Ventricular Tachycardia and Early Fibrillation in Patients With Brugada Syndrome and Ischemic Cardiomyopathy Show Predictable Frequency-Phase Properties on the Precordial ECG Consistent With the Respective Arrhythmogenic Substrate

Supplemental Digital Content is available in the text.

Background—

Ventricular fibrillation (VF) has been proposed to be maintained by localized high-frequency sources. We tested whether spectral-phase analysis of the precordial ECG enabled identification of periodic activation patterns generated by such sources.

Methods and Results—

Precordial ECGs were recorded from 15 ischemic cardiomyopathy and 15 Brugada syndrome (type 1 ECG) patients during induced VF and analyzed in the frequency-phase domain. Despite temporal variability, induced VF episodes lasting 19.6±7.9 s displayed distinctly high power at a common frequency (shared frequency, 5.7±1.1 Hz) in all leads about half of the time. In patients with Brugada syndrome, phase analysis of shared frequency showed a V₁–V₆ sequence as would be expected from patients displaying a type 1 ECG pattern (P<0.001). Hilbert-based phases confirmed that the most stable sequence over the whole VF duration was V₁–V₆. Analysis of shared frequency in ischemic cardiomyopathy patients with anteroseptal (n=4), apical (n=3), and inferolateral (n=4) myocardial infarction displayed a sequence starting at V₁–V₂, V₃–V₄, and V₅–V₆, respectively, consistent with an activation origin at the scar location (P=0.005). Sequences correlated with the Hilbert-based phase analysis (P<0.001). Posterior infarction (n=4) displayed no specific sequence. On paired comparison, phase sequences during monomorphic ventricular tachycardia correlated moderately with VF (P<0.001). Moreover, there was a dominant frequency gradient from precordial leads facing the scar region to the contralateral leads (5.8±0.8 versus 5.4±1.1 Hz; P=0.004).

Conclusions—

Noninvasive analysis of ventricular tachycardia and early VF in patients with Brugada syndrome and ischemic cardiomyopathy shows a predictable sequence in the frequency-phase domain, consistent with anatomic location of the arrhythmogenic substrate.

WHAT IS KNOWN

• The mechanisms of ventricular fibrillation in patients with and without structural abnormalities of the ventricles are poorly understood.

• Phase and Frequency analyses of fibrillatory signals have elucidated arrhythmia mechanisms.

WHAT THE STUDY ADDS

• During human ventricular fibrillation the spatial location of the arrhythmogenic substrate implies a hierarchical organization in the phase and frequency domain of electric activity recorded from the body surface that is not consistent with random cardiac activation.

• The method and findings in this study pave the way to deriving further mechanistic insights from noninvasive ECG recordings of ventricular fibrillation and tachycardia, with the potential to improve therapy.

On a single-lead ECG, the activation of the ventricles during ventricular fibrillation (VF) appears to be highly complex and continuously changing, conjuring up the idea that VF results from random electric excitation. However, studies in animals¹ and computational models^2,3 in which electric wave propagation has been studied at high resolution have demonstrated that the characteristics of VF in the structurally normal heart are both deterministic and quantifiable, and that they depend on the activation rate of highly periodic reentrant sources that result in fibrillatory conduction with spatial gradients of local activation frequency. Correspondingly, studies in humans using a relatively small number of body surface electrode recordings demonstrated spatiotemporal organization during VF.^4,5 However, it is still unclear whether highly periodic sources also maintain VF in humans and how they relate to the substrate in individual patients with distinct pathology.

As a first approximation, patients with Brugada syndrome (BrS) and ischemic cardiomyopathy (IC) may be considered to be 2 extremes of a continuum in which the structure of the ventricles changes from relatively homogeneous (BrS) to highly heterogeneous (IC), ie, whereas in BrS the ventricles are for the most part structurally uniform, in IC they are substantially remodeled, nonuniform, and variably scarred. Here, we study the possible mechanistic link between the ventricular substrate and the ensuing ventricular tachycardia (VT) and early VF in patients using a novel noninvasive analysis based on the standard precordial ECG leads. We tested the hypothesis that early VF in patients with BrS or IC is spatiotemporally organized by high-frequency sources localized stably at the pathophysiological substrate. We demonstrate that, in the human ventricles, the spatial location of the arrhythmogenic substrate yields a hierarchical organization in the phase and frequency domains that is consistent with a deterministic mechanism of fibrillation.

Methods

Detailed Methods are available in the Data Supplement.

Patients

Patients were admitted for electrophysiology study with either BrS type I ECG or IC with healed myocardial infarction (MI; >6 months). The study was approved by the Ethics Committee, and subjects gave informed consent.

Electrophysiology and Recording Protocol

Standard ECG was recorded continuously at a sampling rate of 1 kHz and band-pass filtered at 0.05 to 150 Hz. The electrophysiology strategy varied according to the clinical indication.

Patients With BrS

We induced VF by programmed stimulation through a transvenous catheter electrode at the right ventricular (RV) apex or at the RV outflow tract. We included only those patients in whom the immediately preceding ECG displayed a type I pattern on the right precordial leads.

Patients With IC

Programmed RV apex stimulation was used for VT induction. We placed another catheter in the left ventricle (LV) for endocardial mapping and ablation as follows: 3-dimensional (3D) endocardial reconstructions were created by CARTO (Biosense Webster) mapping system delineating endocardial scar and dense scar tissues as areas displaying bipolar electrogram amplitudes of <1.5 and <0.5 mV, respectively.⁶ Ablation followed mapping of induced or spontaneous sustained VT. Sites displaying presystolic electrograms (within 50 ms before the QRS onset) and fulfilling the criteria of concealed entrainment with equally spike-QRS to electrogram-QRS intervals were marked as the tachycardia exit sites. Pacing was performed for validation. For ethical reasons, the patients with IC included in the study were only those in whom sustained VF was unintentionally induced during the procedure.

Processing of Precordial Signals

Signals from precordial leads were exported, processed by a Hanning window and with a nonbiased 1 to 20 Hz band-pass filter. Thereafter, a 4096-point fast Fourier transformation (FFT; resolution 0.24 Hz) was obtained for frequency-phase analysis. Power and phase versus frequency were calculated for sliding episodes with 15-ms leaps. Dominant frequency (DF) was defined as the frequency with highest power in the spectrum. Shared frequency (SF) was defined as the frequency with power peak lasting >0.5 s simultaneously in all leads (Figure 1). The instantaneous phase was calculated by 2 methods: (1) the ratio of the imaginary to the real components of the FFT at the SF frequency and (2) the Hilbert transform of the entire episode.⁷

Figure 1.

Shared frequency (SF) during ventricular fibrillation (VF) in a patient with Brugada syndrome (BrS). A, ECG tracings during SR (left, BrS type I at V₁ and V₂) and induced VF (right). B, Periodograms of 6 precordial signals. Despite temporal variability in A, all signals reveal simultaneous and relatively stable power at SF of 6.1 and 5.6 Hz for periods lasting several seconds. FFT indicates Fourier transformation.

Location of the Pathological Substrates, Exit Sites, and Precordial Leads

In patients with BrS, the precordial signals displaying type I morphology marked the location of the functional abnormalities.⁸ In a subset of 6 patients with IC, we merged computed tomographic scan images with the CARTO maps for better anatomic reference and marked the positions of the precordial ECG leads on the thorax surface by acquiring the corresponding locations of a roving catheter tip on the CARTO (Figure I in the Data Supplement). We classified dense scars as those being located on the anteroseptal, apical, posterior, or inferolateral LV.

Computer Simulations

We used a realistic 3D computer model of the human heart and torso to simulate body surface potentials during ventricular pacing.

Statistical Analysis

Continuous variables are reported as mean±SD or as 95% confidence interval where noted. In case of a varying number of observations drawn from the subjects, the resampling methods of naive bootstrap⁹ and the general bootstrap algorithm¹⁰ were used to make inferences (Data Supplement).

Results

Patients and VF

We studied early sustained VF in 15 patients with BrS and 15 patients with IC (Table). We induced VF in patients with BrS by RV apex (n=14) or RV outflow tract (n=1) pacing. The IC cohort consisted of patients with previous MI of the anteroseptal (n=4), posterior (n=4), apex (n=3), and inferolateral (n=4) aspects of the LV. The IC patients included those in which VF was unintentionally induced while pacing from the RV apex (n=11), delivering antitachycardia pacing (n=2) or moving the catheters (n=2).

Table.

Patients Characteristics

Precordial Frequency-Phase Domain Organization of VF in Patients With BrS and IC

Figure 1 shows sample VF data from a patient with BrS. The tracings in Figure 1A are simultaneous precordial lead recordings during sinus rhythm and induced VF. The sinus rhythm signals show a type I BrS ECG at V₁ and V₂ (left) indicating an abnormal arrhythmogenic substrate at the RV outflow tract.¹¹ Figure 1B shows the power spectra of sliding periods of signals from V₁–V₆ during VF in the same patient. Despite signal variability in time and between leads, the periodogram of each precordial clearly shows narrow peak power at SFs of 6.1 and 5.6 Hz. Overall, VF episodes recorded for 19.6±7.9 s in the 30 patients with BrS and IC contained 57 long-lasting epochs (3.4±2.1 s; range 0.7–8.7 s) of SFs corresponding to 44.5±24.5% of the time through the whole spectrogram. In the other, 55.5% VF was characterized by power distributed at different frequencies on each precordial lead, possibly reflecting more complex activity when compared with episodes with a SF. Both the duration and the number of SFs were similar in patients with BrS and IC (3.6±1.9 versus 3.3±2.2 s; P=0.591 and 1.8±0.8 versus 1.7±0.9 s; P=0.667, respectively). The overall SF in all leads was 5.7±1.1 Hz; however, SFs in patients with BrS tended to be higher than patients with IC (6.1±0.7 versus 5.3±1.3 Hz, respectively; P=0.065).

We next investigated whether the VF activity recorded in the precordial leads maintained predictable phase sequences. Figure 2A shows the time-course of the phases at the 2 distinctive SFs in the 6 precordials in the same patient with BrS during the VF episode shown in Figure 1A. The phase distribution (black arrow) and propagation (red arrow) between precordials occurred in well-demarcated clusters for the 57 SFs detected in patients with both BrS (27 SFs) and IC (30 SFs). The width of the clusters in those 57 sequences (measured by the length of the vertical black arrow) was 134.3°±46.2°, which was significantly narrower than a nonclustering distribution (P<0.001 versus 180° with similar SD; ie, a range including antiphases). The order of phases in the clusters varied depending on the location of the specific arrhythmogenic substrate responsible for the arrhythmia. All patients with BrS were characterized by a type I ECG pattern with typical ST-segment elevation in V₁ and V₂, which corresponded to an anteroseptal location of the arrhythmogenic substrate. In these patients, the sequence of the phases (either distribution or propagation) was from V₁–V₂ to V₅–V₆ in 21 of the 27 SFs (77.8%), whereas a sequence originating in V₃–V₄ or V₅–V₆ was observed in 2 (7.4%) and 4 (14.8%) SFs, respectively (P<0.001). In Figure 2B, cumulative data show that the phases of V1 were of higher degree (early) in the cluster than V₆ (late), with a monotonic decrease from right (V₁) to left (V₆; P<0.001) precordials, according to their location in the frontal axis (Figure I in the Data Supplement).

Figure 2.

Phase’s analysis during induced ventricular fibrillation (VF) in Brugada syndrome. A, The time course of phases of the sample precordials shown in Figure 1 is characterized by well-demarcated clusters. Black arrow, instantaneous phase distribution across leads; red arrow, phase propagation across leads. B, Cumulative fast Fourier transformation (FFT)–based phases at SF demonstrate a consistent sequence from V₁–V₂ (early) to V₅–V₆ (late). Time for phase determination was at phase ≈0 in the middle of the cluster. Error bars: 95% CI. C, Hilbert transform provides the time course of phases for all frequencies. The phase distribution of the signals shown in Figure 1 denotes a sequence from early to late (black arrow) similar to the FFT-based phases at specific SF shown in A. Time in all panels is measured from initiation of VF.

Because the sequence and phase distribution shown in Figure 1B are necessarily time-invariable by virtue of the FFT at a particular frequency (ie, the SF), we investigated the temporal stability of the phases using also the Hilbert transform, which yields the instantaneous phase of the activity, considering the entire spectral content. Figure 2C shows the Hilbert-based phases of the 6 precordial leads during the same VF episode as in Figure 1. Similar to the FFT-based phases, Hilbert-based phases clustered in well-defined groups. Most importantly, their temporal intrabeat sequence (ie, from V₁ and V₂ [early] to V₅ and V₆ [late]) was remarkably stable, which is in accordance with both the FFT analysis and the stability of the SF in this patient with BrS. Considering all patients with BrS together, the Hilbert-based phase sequences were found to be more frequently originating from V₁ to V₂ than from V₅ to V₆ (88.9% versus 11.1%, respectively; P<0.001), showing substantial agreement with the sequence of FFT-based phases (Figure II in the Data Supplement ).

Precordial Phase Sequences During VF in Patients With IC

In patients with IC, the phase distribution and propagation of each of the 30 SFs during VF sequenced in relation with the location of the scar tissue in 63.3% (P=0.038). Figure 3 shows VF data and phase analysis from an IC patient with a septal MI. As shown in A and B, early VF in that patient was characterized by an SF of ≈5.8 Hz. The time-course of the precordial phases shown in C demonstrates clustered activity with an intrabeat sequence of V₁ (early) to V₆ (late). In D, the intrabeat phase distribution across the precordials is displayed on the 3D torso map (Figures I–III in the Data Supplement) to confirm that the earliest beat-phase is at V₁ (red), closest to the MI location, and the latest beat-phase is at V₆ (purple). In Figure 4A and 4B, we use CARTO maps to illustrate 2 additional examples of distinct relationships between the scar location and the distribution of precordial phases of SFs during VF in patients with IC. In A, a patient with an inferolateral MI presented a VF episode with SF of 5.8 Hz and a V₆–V₁ intrabeat phase sequence. In B, another patient with an LV apex MI developed VF with SF of 4.8 Hz with the earliest phase at V₃–V₄ and spread toward V₁ and V₆. However, the latter behavior was not representative of all patients with either LV apex or posterior LV MI, as the phase sequence varied substantially from one patient to the next. Figure 4C and 4D shows phase distributions across the precordial leads of patients with anteroseptal and inferolateral scars. Data from 8 SFs in 4 patients with anteroseptal scars showed a monotonic distribution from early V₁ to late V₆ (P<0.001; Figure 4C), consistent with the precordials location along the frontal axis. In contrast, Figure 4D shows that in 9 SFs from 4 patients with inferolateral scar the distribution of phases was such that V₆ was earlier than V₁ (P<0.001), V₂ (P<0.001), and V₃ (P=0.048), with V₁ having the lowest phases in the cluster`s alignment (P<0.001). The monotonic decrease of phases from V₆ to V₁ was not statistically significant when considering all precordials (P=0.245), probably because of large variability at V₄. Yet the differences between lateral (V₆) and right (V₁, V₂, and V₃) precordial leads demonstrate that the distribution of phases reverses in patients with inferolateral scars relative to those with anteroseptal scars (see also below).

Figure 3.

Ischemic cardiomyopathy; phases analysis during ventricular fibrillation (VF) in patient with a septal myocardial infarction (MI). A, Precordial signals during VF. B, Periodogram of V₂ showing high power around 5.6, 6.3, and 5.8 Hz. C, The phase’s time course at 5.8 Hz reveals a sequence from V₁ to V₆ (black arrow from early to late). The same alignment was observed for 5.6 and 6.3 Hz. D, Phase distribution superimposed on the thorax map illustrates that the sequence from V₁ to V₆ is consistent with activation originating close to the MI scar location (inset, red is <0.5 mV). FFT indicates fast Fourier transformation.

Figure 4.

Ischemic cardiomyopathy; phases during ventricular fibrillation (VF) and localized myocardial infarctions (MIs). A, Electroanatomic map showing an left ventricular (LV) inferolateral myocardial infarction (MI), sample VF recording and corresponding VF phases time course at shared frequency (SF) = 5.8 Hz. The earliest lead is V₆. B, Electroanatomic map showing an LV apex MI, sample VF recording, and corresponding VF phases time course at SF = 4.8 Hz. The earliest lead is V₄. C, Patients with anteroseptal scar display a steady sequence of propagation from V₁ to V₆. D, In contrast, inferolateral scar patients display a sequence from V₆ to V₁. Error bars, 95% CI.

VF Phase Sequence and Wave Origin in Patients With IC

Figure 5A shows the location of the precordials (Figures I and III in the Data Supplement) relative to the heart to provide a geometric reference of the VF phase distribution across leads. In B, we show the FFT-based time-space plot of phases of SF in 3 sample VF cases with 3 different MI locations. The color-coded phases (white arrows) are seen to propagate across leads (black arrows) in distinct sequences, consistent with what is shown in Figures 3 and 4: for anteroseptal and inferolateral MIs, the phases propagate from V₁ to V₆ and V₆ to V₁, respectively, whereas for a LV apex MI, the phases spread from V₄ to V₁ and V₆.

Figure 5.

Propagation of precordial phases during ventricular fibrillation (VF). A, Precordial leads diagram superimposed on a sample CARTO map of the left ventricular (LV) across the coronal plane. B, Dependency of the directionality of fast Fourier transformation (FFT)–based phase propagation on the location of the scar; in 3 sample patients the progression of the phase reverses from V₁ to V₆ (left) to V₆ to V₁ (right) as the location of the scar changes from the anteroseptal to the inferolateral aspects of the LV. C, Equal sequences are obtained by Hilbert-based phases that account for all frequency components of each signal.

To further validate the correlation between the precordial sequences and the ventricular origin of waves, we used a computer model of the ventricles inside the torso to simulate paced activity and calculate the phase distribution on the precordial body surface locations. In Figure IV in the Data Supplement, pacing the septum, the LV apex, and the lateral LV yielded similar sequences of phase propagation in the precordial leads as in Figure 5B, confirming our interpretation that the patterns of sequences depended on the directionality of ventricular wave propagation during VF.

To determine how precordial phase propagation patterns were affected by the scars exit sites, we performed simulations in which we paced at opposite sides of conduction block lines representing scars similar to those of patients with IC in Figures 3 and 4. Figure VA in the Data Supplement shows that contralateral exit sites from scars that span 30% to 40% of the long axis of the LV and are located on the anteroseptal and inferolateral LV walls produce waves that move transiently in opposite directions, but subsequently follow propagation patterns that are similar for each block line, irrespective of the initial propagation direction. In Figure VB in the Data Supplement, phase propagation graphs are somewhat sensitive to the initial directionality of the exiting wave, but the sequence of the phases is mainly determined by the spatial location of the pacing site; ie, V₆ and V₅ precede other precodials when pacing the lateral wall; V₁ and V₂ precede other precordials when pacing the septum. Thus, the sequence of precordial phases in the patients with IC analyzed in Figures 3 and 4 depends more on the scar location than on the precise exit site from that scar.

Because VF may contain frequency components other than SFs, each with its own phase sequence, we calculated the Hilbert-based phases of the precordial leads recorded during VF and compared their sequence with the FFT-based phase at the SFs contributing to the same ECGs. Figure 5C demonstrates that when all the frequency components of the 3 sample VF signals with similar MI location as in Figure 5B are considered, the propagation of phases remains generally similar to when only a single SF is considered. Further comparison of the leads at which the phases originated in the FFT-based phase analysis with those in the Hilbert-base analysis revealed a highly significant concordance (ƙ concordance coefficient 0.741 [95% confidence interval, 0.497–0.938]), mainly when the origin was at V₁–V₂ or V₅-V₆ (Figure II in the Data Supplement). Overall, patients with anteroseptal, apical, and inferolateral LV MIs tend to display a steady sequence of phases originating at V₁–V₂, V₃–V₄, or V₅–V₆, respectively (Figures 3 and 4). Also, as shown in Figure VI in the Data Supplement, pacing from 2 different areas in the RV (outflow tract and apex; 5 patients) displayed different phase propagation patterns depending on the location of the stimuli and the direction of electric propagation.

Correlation Among Phase Sequences in VT and VF

We further tested the hypothesis that the arrhythmogenic substrate in the ventricles determines the activation directionality during both VT and VF. We used paired comparisons between the mapped VTs and sustained VFs in the group of patients who developed both arrhythmias (n=10). As shown in Figure 6, phase propagation in VF was similar to VT in a patient with inferolateral MI. Figure 6A shows a monomorphic VT that subsequently transformed into VF. Despite the complexity of the precordial VF signals in the time domain, the sampled V₁ and V₆ periodograms in B reveal that a peak power at SF ≈5.8 Hz is present for a significant portion of the VF episode analyzed. Figure 6C shows the phase analyses during VT (left) and VF (right) in the same patient. The phase propagation sequence across the precordials was similar despite the different SFs that characterized each of the 2 arrhythmias; the sequence was from V₆ to V₁ in both.

Figure 6.

Phase distribution during ventricular tachycardia (VT) and ventricular fibrillation (VF) in an ischemic cardiomyopathy patient. A, ECG tracings during VT (left, frequency of 3.2 Hz) and VF (right) in a patient with an inferolateral scar. B, The periodogram during VF shows that 5.8 Hz is the highest frequency to contain significant power for most of the recording time in V₆ and V₁. C, Phase propagation sequence is similar for VT at 3.2 Hz and VF (shared frequency) at 5.8 Hz (V₆–V₁).

Figure 7 summarizes the relationships among the sequences of the precordial phases during VF and VT. In A, data from 10 patients show that the VT and VF phases at SF of the same leads correlate moderately but significantly (linear correlation coefficient, 0.64; P<0.001). The relationship between the VT and VF phases is further quantified in B by a histogram of their differences (δ phase) along with the kernel distribution density estimation: the distribution of δ phase is well concentrated around a mean close to 0 (peak, −6.01; 95% confidence interval, −19.06, 5.57°), with 60% of the phases during VF being ≤30° different from the phases during VT in the same patient (95% confidence interval, 49%–71%). Figure 7C and 7D illustrates the coarse relationship that exists between leads at which the phase originates during VT/VF and the location of the arrhythmogenic scar in patients with IC. The VT graph in C demonstrates that as the arrhythmogenic site shifts from the right (anteroseptal) to the middle (apex) and to the left (inferolateral) of the LV of different patients, a concomitant shift in the earliest VT phase of V₁–V₂, to V₃–V₄ and to V₅–V₆ is observed (P<0.001). In D, the relationship between leads with the earliest phase and the site of the substrate during VF is less pronounced than in VT, mainly because the earliest lead is not uniquely determined for VF with apical and posterior substrate location. Nevertheless, when considering all the analyzed cases together, also during VF the earliest phases at the leads are significantly associated with the substrate location (P=0.005). Comparison between C and D leads us to conclude that the phase propagation of high-frequency periodicities during VF is equivalent to the behavior of monomorphic VT phase propagation and its relation with the scar location.

Figure 7.

Phase sequences in ventricular fibrillation (VF) and ventricular tachycardia (VT). A, Correlation between VT and VF phases in similar leads in 10 patients. Time for determination of phase distribution was selected as the time when the earliest lead was at ≈120°. B, Histogram (bin = 20°) of the differences between the VF and the VT phases (δ phase) of data from A along with the kernel distribution density estimation (see text). C, Origin of VT phase sequence and the exit site or the arrhythmogenic scar in ischemic cardiomyopathy (IC) patients: the phase origin reverses from V₁–V₂ to V₅–V₆ as the exit site of the VT changes from the anteroseptal (right) to the inferolateral (left) aspects of the left ventricular (LV). D, Origin of VF phase sequence in IC patients displayed similar dependency of phase origin on site of LV substrate as in VT in C. SF indicates shared frequency.

On Hilbert-base phase calculation (Figure II in the Data Supplement) VF was characterized by a similar phase activity on the precordial leads as detected at SFs, which further established that the correlation between the substrate and the precordial phases is independent of the SFs. Finally, to confirm the tight geometric relationship between the sequential distribution of phases on the precordial leads and the arrhythmogenic substrate, we compared their relative position on the body surface and on the heart. Angular analysis presented in Figure VII in the Data Supplement demonstrates a good correlation between the azimuths of the precordial with earliest phase, the arrhythmogenic substrate, and the VT exit sites within a common short-axis plane of the torso.

Hierarchical Precordial DFs Localizes the Arrhythmogenic Area

The SFs across the precordials during VF were not necessarily the DFs in the spectrogram for those leads; however, they were usually the DFs on the leads that were closest to the substrate. Figure 8A shows power spectra of the precordials in an IC patient with inferior MI. A 5.85 Hz peak is present in the spectra of all leads, but it is the DF only at V₅ and V₆. It is the highest frequency peak among all leads and is located closest to the scar in this patient. To study the relationship between the hierarchy of the DFs and the location of the scar, we compared the DFs in the leads near the scar and those in the contralateral leads. Figure 8B summarizes the DF data in 11 IC patients with VF (patients with posterior scar were excluded). In 6 of 11 patients, a precordial DF gradient was observed between the scar and the contralateral locations. Considering all the patients with IC, with and without a gradient, we found the DF closest to the substrate to be higher than the DF contralateral to the substrate (5.8±0.8 versus 5.4±1.1 Hz; P=0.004).

Figure 8.

Dominant frequency (DF) gradient at precordials during ventricular fibrillation (VF). A, Representative example of DF gradient in a patient with ischemic cardiomyopathy (IC). Power at a shared frequency (SF) of 5.85 Hz is noted. However, this frequency is dominant only at V₆ and V₅ and lower frequencies are dominant in the other precordials; the lowest DF of 3.17 Hz is in V₁. B, DF during VF in patients with IC (n=10; 18 SF) shows statistically higher values in precordial leads closest to substrate location. C, LV ejection fraction (LVEF) was significantly higher in patients without (n=18) than with (n=8) a DF gradient.

The precordial DF distribution in patients with BrS was found to be more uniform than the IC patients, with only 2 of the 15 patients displaying a DF gradient (P<0.001). The SF value at each precordial was the same as the DF value, regardless of whether it was close to the substrate location (SF 6.1±0.7 Hz versus DF 6.1±0.7 Hz; P=0.327) or more distally (SF 6.1±0.7 Hz versus DF 6.1±0.7 Hz; P=0.476). Obviously, at a mean uniform value of 6.1 Hz, a DF gradient in our BrS cohort is nonexistent (P=0.648).

Cardiac Function and VF Dynamics

Interestingly, the cardiac function of patients during sinus rhythm was related to electric dynamics during VF. Considering the precordial DF values during VF, both BrS and IC patients with no DF gradient had higher LV ejection fraction (60.6±8.9%; n=18) than patients with a DF gradient (44.7±14.7%; P=0.002; n=8; Figure 8C). As expected, patients with BrS had no cardiac structural abnormalities as quantified by echocardiogram. In patients with IC, we observed a trend toward significantly higher LV ejection fraction in patients with no DF gradient (49.2±10.6%) than in those with a DF gradient (38±9.1%; P=0.091). Lack of statistical significance at P<0.05 in the patients with IC is because of low power, possibly related to their generally lower LV ejection fraction than that in the patients with BrS, requiring further study.

Discussion

The main findings of this study are that the signals of the 6 precordial ECG leads during VT and VF in patients with BrS and IC follow a sequence in the phase-domain that is distinctly related to the location the pathophysiologic substrate. Using frequency-phase domain analysis across the standard precordial leads, we demonstrate hierarchical distributions of frequencies and phases suggesting that in many individual patients the arrhythmias are driven from a substrate location, with higher source frequency in VF versus VT. Thus, our results are consistent with a nonrandom mechanism of VF and support the idea that a relatively small number of stable sources maintain VF in both the structurally normal and the diseased human heart.

Deterministic Activation During Fibrillation

Early experimental studies in structurally normal rabbit and guinea-pig hearts demonstrated stable and fast rotors driving VF^1,12 and generating high-frequency waves interacting with the surrounding myocardium giving rise to fibrillatory conduction and irregular ECG.^13,14 In the structurally normal heart of the sheep, fibrillatory conduction is reflected as excitation organized in DF domains connected by Wenckebach-like conduction block patterns at their boundaries.¹⁵ In both animals¹⁶ and humans,¹⁷ the DFs of excitation waves have been found to underlie the frequency domain of specific ECG leads on the thorax. During VF in our patients, there were both surface spectral components in the form of SF across the precordial leads and sequential phase distributions (Figures 6 and 7) that were consistent with patterns of waves arising from a region containing an arrhythmogenic substrate (Figures III and IV in the Data Supplement).

Our findings are consistent with the pioneering work of Clayton et al^4,5 who analyzed VF recordings and found a similar frequency content at 3 ECG orthogonal leads 58% of the time, with small variability in the phase relationship between leads.

Others have found rotors in the human ventricle under different conditions.^18,19 However, whether these highly periodic sources drive the spatiotemporal organization that we found in our patients or simply coexist as epiphenomena embedded in a multiple-wavelet self-sustained system remains controversial. Recently, Bradley et al²⁰ studied in vivo human VF and found that the global ischemia produced by the arrhythmia decreased DFs while increasing the number of wavebreaks, the hallmark of rotor formation. Consistent with our study, Nair et al¹⁹ used plunge needles in Langendorff-perfused human hearts and identified 3D rotors during early VF to localize in regions of greater intramural fibrosis.

Regional Ischemia and Organization of VF

Experimental studies have demonstrated that the epicardial border zone of the myocardial scar is prone to sustain reentrant sources of fast VT in animal models.²¹ Coromilas et al²² confirmed that dynamic anisotropy in the periphery of the scar supports short-cycle reentry after action potential shortening by pinacidil. In a pig model of regional ischemia, Zaitsev et al²³ reported that the distribution of wavebreaks during VF is particularly high at the border zone, which exhibits also the largest heterogeneity of excitation frequencies, including highest local DFs. Although no sustained reentry driving VF was registered in those studies, the conjecture was that the wavebreaks in the border zone contributed to the VF by providing a perpetuation mechanism for source formation and maintenance. Overall, the regional scar and ischemia conditions in the experimental studies reflect the importance of particular myocardial areas in facilitating reentry during fibrillation,^24–26 giving credence to the idea that regular firing sources at the periphery of the scar drive VF in our patients.

Local high-frequency periodicity during VF points to a potential driver location similar to that observed during monomorphic VT at a slower frequency.¹⁵ Because mapping and ablation determined that the exit site of the VT was closely related to the scar, we suggest that periodicities observed during VF reflect centrifugal activity from a similar scar region. Previous studies in human hearts demonstrated that fibrosis increased wavebreaks and turbulent propagation, but also stabilized reentry.^19,27 Our data demonstrate that an intraventricular DF gradient exists during early VF in patients with the lowest LV ejection fraction during sinus rhythm (Figure 8). These observations support the notion that VF in our patients originates from high-frequency sources localized to a regional arrhythmogenic substrate and further include widespread, slower fibrillatory activity.

Limitations

All VF episodes were induced and therefore the mechanisms identified may not apply to spontaneous VF. However, similar to the atria, induced VF likely mimics spontaneous VF.²⁸ Some patients with IC were under β-blocker treatment, which could have converted their VF to be driven by a single source.²⁹ Nevertheless, the fact that similar sequences of phases were observed during VF in patients with BrS that were devoid of β-blockers suggests that the location of the VF source(s) in all patients was related to the location of their scar. As for the data acquisition, the small number of electrodes and their position on the body surface limit the sensitivity of our analysis for possible complex activation patterns and may bias toward a localized source mechanism of VF. In addition, the nonuniqueness of the relation between epicardial potentials and the myocardial waves limits our interpretation. Indeed, scarcity of intracardiac data precludes exclusion of scenarios in which there is no relationship between origin of VT/VF waves and precordial sequences. However, the match between epicardial and nearest body surface potentials demonstrated by Burnes et al,³⁰ as well as the fact that both patients and simulations demonstrate a clear relationship between pacing sites and the precordial sequences, diminishes the impact of such limitations. Finally, despite significant positive relationship between phases and scar locations in patients with anteroseptal and inferolateral scars, the configuration of the 6 precordials precluded us from establishing specific phase sequences in patients with either LV apex or posterior LV MI.

Conclusions and Clinical Implications

Ablating substrates to prevent VT³¹ and VF^11,32 recurrences in humans is now common. Here, we demonstrate that early VF in patients with BrS and IC is characterized by a reproducible phase sequence in the precordial leads consistent with waves originating from the location of the arrhythmogenic substrate. In addition, we demonstrate that there is a hierarchical precordial lead distribution of DF with values decreasing away from the lead closest to the arrhythmogenic substrate location. Together, the hierarchical distribution of precordial phases and DFs suggests that the mechanism that sustains early VF can be attributed to relatively small number of high-frequency sources that could be targeted more effectively with noninvasive guidance.

Sources of Funding

This study was supported by the National Heart, Lung and Blood Institute (P01-HL039707, P01-HL087226 R01-HL118304); the Spanish Society of Cardiology, Arrhythmia and Electrophysiology Section; the Leducq Foundation, Paris, France; Centro Nacional de Investigaciones Científicas, Madrid, Spain; Generalitat Valenciana, Valencia, Spain (PROMETEO/2012/030); VI Plan Nacional de Investigación Científica, Desarrollo e Innovación Tecnológica from the Ministerio de Economía y Competitividad of Spain (TIN2012-37546-C03-01); and the European Commission (European Regional Development Funds [ERDF]-FEDER).

Disclosures

Dr Jalife is on the Scientific Advisory Board of Topera, Inc. Dr Berenfeld is the Scientific Officer of Rhythm Solutions, Inc. Dr Atienza is on the advisory board of Medtronic, Inc. The other authors report no conflicts.

Supplementary Material

hae-8-1133-s001.pdf ^{(1.3MB, pdf)}

hae-8-1133-s002.docx ^{(20.4KB, docx)}

hae-8-1133-s003.pdf ^{(4.5MB, pdf)}