Indices of bipolar complex fractionated atrial electrograms correlate poorly with each other and atrial fibrillation substrate complexity Dennis H. Lau, MBBS, PhD, FHRS,*† Bart Maesen, MD,*‡ Stef Zeemering, MSc,* Pawel Kuklik, PhD,*† Arne van Hunnik, BSc,* Theodorus A.R. Lankveld, MD,*§ Elham Bidar, MD,*‡ Sander Verheule, PhD,* Jan Nijs, MD,‡ Jos Maessen, MD, PhD,‡ Harry Crijns, MD, PhD,§ Prashanthan Sanders, MBBS, PhD, FHRS,† Ulrich Schotten, MD, PhD* From the *Department of Physiology, Maastricht University, Maastricht,The Netherlands, †Centre for Heart Rhythm Disorders, South Australian Health and Medical Research Institute, University of Adelaide and Royal Adelaide Hospital, Adelaide, South Australia, Australia, ‡Department of Cardiothoracic Surgery, Maastricht University Medical Centre, Maastricht, The Netherlands, and §Department of Cardiology, Maastricht University Medical Centre, Maastricht, The Netherlands. BACKGROUND The pathophysiological relevance of complex fractionated atrial electrograms (CFAE) in atrial fibrillation (AF) remains poorly understood. OBJECTIVE The aim of this study was to comprehensively investigate how bipolar CFAE correlates with unipolar electrogram fractionation and the underlying electrophysiological substrate of AF. METHODS Ten-second unipolar AF electrograms were recorded using a high-density electrode from the left atrium of 20 patients with AF (10 with persistent AF and 10 with paroxysmal AF) undergoing cardiac surgery. Semiautomated bipolar CFAE algorithms: complex fractionated electrogram-mean, interval confidence interval, continuous electrical activity, average complex interval, and shortest complex interval were evaluated against AF substrate complexity measures following fibrillation wave reconstruction derived from local unipolar activation time. The effect of
Dr Lau and Dr Maesen contributed equally to this work. This work was supported by grant funding from the Center for Translational Molecular Medicine (CTMM COHFAR project) and the European Network for Translational Research in Atrial Fibrillation (EUTRAF - 261057). Dr Lau was supported by a postdoctoral fellowship from the National Health and Medical Research Council of Australia and the CRB Blackburn Overseas Travelling Fellowship from the Royal Australasian College of Physicians. Dr Kuklik and Dr Sanders were funded by the National Heart Foundation of Australia. Dr Sanders was supported by a practitioner fellowship from the National Health and Medical Research Council. Dr Sanders has served on the advisory board of St Jude Medical, Bard Electrophysiology, Biosense Webster, Medtronic, Sanofi-Aventis, and Merck; he has received lecture fees from St Jude Medical, Bard Electrophysiology, Biosense Webster, Medtronic, and Merck and has also received research funding from Biosense Webster, Biotronik, Boston Scientific, Sorin Medical, St Jude Medical, and Medtronic. Dr Schotten has received grants from the Leducq Foundation, the European Union, the Centre for Translational Molecular Medicine, Medtronic, and Roche; he has served as an advisor to Roche and Bayer AG. Address reprint requests and correspondence: Dr Ulrich Schotten, Department of Physiology, Maastricht University, PO Box 616, 6200 MD Maastricht, The Netherlands. E-mail address: [email protected].
1547-5271/$-see front matter B 2015 Heart Rhythm Society. All rights reserved.
interelectrode spacing and electrode orientation on bipolar CFAE was also examined. RESULTS All 5 semiautomated bipolar CFAE algorithms showed poor correlation with each other and AF substrate complexity measures (conduction velocity, number of waves or breakthroughs per AF cycle, and electrical dissociation). Bipolar CFAE also correlated poorly with fractionation index derived from unipolar electrograms. Increased interelectrode spacing resulted in an increase in bipolar CFAE detected except for the interval confidence interval algorithm. CFAE appears unaffected by bipolar electrode orientation (vertical vs horizontal). By contrast, unipolar fractionation index correlated well with AF substrate complexity measures and can be regarded as a marker for conduction block. CONCLUSION The lack of pathophysiological relevance of bipolar CFAE analysis may in part contribute to the divergent and limited success rates of catheter ablation strategies targeting CFAE. KEYWORDS Atrial fibrillation; Catheter ablation; Mapping; Electrogram fractionation ABBREVIATIONS ACI ¼ average complex interval; AF ¼ atrial fibrillation; CEA ¼ continuous electrical activity; CFAE ¼ complex fractionated atrial electrogram; CFE-m ¼ complex fractionated electrogram-mean; ICL ¼ interval confidence level; SCI ¼ shortest complex interval (Heart Rhythm 2015;12:1415–1423) All rights reserved.
I
2015 Heart Rhythm Society.
Introduction Sites with complex fractionated atrial electrograms (CFAE) are of potential pathophysiological relevance to the atrial fibrillation (AF) substrate.1 Numerous studies have demonstrated that progressive atrial remodeling with AF persistence was associated with increasing atrial http://dx.doi.org/10.1016/j.hrthm.2015.03.017
1416 substrate complexity including electrogram fractionation.2–8 Nademanee et al9 were first to report high AF termination rate after CFAE ablation in 2004. However, this initial success has not been reproducible, with a recent randomized trial (STAR AF 2) demonstrating no extra benefits of additional CFAE ablation over pulmonary vein isolation alone in patients with persistent AF.10 A myriad of factors may account for the variable and curtailed outcomes of CFAE ablation we have seen to date.11 We have recently highlighted that both algorithms and mapping electrode configurations may have significant impact on CFAE determinations.12,13 Furthermore, how CFAE are related to the pathophysiological mechanisms of AF remains poorly understood. At present, CFAE can be defined using semiautomated mapping software incorporated with 3dimensioanl electroanatomic mapping systems. The CARTO system (Biosense Webster, Inc, Diamond Bar, CA) accommodates 4 bipolar CFAE measures: interval confidence level (ICL), average complex interval (ACI), shortest complex interval (SCI), and continuous electrical activity (CEA), while the EnSite NavX system (St Jude Medical, Inc, St Paul, MN) uses only 1: complex fractionated electrogram-mean (CFE-m). In this study, our primary aim was to evaluate these bipolar CFAE algorithms against electrophysiological characteristics of the underlying atrial substrate derived from high-density activation mapping of unipolar human AF electrograms. Furthermore, we aimed to comprehensively evaluate the impact of algorithms, interelectrode spacing, and electrode orientation on CFAE determinations. A comparative analysis of these semiautomated bipolar CFAE measures was also done with correlation to their unipolar activation.
Methods Study population This study included 20 patients with AF (10 with paroxysmal AF and 10 with persistent AF) undergoing cardiac surgery for coronary or valvular heart disease and/or AF at Maastricht University Medical Centre. The institutional ethics review committee approved this study protocol, and all patients gave written informed consent. Intraoperative mapping was performed under general anesthesia. After median sternotomy, a single-use custom-made high-density multiple-electrode array (FlexMEA, Maastricht Instruments BV, Maastricht, The Netherlands - 256 or 64 electrodes, electrode diameter 0.1 mm, and interelectrode spacing 1.5 mm) was used for direct contact epicardial mapping.14 Incremental atrial pacing was used to induce AF if necessary. Unipolar AF electrograms were recorded from the posterior wall of the left atrium and the left atrial appendage using a custom-made 256-channel mapping amplifier (filtering bandwidth 0.1–400 Hz, sampling rate 1 kHz, and analog-todigital resolution 16 bits). A silver plate was secured in the thoracic cavity and acted as an indifferent electrode.
Semiautomated CFAE algorithms Unipolar atrial electrograms were first filtered (bandwidth 1– 400 Hz) before conversion to bipolar electrograms (band-pass
Heart Rhythm, Vol 12, No 7, July 2015 filter 30–400 Hz) using custom software programmed in the MATLAB environment (The MathWorks Inc, Natick, MA). To evaluate the effect of interelectrode spacing and electrode orientation on CFAE determinations, bipolar electrograms were constructed from unipolar signals 1.5, 3.0, 4.5, and 6 mm apart in both horizontal and vertical orientations. Comparisons of CFAE between horizontal and vertical electrode configurations were facilitated by linear interpolation of each fractionation map into 961 nodes. Point-by-point correlations of CFAE determined by different algorithms were analyzed using bipolar electrograms of 1.5-mm interelectrode spacing only. We used AF episodes of 10-second duration to ensure accurate CFAE characterization.15 Electrograms or deflections were first tagged according to the system-specific amplitude and timing criteria, which were designed to avoid noise or far-field detections. For CARTO algorithms, the electrogram width was 15–30 ms with an amplitude range of 0.03–0.15 mV. For the EnSiteNavX CFE-m algorithm, the refractory period was set at 50 ms with a minimum electrogram width of 10 ms and a peak-to-peak sensitivity of 0.03–0.1 mV. All electrograms were verified manually to ensure accurate annotation and determination of CFAE by the custom analysis software according to the individual algorithm at factory default or previously published values13: 1. ICL: It refers to the number of 70–120 ms intervals between tagged intrinsic local activations per 2.5 seconds of AF. CFAE is defined by ICL Z5. 2. ACI: It refers to the average of all intervals between 2 successive tagged deflections within the range of 70–500 ms. CFAE is defined by ACI r100 ms. 3. SCI: It refers to the shortest interval between 2 successive tagged deflections within the range of 70–500 ms. CFAE is defined by SCI r120 ms. 4. CEA: It was defined by the presence of 2 or more successive tagged deflections with a o50 ms interval and expressed as percentage of continuous activity. CFAE is defined by CEA Z75%. 5. CFE-m: It measures the mean of time intervals between the marked deflections. CFAE is defined by CFE-m o120 ms.
Characterization of the atrial substrate by highdensity activation mapping An established wave-mapping technique based on unipolar local activation time was used to compute the following substrate complexity measures: AF wave conduction velocity, number of waves per AF cycle, electrical dissociation, and incidence of breakthrough waves per AF cycle.2–6,16 Detailed methods of unipolar electrogram deflection detection, AF wave reconstruction, and computation of these substrate parameters can be found in the Online Supplement. Furthermore, the automated deflection detection algorithm can differentiate local from nonlocal activations (Figure 1). The downstroke of a unipolar electrogram has been shown to represent local activation that coincides with the upstroke of the
Lau et al
CFAE Algorithms and AF Substrate
Unipolar Fractionation Index =
1417
= 1.55
Figure 1 Fractionation index according to local unipolar activations. Both local (green) and nonlocal (red) deflections are marked by automated deflection detection algorithm. In this example, unipolar fractionation index taken as the ratio of nonlocal to local deflections is 1.55.
action potential of the underlying myocyte.17 Any additional deflections are therefore due to fractionation or nonlocal activations. We therefore defined unipolar fractionation index (FI) as the ratio of nonlocal to local deflections. Bipolar CFAE algorithms were assessed against the FI of the respective unipolar electrograms (mean FI at both unipoles) and each of the AF substrate complexity measures listed earlier in this section.
Statistical analysis All continuous variables are reported as mean ⫾ SD. Data were compared using the Student t test or Wilcoxon rank sum test according to their distribution. Categorical data were compared using the Fisher exact test. To investigate the effect of algorithm, interelectrode spacing, and electrode configuration on CFAE detection as well as the percentage of match between CFAE algorithms, a linear mixed-effects model was used. In this model, each of the above factors was entered as the dependent variable, with AF type entered as a fixed factor and patient ID as a random factor. One-way analysis of variance was used to compare bipolar CFAE across different interelectrode spacings. Correlations between various time domain algorithms and AF complexity measures were also assessed. Statistical significance was established at P o .05.
Results Both groups of patients were well matched except for larger left atrial dimension and higher CHA2DS2-VASc score in those with persistent AF (Table 1). All posterior left atrial mapping (n ¼ 16) was performed with the 256-electrode plaque, while 5 of 15 (4 paroxysmal and 1 persistent) left atrial appendages were mapped with the 64-electrode plaque owing to the smaller surface area. In total, 45,126 ten-second bipolar AF electrograms were used for bipolar CFAE algorithm, interelectrode spacing, and electrode configuration comparisons. CFAE correlations between various algorithms were performed using a subset of 11,900 ten-second bipolar electrograms at 1.5 mm spacing.
Bipolar CFAE and AF substrate complexity Figure 2A shows an example of AF wave map constructed from unipolar activation time map, as described previously.2 Bipolar electrograms 1–3 are all taken along the same line of block, yet a different degree of fractionation is evident: from low amplitude highly fractionated signals in electrograms
1 and 2 to minimally fractionated signal in electrogram 3. Bipolar electrogram 4 taken from an area between a line of block and waves collision also demonstrates a high degree of fractionation. In area within a wave where fractionation is not expected, both fractionated and nonfractionated signals are found at locations 5 and 6, respectively. These demonstrate that the bipolar electrogram fractionation is variable and not always reflective of underlying conduction characteristics. Overall, mean bipolar CFAE measure of each 10-second AF recording (n ¼ 31 from 20 patients) was correlated with each established AF substrate complexity measure. Examples of correlation between AF complexity as quantified by the number of AF waves per cycle with 2 bipolar CFAE measures are shown in Figure 2B to further illustrate the discrepancy between algorithms. A wide range in the number of AF waves is seen at the same SCI of 70 ms with poor overall correlation (left panel), although the relationship between AF wave numbers and CFE-m was marginally better (right panel). Figure 2C shows the entire array of correlations between the various AF substrate characteristics and all bipolar CFAE indices. None of the bipolar CFAE measures demonstrated significant correlation with AF wave conduction velocity. Overall correlations between bipolar CFAE and AF substrate complexity measures were poor, with the CEA algorithm showing the best correlation with the number of AF waves and breakthroughs per AF cycle and electrical dissociation (r ¼ 0.70, 0.70, and 0.69 respectively; all P o .0006) The remaining 4 algorithms (CFE-m, ACI, ICL, and SCI) demonstrated poorer correlations with these AF substrate complexity measures. Furthermore, poor correlations were seen between bipolar CFAE algorithms and AF cycle length.
Point-by-point correlations of bipolar CFAE Figure 3 shows the comparison of the degree of fractionation for each individual 10-second AF recording on the basis of Table 1
Patient and electrogram characteristics
Characteristic
Paroxysmal AF Persistent AF (n ¼ 10) (n ¼ 10)
Age (y) 69 ⫾ 5 Sex: male, n(%) 8 (80) 26.6 ⫾ 4.6 Body mass index (kg/m2) Hypertension, n(%) 7 (70) Diabetes, n(%) 1 (10) Duration of atrial 7⫾7 fibrillation (y) Antiarrhythmic drugs, n(%) 6 (60) Surgery indications, n(%) Coronary heart disease 6 (60) Valvular heart disease 3 (30) Atrial fibrillation 1 (10) Left ventricular ejection 62 ⫾ 8 fraction (%) Left atrial size: 43 ⫾ 8 parasternal (mm) 2.3 ⫾ 0.9 CHA2DS2-VASc score Values are presented as mean ⫾ SD. AF ¼ atrial fibrillation.
P
70 ⫾ 6 6 (60) 27.0 ⫾ 3.0 8 (80) 4 (40) 9⫾6
.7 .6 .9 1.0 .3 .4
4 (40)
.7
4 (40) 5 (50) 1 (10) 62 ⫾ 5
.6 .9
53 ⫾ 6
.006
3.8 ⫾ 1.7
.03
Heart Rhythm, Vol 12, No 7, July 2015
1
2
3
4
Waves per AF cycle (n)
1418
r = - 0.40
25
20
15
15
10
10
5
5 0
0 50
100
150
SCI (ms)
6
r = - 0.57
25
20
200
250
50
100
150
200
250
CFE-m (ms)
5
Figure 2 Bipolar CFAE and AF substrate complexity. A: Example wave map from the posterior left atrium of a patient with paroxysmal AF. Thin arrows demonstrate wavefront propagation direction. Electrograms from sites 1–6 within the wave map demonstrate inconsistent relationship between fractionation and underlying conduction characteristics (see text). B: Correlation graphs for CFE-m and SCI with number of AF waves per cycle. C: Overall correlations between various AF substrate characteristics and all bipolar CFAE indices. ACI ¼ average complex interval; AF ¼ atrial fibrillation; CEA ¼ continuous electrical activity; CFAE ¼ complex fractionated atrial electrogram; CFE-m ¼ complex fractionated electrogram-mean; ICL ¼ interval confidence level; SCI ¼ shortest complex interval.
different algorithms. Correlations between the ICL algorithm and the other 4 algorithms were poor (Figure 3, top row: r ¼ 0.40, 0.44, 0.64, and 0.53; all P o .001). A paradoxical decrease in ICL was noted when fractionation was highest with CEA, CFE-m, and ACI. Similarly, low ICL was seen when SCI was lowest at the cutoff of 70 ms. Correlations between SCI and all other algorithms were also mostly poor (Figure 3, right column: r ¼ 0.53, 0.50, 0.46, and 0.58; all P o .001). Moderate correlations were seen in the remaining CEA, CFE-m, and ACI algorithms (Figure 3; r ¼ ranges between 0.76 and 0.87; all P o .001). A considerable variation in point-by-point correlations of bipolar CFAE is evident. Furthermore, using the respective threshold for each bipolar CFAE algorithm, the resultant percentage of 10-second AF electrograms detected as CFAE ranged from 11% to 88% and from 23% to 97% for the paroxysmal and persistent AF group, respectively (Figure 4A). Specifically, patients with persistent AF demonstrated more fractionated atrial electrograms than those with paroxysmal AF and the CEA algorithm stood out with the lowest CFAE detection. An example of the variation in CFAE detection is
shown in Figure 4B, in which the percentage of detected CFAE in the same 10-second AF recording ranged from 7% to 97%, with poor agreement among all 5 algorithms.
Effect of electrode spacing and orientation on bipolar CFAE detection Increasing interelectrode spacing from 1.5 to 6.0 mm resulted in a corresponding increase in the percentage of AF electrograms deemed as fractionated according to each algorithm with the exception of ICL, whereby a paradoxical decrease in CFAE was seen (Figure 5A). When the effect of increasing interelectrode spacing was analyzed within each individual algorithm, statistical significance was maintained for CEA, ACI, and CFE-m (Figure 5B). The stepwise increase in CFAE detection (except for ICL) with increasing interelectrode spacing is further illustrated in Figure 5C. No significant differences were seen in bipolar CFAE determinations when comparing horizontal vs vertical electrode configurations (Online Supplemental Figure 1). Overall match in CFAE detection was very good between horizontal
Lau et al
CFAE Algorithms and AF Substrate
1419
Figure 3 Point-by-point correlations of bipolar CFAE. Moderately good point-by-point correlations were seen between the CEA/CFE-m/ACI algorithms (column 2, row 2 and column 3, and rows 2 and 3). However, the remaining correlations were at most fair to moderate. ACI ¼ average complex interval; CEA ¼ continuous electrical activity; CFAE ¼ complex fractionated atrial electrogram; CFE-m ¼ complex fractionated electrogram-mean; ICL ¼ interval confidence level; SCI ¼ shortest complex interval.
and vertical configured bipolar electrograms (ICL: 90.4% ⫾ 7.6%; CEA: 94.9% ⫾ 9.0%; CFE-m: 90.1% ⫾ 6.9%; ACI: 91.9% ⫾ 6.2%; SCI: 97.9% ⫾ 7.5%).
Correlation between bipolar and unipolar CFAE in relation to AF wavefronts Figure 6 illustrates the differences in bipolar and unipolar fractionation in relation to conduction patterns. Clean and sharp unipolar electrograms are usually seen with broad unidirectional AF wavefront, while bipolar counterparts demonstrate variable complexity (top left). With colliding wavefronts, a split or double potential is clearly seen in unipolar signals, which is not apparent in all bipolar examples (bottom left). In the case of conduction block, unipolar electrograms demonstrate fractionated potentials along the block line with sharper local deflection away from the site of block. This specificity is not as clear with bipolar recordings demonstrating more continuous fractionation (top right). Unipolar electrograms are most fractionated at the core of rotating wavefront than at the periphery, whereas the differentiation is not as distinct with bipolar electrograms (bottom right). Figure 7 shows the correlation between unipolar and bipolar fractionation. All correlations with unipolar FI were significant. However, correlation was poor for
the ICL and SCI algorithms (r ¼ 0.42 and 0.44) and only fair with the remaining 3 algorithms (r ¼ 0.65, 0.68, and 0.68, respectively). Interestingly, unipolar FI demonstrated very good correlations with AF substrate complexity such as the number of AF waves and breakthroughs per AF cycle and electrical dissociation (Figure 7 bottom panel; r ¼ 0.91, 0.90, and 0.92, respectively). No correlation was seen between unipolar FI and conduction velocity.
Discussion This work provides a comprehensive electrophysiological evaluation of bipolar CFAE algorithms and the impact of mapping electrode configurations on bipolar CFAE. Our main findings are that bipolar CFAE derived from semiautomated algorithms (1) correlate poorly with established AF substrate complexity measures and with each other, (2) are sensitive to interelectrode spacing with an increase in CFAE detected with increased interelectrode distance, and (3) do not correlate well with their constituent unipolar activation–derived FI. This study indicates that CFAE determined by semiautomated bipolar algorithms are highly variable and poor markers for the underlying AF substrate.
1420
Heart Rhythm, Vol 12, No 7, July 2015
Paroxysmal AF Persistent AF p
Dr Lau and Dr Maesen contributed equally to this work. This work was supported by grant funding from the Center for Translational Molecular Medicine (CTMM COHFAR project) and the European Network for Translational Research in Atrial Fibrillation (EUTRAF - 261057). Dr Lau was supported by a postdoctoral fellowship from the National Health and Medical Research Council of Australia and the CRB Blackburn Overseas Travelling Fellowship from the Royal Australasian College of Physicians. Dr Kuklik and Dr Sanders were funded by the National Heart Foundation of Australia. Dr Sanders was supported by a practitioner fellowship from the National Health and Medical Research Council. Dr Sanders has served on the advisory board of St Jude Medical, Bard Electrophysiology, Biosense Webster, Medtronic, Sanofi-Aventis, and Merck; he has received lecture fees from St Jude Medical, Bard Electrophysiology, Biosense Webster, Medtronic, and Merck and has also received research funding from Biosense Webster, Biotronik, Boston Scientific, Sorin Medical, St Jude Medical, and Medtronic. Dr Schotten has received grants from the Leducq Foundation, the European Union, the Centre for Translational Molecular Medicine, Medtronic, and Roche; he has served as an advisor to Roche and Bayer AG. Address reprint requests and correspondence: Dr Ulrich Schotten, Department of Physiology, Maastricht University, PO Box 616, 6200 MD Maastricht, The Netherlands. E-mail address: [email protected].
1547-5271/$-see front matter B 2015 Heart Rhythm Society. All rights reserved.
interelectrode spacing and electrode orientation on bipolar CFAE was also examined. RESULTS All 5 semiautomated bipolar CFAE algorithms showed poor correlation with each other and AF substrate complexity measures (conduction velocity, number of waves or breakthroughs per AF cycle, and electrical dissociation). Bipolar CFAE also correlated poorly with fractionation index derived from unipolar electrograms. Increased interelectrode spacing resulted in an increase in bipolar CFAE detected except for the interval confidence interval algorithm. CFAE appears unaffected by bipolar electrode orientation (vertical vs horizontal). By contrast, unipolar fractionation index correlated well with AF substrate complexity measures and can be regarded as a marker for conduction block. CONCLUSION The lack of pathophysiological relevance of bipolar CFAE analysis may in part contribute to the divergent and limited success rates of catheter ablation strategies targeting CFAE. KEYWORDS Atrial fibrillation; Catheter ablation; Mapping; Electrogram fractionation ABBREVIATIONS ACI ¼ average complex interval; AF ¼ atrial fibrillation; CEA ¼ continuous electrical activity; CFAE ¼ complex fractionated atrial electrogram; CFE-m ¼ complex fractionated electrogram-mean; ICL ¼ interval confidence level; SCI ¼ shortest complex interval (Heart Rhythm 2015;12:1415–1423) All rights reserved.
I
2015 Heart Rhythm Society.
Introduction Sites with complex fractionated atrial electrograms (CFAE) are of potential pathophysiological relevance to the atrial fibrillation (AF) substrate.1 Numerous studies have demonstrated that progressive atrial remodeling with AF persistence was associated with increasing atrial http://dx.doi.org/10.1016/j.hrthm.2015.03.017
1416 substrate complexity including electrogram fractionation.2–8 Nademanee et al9 were first to report high AF termination rate after CFAE ablation in 2004. However, this initial success has not been reproducible, with a recent randomized trial (STAR AF 2) demonstrating no extra benefits of additional CFAE ablation over pulmonary vein isolation alone in patients with persistent AF.10 A myriad of factors may account for the variable and curtailed outcomes of CFAE ablation we have seen to date.11 We have recently highlighted that both algorithms and mapping electrode configurations may have significant impact on CFAE determinations.12,13 Furthermore, how CFAE are related to the pathophysiological mechanisms of AF remains poorly understood. At present, CFAE can be defined using semiautomated mapping software incorporated with 3dimensioanl electroanatomic mapping systems. The CARTO system (Biosense Webster, Inc, Diamond Bar, CA) accommodates 4 bipolar CFAE measures: interval confidence level (ICL), average complex interval (ACI), shortest complex interval (SCI), and continuous electrical activity (CEA), while the EnSite NavX system (St Jude Medical, Inc, St Paul, MN) uses only 1: complex fractionated electrogram-mean (CFE-m). In this study, our primary aim was to evaluate these bipolar CFAE algorithms against electrophysiological characteristics of the underlying atrial substrate derived from high-density activation mapping of unipolar human AF electrograms. Furthermore, we aimed to comprehensively evaluate the impact of algorithms, interelectrode spacing, and electrode orientation on CFAE determinations. A comparative analysis of these semiautomated bipolar CFAE measures was also done with correlation to their unipolar activation.
Methods Study population This study included 20 patients with AF (10 with paroxysmal AF and 10 with persistent AF) undergoing cardiac surgery for coronary or valvular heart disease and/or AF at Maastricht University Medical Centre. The institutional ethics review committee approved this study protocol, and all patients gave written informed consent. Intraoperative mapping was performed under general anesthesia. After median sternotomy, a single-use custom-made high-density multiple-electrode array (FlexMEA, Maastricht Instruments BV, Maastricht, The Netherlands - 256 or 64 electrodes, electrode diameter 0.1 mm, and interelectrode spacing 1.5 mm) was used for direct contact epicardial mapping.14 Incremental atrial pacing was used to induce AF if necessary. Unipolar AF electrograms were recorded from the posterior wall of the left atrium and the left atrial appendage using a custom-made 256-channel mapping amplifier (filtering bandwidth 0.1–400 Hz, sampling rate 1 kHz, and analog-todigital resolution 16 bits). A silver plate was secured in the thoracic cavity and acted as an indifferent electrode.
Semiautomated CFAE algorithms Unipolar atrial electrograms were first filtered (bandwidth 1– 400 Hz) before conversion to bipolar electrograms (band-pass
Heart Rhythm, Vol 12, No 7, July 2015 filter 30–400 Hz) using custom software programmed in the MATLAB environment (The MathWorks Inc, Natick, MA). To evaluate the effect of interelectrode spacing and electrode orientation on CFAE determinations, bipolar electrograms were constructed from unipolar signals 1.5, 3.0, 4.5, and 6 mm apart in both horizontal and vertical orientations. Comparisons of CFAE between horizontal and vertical electrode configurations were facilitated by linear interpolation of each fractionation map into 961 nodes. Point-by-point correlations of CFAE determined by different algorithms were analyzed using bipolar electrograms of 1.5-mm interelectrode spacing only. We used AF episodes of 10-second duration to ensure accurate CFAE characterization.15 Electrograms or deflections were first tagged according to the system-specific amplitude and timing criteria, which were designed to avoid noise or far-field detections. For CARTO algorithms, the electrogram width was 15–30 ms with an amplitude range of 0.03–0.15 mV. For the EnSiteNavX CFE-m algorithm, the refractory period was set at 50 ms with a minimum electrogram width of 10 ms and a peak-to-peak sensitivity of 0.03–0.1 mV. All electrograms were verified manually to ensure accurate annotation and determination of CFAE by the custom analysis software according to the individual algorithm at factory default or previously published values13: 1. ICL: It refers to the number of 70–120 ms intervals between tagged intrinsic local activations per 2.5 seconds of AF. CFAE is defined by ICL Z5. 2. ACI: It refers to the average of all intervals between 2 successive tagged deflections within the range of 70–500 ms. CFAE is defined by ACI r100 ms. 3. SCI: It refers to the shortest interval between 2 successive tagged deflections within the range of 70–500 ms. CFAE is defined by SCI r120 ms. 4. CEA: It was defined by the presence of 2 or more successive tagged deflections with a o50 ms interval and expressed as percentage of continuous activity. CFAE is defined by CEA Z75%. 5. CFE-m: It measures the mean of time intervals between the marked deflections. CFAE is defined by CFE-m o120 ms.
Characterization of the atrial substrate by highdensity activation mapping An established wave-mapping technique based on unipolar local activation time was used to compute the following substrate complexity measures: AF wave conduction velocity, number of waves per AF cycle, electrical dissociation, and incidence of breakthrough waves per AF cycle.2–6,16 Detailed methods of unipolar electrogram deflection detection, AF wave reconstruction, and computation of these substrate parameters can be found in the Online Supplement. Furthermore, the automated deflection detection algorithm can differentiate local from nonlocal activations (Figure 1). The downstroke of a unipolar electrogram has been shown to represent local activation that coincides with the upstroke of the
Lau et al
CFAE Algorithms and AF Substrate
Unipolar Fractionation Index =
1417
= 1.55
Figure 1 Fractionation index according to local unipolar activations. Both local (green) and nonlocal (red) deflections are marked by automated deflection detection algorithm. In this example, unipolar fractionation index taken as the ratio of nonlocal to local deflections is 1.55.
action potential of the underlying myocyte.17 Any additional deflections are therefore due to fractionation or nonlocal activations. We therefore defined unipolar fractionation index (FI) as the ratio of nonlocal to local deflections. Bipolar CFAE algorithms were assessed against the FI of the respective unipolar electrograms (mean FI at both unipoles) and each of the AF substrate complexity measures listed earlier in this section.
Statistical analysis All continuous variables are reported as mean ⫾ SD. Data were compared using the Student t test or Wilcoxon rank sum test according to their distribution. Categorical data were compared using the Fisher exact test. To investigate the effect of algorithm, interelectrode spacing, and electrode configuration on CFAE detection as well as the percentage of match between CFAE algorithms, a linear mixed-effects model was used. In this model, each of the above factors was entered as the dependent variable, with AF type entered as a fixed factor and patient ID as a random factor. One-way analysis of variance was used to compare bipolar CFAE across different interelectrode spacings. Correlations between various time domain algorithms and AF complexity measures were also assessed. Statistical significance was established at P o .05.
Results Both groups of patients were well matched except for larger left atrial dimension and higher CHA2DS2-VASc score in those with persistent AF (Table 1). All posterior left atrial mapping (n ¼ 16) was performed with the 256-electrode plaque, while 5 of 15 (4 paroxysmal and 1 persistent) left atrial appendages were mapped with the 64-electrode plaque owing to the smaller surface area. In total, 45,126 ten-second bipolar AF electrograms were used for bipolar CFAE algorithm, interelectrode spacing, and electrode configuration comparisons. CFAE correlations between various algorithms were performed using a subset of 11,900 ten-second bipolar electrograms at 1.5 mm spacing.
Bipolar CFAE and AF substrate complexity Figure 2A shows an example of AF wave map constructed from unipolar activation time map, as described previously.2 Bipolar electrograms 1–3 are all taken along the same line of block, yet a different degree of fractionation is evident: from low amplitude highly fractionated signals in electrograms
1 and 2 to minimally fractionated signal in electrogram 3. Bipolar electrogram 4 taken from an area between a line of block and waves collision also demonstrates a high degree of fractionation. In area within a wave where fractionation is not expected, both fractionated and nonfractionated signals are found at locations 5 and 6, respectively. These demonstrate that the bipolar electrogram fractionation is variable and not always reflective of underlying conduction characteristics. Overall, mean bipolar CFAE measure of each 10-second AF recording (n ¼ 31 from 20 patients) was correlated with each established AF substrate complexity measure. Examples of correlation between AF complexity as quantified by the number of AF waves per cycle with 2 bipolar CFAE measures are shown in Figure 2B to further illustrate the discrepancy between algorithms. A wide range in the number of AF waves is seen at the same SCI of 70 ms with poor overall correlation (left panel), although the relationship between AF wave numbers and CFE-m was marginally better (right panel). Figure 2C shows the entire array of correlations between the various AF substrate characteristics and all bipolar CFAE indices. None of the bipolar CFAE measures demonstrated significant correlation with AF wave conduction velocity. Overall correlations between bipolar CFAE and AF substrate complexity measures were poor, with the CEA algorithm showing the best correlation with the number of AF waves and breakthroughs per AF cycle and electrical dissociation (r ¼ 0.70, 0.70, and 0.69 respectively; all P o .0006) The remaining 4 algorithms (CFE-m, ACI, ICL, and SCI) demonstrated poorer correlations with these AF substrate complexity measures. Furthermore, poor correlations were seen between bipolar CFAE algorithms and AF cycle length.
Point-by-point correlations of bipolar CFAE Figure 3 shows the comparison of the degree of fractionation for each individual 10-second AF recording on the basis of Table 1
Patient and electrogram characteristics
Characteristic
Paroxysmal AF Persistent AF (n ¼ 10) (n ¼ 10)
Age (y) 69 ⫾ 5 Sex: male, n(%) 8 (80) 26.6 ⫾ 4.6 Body mass index (kg/m2) Hypertension, n(%) 7 (70) Diabetes, n(%) 1 (10) Duration of atrial 7⫾7 fibrillation (y) Antiarrhythmic drugs, n(%) 6 (60) Surgery indications, n(%) Coronary heart disease 6 (60) Valvular heart disease 3 (30) Atrial fibrillation 1 (10) Left ventricular ejection 62 ⫾ 8 fraction (%) Left atrial size: 43 ⫾ 8 parasternal (mm) 2.3 ⫾ 0.9 CHA2DS2-VASc score Values are presented as mean ⫾ SD. AF ¼ atrial fibrillation.
P
70 ⫾ 6 6 (60) 27.0 ⫾ 3.0 8 (80) 4 (40) 9⫾6
.7 .6 .9 1.0 .3 .4
4 (40)
.7
4 (40) 5 (50) 1 (10) 62 ⫾ 5
.6 .9
53 ⫾ 6
.006
3.8 ⫾ 1.7
.03
Heart Rhythm, Vol 12, No 7, July 2015
1
2
3
4
Waves per AF cycle (n)
1418
r = - 0.40
25
20
15
15
10
10
5
5 0
0 50
100
150
SCI (ms)
6
r = - 0.57
25
20
200
250
50
100
150
200
250
CFE-m (ms)
5
Figure 2 Bipolar CFAE and AF substrate complexity. A: Example wave map from the posterior left atrium of a patient with paroxysmal AF. Thin arrows demonstrate wavefront propagation direction. Electrograms from sites 1–6 within the wave map demonstrate inconsistent relationship between fractionation and underlying conduction characteristics (see text). B: Correlation graphs for CFE-m and SCI with number of AF waves per cycle. C: Overall correlations between various AF substrate characteristics and all bipolar CFAE indices. ACI ¼ average complex interval; AF ¼ atrial fibrillation; CEA ¼ continuous electrical activity; CFAE ¼ complex fractionated atrial electrogram; CFE-m ¼ complex fractionated electrogram-mean; ICL ¼ interval confidence level; SCI ¼ shortest complex interval.
different algorithms. Correlations between the ICL algorithm and the other 4 algorithms were poor (Figure 3, top row: r ¼ 0.40, 0.44, 0.64, and 0.53; all P o .001). A paradoxical decrease in ICL was noted when fractionation was highest with CEA, CFE-m, and ACI. Similarly, low ICL was seen when SCI was lowest at the cutoff of 70 ms. Correlations between SCI and all other algorithms were also mostly poor (Figure 3, right column: r ¼ 0.53, 0.50, 0.46, and 0.58; all P o .001). Moderate correlations were seen in the remaining CEA, CFE-m, and ACI algorithms (Figure 3; r ¼ ranges between 0.76 and 0.87; all P o .001). A considerable variation in point-by-point correlations of bipolar CFAE is evident. Furthermore, using the respective threshold for each bipolar CFAE algorithm, the resultant percentage of 10-second AF electrograms detected as CFAE ranged from 11% to 88% and from 23% to 97% for the paroxysmal and persistent AF group, respectively (Figure 4A). Specifically, patients with persistent AF demonstrated more fractionated atrial electrograms than those with paroxysmal AF and the CEA algorithm stood out with the lowest CFAE detection. An example of the variation in CFAE detection is
shown in Figure 4B, in which the percentage of detected CFAE in the same 10-second AF recording ranged from 7% to 97%, with poor agreement among all 5 algorithms.
Effect of electrode spacing and orientation on bipolar CFAE detection Increasing interelectrode spacing from 1.5 to 6.0 mm resulted in a corresponding increase in the percentage of AF electrograms deemed as fractionated according to each algorithm with the exception of ICL, whereby a paradoxical decrease in CFAE was seen (Figure 5A). When the effect of increasing interelectrode spacing was analyzed within each individual algorithm, statistical significance was maintained for CEA, ACI, and CFE-m (Figure 5B). The stepwise increase in CFAE detection (except for ICL) with increasing interelectrode spacing is further illustrated in Figure 5C. No significant differences were seen in bipolar CFAE determinations when comparing horizontal vs vertical electrode configurations (Online Supplemental Figure 1). Overall match in CFAE detection was very good between horizontal
Lau et al
CFAE Algorithms and AF Substrate
1419
Figure 3 Point-by-point correlations of bipolar CFAE. Moderately good point-by-point correlations were seen between the CEA/CFE-m/ACI algorithms (column 2, row 2 and column 3, and rows 2 and 3). However, the remaining correlations were at most fair to moderate. ACI ¼ average complex interval; CEA ¼ continuous electrical activity; CFAE ¼ complex fractionated atrial electrogram; CFE-m ¼ complex fractionated electrogram-mean; ICL ¼ interval confidence level; SCI ¼ shortest complex interval.
and vertical configured bipolar electrograms (ICL: 90.4% ⫾ 7.6%; CEA: 94.9% ⫾ 9.0%; CFE-m: 90.1% ⫾ 6.9%; ACI: 91.9% ⫾ 6.2%; SCI: 97.9% ⫾ 7.5%).
Correlation between bipolar and unipolar CFAE in relation to AF wavefronts Figure 6 illustrates the differences in bipolar and unipolar fractionation in relation to conduction patterns. Clean and sharp unipolar electrograms are usually seen with broad unidirectional AF wavefront, while bipolar counterparts demonstrate variable complexity (top left). With colliding wavefronts, a split or double potential is clearly seen in unipolar signals, which is not apparent in all bipolar examples (bottom left). In the case of conduction block, unipolar electrograms demonstrate fractionated potentials along the block line with sharper local deflection away from the site of block. This specificity is not as clear with bipolar recordings demonstrating more continuous fractionation (top right). Unipolar electrograms are most fractionated at the core of rotating wavefront than at the periphery, whereas the differentiation is not as distinct with bipolar electrograms (bottom right). Figure 7 shows the correlation between unipolar and bipolar fractionation. All correlations with unipolar FI were significant. However, correlation was poor for
the ICL and SCI algorithms (r ¼ 0.42 and 0.44) and only fair with the remaining 3 algorithms (r ¼ 0.65, 0.68, and 0.68, respectively). Interestingly, unipolar FI demonstrated very good correlations with AF substrate complexity such as the number of AF waves and breakthroughs per AF cycle and electrical dissociation (Figure 7 bottom panel; r ¼ 0.91, 0.90, and 0.92, respectively). No correlation was seen between unipolar FI and conduction velocity.
Discussion This work provides a comprehensive electrophysiological evaluation of bipolar CFAE algorithms and the impact of mapping electrode configurations on bipolar CFAE. Our main findings are that bipolar CFAE derived from semiautomated algorithms (1) correlate poorly with established AF substrate complexity measures and with each other, (2) are sensitive to interelectrode spacing with an increase in CFAE detected with increased interelectrode distance, and (3) do not correlate well with their constituent unipolar activation–derived FI. This study indicates that CFAE determined by semiautomated bipolar algorithms are highly variable and poor markers for the underlying AF substrate.
1420
Heart Rhythm, Vol 12, No 7, July 2015
Paroxysmal AF Persistent AF p
