Long-term Temporal Patterns of Ventricular Tachyarrhythmias
Background Technological limitations have precluded investigation of long-term temporal patterns of ventricular tachyarrhythmia recurrences. Newer implantable cardioverter-defibrillators permit such analyses by accurately recording the time and date of tachycardia detections during long-term follow-up. This study tests the hypothesis that ventricular tachycardia occurrences are randomly distributed over time in individual patients.
Methods and Results The time and date of 727 episodes of ventricular tachyarrhythmias were recorded from the data logs of 31 patients with implantable cardioverter-defibrillators followed for a median of 177 days (range, 7 to 782 days). All patients had three or more ventricular tachycardia detections and no detections from causes other than ventricular arrhythmias. In 28 of 31 patients, the distribution of the interdetection time intervals during follow-up differed significantly (all P<.01) from an exponential model distribution of interdetection intervals that assumed that detections were equally likely to occur at any time during follow-up (random). The Kolmogorov-Smirnov goodness-of-fit test was used to compare sample and model distributions. In each patient, the nonrandom distributions resulted from a preponderance of interdetection time intervals that were shorter than predicted by the random model, resulting in a temporal clustering of arrhythmic events. The interdetection interval was ≤1 hour and ≤91 hours for 55% and 78% of all intervals, respectively. When only those episodes receiving shock or antitachycardia pacing therapy were analyzed, 25 of 29 patients still manifested nonrandom distributions (all P<.01). When only episodes with tachycardia rates >240 beats per minute were analyzed, 11 of 13 patients manifested nonrandom distributions (all P<.01).
Conclusions Ventricular tachycardia detections and delivered antitachycardia therapies by implantable cardioverter-defibrillators are nonrandomly distributed throughout long-term follow-up in the majority of patients. The temporal clustering of these arrhythmic events may allow preemptive antiarrhythmic therapy and should be considered in the design of therapy based on suppression of spontaneous ventricular arrhythmias to statistically derived end points.
Ventricular tachyarrhythmias remain the most common cause of cardiovascular death in most industrialized countries; however, little is known about the long-term temporal patterns of arrhythmia occurrence.1 2 3 Methodological shortcomings have limited previous studies to the use of telemetric or ambulatory ECG monitoring over observational periods of only hours to days.4 5 6 7 8 Late-generation implantable cardioverter-defibrillators (ICDs) now have the ability to accurately record the time and date of each tachycardia detection over years of follow-up, thus providing unique insights into the long-term patterns of tachycardia occurrences.9 10 Accurate description of these temporal distributions may be extremely valuable in the design of new treatment strategies and may directly affect current management strategies that depend on ECG monitoring to document suppression of spontaneous ventricular ectopy by serial drug testing.11 12 This study describes for the first time the long-term temporal patterns of ventricular tachycardia occurrence in humans as documented by chronically implanted ICDs and tests the hypothesis that the tachycardia occurrences are randomly distributed over time.
Description of ICDs
The Ventak PRx (Cardiac Pacemakers, Inc) and Guardian 4210 ATP (Telectronics) ICDs were used in this study. Both devices provide extensive retrievable data-logging capabilities, including storage of the time, date, and tachycardia cycle length and the ICD response to each tachycardia detection. The Guardian 4210 has electrogram storage capacity. Both devices use bipolar endocardial or two epicardial unipolar leads for pacing and sensing. The Guardian 4210 device uses epicardial patch electrodes; the PRx uses epicardial patch electrodes or a transvenous endocardial lead (Endotak, CPI) system for shock delivery. The Guardian 4210 uses programmable fixed gain sensing amplifiers (0.5 to 5.7 mV). The PRx uses automatic gain adjusting amplifiers (maximal sensitivity, 0.2 mV). Detection criteria for the Guardian 4210 require that 8 of 10, 12 of 15, or 16 of 20 sensed ventricular events equal or exceed the programmable tachycardia detection rate (100 to 240 beats per minute [bpm]). For the Guardian 4210, tachycardias faster than 240 bpm require 8 of 10 consecutive sensed intervals at these rates to initiate a detection. The CPI PRx records a tachycardia detection when a rate-counting “bin” is filled to a preprogrammed value (8 to 24 beats). Each sensed interval faster than the programmed rate cutoff increases the bin counter by a value of one; subsequent intervals below the rate cutoff decrease the counter by one toward a minimum value of zero.
The time and date of recorded events are provided by reference to time calibrations from the respective programmers. The programmers were maintained at correct local time. The PRx ICD records time to the hour, minute, and second; the Guardian 4210 records time to the hour and minute. The internal clock accuracy for both devices is ±0.03% of the time elapsed since recorded events. The Telectronics Guardian 4210 ATP stores up to 500 tachycardia episodes in clearable memory. The CPI stores up to 228 episodes.
All patients undergoing implantation of the CPI Ventak PRx or Telectronics Guardian 4210 ATP ICDs at the Medical College of Virginia, McGuire Veterans Affairs Medical Center, or Sentara Norfolk General Hospital between April 10, 1991, and April 21, 1993, were eligible for inclusion in the study. Patients receiving defibrillators had failed previous drug testing or had noninducible arrhythmias after documented sustained ventricular arrhythmias or sudden death. Each patient gave informed written consent to protocols approved by each hospital’s Committee for the Conduct of Human Research. Patients were excluded from the study if the time or date of tachycardia detections was not available for any episode from the data log or if detections known or suspected to result from supraventricular tachyarrhythmias, lead or connector failure, environmental or electromagnetic noise, or generator malfunction occurred. The classification of detections as resulting from ventricular tachyarrhythmias, supraventricular tachycardias, or noise in the absence of ECG documentation followed previously published criteria.13 14 15 16 Detections were classified as ventricular tachyarrhythmias if accompanied by syncope, presyncope, typical prodromal symptoms of ventricular tachyarrhythmias for a given patient, or distinct changes in morphology of stored electrograms. Ventricular fibrillation was diagnosed if the recorded R-R interval was ≤250 milliseconds (>240 bpm) and accompanied by symptoms of hemodynamic compromise or changes in the morphology of stored electrograms. Sinus tachycardia was diagnosed if detections occurred without symptoms during physical exertion likely to cause sinus tachycardia. Detections were classified as atrial fibrillation if irregular predetection R-R intervals (>60-millisecond differences) were recorded without symptoms of ventricular tachyarrhythmias or without change in morphology of available stored electrograms. Other supraventricular tachycardias were diagnosed on the basis of previously documented supraventricular tachycardias, regular R-R interval at the same rate as supraventricular tachycardia intervals, and absence of symptoms of ventricular tachycardia. The difficulty of diagnosing supraventricular arrhythmias in the absence of atrial recordings even with ventricular electrograms was recognized previously.14 Noise detections were classified by irregular, closely coupled R-R intervals (<70 milliseconds) or intervals reproduced by manipulation of the device or lead or associated with other evidence of lead or generator malfunction. For statistical reasons related to power, patients with two or fewer detections during follow-up were also excluded from analysis.
ICD Implantation and Follow-up
All devices were implanted by use of standard surgical techniques and were activated on discharge from the operating suite or within 48 hours. Each patient had appropriate sensing confirmed, and each device was interrogated for recorded tachycardia events before hospital discharge and 1, 2, 4, and 6 months after implantation. Follow-up after 6 months from implantation occurred at intervals of 2 to 6 months. Arrhythmia induction for device testing was performed before discharge and 2 to 4 months after discharge. Tachycardia detections resulting from induced arrhythmias were not included in the data analysis.
For each patient, the time intervals between each consecutive pair of tachycardia detections were plotted against the cumulative relative frequency of interdetection intervals during follow-up (Figs 1 through 3⇓⇓⇓). All spontaneous tachycardia detections, whether self-terminating or requiring shock or antitachycardia pacing therapies from the device, were analyzed. Tachycardia episodes requiring multiple therapy deliveries to terminate were considered a single detection. For each patient, a model distribution was generated for the expected random distribution of the same number of interevent intervals as occurred spontaneously during the patient’s follow-up time period. This model assumed that each tachycardia detection had an equal likelihood of occurring at any time during follow-up. Statistically, this model of independent and random detection occurrence would be described by an exponential distribution of the interdetection intervals. For each patient, the cumulative relative frequency for this random model of tachycardia detections over the follow-up duration is given by
where F(t) is the cumulative relative frequency of detections with interdetection intervals ≤t and β is the estimated mean time between episodes for the patient (patient’s total number of episodes divided by duration follow-up). The model of random distribution for each patient was also plotted against the predicted cumulative relative frequency of interdetection intervals during follow-up. The Kolmogorov-Smirnov goodness-of-fit test was used to compare each patient’s observed (sample) distribution with his or her own model of random distribution. The null hypothesis assumed that the interdetection intervals were independent and randomly distributed throughout follow-up for each patient. The null hypothesis was rejected if the patient’s sample distribution differed significantly from his or her model of random distribution. The sign test was used to evaluate the significance of the proportion of patient groups demonstrating nonrandom distributions of tachycardia events. Comparisons between continuous and categorical data were made with Student’s t test and χ2 tests, respectively.
The level of statistical significance was set at P<.01. All values are expressed as mean±SD when appropriate.
Of 114 patients receiving the ICDs used in this study, 31 met inclusion criteria. A total of 83 patients were excluded because of incomplete data logs (n=13), detections caused by supraventricular tachycardias or noise (n=22), or two or fewer detections during follow-up (n=34 patients with no detections; n=12 patients with one or two detections). One patient was excluded because the data storage capacity was completely filled (Guardian 4210), and 1 patient was excluded for multiple ventricular tachycardia detections associated with acute myocardial infarction.
The study group comprised 26 men and 5 women. The mean age was 62.6±14.4 years, and average left ventricular ejection fraction was 34±9%. Twenty-five patients had coronary artery disease, and 6 patients had idiopathic cardiomyopathy. The indication for ICD implantation was ventricular tachycardia in 24 patients, ventricular fibrillation in 6 patients, and sudden cardiac death (primary rhythm uncertain) in 1 patient. Fifteen patients received antiarrhythmic drugs during follow-up. Of these, 12 received antiarrhythmic drugs continuously throughout follow-up, and 4 had tachycardia episodes recorded off medications initially and also after starting amiodarone (n=3) or procainamide (n=1). Three patients on continuous drug therapy changed antiarrhythmic agents (without drug-free intervals) owing to drug intolerance or frequent detections. Overall (including drug changes), 3 patients received procainamide, 2 received mexiletine, 2 received propafenone, 1 received disopyramide, and 7 received amiodarone during follow-up. In addition, 1 patient received combination therapy with procainamide and mexiletine and 1 patient received amiodarone and procainamide. The mean duration of follow-up was 227±183 days (range, 7 to 782 days; median, 177 days). Ten patients had the Telectronics Guardian 4210; 21 patients had the CPI Ventak PRx. Detection criteria at implantation for patients with the Telectronics device were 8 of 10 consecutive beats and 12 of 15 consecutive beats above the rate cutoff in 7 and 3 patients, respectively. All patients with the CPI device had detection criteria set between 10 and 16 beats above the rate cutoff.
Patterns of Tachycardia Detections
A total of 727 detections were recorded from 31 study patients during follow-up (range, 3 to 101 detections per patient; mean, 23±35 detections per patient; median, 12 detections per patient). Fig 2⇑ shows an example of the distribution of tachycardia detections during follow-up for a single patient. For 28 of the 31 patients, the distribution of interdetection intervals during follow-up differed significantly from the individuals’ respective models of random distribution (all P<.01) (Fig 3⇑). The proportion of patients (28 of 31) with nonrandom distributions was highly significant for the group as a whole (P<.001). All 28 patients demonstrated a preponderance of interdetection intervals that were shorter than predicted by their models of random distribution. For the entire study group of 31 patients, 55% of all interdetection intervals were ≤1 hour, 69% were ≤24 hours, and 78% were ≤91 hours (Fig 4⇓).
Table 1⇓ shows the clinical data, follow-up data, shortest recorded interdetection interval, and longest duration during follow-up without detection for all 31 study patients. For 23 of 31 patients (74%), the shortest interdetection interval recorded was ≤14.4 minutes. Table 1⇓ also shows the 50th percentile for interdetection intervals for each patient. The median value for the 50th percentile interdetection intervals among the 31 patients was 15.8 hours. The longest documented intervals between detections for these patients ranged up to 493.3 days (patient 20). The 3 patients (patients 4, 15, and 31) with random detection distributions are identified in Table 1⇓. Of these 3 patients, 2 received antiarrhythmic drugs (patient 4 received procainamide; patient 31, amiodarone).
Data Analysis by Patient Characteristics
Subgroup analysis of the 31 study patients revealed that the proportion of patients with nonrandom detection distributions was significant for patients with coronary artery disease (22 of 25 patients), patients receiving antiarrhythmic drugs (13 of 15 patients), patients receiving no antiarrhythmic drugs (15 of 16 patients), patients with ventricular tachycardia (21 of 24 patients), and patients with ventricular fibrillation or sudden death (7 of 7) as implantation indications (all P<.01 for the groups). The proportion of patients with nonischemic cardiomyopathy and nonrandom distributions (6 of 6 patients) reached the highest achievable significance for a group of only 6 patients at P=.016.
Data Analysis of Detections Receiving Therapy
The analysis for nonrandom detection distributions was performed for all 31 study patients by use of only those detections resulting in delivery of shock or antitachycardia pacing therapy. Two patients were excluded from this subgroup analysis because they had <3 treated detections. Of 31 patients, 29 had ≥3 treated tachycardia detections and were included in this subgroup. The total number of treated episodes was 489 (191 antitachycardia pacing and 298 shocks) from these 29 patients. Of these 29 patients, 25 had distributions of treated detections that differed significantly from the models of random distribution for the treated detections (all P<.01). The proportion of patients with nonrandom distributions (25 of 29) was highly significant for the group (P=.001).
Data Analysis by Tachycardia Rate
The analysis was performed for tachycardia episodes with cycle lengths ≤250 milliseconds (>240 bpm) and separately for episodes with cycle lengths of >250 milliseconds. All 727 detections from the 31 total study patients were included in this subgroup analysis by tachycardia rate. Eleven of 13 patients with ≥3 detections with a cycle length ≤250 milliseconds (108 total detections) had nonrandom distributions of these events (all P<.01; P=.0001 for the group). Of 29 patients with ≥3 tachycardia detections with cycle lengths >250 milliseconds (619 detections), 25 had nonrandom distributions of these events (all P<.01; P<.0001 for the group).
Data Analysis of Detections >7 Days After Implant
Previous studies demonstrated an enhanced likelihood of tachycardia episodes in the period immediately after ICD implantation, possibly caused by mechanical irritation of the heart or the physiological stresses of surgery.17 18 To control for the possibility that multiple postimplantation tachycardia detections would skew the distributions, the analysis was repeated for each of the 31 study patients with all detections except those occurring in the first 7 days after ICD implantation. Four patients were excluded from this subgroup analysis because of <3 tachycardia episodes occurring more than 7 days after implantation. A total of 670 episodes occurred >7 days after ICD implantation in the remaining 27 patients. Of 27 patients, 24 still demonstrated nonrandom distributions of these tachycardia detections (all P<.01). The proportion of patients with nonrandom distributions (24 of 27) was highly significant for the group (P=.001).
Patients With Fewer Than Two Detections
Table 2⇓ gives clinical data, follow-up data, and time to first shock for patients with one (n=8) or two (n=4) detections. The median interdetection interval for the 4 patients with two detections was 19.9 hours. The median time to first shock for all 12 patients was 60 days. The longest interval without a detection in these 12 patients was 697.5 days (patient 10). For these 12 patients with two or fewer detections, the average follow-up was 257±203 days compared with 227±183 days for the 31 patients with three or more detections (P=.64). There were no significant differences in the proportions of patients with three or more or with two or fewer detections with respect to the occurrence of coronary artery disease (25 of 31 versus 9 of 12, respectively), antiarrhythmic drug use during follow-up (15 of 31 versus 6 of 12), or indication for defibrillator implantation (ventricular tachycardia in 24 of 31 versus 9 of 12; all P≥.8).
The description of the long-term temporal patterns of ventricular tachycardia recurrence was not technologically feasible before ICDs with data storage capabilities became available. Previous studies relied on serial Holter monitoring, rarely exceeding 96 hours, or the patients’ own subjective recognition of symptoms to record arrhythmic episodes.2 3 4 5 6 7 8 19 The devices used in this study accurately recorded the time and date of each tachycardia episode that met detection criteria during continuous monitoring up to 782 days. Previous work showed that such long-term follow-up is necessary to assess more accurately the frequency and variance of spontaneous arrhythmia recurrences in individual patients.3 5 20 21
This study demonstrates a nonrandom distribution of ventricular tachycardia episodes over time with a distinct tendency toward temporal clustering of these episodes in 28 of 31 study patients. The tachycardia detections were infrequently isolated events with 55% and 69% of all interdetection intervals ≤1 hour and ≤24 hours, respectively. The median value for the 50th percentile interdetection interval for all 31 patients was 15.8 hours. This nonrandom, clustered pattern persisted for a highly significant majority of patients after the exclusion of potentially surgically related arrhythmias and in separate analyses of episodes by tachycardia rate; by those treated with the device; and by patient characteristics of heart disease, drug therapy, and indications for implantation.
Previous studies of the temporal patterns of ventricular arrhythmias focused on the variability of ectopic activity frequency over time; however, analyses of the time intervals between individual ventricular tachyarrhythmic events are lacking.2 3 5 6 7 8 Marked variation in the frequency of ventricular ectopic activity over observational periods of minutes, hours, and days has been well documented. Winkle22 described up to 100% variation in the frequency of ventricular ectopy during short-term monitoring, while Pratt et al7 found that 90% of 40 patients with ventricular tachycardia on any of four serial 24-hour Holter monitors also had a complete 24-hour recording without ventricular tachycardia. Many studies stressed the importance of long-term follow-up by demonstrating that the assessed variability in ectopy frequency is dependent on the duration of monitoring.3 5 20 21 Anastasiou-Nana et al3 found that 98% suppression of repetitive ventricular ectopic forms is necessary to exceed the 95% CIs for spontaneous variability in individual patients when Holter monitors are performed ≥1 year apart. In this same study, a >4000% increase in total ventricular ectopy was needed to establish proarrhythmia. The temporal clustering of arrhythmic events in our study may explain the extreme variance in repetitive ventricular ectopy and the greater appreciation for the ranges in spontaneous variability with longer follow-up that is more likely to include a period of clustered activity. Stein et al23 also demonstrated nonuniform, clustered distributions of single premature ventricular beats during short-term monitoring.
The mechanism responsible for the clustering of the tachycardia detections is uncertain. The grouping possibly results from relatively persistent states of myocardial ischemia, autonomic imbalance or activation, or electrolyte abnormalities that favor tachycardia induction.24 25 These conditions alone or in combination may linger for hours or days, allowing frequent tachycardia arrhythmias. Our group also described circadian patterns of tachyarrhythmia detections by ICDs.10 This study of 43 patients demonstrated a distribution of detections fitting a sine wave model with 24-hour periodicity and peak frequency at approximately 3 pm. The contribution of circadian distributions to the clustering of arrhythmic events requires further evaluation. Another explanation for clustering events may be that alterations in myocardial repolarization patterns, oxygen demand, or autonomic tone that follow an initial episode of ventricular tachycardia may facilitate closely timed subsequent arrhythmias.26 27 This may also explain the enhanced likelihood of tachycardia episodes in the days after ICD implantation, which necessitates multiple arrhythmic indications. Waning drug effects is an unlikely explanation for event clustering, given the occurrence of the same temporal pattern in patients receiving no antiarrhythmic drugs.
This report also includes data on 12 patients with rare (two or fewer) detections despite a similar average duration of follow-up to those with more frequent (three or more) events. It is unclear why some patients have very infrequent events despite similar use of antiarrhythmic drugs, indications for implantation, and frequency of coronary artery disease compared with those with frequent events. Perhaps this finding identifies a patient subgroup with a different pattern or trigger for ventricular arrhythmia.
The ability to anticipate the occurrence of arrhythmic events even over short periods of time may have important implications for intermittent antiarrhythmic therapy. After an initial tachycardia episode and identification of a “high-risk” period for tachycardia recurrence, additional drug therapy could be taken orally by the patient or conceivably administered by an automatic pump through a drug reservoir in the defibrillator. The device could possibly institute arrhythmia-suppressing pacing algorithms temporarily to prevent subsequent episodes.28 29 Further studies into the pathophysiological factors that contribute to the clustering of arrhythmic events may greatly advance the understanding of spontaneous ventricular tachyarrhythmias. In a previous report,10 the circadian pattern of ventricular tachycardia episodes detected by defibrillators was apparent in patients not taking antiarrhythmic drugs but was not observed in patients on long-term antiarrhythmic drug therapy. In contrast, the clustering of tachycardia episodes over time was noted in patients both on and off drug therapy in this study. This suggests that different factors influence these temporal patterns.
The demonstration of long interdetection intervals (up to 493 days) suggests that very extensive follow-up may be needed to determine the long-term success of pharmacological and nonpharmacological therapy (such as radiofrequency ablation) for ventricular arrhythmias. Short-term suppression of arrhythmias after a cluster of events may not be reassuring that the arrhythmic substrate has been suppressed.
Finally, knowledge of systematic variations in the distribution of ventricular arrhythmias raises a question about the appropriateness of treatment strategies based on the pharmacological suppression of spontaneous ventricular arrhythmic events to statistically derived end points.11 12 This strategy assumes relatively constant mean and variance to the frequency of arrhythmic events such that periods of observation with event frequency above or below statistically derived limits are presumed unlikely to occur by chance and therefore can be attributed to a pharmacological intervention. The demonstrated clustering of arrhythmic events into periods of hours or days greatly confounds the ability to define spontaneous variability and thus calculate reliable treatment end points. Others2 3 4 5 raised similar concerns by suggesting that the frequency of ventricular ectopy may follow systematic fluctuations over time. Our data provide support for this hypothesis.
Although the use of stored electrograms, data logging, and clinical history allows discrimination of ventricular tachyarrhythmic events from noise or supraventricular tachycardias with a high degree of accuracy, it is possible that some arrhythmic events included in this analysis resulted from other causes than ventricular arrhythmias.14 16 All current defibrillators can store electrograms for only a limited number of detections. Thus, it is not always possible to obtain electrograms of every event in patients with very frequent episodes. The devices recognize only the subset of ventricular arrhythmic events of sufficient rate and duration to meet programmed detection criteria. Although considered “sustained” events, the true duration of treated episodes if shock or pacing therapy had not been delivered cannot be known. Factors such as silent myocardial infarction or electrolyte abnormalities may have contributed to the clustering of detections in some patients. The study group was relatively small (31 patients); thus, sample biases may have occurred, and the statistical power of patient subgroup analyses may be limited. Antiarrhythmic drug therapy was not randomized, so selection bias is possible. Previous studies examined primarily the frequencies of isolated or brief ventricular ectopic activities. Conclusions drawn by comparison to these studies must be considered speculative.
Reprint requests to Mark A. Wood, MD, Medical College of Virginia, PO Box 980053, Richmond, VA 23298.
- Received September 12, 1994.
- Revision received November 14, 1994.
- Accepted November 26, 1994.
- Copyright © 1995 by American Heart Association
Anderson JL, Anastasiou-Nana MI, Menlove RL, Moreno FL, Nanas JN, Barker AH. Spontaneous variability in ventricular ectopic activity during chronic antiarrhythmic therapy. Circulation. 1990;82:830-840.
Anastasiou-Nana MI, Menlove RL, Nanas JN, Anderson JL. Changes in spontaneous variability of ventricular ectopic activity as a function of time in patients with chronic arrhythmias. Circulation. 1988;78:286-295.
Morganroth J, Michelson EL, Horowitz LN, Josephson ME, Pearlman AS, Dunkman WB. Limitations of routine long-term electrocardiographic monitoring to assess ventricular ectopic frequency. Circulation. 1978;58:408-414.
Schmidt G, Ulm K, Barthel P, Goedel-Meinen L, Jahns G, Baedeker W. Spontaneous variability of simple and complex ventricular premature contractions during long term intervals in patients with severe organic heart disease. Circulation. 1988;78:296-301.
Michelson EL, Morganroth J. Spontaneous variability of complex ventricular arrhythmias detected by long-term electrocardiographic recording. Circulation. 1980;61:690-695.
Kim SG, Felder SD, Figura I, Johnston DR, Waspe LE, Fisher JD. Value of Holter monitoring in predicting long-term efficacy and inefficacy of amiodarone used alone and in combination with class 1A antiarrhythmic agents in patients with ventricular tachycardia. J Am Coll Cardiol. 1987;9:169-174.
Hook BG, Callans DJ, Kleinman RB, Flores BT, Marchlinkski FE. Implantable cardioverter defibrillator therapy in the absence of significant symptoms. J Am Coll Cardiol. 1993;21:895-900.
Böcher D, Block M, Isbruch F, Wietholt D, Hammel D, Scheld HH, Borggrefe M, Breithardt G. Comparison of frequency of aggravation of ventricular tachyarrhythmias after implantation of automatic defibrillators using epicardial versus non-thoracotomy lead systems. Am J Cardiol. 1993;72:1064-1068.
Winkle RA. Antiarrhythmic drug effect mimicked by spontaneous variability of ventricular ectopy. Circulation. 1978;57:1116-1120.
Stein KM, Borer JS, Hochreiter C, Kligfield P. Fractal clustering of ventricular ectopy and sudden death in mitral regurgitation. J Electrocardiol. 1992;25(suppl):178-181.
Franz MR, Bargheer K, Raffenbeul W, Haverich A, Lichtten PR. Monophasic action potential mapping in human subjects with normal electrocardiograms: direct evidence for the genesis of the T wave. Circulation. 1987;75:379-386.
Johnson RA, Hutter AM, DeSanctis RW. Chronic overdrive pacing in the control of refractory ventricular arrhythmias. Ann Intern Med. 1974;80:380-383.
Marchlinski FE, Buxton AE, Miller JM, Josephson ME. Prevention of ventricular tachycardia induction during right ventricular programmed stimulation by high current strength pacing at the site of origin. Circulation. 1987;76:332-343.