. Because there is only 1 TD, the TD_max in this segment is 60 m/s². In AF patient 3, the anterior wall has no LVA and the CV_min in the propagating vector is 0.4 m/s. CV slows in 2 regions, each representing a TD. Notably, CV increases between the first TD and second TD. The second TD has the TD_max for the anterior segment of 50 m/s².

Figure 2.

Slowest regional conduction velocity (CV_min) and maximum true deceleration (TD_max) in control and atrial fibrillation (AF) patients. One control patient and 3 AF patients are presented column wise. (Top) Activation maps of the left atrium (LA) during high right atrial (RA) pacing at 750 ms (10-ms isochrones). The white dotted arrow indicates the propagating vector with the slowest CV across the highlighted LA region (control: anterior; AF1: roof; AF2: inferior; AF3: anterior wall). The propagating vector is depicted below each activation map with corresponding isochronal colors, and their widths are measured to compute CV per isochrone. (Bottom) Plots of CV for each isochrone. The control and AF1 patients both have no TD. The AF2 patient has 1 TD over 2 isochrones (TD_max 60 m/s², CV_min 0.17 m/s). The AF3 patient has 2 TD, the first over one isochrone and the second over 2 isochrones (TD_max 50 m/s², CV_min 0.39 m/s).

The reproducibility of TD_max measurement was assessed in 10 randomly selected patients, in which the repeat analysis was performed blinded. The intraclass correlation coefficient was 99.7% (95% CI 98.8%–99.9%, P < .001), demonstrating that TD_max measurement was consistent and reliable.

Among all AF patients and control patients, a summary of the LVA, CV_min, and TD_max for each LA region are provided in Table 2. For AF patients, the overall LVA burden was low (3.7% [IQR 1.0%–7.3%]). The median LVA was greatest in the anterior wall at 1.3% (IQR 0%–6.4%) and least in the posterior wall at 0% (IQR 0%–1.2%) (P < .001). This corresponded to the anterior wall having the lowest CV_min at 0.19 (IQR 0.11–0.31) m/s and the posterior wall having the highest CV_min at 0.54 (IQR 0.36–0.70) m/s (P < .001). Among all 4 LA regions (anterior, roof, posterior, inferior wall) in AF patients, the prevalence of CV_min <0.5, 0.4, 0.3, 0.2, and 0.1 m/s was 98%, 98%, 91%, 76%, and 51%, respectively. The median number of TDs was 1 or 2 in each of these LA region, with the inferior wall having the highest TD_max at 68 (IQR 39–120) m/s² and the posterior wall having the lowest TD_max at 41 (IQR 25–81) m/s² (P < .001). Notably, there was weak correlation between CV_min and TD_max on the anterior wall (r = –0.32, P = .005), roof (r = –0.15, P = .2), posterior wall (r = –0.40, P < .001), and inferior wall (r = –0.28, P = .015) in AF patients.

Table 2.

LVA, CV, and TD by LA region

Patient group	LA region	LVA (%)	Regional CV along slow propagating vector (m/s)	CVmin within propagating vector (m/s)	No. of TD within propagating vector	TDmax within propagating vector (m/s2)
AF (n = 80)	Anterior	1.3 (0–6.4)	0.67 ± 0.20	0.19 (0.11–0.31)	2 (1–2)	63 (41–90)
	Roof	0 (0–1.8)	0.67 ± 0.23	0.37 (0.14–0.58)	1 (1–1)	58 (27–86)
	Posterior	0 (0–1.2)	0.81 ± 0.24	0.54 (0.36–0.70)	1 (1–2)	41 (25–81)
	Inferior	0.89 (0–4.3)	0.79 ± 0.20	0.36 (0.17–0.52)	2 (1–2)	68 (39–120)
Control (n = 6)	Anterior	0 (0–0.20)	0.82 ± 0.18	0.38 (0.23–0.49)	1 (1–2)	40 (30–42)
	Roof	0	0.70 ± 0.18	0.39 (0.26–0.59)	1 (1–1)	49 (44–70)
	Posterior	0	0.99 ± 0.19	0.69 (0.63–1.1)	1 (1–1)	71 (48–82)
	Inferior	0.05 (0–0.56)	0.95 ± 0.48	0.36 (0.090–0.38)	1 (1–2)	42 (29–88)

Values are mean ± SD or median (interquartile range).

CV = conduction velocity; CV_min = slowest regional conduction velocity; LA = left atrial; LVA = low-voltage area; TD = true deceleration; TD_max = maximum true deceleration.

To investigate whether LVA, CV_min, and TD_max were influenced by the classification of AF, we compared patients with paroxysmal vs persistent AF for each LA region as shown in Supplemental Table 1. Patients with persistent AF had a greater LVA on the posterior (P = .030) and inferior (P = .014) walls, as well as a higher TD_max on the anterior wall (P = .028) compared with patients with paroxysmal AF. No difference in CV_min was observed.

Relationship of LVA, CV_min, and TD_max to AA recurrence

After 12-month follow-up, 26 (33%) AF patients had AA recurrence post-PVI. Their clinical characteristics are shown in Supplemental Table 2. There was a trend toward persistent AF being more prevalent in patients with AA recurrence compared with those without AA recurrence (42% vs 24%, P = .074), but other clinical characteristics were similar between these groups.

For each LA region, LVA, CV_min, and TD_max were compared between AF patients with and without AA recurrence as shown in Table 3. In the anterior wall, patients with AA recurrence had greater LVA (P = .039), lower CV_min (P = .035), and higher TD_max (P = .028) than those without AA recurrence. In contrast, the roof, posterior wall, and inferior wall showed no difference in CV_min or TD_max between groups.

Table 3.

Relationship between electroanatomic features and AA recurrence

LA region	Electroanatomic metric	No AA recurrence (n = 54)	AA recurrence (n = 26)	P value
Anterior	LVA%	1.0 (0–4.7)	3.0 (0.4–30)	.039
	CVmin, m/s	0.22 (0.13–0.37)	0.14 (0.07–0.27)	.035
	TDmax, m/s2	50 (30–70)	80 (50–100)	.028
Roof	LVA%	0 (0–0.42)	0 (0–19)	.103
	CVmin, m/s	0.36 (0.14–0.59)	0.38 (0.13–0.52)	.696
	TDmax, m/s2	60 (20–90)	50 (30–70)	.722
Posterior	LVA%	0 (0–1)	0 (0–4)	.240
	CVmin, m/s	0.57 (0.39–0.70)	0.42 (0.21–0.67)	.120
	TDmax, m/s2	40 (20–90)	40 (30–80)	.642
Inferior	LVA%	0.25 (0–3.6)	2.4 (0–12)	.048
	CVmin, m/s	0.40 (0.18–0.57)	0.24 (0.11–0.47)	.066
	TDmax, m/s2	80 (50–110)	60 (30–130)	.503

Values are median (interquartile range).

AA = atrial arrhythmia; CV_min = slowest regional conduction velocity; LVA = low-voltage area; TD_max = maximum true deceleration.

The CV and TD metrics, as well as established clinical characteristics (age, sex, persistent AF, LA volume index, and LVA) were evaluated with Cox regression analysis for prediction of AA recurrence following PVI. The CV_min and TD_max across the 4 LA regions (anterior wall, roof, posterior wall, inferior wall) were evaluated as dichotomous predictors. The TD_max was dichotomized based on its median value of 110 m/s² among the 4 LA regions, while the CV_min variable was dichotomized at 0.2 m/s because CVs below this level have previously been associated with both the presence of abnormal atrial substrate⁹ and AF recurrence after PVI.¹⁰ Notably, none of the control subjects had a TD_max >110 m/s² or CV_min <0.2 m/s.

Three sex- and age-adjusted multivariable models were evaluated: (1) LVA without CV or TD, (2) LVA with CV_min <0.2 m/s, and (3) LVA with TD_max >110 m/s². This was performed for any type of AA recurrence (n = 26), AF as the recurrent arrhythmia (n = 17), and finally, AFL or AT as the recurrent arrhythmia (n = 9). The results in full are shown in Table 4. The main findings are, first, when any type of AA recurrence was considered, model 3 achieved the highest C-statistic of all the models (0.75), with both LVA (HR 1.26, 95% CI 1.09–1.46, P = .002) and TD_max >110 m/s² (HR 4.64, 95% CI 1.66–13.0, P = .003) being significant predictors of recurrence. Second, when only AF recurrence was considered, model 3 achieved the highest C-statistic (0.67) with only TD_max >110 m/s² (HR 3.44, 95% CI 1.1–10.7, P = .033) being the significant predictor. Finally, when only AFL or AT recurrence was considered, model 3 achieved the highest C-statistic (0.84); however, only LVA (HR 1.5, 95% CI 1.09–2.08, P = .013), and not TD_max >110 m/s² (HR 18.1, 95% CI 0.96–342, P = .053), was predictive.

Table 4.

Cox regression analysis for prediction of AA recurrence post-PVI

		Univariable analysis		Model 1: LVA only		Model 2: LVA + CVmin		Model 3: LVA + TDmax
		HR (95% CI)	P	HR (95% CI)	P	HR (95% CI)	P	HR (95% CI)	P
All AA recurrence (26 events)	Age, per 5 y	1.15 (0.95–1.40)	.146	1.03 (0.83–1.28)	.802	0.98 (0.78–1.24)	.865	0.98 (0.78–1.24)	.875
	Female	1.65 (0.71–3.79)	.242	1.14 (0.48–2.73)	.763	1.20 (0.49–2.92)	.687	1.47 (0.60–3.60)	.393
	Persistent AF	1.54 (0.70–3.38)	.282	—	—	—	—	—	—
	LAVI, per 5 mL/m2	1.08 (0.95–1.23)	.230	—	—	—	—	—	—
	%LVA (<0.5 mV), per 5%	1.17 (1.06–1.28)	.001	1.15 (1.02–1.3)	.024	1.15 (1.01–1.3)	.035	1.26 (1.09–1.46)	.002
	CVmin <0.2 m/s	2.81 (0.84–9.38)	.093	—	—	2.34 (0.65–8.4)	.192	—	—
	TDmax >110 m/s2	2.46 (1.06–5.70)	.036	—	—	—	—	4.64 (1.66–13.0)	.003
	Harrel’s C-statistic	-		0.64		0.66		0.75
AF recurrence (17 events)	Age, per 5 y	1.11 (0.89–1.41)	.354	1.03 (0.79–1.34)	.812	0.98 (0.73–1.29)	.852	1.01 (0.77–1.32)	.958
	Female	2.10 (0.78–5.67)	.145	1.62 (0.56–4.69)	.375	1.83 (0.60–5.57)	.285	2.10 (0.70–6.26)	.185
	Persistent AF	1.18 (0.44–3.21)	.740	—	—	—	—	—	—
	LAVI, per 5 mL/m2	1.06 (0.91–1.24)	.453	—	—	—	—	—	—
	%LVA (<0.5 mV), per 5%	1.14 (1.00–1.29)	.053	1.10 (0.93–1.31)	.265	1.09 (0.91–1.30)	.365	1.17 (0.97–1.42)	.105
	CVmin <0.2 m/s	2.77 (0.63–12.1)	.177	—	—	2.71 (0.56–13.2)	.216	—	—
	TDmax >110 m/s2	2.13 (0.79–5.77)	.136	—	—	—	—	3.44 (1.10–10.7)	.033
	Harrel’s C-statistic	—		0.59		0.65		0.67
AFL/AT recurrence (9 events)	Age, per 5 y	1.24 (0.87–1.77)	.228	1.02 (0.68–1.52)	.926	0.98 (0.64–1.48)	.914	0.90 (0.57–1.41)	.646
	Female	0.98 (0.20–4.77)	.983	0.62 (0.13–3.08)	.560	0.61 (0.12–3.06)	.544	0.71 (0.13–3.78)	.687
	Persistent AF	2.51 (0.66–9.61)	.178	—	—	—	—	—	—
	LAVI, per 5 mL/m2	1.11 (0.90–1.38)	.307	—	—	—	—	—	—
	%LVA (<0.5 mV), per 5%	1.22 (1.06–1.40)	.007	1.22 (1.01–1.47)	.043	1.22 (1.01–1.47)	.047	1.5 (1.09–2.08)	.013
	CVmin <0.2 m/s	2.91 (0.36–23.3)	.315	—	—	2.09 (0.23–19.1)	.513	—	—
	TDmax >110 m/s2	3.43 (0.69–17.0)	.132	—	—	—	—	18.1 (0.96–342)	.053
	Harrel’s C-statistic	—		0.83		0.75		0.84

AA = atrial arrhythmia; AFL = atrial flutter; AF = atrial fibrillation; AT = atrial tachycardia; CI = confidence interval; CV = conduction velocity; CV_min = slowest regional conduction velocity; HR = hazard ratio; LA = left atrial volume index; LVA = low-voltage area; TD_max = maximum true deceleration.

Relationship of low bipolar voltage to clinical TD in AF patients

Based on model 3 and the control dataset, we defined a clinically significant TD_max as >110 m/s². A total of 51 clinical TDs were identified in 40 AF patients, with the inferior wall having the largest proportion of clinical TD sites (n = 25 of 51 [49%]). Of the clinical TDs, 17 (33%) of 51 were the result of a deceleration to a CV_min of <0.2 m/s and therefore would result in conduction block. The remaining 34 (67%) of 51 resulted in a deceleration to >0.2 m/s; however, they were characterized by an initial acceleration to supraphysiological values (ie, >1.0 m/s). The median CVs producing clinical TD were from 1.85 m/s to 0.5 m/s over a single 10-ms isochrone.

Among all AF patients with clinical TD, we evaluated bipolar voltage at clinical TD sites and the results are summarized in Supplemental Table 3. The median bipolar voltage at clinical TD sites ranged from 1.9 (IQR 1.1–2.3) mV at the roof to 2.3 (IQR 1.7–3.0) mV on the posterior wall (P = .286). The proportion of clinical TD sites with bipolar voltage <0.5 mV ranged from 13% on the posterior wall to 44% on the anterior wall, and the proportion of clinical TD sites with bipolar voltage <1.0 mV ranged from 38% on the posterior wall to 67% on the anterior wall. Thus, less than half of clinical TD sites were associated with LVA <0.5 mV, while the concordance increased 1.5-fold (anterior wall) to 3-fold (roof and posterior wall) with LVA <1.0 mV. Two patient-level examples are shown in Figure 3.

Figure 3.

Relationship of bipolar voltage to clinical true deceleration (TD) in atrial fibrillation (AF) patients. A, B: Two examples of patients in the AF cohort with clinical TD (maximum TD [TD_max] >110 m/s²), demonstrating the relationship between TD and bipolar voltage. i: An isochronal left atrial (LA) activation map with the corresponding dimension of each isochrone in mm along the propagating vector (A→B) on the anterior wall. ii: A graphical representation of conduction velocity (CV) against time separated into 10-ms isochrones. Areas demonstrating TD are highlighted. Clinical TD sites are characterized by sudden acceleration over 1 isochrone followed by rapid deceleration over a second isochrone. iii: The relationship between TD sites and bipolar voltage, using <0.5 mV cutoffs. iv: The relationship between TD sites and bipolar voltage, using a higher <1-mV cutoff. In both patient examples, isochrones exhibiting acceleration and clinical TD occur in a low voltage area defined by <1 mV but not by <0.5 mV. BI = bipolar voltage; CV_min = slowest regional conduction velocity; LAT = local activation time.

In silico modeling of conduction block at TD sites

Using the simple cable model, 900 simulations were run, with varying sodium conductance and cable conductivity. Of these, 60 (7%) were excluded as they failed to produce a normal-fast-slow-normal conduction pattern reflective of TD. Propagation failure (safety factor <1) occurred in 266 (32%) of 840 cables, of which 122 (46%) occurred in the slow region and 144 (54%) occurred at the slow-normal interface. The primary determinant of propagation failure in the slow region was a 90% reduction in sodium conductance (Na_s = 0.1) and at the slow-normal interface was a >80% reduction in cable conductivity (G_s = 0.1 or 0.2). An example of the effect of low sodium conductance and low cable conductivity on propagation is shown in Figure 4.

Figure 4.

Cable model of propagation failure at true deceleration (TD) sites. A: An overview of the simulated cable demonstrating the 4 regions and the interface between them, which were used to determine the site of propagation failure. Normal regions are represented by a relative sodium conductance (Na_s) of 1 and a relative cable conductivity (G_s) of 1. B: The fast region has a Na_s of 1.4 (40% increase above normal) and a G_s of 1.5. The slow region has a fixed G_s of 0.2 (80% reduction below normal) and the Na_s is varied from 0.1 to 1.0. C: The fast region has identical parameters; however, Na_s is fixed at 0.2 and G_s is varied from 0.1 to 1.0. In both panels (B and C), propagation failure (block) occurs when the safety factor is <1.

Discussion

The key findings from this study are the following:

• 1.TD is a novel metric to estimate the change in CV over time along the propagating vector in the LA during fixed rate pacing.
• 2.In patients with AF, a median of 1 to 2 TDs were found in each LA region, and the TD_max only weakly correlated with the regional CV_min.
• 3.Patients with AA recurrence post-PVI had a significantly larger LVA, lower CV_min, and greater TD_max on the anterior wall.
• 4.Multivariate modeling demonstrated that clinical TD (a TD_max >110 m/s²) and not the CV_min (when <0.2 m/s) predicted any AA recurrence with a C-statistic of 0.74. On subanalysis, a TD_max >110 m/s² was the main predictor of AF as the arrhythmia at recurrence (C-statistic = 0.67) and LVA was the main predictor of AFL/AT (C-statistic = 0.84).
• 5.Clinical TD sites only colocalized with LVA sites in a minority of LA segments (range 13%–44%) and were absent from the control population, suggesting that TD is a more sensitive marker for early atrial structural remodeling.
• 6.In silico modeling suggests that TDs are sites with significant reductions in sodium conductance, cell-cell conductivity, or both that lead to propagation failure during abrupt transitions in CV.

TD: A novel marker for atrial remodeling

The dominant paradigm that forms our current understanding of the pathophysiological basis of AF is based on the original multiple wavelet hypothesis,¹¹ in which triggers can induce fibrillation in a vulnerable atrial substrate. The traditional markers for quantifying structural remodeling clinically are LVA and CV, and this study demonstrates the additional benefit of assessing TD. TD is a novel marker for quantifying remodeling that looks at the change in CV over time rather than the local CV at discrete locations in the atria. Our methodology of quantifying TD evaluated the maximum deceleration along the slowest propagating vector for each LA region, highlighting the most vulnerable area within each region expressed as TD_max.. Clinical TD sites (TD_max >110 m/s²) correspond to rapid deceleration, in which a greater TD_max value represents greater deceleration and likely a greater amount of pathological substrate present that can support AF.

Clinical TD sites colocalized to LVA in a minority of cases, meaning that it is a novel marker for detecting significant regions of atrial structural remodeling that are not readily apparent on bipolar voltage maps. This may help to explain why ablation of LVA in routine clinical practice has not shown long-term benefit,^12,13 as LVA fails to accurately reflect the extent of the underlying atrial substrate.

Clinical TD sites are nonuniform across the LA, with only those on the anterior wall being associated with AA recurrence. Structural remodeling in AF is recognized to be heterogeneous in nature, and our data support functional changes on the anterior wall as being associated with a greater degree of remodeling, tendency to persistent AF, and recurrence post-PVI.^5,6,14,15

TD outperforms CV for predicting recurrence

The role of LA CV assessment has been extensively studied previously, in which it has been reported that either a slowed regional or global CV is associated with AF recurrence post-PVI.^6,14,16 The cutpoints of CV which best predict recurrence vary dependent on the methodology used; however, slower CVs within individual studies are associated with more advanced substrate, either in the form of LVA¹⁷ or late gadolinium enhancement on magnetic resonance imaging.¹⁸

Atrial functional substrate mapping in the form of ILAM has seen increasing interest recently. It is important to stress TD_max cannot be identified by ILAM mapping alone, as the “DZ” (>3 isochrones within a 1-cm radius) is constrained by a fixed time period and does not measure the change in CV with respect to time. The use of ILAM in the atrium has mainly been focused on single macro–re-entrant circuits, rather than on AF, in which the DZ in sinus or a paced rhythm can be used to colocalize the location of the critical isthmus.^19,20 One retrospective study by Kuo and colleagues¹⁰ found DZ were able to predict AF recurrence post-PVI, and a more stringent version of DZ (≥4 isochrones in 1 cm) was a better predictor. However, ILAM mapping (dependent on the definition of the DZ chosen) is a visual representation of heterogenous CV across the atrium and therefore is simply regional CV measurement with a predefined threshold for CV_min that colocalizes with LVA.⁵

Our study has demonstrated that TD_max was a better predictor of AA recurrence than CV_min after PVI, and given that TD_max correlated weakly with CV_min, it has incremental value in its assessment rather than calculating CV in isolation. We did not directly compare ILAM with TD_max sites in our patients, given that there are still limitations in its routine use in the atrium, including annotation of low-amplitude late EGMs and high interobserver variability. Given that ILAM is a visual representation of CV, we did not expect it to perform better than CV_min in predicting AF recurrence.

The mechanistic relationship between TD and AF

Clinical TDs (>110 m/s²) were characterized by initial supraphysiological acceleration followed by sudden deceleration, with two-thirds demonstrating a CV_min >0.2 m/s, which is the expected threshold for conduction block (Figure 3). In contrast, nonclinical TDs (≤110 m/s²) did not display significant initial acceleration. Supraphysiological CVs have previously been reported in patients with AF with the upper limits of the range in the order of 1.6 to 1.8 m/s, in which CV was either estimated by LA regions²¹ or locally between 2 adjacent LAT points.^17,22 Overall, it is uncertain whether these high CVs reflect genuine conduction heterogeneity, given that we lack a high-resolution 3-dimensional mapping gold standard to measure local CV against, or whether it is a reflection of our methodology wherein we assumed that the wavefront followed the line of velocity measurement. Based on our modeling data, we hypothesize that clinical TDs are localized areas of CV heterogeneity with large, abrupt transitions in CV. These clinical TD sites may then serve as regions of source-sink mismatch leading to localized conduction block⁸ and form regions of refractory tissue, about which functional re-entry and fibrillation may initiate.²³ A similar mechanism has been observed in the presence of abnormal CV restitution, when rate-dependent conduction slowing and block provides anchors for rotational activity duration AF.¹⁷ Upon subanalysis of the type of AA recurrence, we only observed a relationship between clinical TD and AF, and not AFL/AT, which was driven by LVA. This suggests that AF recurrence is primarily driven by the presence of functional substrate (TD_max >110 m/s²), rather than by fixed substrate within LVA.

Clinical implications

The main clinical benefit of TD_max assessment is detection of atrial remodeling not readily apparent by voltage mapping and CV measurement. Utilizing a threshold of TD_max >110 m/s² can identify patients who are most likely to have a future AA recurrence. In the future, automated TD_max measurement from isochronal maps would allow for optimized real-time assessment to avoid significantly increasing the LA dwell time. The LAT velocity vector algorithm in some electroanatomic mapping systems (eg, CARTO3 V8), which displays the activation direction and CV mapped during sinus or a paced rhythm, may streamline this workflow and warrants further study. At present, it is uncertain how to approach patients with clinical TD sites during the index ablation. Future studies should focus on whether these areas may serve as targets for ablation, particularly in redo procedures in which the pulmonary veins are already isolated. Its superiority in predicting recurrence over CV_min and LVA may help guide an individualized approach to ablation.

Limitations

The main strengths of this study design are its prospective follow-up and validation of TD_max >110 m/s² being absent from a control population without AF; however, the sample size was small, and future studies are needed to validate the TD_max in a larger cohort, particularly with persistent AF. There is no gold standard for the assessment of regional CV, and our methodology focused on creating a workflow that was reproducible and intuitive in its approach. However, the methodology applied a 2-dimensional assessment to a linear wavefront, which can only be considered an estimate of true anisotropic conduction across a 3-dimensional atrium. As a result, this generated the supraphysiological CV seen, which we argue is a marker of substrate, rather than an absolute value. TD_max was assessed only from a single pacing site (RA) and at a single rate (750 ms). Based on previous CV studies, changing either variable is expected to change the TD_max threshold,²⁴ and whether this changes its predictive power is uncertain. In addition, interindividual variation in interatrial conduction time and the location of interatrial connections may have influenced the direction of the wavefront in the LA; however, this was the reasoning behind identifying the propagating vector first prior to TD measurement. TD sites were not correlated with cross-sectional imaging to assess for scar; therefore, we could not corroborate whether these were sites with structural abnormalities. There is no accepted standard for the assessment of AA recurrence, and it is possible that some cases of asymptomatic AA may have been missed in follow-up, particularly given our low recurrence rates. We did not repeat LA mapping in the group who had AA in follow-up; therefore, TD_max is unlikely to be the sole factor for recurrence.

Conclusion

TD is a novel, pragmatic metric for quantifying LA structural remodeling in patients with AF at the time of PVI and detects abnormalities prior to apparent changes in low bipolar voltage. Unlike patients with local LA CV_min <0.2 m/s, those with ≥1 clinical TD site (defined by a TD_max >110 m/s²) are more likely to have a recurrence of AA in follow-up, and this should help inform future trials focusing on LA substrate-based ablation after the pulmonary veins have been successfully isolated.