Relationship between the electrocardiographic atrial fibrillation cycle length and left atrial remodeling: A detailed electroanatomic mapping study Tomos E. Walters, MBBS,*† Andrew W. Teh, PhD,‡ Steven Spence, GradDipSci,* Peter M. Kistler, PhD,§ Joseph B. Morton, PhD,*† Jonathan M. Kalman, PhD, FHRS*† From the *Department of Cardiology, Royal Melbourne Hospital, Melbourne, Australia, †Department of Medicine, University of Melbourne, Melbourne, Australia, ‡Department of Cardiology, Austin Hospital, Melbourne, Australia, and §Department of Cardiology, Alfred Hospital and Baker IDI, Melbourne, Australia. BACKGROUND/OBJECTIVE The purpose of this study was to characterize the association between the extent of left atrial electroanatomic remodeling in atrial fibrillation and the atrial fibrillation cycle length derived from lead V1 of the surface ECG (V1AFCL).
proximal and distal coronary sinus recordings. A longer V1AFCL was associated with slower left atrial conduction velocity and greater signal complexity but not with other remodeling variables, including left atrial size, atrial refractoriness, and mean endocardial voltage.
METHODS Twenty-three patients in atrial fibrillation (AF) who presented for AF ablation underwent detailed electroanatomic mapping of the left atrium. The digital 12-lead ECG was exported for offline analysis, with signal filtering and QRST subtraction used to reveal the fibrillatory baseline in lead V1. Mean V1AFCL was determined by direct annotation of the fibrillatory baseline, and the corresponding dominant V1AFCL was determined by Fourier transformation to derive the dominant frequency from the frequency power spectrum. The simultaneous AFCL from proximal and distal coronary sinus recordings was determined using the same methods. The strength of the association between various left atrial remodeling variables and V1AFCL was determined.
CONCLUSION A longer atrial fibrillatory cycle length in surface ECG lead V1 is significantly associated with parameters of more advanced left atrial electroanatomic remodeling, specifically slower atrial conduction and more extensive electrogram fractionation.
RESULTS The 2 methods of deriving V1AFCL and intracardiac AFCL were found to produce highly equivalent results. V1AFCL showed significant correlation with intracardiac AFCL derived from both
Introduction Atrial fibrillation (AF) is associated with progressive changes in left atrial (LA) electrophysiology, the extent of which can be assessed by invasive techniques to characterize refractoriness, conduction, voltage amplitude, and electrogram complexity.1,2 Previous studies have used the cycle length of atrial activation (AFCL) in the electrocardiogram (ECG) as a noninvasive indirect marker of atrial remodeling. These studies have observed that a shorter AFCL is associated Dr. Walters is a recipient of a research scholarship from the National Health and Medical Research Council of Australia and the National Heart Foundation of Australia. Address reprint requests and correspondence: Dr. Jonathan M. Kalman, Royal Melbourne Hospital, Melbourne Heart Centre; Melbourne, Victoria 3050, Australia. E-mail address: jon.kalman@ mh.org.au.
1547-5271/$-see front matter B 2014 Heart Rhythm Society. All rights reserved.
KEYWORDS Atrial fibrillation; Cycle length; Left atrium; Remodeling ABBREVIATIONS AF ¼ atrial fibrillation; AFCL ¼ atrial fibrillation cycle length; CS ¼ coronary sinus; CSd ¼ distal coronary sinus recording bipole; CSp ¼ proximal coronary sinus recording bipole; DF ¼ dominant frequency; ECG ¼ electrocardiogram; ERP ¼ effective refractory period; LA ¼ left atrium; V1AFCL ¼ atrial fibrillation cycle length in ECG lead V1 (Heart Rhythm 2014;11:670–676) I 2014 Heart Rhythm Society. All rights reserved.
with a lower likelihood of reversion from persistent AF to sinus rhythm, whether this occurs spontaneously,3 as a result of pharmacologic reversion,4,5 or with catheter ablation.6 Other studies have also shown that a shorter AFCL is associated with a higher rate of recurrent AF.7,8 Set against this indirect evidence that a shorter noninvasive AFCL may reflect more advanced atrial remodeling a longer LA AFCL has been directly related to a greater extent of atrial fibrosis in a postcardiac surgical population with AF.9 However, it is likely that there are multiple determinants of change in both the local intracardiac AFCL and the AFCL in the ECG in association with atrial remodeling. Simple wavelength considerations dictate that either or both refractoriness and conduction velocity will influence the local AFCL. Previous studies have shown a high correlation between the refractory period measured by programmed stimulation and the minimum local AFCL during fibrillation.10 In addition, http://dx.doi.org/10.1016/j.hrthm.2013.12.034
Walters et al
Relationship Between ECG AFCL and LA Remodeling
ablation studies have suggested that intracardiac AFCL may be influenced by the number and rate of AF sources.11 No study has yet attempted a direct comparison between the AFCL derived from surface ECG and the extent of atrial remodeling in AF, so the ability of this relatively simple measure to estimate the complexities of atrial remodeling remains uncertain. Therefore, we sought to directly compare the AFCL of ECG lead V1 (V1AFCL) with the extent of such remodeling, as assessed by detailed electroanatomic mapping in a cohort of patients with AF undergoing catheter ablation. We hypothesized that a longer V1AFCL would reflect more advanced remodeling manifest by slower intraatrial conduction, more fractionated electrograms, and lower endocardial bipolar voltage.
Methods The study population consisted of 23 patients undergoing a first catheter ablation of symptomatic drug-refractory AF who presented to the electrophysiology laboratory in AF. All patients provided written informed consent before enrollment. The study protocol was approved by the appropriate Human Research and Ethics Committees.
Electroanatomic mapping protocol The procedure was performed with patients in the fasting state under general anesthesia. All antiarrhythmic medications were withheld for at least 5 half-lives before the procedure. Amiodarone was ceased at least 3 months earlier. All patients underwent transthoracic echocardiography to define cardiac structure and function and then underwent transesophageal echocardiography on the day of the procedure to exclude LA thrombus. A 10-pole catheter (2-5-2mm interelectrode spacing) was positioned in the coronary sinus (CS) with the proximal bipole at the ostium, and a 6-pole catheter was placed at the His-bundle/right ventricular recording position. Double transseptal access was obtained with standard techniques. After establishing LA access, intravenous heparin was administered with a target activated clotting time of 350 seconds. Both an externally irrigated 3.5-mm ablation catheter (2-5-2-mm interelectrode spacing; Biosense Webster, Johnson & Johnson, Diamond Bar, CA) and a multipolar circular mapping catheter were introduced into the LA. Bipolar intracardiac electrograms and 12-lead surface ECG were recorded simultaneously on a computerized digital amplifier system (EPMed Systems, West Berlin, NJ). Intracardiac electrograms were filtered between 30 and 500 Hz, and the surface ECG was filtered between 0.05 and 40 Hz. All procedures were performed with the assistance of a 3-dimensional electroanatomic mapping system (EnSite NavX, St. Jude Medical, St Paul, MN). Cardiac computerized tomography or magnetic resonance images acquired preoperatively were integrated into this system. Using a methodology that we reported in previous studies, voltage maps were created during AF.1,12–14 After electrical cardioversion, voltage and activation maps again were
671
created during pacing from the distal coronary sinus (CSd) at a constant 600-ms cycle length so as to standardize the speed and direction of wavefront propagation. The complexity of intracardiac electrograms was analyzed during paced rhythm. The ablation catheter was used to collect electroanatomic datapoints, with all data collected before tissue ablation. Points were manually acquired after careful evaluation of tissue contact based on tactile catheter pressure, the motion of the catheter during fluoroscopic imaging, the catheter icon-to-surface feature of the mapping system, and the presence of constant electrogram characteristics. Surface color projection with an interpolation fill threshold of 10 mm was used to ensure a minimum number of points in an even distribution throughout the LA. Each individual electrogram was manually verified offline. When mapping during AF, each site was sampled over a 5-second recording period, consistent with prior publications demonstrating that this interval allows accurate characterization of atrial activation during AF.15,16 When mapping during pacing, points acquired after ectopic beats were excluded from the analysis. Voltage mapping Local voltage during AF was defined as the amplitude between the maximum peak positive and minimum peak negative deflections over the 5-second recording period. Local voltage during paced rhythm was defined as the amplitude between the absolute peak positive and peak negative deflections of the electrogram. Bipolar voltage was annotated with the assistance of automated algorithms, followed by manual verification of each individual point. Appropriate time windows were applied during paced rhythm to exclude stimulus artifact and far-field ventricular electrograms. A baseline noise threshold of 0.05 mV was applied to exclude background system noise. The LA was considered globally, with mean LA voltage determined from all datapoints. Low-voltage zones were defined as those with bipolar voltage o0.5 mV.17 The proportion of low-voltage signals in the LA was calculated by dividing the number of low-voltage signals by the total number of datapoints. Atrial conduction Local activation time for each point on the pace-map was annotated to the earliest sharp deflection. Total LA activation time was determined from the pacing stimulus to the latest point of LA activation. The surface area of the LA was measured offline using NavX system software, and an activation index was calculated to adjust for different LA surface areas as shown in Equation 1: LA activation time cm2 =ms ð1Þ Activation index ¼ LA surface area
Signal fractionation Complex fractionated electrograms during paced rhythm were defined as those with Z3 deflections and 450-ms duration or those with 2 separate deflections separated by an
672 isoelectric interval.18 Analysis of LA signal complexity was performed globally, with the percentage of complex signals calculated by dividing the number of complex signals by the total number of datapoints. Atrial refractoriness Effective refractory period (ERP) testing was performed after voltage and activation mapping. ERPs were measured at the posterior LA wall, the LA appendage, and the distal CS using a pacing drive train of 8 beats followed by a single extrastimulus commencing at a coupling interval of 150 ms and incrementing by 10 ms until local capture was demonstrated. The local ERP was defined as the longest extrastimulus that failed to capture the pacing site. ERPs were measured at drive train cycle lengths of 600 and 450 ms and were repeated at least twice at each site to ensure consistency.
ECG V1 AF cycle length analysis A continuous 12-lead ECG was recorded throughout the procedure and was exported for offline analysis using the customized signals analysis software Cardiac ElectroPhysiology Analysis System (CEPAS, Cuoretech Pty Ltd, Sydney, Australia). The AF signal was analyzed in 10-second time aliquots, with the mean V1AFCL determined over 90 seconds. Simultaneous intracardiac electrograms recorded from the proximal and distal bipoles of the CS catheter also were exported. Digital ECG signals were filtered with a notch filter at 50 Hz, a low-pass elliptic filter with pass frequency of 40 Hz and stop frequency of 80 Hz, and a baseline wavelet filter. To isolate the fibrillatory baseline from lead V1, a templatematching QRST cancellation algorithm similar to that used in previous publications was applied.5,19 An ECG channel with a clear QRST segment was chosen for automated annotation of Q-wave onset and T-wave offset positions throughout the 10-second recording, with these annotations manually verified. The average QRST template so derived was then used for automated QRST subtraction from the lead V1 ECG channel in order to derive the atrial fibrillatory baseline in this lead (Figure 1). The power spectrum of the fibrillatory baseline was subsequently derived by frequency analysis. The signal was half-wave rectified and edge-tapered with a Hanning window before Fourier transformation using zero padding. The dominant frequency (DF) of the fibrillatory baseline in ECG lead V1 was defined as the frequency demonstrating the highest power within the frequency domain of 3 to 15 Hz. The inverse of the DF was taken as an estimate of the dominant AF cycle length in lead V1 (V1AFCL), calculated using Equation 2: 1000 ðmsÞ ð2Þ V1AFCL ¼ DF The mean AFCL in lead V1 was also determined directly by direct annotation of the QRST-subtracted fibrillatory baseline within CEPAS.
Heart Rhythm, Vol 11, No 4, April 2014 The bipolar electrograms recorded by the CS catheter were filtered with a 50-Hz notch filter. The mean AFCL from each 10-second aliquot of intracardiac atrial signal was determined by direct annotation of the atrial signal using an automated algorithm within CEPAS, which annotated the steepest negative deflections within each electrogram. Similar to commercially available software, CEPAS has specific user-defined functions for accurate identification of electrogram activations that include a baseline noise threshold, a maximum electrogram width criterion to avoid detection of broad far-field activations, an electrogram slope criterion, and an electrogram “refractory” period to avoid multiple detections within the same activation. The baseline noise threshold was adjusted according to the baseline noise of the electrogram recordings, and an electrogram width criterion of 10 ms with a refractory period of 50 ms were used. Automated electrogram annotations were verified and edited manually as necessary. The power spectrum of the atrial signal from each extracted 10-second aliquot was also derived by frequency analysis, as described for the surface ECG signal, with the inverse of the DF taken as an estimate of the dominant intracardiac AFCL.
Statistical analysis All statistical analysis was performed using STATA software (version 12.1, StataCorp, College Station, TX). Continuous variables are expressed as mean ⫾ SD and categorical variables as number of subjects (%). Normality of distribution was tested with the Shapiro-Wilk method. Correlation between normally distributed variables was tested using the Pearson product-moment correlation and that between nonparametric variables was tested using the Spearman rank correlation. Simple linear regression was used to further define the relationship between independent and dependent variables. Bland-Altman plots of the measurement mean against the difference were used to compare 2 methods of measuring the same variable. Statistical significance was tested at the .05 level.
Results Data from 23 subjects (91% male; mean age 58 ⫾ 8 years; Table 1) with persistent AF were analyzed. Median symptom duration was 24 months (interquartile range 12–48), and mean LA area was 27.2 ⫾ 5.9 cm2. There was a relatively low rate of cardiovascular comorbidities and exposure to antiarrhythmic drugs.
V1AFCL measurement Mean V1AFCL over 90 seconds derived by frequency power spectrum analysis was 169 ⫾ 17 ms and that derived by direct annotation was 167 ⫾ 17 ms. A Bland-Altman comparison of the V1AFCL values generated by each method revealed an average difference of 1.8 ms, with the 95% limits of agreement at –4.8 and 8.3 ms. There was a similarly high agreement between the intracardiac AFCL determined from CS recordings using direct
Walters et al
Relationship Between ECG AFCL and LA Remodeling
673
Figure 1 Signal processing method used to determine the dominant frequency in a 10-second sample of atrial fibrillation. A: Lead V1 recording, with a prominent fibrillatory baseline and an indistinct QRST segment. B: Lead V5 recording, with a clear QRST complex allowing accurate annotation. C: Fibrillatory baseline after digital QRST subtraction, used for frequency power spectrum analysis and for direct annotation to derive the mean atrial fibrillation cycle length in ECG lead V1 (V1AFCL). D: Frequency power spectrum of the signal in panel C, with the power concentrated in a single relatively narrow peak at 7.3 Hz.
annotation of the atrial signal and analysis of the frequency power spectrum. For the distal CS bipole there was a mean difference of 1.3 ms, with 95% limits of agreement at –1.7 and 4.2 ms. For the proximal CS bipole, the mean difference was 3.1 ms, with 95% limits of agreement at –7.4 and 1.3 ms.
annotation to derive both values, the correlation for the proximal CS was 0.78 (P ¼ .0001) and for the distal CS was 0.57 (P ¼ .01). Using power spectrum analysis, the equivalent paired correlation coefficients were 0.72 (P ¼ .001) and 0.60 (P ¼ .01), respectively.
Relationship between V1AFCL and intracardiac AFCL
Relationship between V1AFCL and parameters of LA remodeling
A significant relationship was found between the V1AFCL over 90 seconds and the simultaneous mean AFCL values determined from both the proximal and distal bipoles of the CS recording catheter (Table 2 and Figure 2). Using direct Table 1
Baseline clinical data (n ¼ 23)
Age (years) Male gender Atrial fibrillation duration (months) Left ventricular ejection fraction (%) Left atrial area (cm2) Hypertension Diabetes Heart failure Stroke/transient ischemic attack Warfarin Beta-blocker Calcium channel blocker Sotalol Flecainide Amiodarone Values are given as number (percent) or mean ⫾ SD.
57.6 ⫾ 8.2 21 (91) 34 ⫾ 31 58 ⫾ 7 27.2 ⫾ 5.9 3 (12) 1 (4) 3 (12) 2 (8) 14 (56) 9 (36) 5 (20) 4 (16) 3 (12) 1 (4)
V1AFCL and LA size There was no association between mean V1AFCL and symptom duration (r ¼ 0.02, P ¼ .9 for V1AFCL derived by frequency analysis; r ¼ 0.002, P ¼ .99 for V1AFCL measured directly). There was no significant association between mean V1AFCL and either LA volume (r ¼ –0.18, P ¼ .5 for V1AFCL derived by frequency analysis; r ¼ – 0.26, P ¼ .3 for V1AFCL measured directly) or LA area (r ¼ 0.003, P ¼ .9 for V1AFCL derived by frequency analysis; r ¼ 0.06, P ¼ .8 for V1AFCL measured directly) (Table 2). V1AFCL and atrial refractoriness No significant association was observed between V1AFCL and any LA ERP variable. For example, at a drive train cycle length of 600 ms, there was no association with ERP at the posterior LA wall (r ¼ –0.20, P ¼ 0.7 for V1AFCL derived by frequency analysis; r ¼ –0.16, P ¼ .8 for V1AFCL measured directly), LA appendage (r ¼ 0.80, P ¼ .2 for V1AFCL derived by frequency analysis; r ¼ 0.90, P ¼ .1 for
674 Table 2
Heart Rhythm, Vol 11, No 4, April 2014 Relationship between V1AFCL and parameters of LA electroanatomic remodeling
CSp AFCL CSd AFCL LA bipolar voltage* LA percent low voltage LA area LA volume LA total activation time LA activation index LA complex fractionated electrogram
V1AFCL derived by frequency spectrum analysis
V1AFCL measured directly
Adjusted R2
t
P value
Adjusted R2
t
P value
0.37 0.27 0.06 0.12 –0.09 0.03 0.33 0.22 0.35
3.22 2.75 –1.4 1.6 0.01 –0.7 2.8 2.3 2.8
.001 .01 .2 .1 .9 .5 .01 .01 .02
0.56 0.33 0.03 0.10 –0.09 0.01 0.36 0.28 0.34
4.92 3.24 –1.2 1.5 0.2 –1.1 3.0 2.6 2.8
o.001 .01 .3 .2 .8 .3 .01 .007 .02
AFCL ¼ atrial fibrillation cycle length; CSd ¼ coronary sinus distal; CSp ¼ coronary sinus proximal; LA ¼left atrial; V1AFCL ¼ atrial fibrillation cycle length in ECG lead V1. * Bipolar voltage recorded during stable pacing from the distal coronary sinus.
V1AFCL measured directly), or distal CS (r ¼ –0.58, P ¼ .4 for V1AFCL derived by frequency analysis; r ¼ –0.48, P ¼ .5 for V1AFCL measured directly).
V1AFCL and LA voltage There was no significant association between V1AFCL and any of the following: (1) mean LA voltage recorded during paced rhythm (r ¼ –0.35, P ¼ .2 for V1AFCL derived by frequency analysis; r ¼ –0.31, P ¼ .3 for V1AFCL measured directly) (Table 2); (2) proportion of low-voltage electrograms recorded during paced rhythm (r ¼ 0.44, P ¼ .1 for V1AFCL derived by frequency analysis; r ¼ 0.42, P ¼ .2 for V1AFCL measured directly); or (3) mean LA voltage recorded during AF (r ¼ 0.04, P ¼ .9 for V1AFCL derived by frequency analysis; r ¼ –0.05, P ¼ .9 for V1AFCL measured directly). V1AFCL and LA conduction There was a significant association between V1AFCL and LA total activation time (Table 2 and Figure 3), with slower LA conduction reflected in significantly longer AF cycle lengths and lower DF. The association persisted after derivation of the LA activation index to adjust for the influence of LA size. V1AFCL and LA signal complexity There was a significant association between V1AFCL and the extent of signal fractionation in the LA, with a higher proportion of complex signals being reflected in a significantly longer V1AFCL (Table 2 and Figure 4).
Discussion The key findings of this study are as follows:
Figure 2 Relationship between mean atrial fibrillation cycle length (AFCL) recorded from the coronary sinus (CS) recording catheter and mean AFCL in ECG lead V1 (V1AFCL) determined from ECG lead V1. A: Distal recording bipole. B: Proximal recording bipole.
1. Using a QRST subtraction algorithm and either direct annotation or frequency power spectrum analysis of the fibrillatory baseline as equivalent techniques, there is a significant association between V1AFCL and the intracardiac AFCL measured from CS recordings. 2. V1AFCL provides a noninvasive reflection of important parameters of electroanatomic LA remodeling, including both conduction and signal fractionation, with a longer V1AFCL associated with more advanced remodeling
Walters et al
Relationship Between ECG AFCL and LA Remodeling
Figure 3 Relationship between atrial fibrillation cycle length in ECG lead V1 (V1AFCL) and left atrial (LA) conduction. A: Total LA activation time. B: LA activation index.
marked by slower conduction and a greater degree of signal fractionation. 3. V1AFCL shows no significant relationship to atrial refractoriness, LA size, or LA endocardial bipolar voltage. Prior studies have demonstrated that the AFCL derived from noninvasive recordings reflects intracardiac atrial cycle length and therefore potentially provides a surrogate marker of atrial remodeling in human AF. A range of techniques has been developed to determine the noninvasive AFCL, ranging from direct measurement of the interval between consecutive unambiguous fibrillatory waves6 to techniques involving digital subtraction of the QRST complex and analysis of the frequency power spectrum contained within the AF signal,5,20 to still more advanced methods that analyze the AF signal in both the frequency and time domains simultaneously in order to appreciate temporal variation in the AFCL.21 Studies using these techniques have suggested that V1AFCL best represents a spatial average of right atrial cycle lengths.5,20,22 However, they also have reported
675
significant correlations between fibrillatory wave frequency characteristics in ECG lead V1 and the fibrillatory rate recorded in the LA,23 the CS,20,22 and even the pulmonary veins.22 The results of the current study, using techniques to analyze the AF signal analogous to those used in previously published studies to demonstrate a relationship with the intracardiac AFCL derived from CS recordings, are consistent with these data. Several studies using advanced signals analysis techniques have demonstrated that a shorter noninvasive AFCL is associated with clinical behavior that might be expected in patients with more advanced atrial remodeling. For example, over a period of months of sustained AF or repetitive paroxysmal AF, the cycle length in the surface ECG has been reported to significantly shorten24,25 and then, in the subset with spontaneous termination, to lengthen immediately before resumption of sinus rhythm.26–28 A shorter cycle length has also been found to predict resistance to spontaneous or pharmacologic AF termination3–5 and a higher rate of recurrent AF after cardioversion7,8 or sequential stepwise AF ablation.6 In the current study, we observed that longer V1AFCL was actually associated with more advanced atrial remodeling marked by conduction slowing and signal complexity when this was measured directly. We suggest that V1AFCL may reflect atrial conduction slowing as a result of the development of atrial fibrosis.29 A recent study of the atrial fibrillatory rate in patients with persistent AF and heart failure demonstrated that AFCL lengthens with age and that a longer AFCL is independently associated with more severe heart failure and higher mortality.30 It has also been reported that after cardiac surgery there is an association between the risk of AF and the extent of atrial fibrosis, and that in those who do develop AF more extensive fibrosis is associated with a longer V1AFCL.9 Also consistent with these observations are recent data demonstrating that intracardiac electrograms have a longer cycle length when recorded from
Figure 4 Relationship between mean atrial fibrillation cycle length in ECG lead V1 (V1AFCL) and the proportion of complex electrograms recorded from the left atrial (LA) endocardium during distal coronary sinus pacing.
676 atrial regions with dense scar demonstrated on atrial magnetic resonance imaging.31 It is possible that a short V1AFCL not only is reflective of atrial wavelength but also is an indicator of the number of atrial sources driving AF. This may explain in part why a short V1AFCL is associated with failure of maintenance of sinus rhythm after DCR or catheter ablation. This concept is consistent with that suggested by previous computer modeling that related a longer AFCL in the atrial appendages to a lower number of independent sources of AF.11 In this study, V1AFCL did not reflect all parameters of LA remodeling. In particular, there was no relationship with LA size, voltage, or refractoriness. Although it is recognized that LA remodeling in AF can affect all variables analyzed in this study, it is quite conceivable that not all will change in parallel with each other during the disease process. Furthermore, it is quite reasonable for noninvasive measurement of V1AFCL, as a single relatively simple measure, not to reflect the full complexity of the mechanisms by which AF is sustained in the remodeled atrial substrate. It is reasonable for AFCL to change as tissue conduction velocity changes even in the absence of change in local tissue refractoriness, the second determinant of wavelength.
Study limitations This study is limited by the requirement for participants to present to the electrophysiology laboratory in AF, thus restricting the extent of the spectrum of atrial disease studied. It may be that conclusions drawn from this population with persistent AF do not extend to those with paroxysmal AF, in whom electrical triggers would be expected to play a relatively larger role than the LA substrate would and in whom the contribution of focal sources and rotors may be different.
Conclusion A longer atrial fibrillatory cycle length derived from ECG lead V1 is associated with more advanced LA electroanatomic remodeling marked by slower atrial conduction and a greater number of complex electrograms.
Heart Rhythm, Vol 11, No 4, April 2014
8.
9.
10.
11.
12.
13. 14.
15.
16. 17.
18.
19.
20.
21.
22.
23.
24.
25.
References 1. Teh AW, Kistler PM, Lee G, et al. Electroanatomic remodeling of the left atrium in paroxysmal and persistent atrial fibrillation patients without structural heart disease. J Cardiovasc Electrophysiol 2012;23:232–238. 2. Nattel S, Burstein B, Dobrev D. Atrial remodeling and atrial fibrillation: mechanisms and implications. Circ Arrhythm Electrophysiol 2008;1:62–73. 3. Nilsson F, Stridh M, Bollmann A, Sörnmo L. Predicting spontaneous termination of atrial fibrillation using the surface ECG. Med Eng Phys 2006;28: 802–808. 4. Husser D, Cannom DS, Bhandari AK, et al. Electrocardiographic characteristics of fibrillatory waves in new-onset atrial fibrillation. Europace 2007;9: 638–642. 5. Bollmann A, Kanuru NK, McTeague KK, Walter PF, DeLurgio DB, Langberg JJ. Frequency analysis of human atrial fibrillation using the surface electrocardiogram and its response to ibutilide. Am J Cardiol 1998;81:1439–1445. 6. Matsuo S, Lellouche N, Wright M, et al. Clinical predictors of termination and clinical outcome of catheter ablation for persistent atrial fibrillation. J Am Coll Cardiol 2009;54:788–795. 7. Husser D, Bollmann A, Kang S, et al. Determinants and prognostic significance of immediate atrial fibrillation recurrence following cardioversion in patients
26.
27.
28.
29.
30.
31.
undergoing pulmonary vein isolation. Pacing Clin Electrophysiol 2005;28: 119–125. Holmqvist F, Stridh M, Waktare JEP, Sörnmo L, Olsson SB, Meurling CJ. Atrial fibrillatory rate and sinus rhythm maintenance in patients undergoing cardioversion of persistent atrial fibrillation. Eur Heart J 2006;27:2201–2207. Swartz MF, Fink GW, Lutz CJ, et al. Left versus right atrial difference in dominant frequency, K(þ) channel transcripts, and fibrosis in patients developing atrial fibrillation after cardiac surgery. Heart Rhythm 2009;6:1415–1422. Kim KB, Rodefeld MD, Schuessler RB, Cox JL, Boineau JP. Relationship between local atrial fibrillation interval and refractory period in the isolated canine atrium. Circulation 1996;94:2961–2967. Haïssaguerre M, Lim KT, Jacquemet V, et al. Atrial fibrillatory cycle length: computer simulation and potential clinical importance. Europace 2007;9(Suppl 6):64–70. Kistler PM, Sanders P, Fynn SP, et al. Electrophysiologic and electroanatomic changes in the human atrium associated with age. J Am Coll Cardiol 2004;44: 109–116. Sanders P, Morton JB, Davidson NC, et al. Electrical remodeling of the atria in congestive heart failure. Circulation 2003;108:1461–1468. Teh AW, Kistler PM, Lee G, et al. The relationship between complex fractionated electrograms and atrial low-voltage zones during atrial fibrillation and paced rhythm. Europace 2011;13:1709–1716. Stiles MK, Brooks AG, John B, et al. The effect of electrogram duration on quantification of complex fractionated atrial electrograms and dominant frequency. J Cardiovasc Electrophysiol 2008;19:252–258. Lin Y-J, Tai C-T, Kao T, et al. Consistency of complex fractionated atrial electrograms during atrial fibrillation. Heart Rhythm 2008;5:406–412. Stiles MK, John B, Wong CX, et al. Paroxysmal lone atrial fibrillation is associated with an abnormal atrial substrate: characterizing the “second factor.”. J Am Coll Cardiol 2009;53:1182–1191. Rostock T, Rotter M, Sanders P, et al. High-density activation mapping of fractionated electrograms in the atria of patients with paroxysmal atrial fibrillation. Heart Rhythm 2006;3:27–34. Bollmann A, Husser D, Mainardi L, et al. Analysis of surface electrocardiograms in atrial fibrillation: techniques, research, and clinical applications. Europace 2006;8:911–926. Holm M, Pehrson S, Ingemansson M, et al. Non-invasive assessment of the atrial cycle length during atrial fibrillation in man: introducing, validating and illustrating a new ECG method. Cardiovasc Res 1998;38:69–81. Stridh M, Sörnmo L, Meurling CJ, Olsson SB. Characterization of atrial fibrillation using the surface ECG: time-dependent spectral properties. IEEE Trans Biomed Eng 2001;48:19–27. Husser D, Stridh M, Cannom DS, et al. Validation and clinical application of time-frequency analysis of atrial fibrillation electrocardiograms. J Cardiovasc Electrophysiol 2007;18:41–46. Petrutiu S, Sahakian AV, Fisher W, Swiryn S. Manifestation of left atrial events and interatrial frequency gradients in the surface electrocardiogram during atrial fibrillation: contributions from posterior leads. J Cardiovasc Electrophysiol 2009;20:1231–1236. Sasaki T, Niwano S, Sasaki S, et al. Long-term follow-up of changes in fibrillation waves in patients with persistent atrial fibrillation: spectral analysis of surface ECG. Circ J 2006;70:169–173. Niwano S, Fukaya H, Sasaki T, Hatakeyama Y, Fujiki A, Izumi T. Effect of oral L-type calcium channel blocker on repetitive paroxysmal atrial fibrillation: spectral analysis of fibrillation waves in the Holter monitoring. Europace 2007;9: 1209–1215. Petrutiu S, Sahakian AV, Swiryn S. Short-term dynamics in fibrillatory wave characteristics at the onset of paroxysmal atrial fibrillation in humans. J Electrocardiol 2007;40:155–160. Petrutiu S, Sahakian AV, Swiryn S. Abrupt changes in fibrillatory wave characteristics at the termination of paroxysmal atrial fibrillation in humans. Europace 2007;9:466–470. Fujiki A, Sakabe M, Nishida K, Mizumaki K, Inoue H. Role of fibrillation cycle length in spontaneous and drug-induced termination of human atrial fibrillation. Circ J 2003;67:391–395. Tanaka K, Zlochiver S, Vikstrom KL, et al. Spatial distribution of fibrosis governs fibrillation wave dynamics in the posterior left atrium during heart failure. Cirs Res 2007;101:839–847. Platonov PG, Stridh M, de Melis M, et al. Analysis of atrial fibrillatory rate during spontaneous episodes of atrial fibrillation in humans using implantable loop recorder electrocardiogram. J Electrocardiol 2012;45:723–726. Jadidi AS, Cochet H, Shah AJ, et al. Inverse relationship between fractionated electrograms and atrial fibrosis in persistent atrial fibrillation: combined magnetic resonance imaging and high-density mapping. J Am Coll Cardiol 2013;62:802–812.
proximal and distal coronary sinus recordings. A longer V1AFCL was associated with slower left atrial conduction velocity and greater signal complexity but not with other remodeling variables, including left atrial size, atrial refractoriness, and mean endocardial voltage.
METHODS Twenty-three patients in atrial fibrillation (AF) who presented for AF ablation underwent detailed electroanatomic mapping of the left atrium. The digital 12-lead ECG was exported for offline analysis, with signal filtering and QRST subtraction used to reveal the fibrillatory baseline in lead V1. Mean V1AFCL was determined by direct annotation of the fibrillatory baseline, and the corresponding dominant V1AFCL was determined by Fourier transformation to derive the dominant frequency from the frequency power spectrum. The simultaneous AFCL from proximal and distal coronary sinus recordings was determined using the same methods. The strength of the association between various left atrial remodeling variables and V1AFCL was determined.
CONCLUSION A longer atrial fibrillatory cycle length in surface ECG lead V1 is significantly associated with parameters of more advanced left atrial electroanatomic remodeling, specifically slower atrial conduction and more extensive electrogram fractionation.
RESULTS The 2 methods of deriving V1AFCL and intracardiac AFCL were found to produce highly equivalent results. V1AFCL showed significant correlation with intracardiac AFCL derived from both
Introduction Atrial fibrillation (AF) is associated with progressive changes in left atrial (LA) electrophysiology, the extent of which can be assessed by invasive techniques to characterize refractoriness, conduction, voltage amplitude, and electrogram complexity.1,2 Previous studies have used the cycle length of atrial activation (AFCL) in the electrocardiogram (ECG) as a noninvasive indirect marker of atrial remodeling. These studies have observed that a shorter AFCL is associated Dr. Walters is a recipient of a research scholarship from the National Health and Medical Research Council of Australia and the National Heart Foundation of Australia. Address reprint requests and correspondence: Dr. Jonathan M. Kalman, Royal Melbourne Hospital, Melbourne Heart Centre; Melbourne, Victoria 3050, Australia. E-mail address: jon.kalman@ mh.org.au.
1547-5271/$-see front matter B 2014 Heart Rhythm Society. All rights reserved.
KEYWORDS Atrial fibrillation; Cycle length; Left atrium; Remodeling ABBREVIATIONS AF ¼ atrial fibrillation; AFCL ¼ atrial fibrillation cycle length; CS ¼ coronary sinus; CSd ¼ distal coronary sinus recording bipole; CSp ¼ proximal coronary sinus recording bipole; DF ¼ dominant frequency; ECG ¼ electrocardiogram; ERP ¼ effective refractory period; LA ¼ left atrium; V1AFCL ¼ atrial fibrillation cycle length in ECG lead V1 (Heart Rhythm 2014;11:670–676) I 2014 Heart Rhythm Society. All rights reserved.
with a lower likelihood of reversion from persistent AF to sinus rhythm, whether this occurs spontaneously,3 as a result of pharmacologic reversion,4,5 or with catheter ablation.6 Other studies have also shown that a shorter AFCL is associated with a higher rate of recurrent AF.7,8 Set against this indirect evidence that a shorter noninvasive AFCL may reflect more advanced atrial remodeling a longer LA AFCL has been directly related to a greater extent of atrial fibrosis in a postcardiac surgical population with AF.9 However, it is likely that there are multiple determinants of change in both the local intracardiac AFCL and the AFCL in the ECG in association with atrial remodeling. Simple wavelength considerations dictate that either or both refractoriness and conduction velocity will influence the local AFCL. Previous studies have shown a high correlation between the refractory period measured by programmed stimulation and the minimum local AFCL during fibrillation.10 In addition, http://dx.doi.org/10.1016/j.hrthm.2013.12.034
Walters et al
Relationship Between ECG AFCL and LA Remodeling
ablation studies have suggested that intracardiac AFCL may be influenced by the number and rate of AF sources.11 No study has yet attempted a direct comparison between the AFCL derived from surface ECG and the extent of atrial remodeling in AF, so the ability of this relatively simple measure to estimate the complexities of atrial remodeling remains uncertain. Therefore, we sought to directly compare the AFCL of ECG lead V1 (V1AFCL) with the extent of such remodeling, as assessed by detailed electroanatomic mapping in a cohort of patients with AF undergoing catheter ablation. We hypothesized that a longer V1AFCL would reflect more advanced remodeling manifest by slower intraatrial conduction, more fractionated electrograms, and lower endocardial bipolar voltage.
Methods The study population consisted of 23 patients undergoing a first catheter ablation of symptomatic drug-refractory AF who presented to the electrophysiology laboratory in AF. All patients provided written informed consent before enrollment. The study protocol was approved by the appropriate Human Research and Ethics Committees.
Electroanatomic mapping protocol The procedure was performed with patients in the fasting state under general anesthesia. All antiarrhythmic medications were withheld for at least 5 half-lives before the procedure. Amiodarone was ceased at least 3 months earlier. All patients underwent transthoracic echocardiography to define cardiac structure and function and then underwent transesophageal echocardiography on the day of the procedure to exclude LA thrombus. A 10-pole catheter (2-5-2mm interelectrode spacing) was positioned in the coronary sinus (CS) with the proximal bipole at the ostium, and a 6-pole catheter was placed at the His-bundle/right ventricular recording position. Double transseptal access was obtained with standard techniques. After establishing LA access, intravenous heparin was administered with a target activated clotting time of 350 seconds. Both an externally irrigated 3.5-mm ablation catheter (2-5-2-mm interelectrode spacing; Biosense Webster, Johnson & Johnson, Diamond Bar, CA) and a multipolar circular mapping catheter were introduced into the LA. Bipolar intracardiac electrograms and 12-lead surface ECG were recorded simultaneously on a computerized digital amplifier system (EPMed Systems, West Berlin, NJ). Intracardiac electrograms were filtered between 30 and 500 Hz, and the surface ECG was filtered between 0.05 and 40 Hz. All procedures were performed with the assistance of a 3-dimensional electroanatomic mapping system (EnSite NavX, St. Jude Medical, St Paul, MN). Cardiac computerized tomography or magnetic resonance images acquired preoperatively were integrated into this system. Using a methodology that we reported in previous studies, voltage maps were created during AF.1,12–14 After electrical cardioversion, voltage and activation maps again were
671
created during pacing from the distal coronary sinus (CSd) at a constant 600-ms cycle length so as to standardize the speed and direction of wavefront propagation. The complexity of intracardiac electrograms was analyzed during paced rhythm. The ablation catheter was used to collect electroanatomic datapoints, with all data collected before tissue ablation. Points were manually acquired after careful evaluation of tissue contact based on tactile catheter pressure, the motion of the catheter during fluoroscopic imaging, the catheter icon-to-surface feature of the mapping system, and the presence of constant electrogram characteristics. Surface color projection with an interpolation fill threshold of 10 mm was used to ensure a minimum number of points in an even distribution throughout the LA. Each individual electrogram was manually verified offline. When mapping during AF, each site was sampled over a 5-second recording period, consistent with prior publications demonstrating that this interval allows accurate characterization of atrial activation during AF.15,16 When mapping during pacing, points acquired after ectopic beats were excluded from the analysis. Voltage mapping Local voltage during AF was defined as the amplitude between the maximum peak positive and minimum peak negative deflections over the 5-second recording period. Local voltage during paced rhythm was defined as the amplitude between the absolute peak positive and peak negative deflections of the electrogram. Bipolar voltage was annotated with the assistance of automated algorithms, followed by manual verification of each individual point. Appropriate time windows were applied during paced rhythm to exclude stimulus artifact and far-field ventricular electrograms. A baseline noise threshold of 0.05 mV was applied to exclude background system noise. The LA was considered globally, with mean LA voltage determined from all datapoints. Low-voltage zones were defined as those with bipolar voltage o0.5 mV.17 The proportion of low-voltage signals in the LA was calculated by dividing the number of low-voltage signals by the total number of datapoints. Atrial conduction Local activation time for each point on the pace-map was annotated to the earliest sharp deflection. Total LA activation time was determined from the pacing stimulus to the latest point of LA activation. The surface area of the LA was measured offline using NavX system software, and an activation index was calculated to adjust for different LA surface areas as shown in Equation 1: LA activation time cm2 =ms ð1Þ Activation index ¼ LA surface area
Signal fractionation Complex fractionated electrograms during paced rhythm were defined as those with Z3 deflections and 450-ms duration or those with 2 separate deflections separated by an
672 isoelectric interval.18 Analysis of LA signal complexity was performed globally, with the percentage of complex signals calculated by dividing the number of complex signals by the total number of datapoints. Atrial refractoriness Effective refractory period (ERP) testing was performed after voltage and activation mapping. ERPs were measured at the posterior LA wall, the LA appendage, and the distal CS using a pacing drive train of 8 beats followed by a single extrastimulus commencing at a coupling interval of 150 ms and incrementing by 10 ms until local capture was demonstrated. The local ERP was defined as the longest extrastimulus that failed to capture the pacing site. ERPs were measured at drive train cycle lengths of 600 and 450 ms and were repeated at least twice at each site to ensure consistency.
ECG V1 AF cycle length analysis A continuous 12-lead ECG was recorded throughout the procedure and was exported for offline analysis using the customized signals analysis software Cardiac ElectroPhysiology Analysis System (CEPAS, Cuoretech Pty Ltd, Sydney, Australia). The AF signal was analyzed in 10-second time aliquots, with the mean V1AFCL determined over 90 seconds. Simultaneous intracardiac electrograms recorded from the proximal and distal bipoles of the CS catheter also were exported. Digital ECG signals were filtered with a notch filter at 50 Hz, a low-pass elliptic filter with pass frequency of 40 Hz and stop frequency of 80 Hz, and a baseline wavelet filter. To isolate the fibrillatory baseline from lead V1, a templatematching QRST cancellation algorithm similar to that used in previous publications was applied.5,19 An ECG channel with a clear QRST segment was chosen for automated annotation of Q-wave onset and T-wave offset positions throughout the 10-second recording, with these annotations manually verified. The average QRST template so derived was then used for automated QRST subtraction from the lead V1 ECG channel in order to derive the atrial fibrillatory baseline in this lead (Figure 1). The power spectrum of the fibrillatory baseline was subsequently derived by frequency analysis. The signal was half-wave rectified and edge-tapered with a Hanning window before Fourier transformation using zero padding. The dominant frequency (DF) of the fibrillatory baseline in ECG lead V1 was defined as the frequency demonstrating the highest power within the frequency domain of 3 to 15 Hz. The inverse of the DF was taken as an estimate of the dominant AF cycle length in lead V1 (V1AFCL), calculated using Equation 2: 1000 ðmsÞ ð2Þ V1AFCL ¼ DF The mean AFCL in lead V1 was also determined directly by direct annotation of the QRST-subtracted fibrillatory baseline within CEPAS.
Heart Rhythm, Vol 11, No 4, April 2014 The bipolar electrograms recorded by the CS catheter were filtered with a 50-Hz notch filter. The mean AFCL from each 10-second aliquot of intracardiac atrial signal was determined by direct annotation of the atrial signal using an automated algorithm within CEPAS, which annotated the steepest negative deflections within each electrogram. Similar to commercially available software, CEPAS has specific user-defined functions for accurate identification of electrogram activations that include a baseline noise threshold, a maximum electrogram width criterion to avoid detection of broad far-field activations, an electrogram slope criterion, and an electrogram “refractory” period to avoid multiple detections within the same activation. The baseline noise threshold was adjusted according to the baseline noise of the electrogram recordings, and an electrogram width criterion of 10 ms with a refractory period of 50 ms were used. Automated electrogram annotations were verified and edited manually as necessary. The power spectrum of the atrial signal from each extracted 10-second aliquot was also derived by frequency analysis, as described for the surface ECG signal, with the inverse of the DF taken as an estimate of the dominant intracardiac AFCL.
Statistical analysis All statistical analysis was performed using STATA software (version 12.1, StataCorp, College Station, TX). Continuous variables are expressed as mean ⫾ SD and categorical variables as number of subjects (%). Normality of distribution was tested with the Shapiro-Wilk method. Correlation between normally distributed variables was tested using the Pearson product-moment correlation and that between nonparametric variables was tested using the Spearman rank correlation. Simple linear regression was used to further define the relationship between independent and dependent variables. Bland-Altman plots of the measurement mean against the difference were used to compare 2 methods of measuring the same variable. Statistical significance was tested at the .05 level.
Results Data from 23 subjects (91% male; mean age 58 ⫾ 8 years; Table 1) with persistent AF were analyzed. Median symptom duration was 24 months (interquartile range 12–48), and mean LA area was 27.2 ⫾ 5.9 cm2. There was a relatively low rate of cardiovascular comorbidities and exposure to antiarrhythmic drugs.
V1AFCL measurement Mean V1AFCL over 90 seconds derived by frequency power spectrum analysis was 169 ⫾ 17 ms and that derived by direct annotation was 167 ⫾ 17 ms. A Bland-Altman comparison of the V1AFCL values generated by each method revealed an average difference of 1.8 ms, with the 95% limits of agreement at –4.8 and 8.3 ms. There was a similarly high agreement between the intracardiac AFCL determined from CS recordings using direct
Walters et al
Relationship Between ECG AFCL and LA Remodeling
673
Figure 1 Signal processing method used to determine the dominant frequency in a 10-second sample of atrial fibrillation. A: Lead V1 recording, with a prominent fibrillatory baseline and an indistinct QRST segment. B: Lead V5 recording, with a clear QRST complex allowing accurate annotation. C: Fibrillatory baseline after digital QRST subtraction, used for frequency power spectrum analysis and for direct annotation to derive the mean atrial fibrillation cycle length in ECG lead V1 (V1AFCL). D: Frequency power spectrum of the signal in panel C, with the power concentrated in a single relatively narrow peak at 7.3 Hz.
annotation of the atrial signal and analysis of the frequency power spectrum. For the distal CS bipole there was a mean difference of 1.3 ms, with 95% limits of agreement at –1.7 and 4.2 ms. For the proximal CS bipole, the mean difference was 3.1 ms, with 95% limits of agreement at –7.4 and 1.3 ms.
annotation to derive both values, the correlation for the proximal CS was 0.78 (P ¼ .0001) and for the distal CS was 0.57 (P ¼ .01). Using power spectrum analysis, the equivalent paired correlation coefficients were 0.72 (P ¼ .001) and 0.60 (P ¼ .01), respectively.
Relationship between V1AFCL and intracardiac AFCL
Relationship between V1AFCL and parameters of LA remodeling
A significant relationship was found between the V1AFCL over 90 seconds and the simultaneous mean AFCL values determined from both the proximal and distal bipoles of the CS recording catheter (Table 2 and Figure 2). Using direct Table 1
Baseline clinical data (n ¼ 23)
Age (years) Male gender Atrial fibrillation duration (months) Left ventricular ejection fraction (%) Left atrial area (cm2) Hypertension Diabetes Heart failure Stroke/transient ischemic attack Warfarin Beta-blocker Calcium channel blocker Sotalol Flecainide Amiodarone Values are given as number (percent) or mean ⫾ SD.
57.6 ⫾ 8.2 21 (91) 34 ⫾ 31 58 ⫾ 7 27.2 ⫾ 5.9 3 (12) 1 (4) 3 (12) 2 (8) 14 (56) 9 (36) 5 (20) 4 (16) 3 (12) 1 (4)
V1AFCL and LA size There was no association between mean V1AFCL and symptom duration (r ¼ 0.02, P ¼ .9 for V1AFCL derived by frequency analysis; r ¼ 0.002, P ¼ .99 for V1AFCL measured directly). There was no significant association between mean V1AFCL and either LA volume (r ¼ –0.18, P ¼ .5 for V1AFCL derived by frequency analysis; r ¼ – 0.26, P ¼ .3 for V1AFCL measured directly) or LA area (r ¼ 0.003, P ¼ .9 for V1AFCL derived by frequency analysis; r ¼ 0.06, P ¼ .8 for V1AFCL measured directly) (Table 2). V1AFCL and atrial refractoriness No significant association was observed between V1AFCL and any LA ERP variable. For example, at a drive train cycle length of 600 ms, there was no association with ERP at the posterior LA wall (r ¼ –0.20, P ¼ 0.7 for V1AFCL derived by frequency analysis; r ¼ –0.16, P ¼ .8 for V1AFCL measured directly), LA appendage (r ¼ 0.80, P ¼ .2 for V1AFCL derived by frequency analysis; r ¼ 0.90, P ¼ .1 for
674 Table 2
Heart Rhythm, Vol 11, No 4, April 2014 Relationship between V1AFCL and parameters of LA electroanatomic remodeling
CSp AFCL CSd AFCL LA bipolar voltage* LA percent low voltage LA area LA volume LA total activation time LA activation index LA complex fractionated electrogram
V1AFCL derived by frequency spectrum analysis
V1AFCL measured directly
Adjusted R2
t
P value
Adjusted R2
t
P value
0.37 0.27 0.06 0.12 –0.09 0.03 0.33 0.22 0.35
3.22 2.75 –1.4 1.6 0.01 –0.7 2.8 2.3 2.8
.001 .01 .2 .1 .9 .5 .01 .01 .02
0.56 0.33 0.03 0.10 –0.09 0.01 0.36 0.28 0.34
4.92 3.24 –1.2 1.5 0.2 –1.1 3.0 2.6 2.8
o.001 .01 .3 .2 .8 .3 .01 .007 .02
AFCL ¼ atrial fibrillation cycle length; CSd ¼ coronary sinus distal; CSp ¼ coronary sinus proximal; LA ¼left atrial; V1AFCL ¼ atrial fibrillation cycle length in ECG lead V1. * Bipolar voltage recorded during stable pacing from the distal coronary sinus.
V1AFCL measured directly), or distal CS (r ¼ –0.58, P ¼ .4 for V1AFCL derived by frequency analysis; r ¼ –0.48, P ¼ .5 for V1AFCL measured directly).
V1AFCL and LA voltage There was no significant association between V1AFCL and any of the following: (1) mean LA voltage recorded during paced rhythm (r ¼ –0.35, P ¼ .2 for V1AFCL derived by frequency analysis; r ¼ –0.31, P ¼ .3 for V1AFCL measured directly) (Table 2); (2) proportion of low-voltage electrograms recorded during paced rhythm (r ¼ 0.44, P ¼ .1 for V1AFCL derived by frequency analysis; r ¼ 0.42, P ¼ .2 for V1AFCL measured directly); or (3) mean LA voltage recorded during AF (r ¼ 0.04, P ¼ .9 for V1AFCL derived by frequency analysis; r ¼ –0.05, P ¼ .9 for V1AFCL measured directly). V1AFCL and LA conduction There was a significant association between V1AFCL and LA total activation time (Table 2 and Figure 3), with slower LA conduction reflected in significantly longer AF cycle lengths and lower DF. The association persisted after derivation of the LA activation index to adjust for the influence of LA size. V1AFCL and LA signal complexity There was a significant association between V1AFCL and the extent of signal fractionation in the LA, with a higher proportion of complex signals being reflected in a significantly longer V1AFCL (Table 2 and Figure 4).
Discussion The key findings of this study are as follows:
Figure 2 Relationship between mean atrial fibrillation cycle length (AFCL) recorded from the coronary sinus (CS) recording catheter and mean AFCL in ECG lead V1 (V1AFCL) determined from ECG lead V1. A: Distal recording bipole. B: Proximal recording bipole.
1. Using a QRST subtraction algorithm and either direct annotation or frequency power spectrum analysis of the fibrillatory baseline as equivalent techniques, there is a significant association between V1AFCL and the intracardiac AFCL measured from CS recordings. 2. V1AFCL provides a noninvasive reflection of important parameters of electroanatomic LA remodeling, including both conduction and signal fractionation, with a longer V1AFCL associated with more advanced remodeling
Walters et al
Relationship Between ECG AFCL and LA Remodeling
Figure 3 Relationship between atrial fibrillation cycle length in ECG lead V1 (V1AFCL) and left atrial (LA) conduction. A: Total LA activation time. B: LA activation index.
marked by slower conduction and a greater degree of signal fractionation. 3. V1AFCL shows no significant relationship to atrial refractoriness, LA size, or LA endocardial bipolar voltage. Prior studies have demonstrated that the AFCL derived from noninvasive recordings reflects intracardiac atrial cycle length and therefore potentially provides a surrogate marker of atrial remodeling in human AF. A range of techniques has been developed to determine the noninvasive AFCL, ranging from direct measurement of the interval between consecutive unambiguous fibrillatory waves6 to techniques involving digital subtraction of the QRST complex and analysis of the frequency power spectrum contained within the AF signal,5,20 to still more advanced methods that analyze the AF signal in both the frequency and time domains simultaneously in order to appreciate temporal variation in the AFCL.21 Studies using these techniques have suggested that V1AFCL best represents a spatial average of right atrial cycle lengths.5,20,22 However, they also have reported
675
significant correlations between fibrillatory wave frequency characteristics in ECG lead V1 and the fibrillatory rate recorded in the LA,23 the CS,20,22 and even the pulmonary veins.22 The results of the current study, using techniques to analyze the AF signal analogous to those used in previously published studies to demonstrate a relationship with the intracardiac AFCL derived from CS recordings, are consistent with these data. Several studies using advanced signals analysis techniques have demonstrated that a shorter noninvasive AFCL is associated with clinical behavior that might be expected in patients with more advanced atrial remodeling. For example, over a period of months of sustained AF or repetitive paroxysmal AF, the cycle length in the surface ECG has been reported to significantly shorten24,25 and then, in the subset with spontaneous termination, to lengthen immediately before resumption of sinus rhythm.26–28 A shorter cycle length has also been found to predict resistance to spontaneous or pharmacologic AF termination3–5 and a higher rate of recurrent AF after cardioversion7,8 or sequential stepwise AF ablation.6 In the current study, we observed that longer V1AFCL was actually associated with more advanced atrial remodeling marked by conduction slowing and signal complexity when this was measured directly. We suggest that V1AFCL may reflect atrial conduction slowing as a result of the development of atrial fibrosis.29 A recent study of the atrial fibrillatory rate in patients with persistent AF and heart failure demonstrated that AFCL lengthens with age and that a longer AFCL is independently associated with more severe heart failure and higher mortality.30 It has also been reported that after cardiac surgery there is an association between the risk of AF and the extent of atrial fibrosis, and that in those who do develop AF more extensive fibrosis is associated with a longer V1AFCL.9 Also consistent with these observations are recent data demonstrating that intracardiac electrograms have a longer cycle length when recorded from
Figure 4 Relationship between mean atrial fibrillation cycle length in ECG lead V1 (V1AFCL) and the proportion of complex electrograms recorded from the left atrial (LA) endocardium during distal coronary sinus pacing.
676 atrial regions with dense scar demonstrated on atrial magnetic resonance imaging.31 It is possible that a short V1AFCL not only is reflective of atrial wavelength but also is an indicator of the number of atrial sources driving AF. This may explain in part why a short V1AFCL is associated with failure of maintenance of sinus rhythm after DCR or catheter ablation. This concept is consistent with that suggested by previous computer modeling that related a longer AFCL in the atrial appendages to a lower number of independent sources of AF.11 In this study, V1AFCL did not reflect all parameters of LA remodeling. In particular, there was no relationship with LA size, voltage, or refractoriness. Although it is recognized that LA remodeling in AF can affect all variables analyzed in this study, it is quite conceivable that not all will change in parallel with each other during the disease process. Furthermore, it is quite reasonable for noninvasive measurement of V1AFCL, as a single relatively simple measure, not to reflect the full complexity of the mechanisms by which AF is sustained in the remodeled atrial substrate. It is reasonable for AFCL to change as tissue conduction velocity changes even in the absence of change in local tissue refractoriness, the second determinant of wavelength.
Study limitations This study is limited by the requirement for participants to present to the electrophysiology laboratory in AF, thus restricting the extent of the spectrum of atrial disease studied. It may be that conclusions drawn from this population with persistent AF do not extend to those with paroxysmal AF, in whom electrical triggers would be expected to play a relatively larger role than the LA substrate would and in whom the contribution of focal sources and rotors may be different.
Conclusion A longer atrial fibrillatory cycle length derived from ECG lead V1 is associated with more advanced LA electroanatomic remodeling marked by slower atrial conduction and a greater number of complex electrograms.
Heart Rhythm, Vol 11, No 4, April 2014
8.
9.
10.
11.
12.
13. 14.
15.
16. 17.
18.
19.
20.
21.
22.
23.
24.
25.
References 1. Teh AW, Kistler PM, Lee G, et al. Electroanatomic remodeling of the left atrium in paroxysmal and persistent atrial fibrillation patients without structural heart disease. J Cardiovasc Electrophysiol 2012;23:232–238. 2. Nattel S, Burstein B, Dobrev D. Atrial remodeling and atrial fibrillation: mechanisms and implications. Circ Arrhythm Electrophysiol 2008;1:62–73. 3. Nilsson F, Stridh M, Bollmann A, Sörnmo L. Predicting spontaneous termination of atrial fibrillation using the surface ECG. Med Eng Phys 2006;28: 802–808. 4. Husser D, Cannom DS, Bhandari AK, et al. Electrocardiographic characteristics of fibrillatory waves in new-onset atrial fibrillation. Europace 2007;9: 638–642. 5. Bollmann A, Kanuru NK, McTeague KK, Walter PF, DeLurgio DB, Langberg JJ. Frequency analysis of human atrial fibrillation using the surface electrocardiogram and its response to ibutilide. Am J Cardiol 1998;81:1439–1445. 6. Matsuo S, Lellouche N, Wright M, et al. Clinical predictors of termination and clinical outcome of catheter ablation for persistent atrial fibrillation. J Am Coll Cardiol 2009;54:788–795. 7. Husser D, Bollmann A, Kang S, et al. Determinants and prognostic significance of immediate atrial fibrillation recurrence following cardioversion in patients
26.
27.
28.
29.
30.
31.
undergoing pulmonary vein isolation. Pacing Clin Electrophysiol 2005;28: 119–125. Holmqvist F, Stridh M, Waktare JEP, Sörnmo L, Olsson SB, Meurling CJ. Atrial fibrillatory rate and sinus rhythm maintenance in patients undergoing cardioversion of persistent atrial fibrillation. Eur Heart J 2006;27:2201–2207. Swartz MF, Fink GW, Lutz CJ, et al. Left versus right atrial difference in dominant frequency, K(þ) channel transcripts, and fibrosis in patients developing atrial fibrillation after cardiac surgery. Heart Rhythm 2009;6:1415–1422. Kim KB, Rodefeld MD, Schuessler RB, Cox JL, Boineau JP. Relationship between local atrial fibrillation interval and refractory period in the isolated canine atrium. Circulation 1996;94:2961–2967. Haïssaguerre M, Lim KT, Jacquemet V, et al. Atrial fibrillatory cycle length: computer simulation and potential clinical importance. Europace 2007;9(Suppl 6):64–70. Kistler PM, Sanders P, Fynn SP, et al. Electrophysiologic and electroanatomic changes in the human atrium associated with age. J Am Coll Cardiol 2004;44: 109–116. Sanders P, Morton JB, Davidson NC, et al. Electrical remodeling of the atria in congestive heart failure. Circulation 2003;108:1461–1468. Teh AW, Kistler PM, Lee G, et al. The relationship between complex fractionated electrograms and atrial low-voltage zones during atrial fibrillation and paced rhythm. Europace 2011;13:1709–1716. Stiles MK, Brooks AG, John B, et al. The effect of electrogram duration on quantification of complex fractionated atrial electrograms and dominant frequency. J Cardiovasc Electrophysiol 2008;19:252–258. Lin Y-J, Tai C-T, Kao T, et al. Consistency of complex fractionated atrial electrograms during atrial fibrillation. Heart Rhythm 2008;5:406–412. Stiles MK, John B, Wong CX, et al. Paroxysmal lone atrial fibrillation is associated with an abnormal atrial substrate: characterizing the “second factor.”. J Am Coll Cardiol 2009;53:1182–1191. Rostock T, Rotter M, Sanders P, et al. High-density activation mapping of fractionated electrograms in the atria of patients with paroxysmal atrial fibrillation. Heart Rhythm 2006;3:27–34. Bollmann A, Husser D, Mainardi L, et al. Analysis of surface electrocardiograms in atrial fibrillation: techniques, research, and clinical applications. Europace 2006;8:911–926. Holm M, Pehrson S, Ingemansson M, et al. Non-invasive assessment of the atrial cycle length during atrial fibrillation in man: introducing, validating and illustrating a new ECG method. Cardiovasc Res 1998;38:69–81. Stridh M, Sörnmo L, Meurling CJ, Olsson SB. Characterization of atrial fibrillation using the surface ECG: time-dependent spectral properties. IEEE Trans Biomed Eng 2001;48:19–27. Husser D, Stridh M, Cannom DS, et al. Validation and clinical application of time-frequency analysis of atrial fibrillation electrocardiograms. J Cardiovasc Electrophysiol 2007;18:41–46. Petrutiu S, Sahakian AV, Fisher W, Swiryn S. Manifestation of left atrial events and interatrial frequency gradients in the surface electrocardiogram during atrial fibrillation: contributions from posterior leads. J Cardiovasc Electrophysiol 2009;20:1231–1236. Sasaki T, Niwano S, Sasaki S, et al. Long-term follow-up of changes in fibrillation waves in patients with persistent atrial fibrillation: spectral analysis of surface ECG. Circ J 2006;70:169–173. Niwano S, Fukaya H, Sasaki T, Hatakeyama Y, Fujiki A, Izumi T. Effect of oral L-type calcium channel blocker on repetitive paroxysmal atrial fibrillation: spectral analysis of fibrillation waves in the Holter monitoring. Europace 2007;9: 1209–1215. Petrutiu S, Sahakian AV, Swiryn S. Short-term dynamics in fibrillatory wave characteristics at the onset of paroxysmal atrial fibrillation in humans. J Electrocardiol 2007;40:155–160. Petrutiu S, Sahakian AV, Swiryn S. Abrupt changes in fibrillatory wave characteristics at the termination of paroxysmal atrial fibrillation in humans. Europace 2007;9:466–470. Fujiki A, Sakabe M, Nishida K, Mizumaki K, Inoue H. Role of fibrillation cycle length in spontaneous and drug-induced termination of human atrial fibrillation. Circ J 2003;67:391–395. Tanaka K, Zlochiver S, Vikstrom KL, et al. Spatial distribution of fibrosis governs fibrillation wave dynamics in the posterior left atrium during heart failure. Cirs Res 2007;101:839–847. Platonov PG, Stridh M, de Melis M, et al. Analysis of atrial fibrillatory rate during spontaneous episodes of atrial fibrillation in humans using implantable loop recorder electrocardiogram. J Electrocardiol 2012;45:723–726. Jadidi AS, Cochet H, Shah AJ, et al. Inverse relationship between fractionated electrograms and atrial fibrosis in persistent atrial fibrillation: combined magnetic resonance imaging and high-density mapping. J Am Coll Cardiol 2013;62:802–812.
