Endocardial Activation Drives Activation Patterns during Long-Duration Ventricular Fibrillation and Defibrillation

Abstract

Background

Understanding the mechanisms that drive ventricular fibrillation is essential for developing improved defibrillation techniques to terminate ventricular fibrillation (VF). Distinct organization patterns of chaotic, regular, and synchronized activity were previously demonstrated in VF that persisted over 1-2 minutes (long duration VF, LDVF). We hypothesized that activity on the endocardium may be driving these activation patterns in LDVF and that unsuccessful defibrillation shocks may alter activation patterns.

Methods and Results

The study was performed using a 64-electrode basket catheter on the left ventricle endocardium and 54 6-electrode plunge needles inserted into the left ventricles of 6 dogs. VF was induced electrically and following short duration VF (SDVF, 10 s) and LDVF (7 minutes), shocks of increasing strengths were delivered every 10 s until VF was terminated. Endocardial activation patterns were classified as chaotic (varying cycle lengths and nonsynchronous activations), regular (highly repeatable cycle lengths), and synchronized (activation that spreads rapidly over the endocardium with diastolic periods between activations).

Conclusions

The results showed that the chaotic pattern was predominant in early VF, but the regular pattern emerges as VF progressed. The synchronized pattern only emerged occasionally during late VF. Failed defibrillation shocks changed chaotic and regular activation patterns to synchronized in LDVF, but not in SDVF. The regular and synchronized patterns of activation were driven by rapid activations on the endocardial surface that blocked and broke up transmurally, leading to an endocardial to epicardial activation rate gradient as LDVF progressed.

Introduction

Sudden cardiac arrest due to ventricular fibrillation (VF) is a significant cause of death in the developed world. The most effective treatment for VF is defibrillation with an electrical shock. Even in areas with the quickest first responders, patients typically are not defibrillated for several minutes.^{1, 2} However, the majority of defibrillation studies have been conducted in animal models of short-duration ventricular fibrillation (SDVF, VF lasting <1 min in duration).³ Due to the effects of the high activation rate of VF and global ischemia, the characteristics of VF evolve as seconds lengthen to minutes (LDVF, long duration VF).^4-7

We have recently demonstrated that 3 distinct patterns of endocardial activation exist in dogs during prolonged VF, and that the incidence of these patterns of activation change with VF duration.^8-10 In this study, we will test the hypothesis that the endocardium plays a critical role in promoting and sustaining these distinct activation patterns during LDVF and defibrillation and that activation of the midwall and epicardium is driven by endocardial activation. We will also evaluate the effect of defibrillation shocks on SDVF and LDVF activation patterns. Development of more appropriate patient and time specific therapies for VF hinge upon increased understanding of the evolving mechanisms of VF.

Methods

The animals were managed in accordance with guidelines established by the American Heart Association on research animal use,¹¹ and the protocol was approved by the Institutional Animal Care and Use Committee at the University of Alabama at Birmingham. The data, analytic methods, and study materials will not be made available to other researchers for purposes of reproducing the results or replicating the procedure.

Animal Preparation

Purpose bred mixed breed hounds (20-25 kg) were fasted overnight and anesthetized with sodium thiopental (25 mg/kg iv), intubated, and mechanically ventilated with 2–3% isoflurane in 100% oxygen. Lead II surface ECG, core body temperature, arterial blood gases, arterial blood pressure, and serum electrolytes were monitored and maintained within normal levels.

Electrode Placement

A 38-mm multielectrode basket with 64 electrodes (Constellation Catheter, Boston Scientific, Natick, MA) was introduced through a femoral artery and passed through the aortic valve into the left ventricle (LV). This catheter contained eight splines with eight electrodes spaced at 3-mm intervals along each spline. A defibrillation catheter with a coil in the right ventricle (RV) and another in the superior vena cava (SVC) with a pacing was inserted into the right ventricular (RV) apex through a jugular vein.

Silver wire, fiberglass reinforced epoxy needles with 6 electrodes at 2 mm spaced intervals were constructed as described elsewhere¹². Needles were tested and chloridized prior to each experiment. After the chest was opened by a sternal thoracotomy, 54 plunge needles were spaced at 5-10 mm intervals and inserted from the epicardial surface of the LV to record transmural activation. Electrograms recorded on the basket and plunge needle electrodes were band pass filtered (0.5-500 Hz) and sampled at 2 kHz.

VF Induction and Defibrillation

VF was induced by delivering 2-3 seconds of 60 Hz stimulation to the pacing tip of the defibrillation catheter in the RV apex. The defibrillation threshold (DFT) was determined with a ramp-up protocol with shocks delivered approximately every 10 s with increasing energy in the following increments: 2, 5, 10, 15, 20, 30, and 50 J until VF was terminated. The DFT was determined following 20 s VF and after at least 15 minutes of recovery period, following 7 min VF. If the 50 J internal shock was unsuccessful at terminating VF, shocks were delivered with spoon paddle electrodes and the DFT was assumed to be 70 J.

Data Analysis

LV Basket Data Analysis

The LV endocardium activation patterns were analyzed and divided into one of three activation patterns; 1) chaotic, 2) regular, or 3) synchronized, as we have done previously.^{9, 13, 14} Further details regarding basket catheter data analysis methods can be found in the data supplement of Li et al, 2013.¹³ In brief, the VF episodes were broken into 2-second epochs and activations were detected by taking the 9-point first temporal derivative and finding the most negative the local minimum that exceeds -0.3 v/s with a 50-ms search window followed by a 50-ms refractory period subsequent to an identified activation. A threshold of -0.5 V/s has been used historically to identify cardiac activations for many years.¹⁵ This threshold provides acceptable discrimination during normal sinus rhythm and during short duration VF. However, in many studies that we have conducted analyzing activation patterns during LDVF, we have used a lower threshold of -0.2 or -0.3 V/s.^{13, 16, 17} During periods of prolonged VF, due to the effects of ischemia and the rapid activation rate of VF, many activations were not being detected by the more stringent -0.5 V/s threshold. The lower threshold of -0.3 V/s produces results more consistent with manual overreading and manual activation picking with fewer missing data point than -0.5 V/s threshold. In order to keep our results comparable to our other previously published work and to more accurately detect the activations during LDVF, we have conducted our analysis using the same threshold as we have used previously of -0.3 V/s. Electrodes that do not consistently record activations that exceed -0.3 V/s were assumed to not be in contact with the LV wall and were excluded from the analysis. Synchronicity and regularity indexes were calculated for each 2 second epoch.¹³ The regularity index quantifies the regularity of activation on each electrode and creates a measure of cycle length variability for the LV endocardium and quantitatively it is the mean of the covariance of the cycle lengths of each channel. The synchronicity index quantifies how quickly the entire LV endocardium activates and whether there are distinct pauses on the endocardium between activations and quantitatively it is the ratio of the standard deviations of activation times and the median cycle length of the episode. A two-second epoch was determined to be synchronized when only the synchronicity index is lower than a threshold of 0.1. The two-second epoch was determined to be regular when the regularity index was below the threshold of 0.1. The two-second epoch was determined to be in the chaotic pattern when both synchronicity and regularity indices are higher than the 0.1 thresholds.

Plunge Needle Data Analysis

Electrograms from plunge needles were analyzed to determine the directionality of transmural propagation during long duration VF. Activations were identified by taking the 9-point first temporal derivative and finding the most negative local minimum that exceeds -0.3 v/s with a 50-ms search window followed by a 50-ms refractory period subsequent to an identified activation. Electrograms were compared to neighboring electrograms on the same plunge needles. Historical articles have demonstrated that propagation in cardiac tissue has a minimum threshold of 0.1 m/s.^18-20 Wavefronts that move slower than this speed block rather than conducting. Since the electrodes on our plunge needles are 2 mm apart, we determined that activations on adjacent electrodes with more than 40 ms between them could not have propagated that slowly. We assumed that activations more than 40 ms apart on adjacent electrodes must have originated from different wavefronts rather than conducting from one electrode to the next.

Activation Rate Calculation

The activation rate was calculated for each channel of the basket and for each electrode of the plunge needles by calculating the power spectrum using 5 second epochs. The peak of the power spectrum was picked for each 5 second epoch and the mean activation rates were calculated for all the basket electrodes and for each level of the plunge needles (electrodes 1-6).

Results

The 64-channel LV basket recordings were analyzed and categorized into three activation patterns. Sample 2-second recordings of each pattern are shown in Figure 1, which also include the visualization of activations in each cycle. In chaotic pattern (Figure 1A), highly unpredictable and unorganized activation sequences were observed. In regular pattern (Figure 1B), highly consistent and repeatable activations that primarily propagated from apex to base electrodes were found. Lastly, for synchronized pattern (Figure 1C), highly organized simultaneous activations were observed throughout the LV endocardium. In addition, the transition between activation patterns may be abrupt, such as an example case showing transition of activation patterns, from regular to synchronous, as shown in Figure 2.

Figure 1.

Electrograms and activation times from 64-channel basket recordings in various activation patterns during a sample 2 second epoch: (A) Chaotic (animal 1, synchronicity index = 0.19, regularity index = 0.16); (B) Regular (animal 6, synchronicity index = 0.21, regularity index = 0.05) and (C) Synchronized (animal 2, synchronicity index = 0.04, regularity index = 0.16). The top group of signals shows the raw electrograms from the 64 electrodes on the basket catheter, while the bottom set of signals shows only the picked activations on each electrode. In the picked activations panels, horizontal lines separate each spline of the basket catheter. The bottom most trace on picked activation plots shows a summation of all activations recorded on the basket. Note that for the chaotic and regular patterns (A and B), activations are recorded almost continually while in the synchronized pattern (C) there are clear pauses in activation between each cycle. Blue vertical lines divide successive cycles.

Figure 2.

An example case observed in Animal 5 showing transition of activation patterns, from regular to synchronous. Electrograms (A) and activation picks (B) for each electrode on each spline are plotted over a 6-second period. Regularity indices calculated for 2-second epochs from 296 to 302 seconds were found to be 0.0871, 0.2032, and 0.2150. In the same manner, the synchronicity indices were calculated to be 0.2613, 0.0806, and 0.0397. Note that for entire Spline 8 and first electrode on Spline 2 were excluded since they were identified to be in bad contact or broken for the study.

Activation patterns at various stage of VF

For the 7-minute basket electrode recordings, different activation patterns were observed as the VF progressed. Figure 3 shows the average incidence of all 6 animals of each endocardial activation pattern averaged by minute since VF induction. In Figure 4, the incidence of the different patterns across the entire 7-minute VF period for each animal is shown as the colored line in each panel at the 2 Hz activation rate level. The chaotic pattern was observed primarily during initial VF which has progressed in to regular pattern as VF persisted. Synchronized pattern was not observed until later stage of VF, i.e. at 5 minutes. A regression analysis was performed to determine the trends of the patterns as VF progressed each minute. A linear trend was characterized for the chaotic pattern with the slope of -0.11 incidence per minute (R² = 0.8168), showing a decrease in incidence as VF duration increased. Conversely for non-chaotic (both regular and synchronized) patterns, the linear regression line was determined to be 0.10 incidence per minute (R² = 0.8328), showing increasing incidence as VF duration increased.

Figure 3. Average incidence of chaotic, regular and synchronized patterns for all 6 animals per min during 7 minutes of LDVF.

Figure 4.

Activation rates at different depths of LV freewall as VF progressed in animals 1-6 (AF). Activation rates were calculated for 7 groups. Group ‘basket’ consists of the average activation rate of the 64 basket electrodes. Groups ‘1’ to ‘6’ consist of the average activation rates of the 1^st electrodes (closest to the endocardium) to the 6^th electrodes (closest to the epicardium) on 54 plunge needles. The basket activation patterns as VF progressed are also shown at bottom (at the 2 Hz activation rate level) of each plot.

Discrepancy in endocardial activations of plunge needle recordings at various depths

Activation rates were calculated as VF progressed for both averaged basket and plunge-needle electrode recordings at LV freewall. Figure 4 demonstrates that in SDVF, the activation rates were consistent between all transmural levels. However, after about two minutes, the activation rate of the endocardium exceeded the activation rate of the epicardium. From the basket data pattern analysis, the activation rate gradient was present during both chaotic and regular patterns; while an activation rate gradient was not observed when the VF was in synchronized pattern (see Figure 4B, minutes 5-6).

Variability of pattern quantification parameters due to transmural distance from the endocardium

As VF persisted, the synchronicity and regularity indices were computed for both averaged basket and plunge-needle electrode recordings at LV freewall, which were used to distinguish VF patterns. As shown in Figure 5A, the regularity index values were grouped closely together at the early stage of VF. However, after one minute, the regularity index of the plunge needle recordings toward the epicardium increased VF progressed. The basket data activation pattern suggests that the LV endocardium was in the regular pattern after two minutes of VF since the index fell below 0.1. If the same threshold was applied to any of the plunge needle data, it would not distinguish the regular pattern.

Figure 5.

Variation of synchronicity and regularity indices at different depths of LV freewall: (A) Regularity index of animal 3 during 7-min of VF; (B) Synchronicity index of animal 5 during 7-min of VF. Both indices were calculated for 7 groups as defined in Figure 3. The basket activation patterns as VF progressed are also shown at the bottom of each plot.

For the synchronicity index, as shown in Figure 5B, all of the plunge needle and the basket data showed predominantly similar values as VF progressed. However, at 5 minutes, the basket activation analysis detected synchronized pattern (the synchronicity index less than 0.1). In similar manner to the regularity index, the synchronicity index would exceed the threshold for the plunge needle data and therefore not able to detect the synchronized pattern.

Transmural propagation blockage as suggested by plunge needle recordings and entire endocardium activation as shown in basket electrode data

Activation propagation from endocardium to epicardium along 6-electrode plunge needles during regular and synchronized pattern are shown in Figure 6. In both activation patterns, the endocardial activations demonstrated blockage between neighboring activation cycles as it propagated towards epicardium. In addition, the activation time map on the endocardium from the basket electrode of the sample period (marked by the blue line) revealed propagation of wave front over the entire endocardium in both patterns despite transmural blockage. The LV endocardial activation propagation followed the pathway in the counterclockwise direction from anterior to septum. Moreover, the results suggested faster endocardial propagation for the synchronized pattern than for the regular pattern.

Figure 6.

Representative examples of transmural activation propagation during (A) regular and (B) synchronized patterns. In A and B, left pane shows the activation time map calculated from the basket electrodes, with electrodes near the apex in the middle and electrodes near the base on the outside, and right panel shows the unipolar recordings from 6 electrodes on one plunge needle. The red circles on the right panels show activations identified by activation criteria. The time interval of the left activation time map is shown as a blue line above the unipolar recordings.

Incidences of blockage between electrodes at various transmural depths

To investigate the propagation blockage between neighboring electrodes of the plunge needles, blockage rate plot was presented (Figure 7) for each animal as VF progressed. Effectively there were no conduction block rate gradient observed in SDVF. However, as VF progressed, block rate increased and a block rate gradient from the endocardium to epicardium was frequently observed with higher epicardial blockage. When compared to basket recording's pattern analysis, reduction in block rate was observed for the endocardial electrode pairs during the synchronized pattern.

Figure 7.

Block rates at different depths of the LV freewall as VF progressed. A-F show the block rates from animals 1-6. For each animal, block rates were calculated for 5 groups. Group ‘1-2’ is between the 1^st (closest to the endocardium) and the 2^nd electrodes on 54 plunge needles. The remaining groups are defined the same way towards the epicardium. The basket activation patterns as VF progressed are shown at the top of each plot

Changes of activation patterns in response to failed defibrillation shocks

Time course representation of the VF activation pattern analysis of the 64-channel basket recordings as the ramp-up defibrillation protocol was applied is shown in Figure 8. During only one ramp-up run following SDVF, activation pattern converted to the synchronized pattern. In the other five episodes, SDVF activation patterns remained unchanged in response to failed shocks.

Figure 8.

Time course of VF activation pattern of 64-channel basket data as shocks were given at 10-second intervals. Panel A shows patterns during shocks delivered during SDVF and panel B shows patterns following shocks delivered during LDVF. * indicates episodes that were unsuccessfully defibrillated at 50 J and subsequently defibrillated with epicardial spoon paddles.

However, for LDVF at a mean failed shock strength of 5.6±4.8 J, the VF activation patterns were converted to the synchronized activation pattern which occurred immediately following failed shocks and persisted until the successful defibrillation shock. The results also suggested that conversion of typical VF activation patterns to synchronized patterns occurred more often following LDVF (5 out of 5) than SDVF (1 out of 6). The Fisher's exact test confirmed that the associations between SDVF and LDVF conversions to synchronized pattern are different (p = 0.0152). The defibrillation threshold for this experiment was determined to be 37±31 J for SDVF and 16±11 J for LDVF (p=0.09).

Discussion

The primary findings of this study are: 1) Distinct activation patterns emerged during LDVF with chaotic dominating the first couple of minutes and regular dominating after 2 minutes, with synchronized emerging sometime in late VF. 2) Regular and synchronized activity is driven by endocardial activations that block and break up transmurally. 3) Failed defibrillation shock change chaotic and regular activation patterns to synchronized in LDVF but not in SDVF.

Distinct activation patterns determined by LV endocardial mapping

Other investigators have categorized VF into different stages depending on mechanical or electrical activation as recorded directly from the surface of the heart or from the EKG. More than 75 years ago, Wiggers conducted high speed cinematographic filming of the fibrillating canine heart.²¹ He proposed 4 stages to VF based on the mechanical motion of the heart: 1) Stage I, which consists of 3-6 undulatory contractions and lasting < 1 second, 2) Stage II, lasting 15-40 seconds, which is the convulsive incoordination stage, 3) Stage III, lasting 2-3 minutes, which was the tremulous incoordination phase and was characterized by progressively smaller, independently contracting areas, and finally 4) Stage IV, which was atonic fibrillation, with little to no contractile motion. Other studies have characterized VF into two,²² four,⁶ five,⁴ or more²³ patterns of activation based on electrical wavefront parameters, but Wiggers stages of VF has been the most enduring characterization of VF activation patterns. The analysis conducted in this study is fundamentally different in that the patterns of VF activation are based on endocardial activation sequences that are not detectable in the midwall, epicardium, or from the body surface. The activation patterns are not necessarily sequential as many of the other stage characterization methods, but can change in every 2 second time period, as evidenced by the VF characterization traces in Figure 4.

Three distinct activation patterns appeared on the LV endocardium. Early VF was dominated by the chaotic pattern, which was demonstrated by irregular endocardial cycle lengths and activity appearing scattered as small wavefronts on different regions of the endocardium. Before 2 minutes of VF, an activation rate gradient (endo- to epicardium) was not typically present. This pattern of activation on the basket and transmural plunge needle electrodes is consistent with intramural reentry or wandering wavelet reentry.^24-27

By the third minute of VF, the regular pattern became dominant on the endocardium and remained the most common pattern until VF was terminated at 7 minutes of VF. This pattern of activation exhibited continuous endocardial activation with regular cycle lengths that repeated around nearly the entire LV endocardial surface. This regular activity broke down and blocked transmurally and an activation rate gradient emerged. This regular endocardial activity is consistent with a stable mother rotor on the endocardial surface.^{24, 27}

As we have reported previously, the synchronized pattern of activation emerged later in VF (>5 min) and was characterized by irregular cycle lengths and nearly synchronized activity on the entire endocardial surface. This activation appears to consist of an activation that enters the excitable Purkinje fiber system, which then spreads the activation rapidly throughout the endocardium through the specialized conduction system. The wavefront appears to travel towards the epicardium, where the ischemic myocardium leads to breakup of the coherent endocardial wavefront. Previous studies have shown that there is heterogeneity in the development of the endo-epicardial activation rate gradient, with some areas conducting to the epicardial surface and other areas of the epicardium not activating hardly at all.^{28, 29} The potential exists for a wavefront to find a region of unidirectional block and the wavefront may follow a pathway back to the endocardium, where it activates the excitable Purkinje system and causes another synchronous activation on the endocardium. A second potential mechanism of activation during the synchronized pattern is that triggered activity in the Purkinje system may lead to focal activations in the Purkinje system, particularly in prolonged VF.^{17, 27} These focal activations may then spread rapidly through the LV Purkinje system before blocking and fractionating transmurally.

The endocardial activation rate during the synchronized pattern is slower than during the chaotic or synchronized patterns (Figure 4B and 4E), which may indicate that the synchronized pattern is overdrive suppressed by the other, more rapid activation types. Failed defibrillation shocks may be sufficient to terminate wandering wavelet reentry or mother rotor reentrant circuits, which made it possible for the synchronized pattern to emerge.

Endocardium activation propagates transmurally to the epicardium

The activations of transmural LV freewall was investigated with the plunge needle electrodes at various depths from endocardium to epicardium. As VF progressed beyond 2 minutes, the results showed the development of an activation rate gradient from endocardial electrodes to the epicardial electrodes. This implies that the activations initiate from the endocardium and then translate to epicardium with transmural block occurring at some locations. The development of an activation rate gradient has been reported after the first 1-2 minutes of VF in rabbit³⁰, canine^{16, 31}, and human³² hearts. Additionally, this was reaffirmed similarly by the high rate of conduction block for the epicardial plunge needle electrodes, reiterating the observation that VF propagated from the endocardium, with blockage en route, to the epicardium. Previous studies have shown that the Purkinje fiber system plays an important role in driving the activation rate in LDVF,^{16, 17} and elimination of the subendocardium and/or Purkinje system significantly alters LDVF conduction patterns,^{16, 33} which is consistent with the driving role of the endocardium described in this work. Heterogeneity in distribution of the endo-epicardial rate gradient was qualitatively observed (as seen in Figure 6), however it was not quantified in this study. A previous study has demonstrated heterogeneity in the development of the rate gradient in different area of the left ventricular wall, and we expect similar heterogeneity would have been found in our recordings.³⁴

Defibrillation shocks shift the activation patterns of LDVF but not SDVF

As demonstrated the failed defibrillation shocks significantly change the activation pattern of LDVF before termination, but not in SDVF. The results of this study demonstrated that there are differences in the response of the heart to defibrillation shocks depending on the duration of VF before the shocks. In addition, the results implied that the mechanism of VF maintenance can be altered from failed shocks. For instance, failed shocks could change the onset of VF initiated by mother rotor which maintains the regular pattern to the synchronized pattern.

Furthermore, investigation is warranted to determine whether the trend of lower defibrillation success is related to the change in activation patterns observed following failed shocks in LDVF. Other studies have shown a trend towards lower DFTs in LDVF compared to SDVF.^35-37 Ultimately, with better understanding of the mechanisms and the differentiation of LDVF to SDVF can potentially lead to customization of clinical therapy based on VF duration and activation patterns.

Study Limitations

The hearts used in this protocol were normal rather than diseased models. However, there are many potential models that could have been used that do not accurately represent the range of conditions known to be associated with VF risk. Acute myocardial infarction, ischemia reperfusion, and heart failure are all associated with VF risk. Patients may also have extensive fibrosis, scar formation, structural heart disease, or valvular dysfunction, or any combination of these conditions. Idiopathic VF affects patients that appear to be free of these conditions. This model used in the present study was chosen because the Purkinje system in dogs is more similar to humans than other models (pigs, goats, sheep, or cows). We chose a healthy model as the potential for loss of the animal during instrumentation with the plunge needles was deemed to be prohibitive. In order to study the temporal progression of the VF, we utilized a model that was as stable as possible prior to the onset of global ischemia with VF induction.

Opening of the chest and insertion of plunge needles could have potentially changed the endocardial VF activation patterns. However, previous studies have shown that insertion of plunge needles do not change epicardial activation patterns in acute studies.^{12, 38}

Plunge needles of uniform length were inserted from the epicardial surface of the heart. Due to the uneven thickness of the LV at different regions, the most endocardial electrodes of the plunge needles were not uniformly located on the endocardium. In a few locations, the most endocardial electrodes may have been in the blood chamber while at other locations the most endocardial electrode may have been in the ventricular midwall. This may explain why there was a substantial difference in activation rate between the most endocardial plunge needles and the endocardial basket electrodes (Figure 4).

While defibrillation shocks following SDVF may be delivered with a transvenous shocking coil as we have done in this study, defibrillation shocks following LDVF would typically be delivered with external paddles or patches. In order to make the testing between SDVF and LDVF defibrillation more comparable, shocks were delivered with internal defibrillation coils in both cases.

What is known?

• Ventricular fibrillation is a leading cause of death due to sudden cardiac arrest.

• The organization of fibrillation changes as time progressed.

What the Study Add?

• Organized patterns in prolonged ventricular fibrillation is driven by activity on the endocardium.

• Failed defibrillation shocks in prolonged but not short duration ventricular fibrillation alter endocardial fibrillation patterns.