Electrophysiological Mechanism of the Characteristic Electrocardiographic Morphology of Torsade de Pointes Tachyarrhythmias in the Long-QT Syndrome
Detailed Analysis of Ventricular Tridimensional Activation Patterns
Background The long-QT syndrome (LQTS) is an electrophysiological (EP) entity characterized by prolongation of cardiac repolarization and the occurrence of polymorphic ventricular tachyarrhythmias (VTs), sometimes with a twisting QRS morphology, better known as torsade de pointes (TdP). In the present study, detailed analysis of ventricular tridimensional activation patterns during nonsustained TdP VT was performed to provide an EP mechanism of the periodic transition in QRS axis.
Methods and Results The studies were conducted with the anthopleurin-A canine model of LQTS. Tridimensional isochronal maps of ventricular activation were constructed from 256 bipolar electrograms obtained from the use of 64 plunge needle electrodes. In 26 episodes of nonsustained TdP VT, detailed activation maps could be accurately constructed during QRS-axis transitions in surface ECGs. The initial beat of all VTs consistently arose as a subendocardial focal activity, whereas subsequent beats were due to reentrant excitation in the form of rotating scrolls. The VT ended when reentrant excitation was terminated. In 22 of 26 episodes, the transition in QRS axis coincided with the transient bifurcation of a predominantly single rotating scroll into two simultaneous scrolls involving both the right ventricle and left ventricle separately. The common mechanism for initiation or termination of bifurcation was the development of functional conduction block between the anterior or posterior right ventricle free wall and the ventricular septum. In 4 of 26 episodes, a fast polymorphic VT, with an apparent shift in QRS axis, was due to a predominantly single localized circuit that varied its location and orientation from beat to beat, with the majority of ventricular myocardium being activated in a centrifugal pattern.
Conclusions The study provides for the first time an EP mechanism for the characteristic periodic transition of the QRS axis during TdP VT in the LQTS.
The LQTS is an EP entity characterized by prolongation of cardiac repolarization, reflected as a long-QT interval in the surface ECG and the frequent occurrence of polymorphic VTs, sometimes with a twisting QRS morphology, better known as TdP.1,2 Since the initial description of TdP by Dessertenne1 >30 years ago, many electrophysiologists have been intrigued by the QRS morphological characteristics of the arrhythmia, whereas little progress has been made toward developing a cohesive EP mechanism for the characteristic morphology.
We developed a surrogate canine model of the LQT3 syndrome by using the neurotoxin AP-A.3 In a recent report using this model, we showed that the initial beat of polymorphic VT consistently arose as focal activity from a subendocardial site, whereas subsequent beats were due to successive subendocardial focal activity, reentrant excitation, or a combination of both mechanisms. Reentrant excitation was due to infringement of a focal activity on the spatial dispersion of repolarization, resulting in functional conduction block and circulating wave fronts.4
In the previous report,4 only a few examples of self-terminating TdP VT were shown, and the EP mechanism or mechanisms of the twisting QRS pattern were not analyzed in detail. In the present study, we conducted new experiments to provide detailed analysis of ventricular tridimensional activation patterns during nonsustained TdP VT and provide an EP mechanism of the periodic transition in QRS axis in surface ECG leads.
The present study was approved by the Animal Studies Subcommittee of the local institutional review board and conformed to the guiding principles of the Declaration of Helsinki. Experiments were performed on eight purpose-bred mongrel puppies 10 to 12 weeks old and weighing 3.5 to 5.0 kg. Puppies were preanesthetized with sodium thiopental (17.5 mg/kg IV) via the cephalic vein. Puppies were then intubated and anesthetized with 1.0% to 2.0% isoflurane (vaporized in 100% O2) via a positive ventilation anesthesia machine (F500; The Forreger Co). Catheters were inserted into the femoral vein for administration of fluids and drugs and into the femoral artery to monitor blood pressure. ECG leads I, aVF, and V1 and blood pressure (Statham transducer; Gould) were continuously monitored with a physiological recorder (VR12; PPG Industries). The heart was exposed through a midsternotomy. Core temperature was maintained constant at 37°C by the use of a thermostatically controlled thermal blanket and heat lamp. AP-A was administered as an intravenous bolus of 25 μg/kg followed by a maintenance dose of 1.0 μg/kg per minute. Wild-type AP-A produced through a bacterial expression system was used in this study.5
In five experiments, vagal stimulation4 was applied to slow the heart rate. In the remaining three experiments, complete AV block was achieved by radiofrequency electrode catheter ablation, resulting in a slow ventricular escape rhythm. On completion of the experiment, the animal was euthanized by electrical induction of ventricular fibrillation, and extirpation of the heart under general anesthesia was performed.
Recording Electrodes and Electrode Localization
Sixty-four plunge needle electrodes were used for tridimensional mapping of the whole ventricle. Details of plunge needle electrode construction and electrode localization were previously reported.4 After insertion of plunge needle electrodes, the chest wall was approximated. After termination of the experiment, each plunge needle was removed and carefully replaced by a labeled maptack pin in the exact electrode site. The heart was removed, fixed in formalin, and the location was noted of the labeled pins in relation to each other and to anatomic landmarks. The pins were replaced by small color-coated plastic brush bristles to facilitate sectioning. The heart was then cut transversely into five slices ≈5 to 7 mm thick. The outline of each slice was traced carefully to show the exact insertion site and direction of each electrode, as well as the site of the most distant bipole pair. The tracings were then enlarged for later tridimensional isochronal map construction.
Data Acquisition and Isochronal Mapping
Bipolar electrograms were acquired using two variable-gain 128-channel multiplexed data acquisition systems (DSC 2000; INET Corp), allowing simultaneous recording of 256 channels. Details of the mapping system were previously reported.4 Activation times were determined according to previously published criteria.6,7 Briefly, in uniphasic and triphasic signals, the peak voltage is a reliable predictor of activation time.7 In biphasic signals, the activation time was selected at the maximum slope. Computer-generated isochrones of activation were derived from the activation time data and delineated by closed contours at 20-ms intervals beginning with the earliest detected site of activation. For the whole ventricle activation maps, zones of functional unidirectional conduction block were identified using previously defined criteria.6,7 A continuous line or surface was drawn through these regions and was defined as a zone of functional conduction block.
In eight experiments, there were a total of 36 episodes of nonsustained VT available for analysis that varied in length from 4 to 39 beats (12±6 beats, mean±SD). The CL of VT ranged from 205 to 295 ms (244±28 ms, mean±SD). Fig 1⇓ illustrates four runs of nonsustained VT from four different experiments. There was no statistical difference in the mean CL of VT in dogs subjected to vagally induced bradycardia (241±26 ms) versus complete AV block (246±31 ms, unpaired Student’s t test). Fig 1A⇓ illustrates a 12-beat run of nonsustained VT at an average CL of 273 ms. In this experiment, complete AV block was present, and the first and last beats in the recording represent slow ventricular escape impulses. The VT had a polymorphic configuration with the distinct undulating pattern especially evident in the second half of the episode in leads I and aVF. Fig 1B⇓ illustrates a 15-beat run of nonsustained VT at an average CL of 235 ms with a characteristic change of the QRS axis as if the complex rotated around the baseline. This is especially evident in lead aVF. Fig 1C⇓ and 1D⇓ illustrate longer runs of nonsustained VT at an average CL of 206 and 213 ms, respectively.
The tridimensional ventricular activation maps showed that the initiating beat of each episode of nonsustained VT (marked by star in Fig 1⇑) arose as focal activity from anywhere in the subendocardium and with no preference to a particular site. This beat was always coupled to the prolonged QT interval of the preceding basic impulse. In the present study, subsequent beats in a run of nonsustained VT were always due to intramural reentrant excitation. Each episode terminated when reentrant excitation failed to continue. After termination of reentrant excitation, one or more beats of varying subendocardial focal origin could occur at irregular and consistently longer CLs compared with preceding reentrant tachyarrhythmia (these are marked by asterisks in Fig 1⇑).
Fig 2A⇓ through 2C illustrate the tridimensional ventricular activation patterns of the 12-beat VT shown in Fig 1A⇑. The isochrones are drawn as closed contours at 20-ms intervals. (Successive isochrones are labeled as 1, 2, 3, and so on rather than 20 ms, 40 ms, 60 ms, and so on to make it easier to follow the activation patterns of subsequent beats of the VT.) Conduction block between electrode sites is represented in the maps by heavy solid lines. The initiating beat of the VT had a focal subendocardial origin (marked by a star in section 1). Multiple sites showed functional conduction block. The activation wave front advanced from section 1 to section 4, where it proceeded in a very slow counterclockwise circular pathway around the LV cavity before reactivating sites in sections 3 and 4 at isochrone 20, initiating the first reentrant cycle. During V2, the activation wave front advanced in a clockwise circular pathway in sections 4 and 5 around the LV cavity with reactivation occurring at isochrone 37 in section 4. During V3 through V6, reentrant excitation continued in clockwise circular pathways in sections 3 and 4 around the LV cavity. However, the reentrant pathway was different from beat to beat. In the tridimensional maps, the circulating wave fronts are more accurately described as rotating scrolls of activation. During V7, the rotating scroll continued in sections 4 and 5. However, the development of functional conduction block in the midseptum in sections 2 and 3 forced the occurrence of a rotating scroll around the RV free wall simultaneous with the scroll in sections 4 and 5. This circulating wave front was extinguished by collision in the septum with a counterclockwise wave front around the LV cavity. During V8, a similar activation pattern continued. During isochrone 97, the activation wave front bifurcated into two distinct simultaneous rotating wave fronts: a clockwise wave front around the LV cavity in section 4, and a clockwise wave front around the RV cavity in section 2. The two wave fronts continued in V9 with reactivation at isochrone 106 in section 5 and reactivation at isochrone 104 in section 2, respectively. The clockwise wave front in section 5 blocked at isochrone 109 while the circular wave front around the RV cavity continued for one more complete cycle in sections 2 and 3. During V10, this activation wave front proceeded quickly from section 2 to section 4 in a basal-to-apical direction. The circular activation around the RV terminated due to the development of functional conduction block between the RV free wall and anterior septum (section 2 in V11), whereas a very slow clockwise circular wave front redeveloped in section 4 and continued during V12. During V12, the clockwise circular wave front blocked in the anterolateral wall of the LV (in sections 2 through 5) and in the anterior wall of the RV (in section 1), ending the reentrant activity and the episode of VT.
Fig 3⇓ shows selected electrograms along the reentrant pathway during the ectopic beat (V1) that initiated the VT shown in Fig 2⇑. The V1 impulse had a long coupling interval compared with the proceeding basic beat (700 ms). The first reentrant cycle was relatively long (20 isochrones, 400 ms). Bipolar electrograms recorded from the slowly conducting pathway in section 4 had a wide multicomponent configuration. The first of the double potentials of electrogram J represents activation (A) during isochrone 10, and the second deflection (E) represents an electrotonic potential synchronous with activation of isochrone 13 in section 5, immediately caudal to electrode site J. On the other hand, the first of the double potentials of electrogram K represents an electrotonic potential (E) synchronous with activation at isochrone 9 across the arc of functional conduction block in section 4, whereas the second potential (A) represents activation during isochrone 13. A similar interpretation applies to the double potentials of electrogram Q.
Fig 4⇓ shows selected electrograms of the two simultaneous reentrant circuits during the interval that coincided with the initial “gradual” transition of the QRS axis in lead aVF. Electrograms A through G were recorded along the clockwise circuit around the RV cavity in section 2, and electrograms H through N illustrate the activation of the simultaneous clockwise circuit around the LV cavity in section 4.
Fig 5⇓ shows selected electrograms during V10 that coincided with the last “abrupt” transition of the QRS axis in lead aVF. As described previously in Fig 3⇑, several bipolar electrograms recorded along the very slow conducting wave front in section 4 had broad multicomponent configurations that reflect electrotonic interaction across zones of functional conduction block6,7 and/or very slow anisotropic conduction transverse to the long axis of myocardial fibers8 (see electrogram D).
Correlation of Transition in QRS Axis and Tridimensional Isochronal Activation Pattern
The twisting of the QRS axis in the second half of the VT episode shown in Fig 1A⇑ and Figs 2 through 5⇑⇑⇑⇑ was more evident in leads I and aVF but not in lead V1. In particular, lead aVF showed a “gradual” twisting from a sharp negative complex in V7 to a sharp positive complex in V10 with two low-amplitude intermediate deflections labeled V8 and V9. On the other hand, the transition from the sharp positive V10 complex to the sharp negative V11 complex was rather “abrupt.” The changes in the QRS configuration in lead aVF correlated with the changes in tridimensional ventricular activation pattern. The sharp negative QRS configuration of V7 correlated with the fast apical-to-basal activation pattern. The gradual transition between V7 and V10 coincided with the change from a single reentrant wave front to two simultaneous circulating wave fronts around the RV and LV cavities. The opposing direction of the activation wave front running predominantly from apex to base in the free RV wall and septum and the activation front running predominantly from base to apex in the free LV wall explains the relatively small-amplitude QRS complex in lead aVF labeled V8. The block of the apical LV circuit in V9 allowed a basal-to-apical activation of the ventricles during V10 that correlates with the positive component of the V10 QRS complex in lead aVF. The sharp negative component of the V10 QRS complex, followed by a largely isoelectric segment preceding the inscription of the sharp negative QRS complex of V11- coincided with the termination of the RV circuit and the development of a very slow circulating wave front in apical section 4. Finally, the negative QRS configuration of V11 and V12 in lead aVF correlated with the predominantly apical-to-basal activation pattern of both beats. Although the twisting of the QRS axis was also quite evident in lead I, the correlation between the QRS configuration in lead I, and the tridimensional activation pattern was more difficult to interpret. This is probably because of the difficulty in correlating the frontal plane activation pattern in each of the five sections with the axial projection of lead I on the body surface compared with lead aVF.
Transient Bifurcation of a Single Rotating Scroll During QRS-Axis Transitions
It was not always possible to analyze in detail the tridimensional ventricular activation pattern during the transition in QRS axis in all of the 36 episodes of nonsustained VT because of limitations of the resolution of the recordings. This limitation applied in particular to longer VT episodes at shorter CLs and with multiple undulations of the QRS axis. However, in 22 of 26 examples in which detailed activation maps could be accurately constructed during transitions in QRS axis, the EP mechanism was similar to that shown in Figs 2 through 5⇑⇑⇑⇑. The occurrence of a gradual shift in the QRS axis was associated with the bifurcation of a single circulating wave front into two simultaneous wave fronts. The termination of one of the two circuits and the reestablishment of one predominant circuit would also usher a transition in the QRS axis. A longer transition covering several cycles was usually associated with the gradual dominance of one of the circulating wave fronts.
Wherever two wave fronts occurred, they consistently involved the RV and LV separately. The RV circuit always rotated around the RV cavity, whereas the LV circuit could develop either around the LV cavity or within the LV free wall. This is shown in Fig 6⇓. The figure illustrates the tridimensional activation maps of beats V6 through V9 of the VT shown in Fig 1B⇑. During these four cycles, the QRS axis in lead aVF shifted from a predominantly negative QRS complex to a predominantly positive QRS complex, with the QRS of V7 through V9 showing a transitional configuration. On the other hand, there was a “sudden” shift of the QRS axis in lead V1 from a predominantly negative QRS complex in V6 and V7 to a predominantly positive QRS complex in V8 and V9. The activation maps show that during V6, there was continuation of a predominantly single clockwise circulating wave front around both the RV and LV due to the presence of functional conduction block in the interventricular septum. During V7, a predominantly single clockwise wave front around the LV cavity continued while the RV wall was activated passively. During V8, two distinct circulating wave fronts developed: a counterclockwise wave front in the LV free wall in sections 2 and 3, and a counterclockwise wave front around the RV cavity in sections 1 through 3. The establishment of the two wave fronts was initiated by the development of functional conduction block between the RV free wall and the posterior septum in section 1 through 3. Both the RV and LV circuits continued in V9. The RV circuit blocked during V10 (not shown), whereas a single intramural circuit in the LV free wall continued. The transitional QRS complexes of V8 and V9 in lead aVF could be explained by the presence of two opposing activation wave fronts. Similarly, the sudden transition of QRS axis in lead V1 could be explained by the sudden shift from a predominantly RV-to-LV activation pattern in V7 to one with predominantly late activation of the RV free wall in V8. Reentrant activation terminated by beat V14. The last impulse (V15) was due to a subendocardial focal activity.
Apparent QRS-Axis Transitions During Single Circulating Wave Fronts
In four episodes of nonsustained VT, the transition in QRS axis was not associated with the transient development of two competing LV and RV circulating wave fronts but rather a localized single reentrant circuit that varied in location from beat to beat, with most of the ventricle being activated in a centrifugal fashion. In these examples, the transition in QRS axis did not resemble the “torsade de pointes” pattern, but instead was described as a polymorphic configuration. One example is shown in Fig 7⇓. The surface ECG leads show a seven-beat run of nonsustained VT with a polymorphic QRS configuration. A transition in QRS axis is seen in all three leads. Fig 8A⇓ and 8B⇓ illustrate the tridimensional activation maps of the sinus beat and V1 through V7. As was always the case, the V1 beat that initiated the VT episode arose as a focal subendocardial activity (marked by the star in section 2) and initiated a localized intramural reentrant circuit in the apical region of the LV wall in sections 4 and 5. Reentrant activation continued during V2 through V6 with continuous shifting of the location and orientation of a single reentrant wave front within a relatively small region in the apical portion of the LV in sections 4 and 5. During V2 through V6, the majority of ventricular myocardium (sections 1 through 3) were activated in a centrifugal pattern from the localized reentrant wave front in the apical region. The beat-to-beat variation in the activation pattern of sections 1 through 3 could explain the changes in QRS configuration in the surface ECG leads. For example, in lead aVF, the positive QRS configuration of V1 correlated with the predominantly basal-to-apical activation pattern. On the other hand, the largely negative QRS configuration of V2 through V4 correlated with a predominantly apical-to-basal activation patterns. The negative small-amplitude (almost isoelectric) QRS configuration of V5 could be explained by the fact that although the major direction of activation of both RV and LV tended to run from apex to base, the majority of ventricular myocardium was activated simultaneously within only 40 ms (isochrones 40 through 42). Reentrant activation terminated in V6 and V7 was due to a subendocardial focal discharge (marked by a star in section 5).
Fig 9⇓ illustrates selected electrograms during V1 and V2 in the seven-beat VT episode shown in Figs 7⇑ and 8⇑. Electrograms A through H illustrate the continuous activation during the first reentrant cycle. As was shown in Fig 3⇑, bipolar electrograms recorded in close proximity to arcs of functional conduction block frequently exhibited double potentials that represented electrotonic (E) and activation (A) potentials (see electrograms E, H, and I). The V2 map and electrograms I through L illustrate an important observation. The first reentrant wave front reactivated a subendocardial site in the LV wall in section 4 at isochrone 11 (electrogram H) and isochrone 12 (electrogram I). This was followed by a rapid activation (within 5 ms) of localized corresponding subendocardial sites in sections 3, 2, and 1 (represented by electrograms J, K, and L, respectively). This fast, localized subendocardial activation strongly suggests retrograde conduction along branches of the left bundle. This observation emphasizes the potential contribution of conduction through the His-Purkinje system to the varying activation pattern during polymorphic VT. Fig 9⇓ also shows that during the focal beat, V7, the electrode sites recorded relatively synchronous electrograms compared with those recorded during reentrant excitation, especially at apical sites (E through H). The onset of V7 focal activity occurred ≈400 ms after termination of reentrant activation.
EP Mechanism of TdP VT in the LQTS
The present study confirms, in more detail, the basic tenets of the EP mechanism of polymorphic VT in the LQTS as previously reported.4 However, an important modification is necessary. In the previous report, we stated that “the initial beat of polymorphic VT consistently arose as a subendocardial focal activity, whereas subsequent beats were due to successive subendocardial focal activity, reentrant excitation, or a combination of both mechanisms.” Nonsustained polymorphic VT in LQTS were separated into two groups according to whether all beats in the run were focal in origin (group A) or a combination of focal activity and reentrant excitation (group B). The CL of group A VT was significantly longer (335±26 ms, mean±SD) compared with group B VT (242±9 ms, mean±SD, P<.001).4 Group A VT never resembled the TdP pattern. In the present study, all fast polymorphic VTs (with or without the classic TdP configuration) were due to successive reentrant excitation initiated by a subendocardial focal activity. As already described here, in some instances, termination of fast reentrant VT was followed by one or more focal beats occurring at a much longer CL. However, during the fast VT, there never was a combination of focal and reentrant excitation. Thus, our previous report of such a combination of the two mechanisms (eg, Figs 17 and 18 in 4 is not confirmed by the present observations. It is fair, however, to mention that the previous conclusion was qualified as follows: “Because of the limitations of the recordings resolution … beats that were interpreted as focal in origin may have been the result of a small reentrant circuit.”4
Mechanisms of Twisting QRS Axis of TdP VT
We have clearly demonstrated in the present report that when tridimensional activation patterns could be accurately constructed during the transition in QRS axis of a TdP VT, the mechanism was consistently due to the transient bifurcation of a predominantly single rotating scroll into two simultaneously rotating scrolls involving both the RV and LV. The RV scroll always rotated around the RV cavity. On the other hand, the LV scroll could develop either around the LV cavity (Fig 2⇑) or in the LV free wall (Fig 6⇑). The initiating mechanism for the bifurcation of the single wave front frequently was the development of functional conduction block between the anterior or posterior RV free wall and the interventricular septum. The termination of the RV wave front was also frequently associated with the development of functional conduction block ahead of the circulating wave front between the RV free wall and the anterior or posterior border of the septum (see V9 map in Fig 6⇑). In other instances, the RV circulating wave front was extinguished through collision with an opposing wave front in the interventricular septum (see V7 map in Fig 2B⇑). The RV circulating wave front usually did not exhibit a localized zone of slow conduction. This may suggest that the conduction block that develops at the border between the thin RV free wall and the much thicker interventricular septum may be, at least in part, secondary to an impedance-mismatch mechanism.9 On the other hand, LV circuits frequently encompassed a varying zone of slow conduction, and conduction block usually developed in this slow zone probably secondary to decremental conduction. Although it was more difficult to correlate accurately, there was evidence that a period of transitional complexes covering more than one cycle was associated with gradual dominance of one of the two circulating wave fronts before termination of the other wave front (see the transitional QRS complexes labeled V8 and V9 in Fig 2B⇑ and 2C⇑).
A tachyarrhythmia with TdP-like morphology in a transmural ECG could be induced by erythromycin, an IKr blocker, in an arterially perfused canine LV wedge preparation.10 Simultaneous transmembrane recordings were highly suggestive of reentry as the underlying mechanism. It is possible that the transient bifurcation of a single impulse into two scroll waves as the basis of QRS transitions of TdP can occur anywhere in the ventricle. The characteristic mechanism involving the RV and LV described here may be partly related to the longer pathway required to accommodate the longer wavelength of reentrant circuits in the AP-A model (average CL, 244 ms).
The accuracy of tridimensional mapping of activation is directly proportional to the resolution of the recordings. In the present study, the recording resolution was ≈4 to 8 mm between needles and 2 mm along the needles. Sometimes, this may not be adequate to resolve small reentrant circuits, which may be misinterpreted as focal activity. This limitation, however, had little bearing on the results in the present study because conclusions regarding the mechanism of transition of QRS axis of TdP VT were obtained from VT runs in which tridimensional activation patterns could be accurately constructed and consistently revealed reentrant excitation. Although in several examples of fast TdP VT with periodic transitions in QRS axis, the resolution of the recordings did not allow accurate analysis of successive transition episodes, there is no reason to believe that a different EP mechanism was operative. The correlation between the isochronal tridimensional activation pattern and the body surface ECG should be considered an approximation. The position of the heart in the thoracic cavity was disturbed by the thoracotomy, even through the chest wall was approximated, and by insertion of multiple plunge needles. Furthermore, there was no way to evaluate with accuracy the relative contribution of the different isochrones of activation to the overall pattern of the surface ECG; the latter represents a vectorial summation of multiple isochrones.
Selected Abbreviations and Acronyms
|LV||=||left ventricular, ventricle|
|RV||=||right ventricular, ventricle|
|TdP||=||torsade de pointes|
This work was supported in part by funds from the Department of Veterans Affairs Medical Research Service (Drs El-Sherif and Restivo) and a Medtronic Japan Fellowship Award (Dr. Chinushi). We wish to thank Victoria Stoyanovsky, Richard Levin, and Drs Jacqueline Chen, Sabu John, and Monir Chaudhry for their technical support and Antoinette Wells and Joyce Ince for their care of the animals.
- Received July 3, 1997.
- Revision received September 4, 1997.
- Accepted September 9, 1997.
- Copyright © 1997 by American Heart Association
Dessertenne F. La tachycardie ventriculaire á deux foyers opposés variables. Arch Mal Coeur. 1966;59:263–272.
El-Sherif N, Zeiler RH, Craelius W, Gough WB, Henkin R. QTU prolongation and polymorphic ventricular tachyarrhythmias due to bradycardia-dependent early afterdepolarizations. Circ Res. 1988;63:286–305.
El-Sherif, N, Caref EB, Yin H, Restivo M. The electrophysiological mechanism of ventricular tachyarrhythmias in the long QT syndrome: tridimensional mapping of activation and recovery patterns. Circ Res. 1996;79:474–492.
Gallagher MJ, Blumenthal KM. Cloning and expression of wild-type and mutant forms of the cardiotonic polypeptide anthopleurin B. J Biol Chem. 1992;267:13958–139636.
Restivo M, Gough WB, El-Sherif N. Ventricular arrhythmias in subacute myocardial infarction period: high-resolution activation and refractory patterns of reentrant rhythms. Circ Res. 1990;66:1310–13277.
Nassif G, Dillon SM, Rayhill S, Wit AL. Reentrant circuits and the effects of heptanol in a rabbit model of infarction with a uniform anisotropic border zone. J Cardiovasc Electrophysiol. 1993;4:112–133.