RR Interval Variability in Irregular Monomorphic Ventricular Tachycardia and Atrial Fibrillation
Background Algorithms to reject irregular tachyarrhythmias are available in implantable cardioverter-defibrillator devices to discriminate ventricular tachycardia (VT) from atrial fibrillation (AF). The hazard of underdetection of irregular monomorphic VTs using these algorithms has not yet been fully evaluated. The purpose of this study was to determine the ability of a commonly used stability algorithm to reject AF and to correctly detect VT with a high RR interval variability.
Methods and Results The electrophysiological studies from 232 patients with induced monomorphic VT (cycle length >250 ms) and 21 with AF were reviewed. A preliminary analysis was performed to classify the VT episodes in irregular (successive RR differences >20 ms after 4 seconds from onset) or regular (otherwise). Three study groups were defined: group 1 (27 patients with irregular VT), group 2 (22 randomly selected patients with regular VT), and group 3 (21 patients with AF). A computer program analyzed the first 50 RR intervals of the induced VT (AF), resetting a VT counter if the interval was greater than a tachycardia detection interval (TDI) or if its absolute difference with the preceding three beats exceeded a programmed stability value (STAB). The VT was detected when the VT counter reached a preset number of intervals (NIDs). Different combinations of TDI, STAB, and NID were analyzed. All VTs in group 2 were correctly detected. In contrast, up to 10 VTs from group 1 were not detected when high NIDs and low STAB parameters were programmed. With usual values (10 to 16 beats and 50 to 60 ms, respectively), only 1 to 2 VTs remained undetected, but 20% to 50% had a detection delay >8 seconds. Undetected VTs were significantly slower than early detected VTs for most STAB and NID combinations. With usual stability and NID values, 10% to 20% of episodes of AF were inappropriately detected. Changes in TDI had a small impact on sensitivity and specificity when currently used values for stability were programmed.
Conclusions An implantable cardioverter-defibrillator tachycardia detection algorithm with a stability criterion of 50 to 60 ms and 12 to 14 RR intervals is able to detect over 90% of monomorphic irregular VTs. Nevertheless, significant VT detection delays may arise, and inappropriate detection of AF cannot be totally prevented.
Third-generation ICDs are able to detect VT and VF and to deliver tiered therapy for these arrhythmias. The recognition criteria usually rely on the duration of the RR intervals, defined as the time between two successive sensed ventricular activations. This detection system does not discriminate ventricular from supraventricular arrhythmias, causing inappropriate ICD therapies due to detection of rapid atrial rhythm disturbances.1 Paroxysmal AF is the leading cause of inappropriate shocks in ICD patients.2 Up to 38% of all delivered shocks might be due to AF.3 Inappropriate shocks cause battery depletion, are painful for the patient, increase the patient’s state of anxiety,4 and may induce dangerous or even fatal ventricular arrhythmias.2 5 6 To minimize inappropriate detections of AF, algorithms have been designed that reject tachycardias with a high degree of RR variability.7 Some third-generation ICDs have implemented an optional stability criterion for slow tachycardias8 9 while others have not.10 11 These algorithms assume that monomorphic VTs are regular, whereas AF consistently shows a higher degree of irregularity. However, significant variations of RR intervals are common during the first beats of monomorphic VTs.12 RR interval differences greater than 20 ms have been observed in 21% of VTs,13 and standard deviations of the first 20 beats as high as 70 ms have been reported.14 The above observations have prompted some concern about failure or significant delay in the detection of VTs by ICDs when stability criteria are applied.15 The aim of this study was to determine the ability of a commonly used stability algorithm to reject AF and to detect VTs with a high RR interval variability.
The electrophysiological studies performed in our institution between 1989 and 1994 were reviewed, and patients with at least one episode of induced monomorphic VT or AF were selected.
Details of the recording technique and stimulation protocols have been described previously.16 17 Briefly, 4F to 6F bipolar or quadripolar catheters were introduced via the femoral vein and positioned in the high right atrium, His bundle, and right ventricular apex. Surface leads I, II, V2, and V6 and filtered (40 to 400 Hz) intracardial electrograms were recorded at 50 to 200 mm/s speed. Programmed stimulation was performed at the right atrium with one premature stimulus and at the right ventricular apex delivering up to 2 premature stimuli during both sinus and paced ventricular rhythms at CLs of 500, 430, 370, and 330 ms. If necessary, a third premature stimulus was delivered during pacing at 500 ms, and the protocol was repeated at the right ventricular outflow tract.
Two hundred thirty-two patients with induced sustained VT and 21 patients with induced sustained AF and normal atrioventricular conduction were identified and considered for inclusion into the study. Only episodes of monomorphic VT with a duration of at least 50 beats and an average CL >250 ms were analyzed. RR intervals were measured from intracardiac electrograms using manual calipers with an accuracy of 10 ms. A preliminary analysis was performed, measuring 10 RR intervals 4 seconds after the onset of the VT. If any difference between two successive beats was greater than 20 ms, the VT was classified as irregular. Otherwise, the VT was considered regular.
Irregular VTs were found in 27 patients, and all of them were included in the study (group 1). From the regular VTs (n=205), a subset of 22 was randomly selected and constituted study group 2. Twenty-one episodes of induced AF constituted study group 3.
For each selected patient, the first 50 RR intervals of the induced VT (AF) were measured and included in a computer data file. The first VT interval was defined as the interval from the last paced beat to the first tachycardia beat.
Analysis of RR Intervals
The mean, standard deviation, and maximum and minimum RR intervals of each episode were determined.
A computer program was developed to simulate a commonly used stability algorithm.18 Each RR interval was classified as a VT beat if it was less than the TDI and as a sinus beat otherwise. Each VT beat was added in a VT counter that was reset to zero by any sinus beat. VT was detected when the count equaled the programmed NID. To reject irregular intervals, each VT interval was compared with the preceding three beats. If the maximum absolute difference exceeded a programmed STAB, the VT counter was also reset to zero. To avoid comparisons between sinus and VT beats, the stability measurement was only performed if the VT interval count was higher than 3.
Some combinations of the above programmable parameters were sequentially tested. A first simulation was performed with the TDI value set to 600 ms for all episodes (TDI-600). Next, TDIs corresponding to 30%, 20%, and 10% below the mean tachycardia rate (TDI-30%, TDI-20%, and TDI-10%, respectively), with a maximum value of 600 ms, were programmed for VT episodes; from the above calculated TDIs, a randomly selected value was assigned to each AF episode. The NID was preset from 6 to 16 in steps of 2 beats. The STAB was programmed to “off” and then changed from 30 to 80 ms in steps of 10 ms. For each combination of parameters, the program determined whether or not the tachycardia was detected as well as the elapsed time since the tachycardia onset to the detection. To simplify the analysis and to avoid possible competition effects between VF and VT zones,19 a VF detection window was not programmed, so that any RR interval under TDI was classified as a VT beat.
Statistical significance was evaluated with the use of the Fisher’s exact test and the unpaired Student’s t test. Continuous data are presented as mean±SD. A value of P<.05 was considered significant. The Bonferroni correction was applied when multiple groups were compared.
The clinical characteristics of the study groups are shown in Table 1⇓. Cardiac diagnosis of patients with VT (groups 1 and 2) was coronary artery disease in 39, dilated cardiomyopathy in 5, hypertrophic cardiomyopathy in 2, right ventricular dysplasia in 1, valvular heart disease in 1, and congenital heart disease in 1. Twenty-two patients were taking antiarrhythmic drugs, in most cases class III (84%). Patients in whom irregular VTs were induced were on antiarrhythmic agents at the time of electrophysiological study more often than patients in whom regular VTs were induced (16 of 27 versus 3 of 22, P<.01). There were no significant differences between them regarding other clinical variables.
RR Intervals During Tachycardia
The mean, standard deviation, and maximum and minimum of the 50 RR intervals for each study group are presented in Table 2⇓. Groups 1 and 2 did not show significant differences. In contrast, the mean, standard deviation, and maximum RR interval were higher in AF than in VT episodes (P<.01 in all cases).
VT Detection and AF Rejection
All regular VTs were correctly detected with any combination of TDI, NID, and STAB parameters. The accuracy of detection of irregular VTs with a fixed TDI-30% and different NID and STAB parameters is presented in Table 3⇓. As expected, VT detection increased as the STAB value was increased and the NID was reduced. With TDI-30% and stability of 50 ms, two VTs (7%) remained undetected. One of them was a very irregular tachycardia, with differences in the CL of successive beats above 100 ms during the entire detection period. This patient did not take any antiarrhythmic drug. The other one was a very slow VT (mean CL, 559 ms) in a patient taking amiodarone. As shown in Fig 1⇓, sinus beats are conducted to the ventricles in a regular pattern, producing fusion beats and leading to irregularities of the CL.
Changes of the TDI parameter did not alter the detection rate as long as a stability criterion was operative. Only when STAB was programmed to “off,” one additional VT was not detected with TDI-10%.
The ability to reject AF was highly dependent on the TDI value when the STAB parameter was set to “off” (Fig 2⇓). With TDI-600, 18 to 21 (86% to 100%) episodes were inappropriately detected, but only 4 to 7 (19% to 33%) were detected when a TDI-10% was selected. The number of inappropriate detections was reduced when a stability criterion was programmed (Table 4⇓). In this case, the changes of TDI had a less pronounced effect on the ability to reject AF. For NID of 12 to 16 and STAB of 30 to 60, the same number of episodes were detected with any TDI value. When lower NIDs or higher STABs were used, one to seven episodes were detected with TDI-30% but not with TDI-10%.
Even with the most strict combination of TDI, NID, and STABs, two AFs (10%) remained inappropriately detected (Fig 3⇓).
Delay in VT Detection
All regular VTs were detected in less than 8 seconds with NID of 8 beats. One to four VTs had a detection delay >8 seconds when the number of intervals for detection was increased to 10 to 12. With NID of 14 beats, 3 (STAB, 80) to 5 (STAB, 30) VTs were detected after 8 seconds and the numbers increased to 6 and 8, respectively, with NID of 16 beats. The TDI value did not affect the detection times in this group.
A greater proportion of detection delays was noted for irregular VTs. Without stability criterion, the time to detect the tachycardia increased with higher NID and lower TDI values (Table 5⇓). When the stability criterion was programmed, the delay was longer as the STAB parameter decreased and as the number of beats required to perform the detection increased (Fig 4⇓). As an example, for a fixed value of TDI-30%, programming STAB of 50 ms and NID of 8 beats, 22 of 27 VTs (81.5%) would be detected in less than 8 seconds and 25 (92.6%) in less than 10 seconds, but the figures would be 11 (40.7%) and 14 (51.8%), respectively, if 16 beats were required to detect VT. For any STAB and NID value, only small differences in the detection delay were observed when TDI was changed to TDI-20% or TDI-10% if the stability criterion was set.
The mean CL was greater for late (>8 seconds) and undetected VTs than for VTs with early detection, and the difference was significant for most combinations of TDI, STAB, and NID. For example, with TDI-30%, stability of 50, and 12 beats to detect, the mean CLs were 419.1±105.1 and 291.9±50.1, respectively (P<.01).
Our results show that, using a stability of 50 to 60 ms combined with NID of 12 to 14 beats, the simulated algorithm is able to detect over 90% of monomorphic VTs with relatively irregular RR intervals. Since all regular VTs are detected, and approximately 90% of all monomorphic VTs have been regular, the probability of nondetection with this algorithm for a randomly selected VT is very low. Nevertheless, significant VT detection delays may arise, and inappropriate detection of AF cannot be totally prevented.
Despite some recently reported good clinical results,8 20 21 the stability function has not been tested extensively, and some concerns remain about the risk of underdetection when these algorithms are applied. Some instances of VT underdetection due to the stability criterion in patients with ICD have been reported,19 but the true incidence of the problem is unknown. Since some episodes may be asymptomatic and RR intervals are not always saved by the ICD, the recognition of such instances may be difficult. Moreover, the storage capacity of ICD devices is limited and the onset of the episode is not always available, preventing a detailed assessment of possible detection delays.
Detectable irregularities in the CL, especially during the first beats, have been reported,12 13 14 22 23 but most studies analyze the variability in periods of 5 or more beats14 22 23 and therefore are not adequate to predict the performance of algorithms based on short-term (4 beats) RR interval differences. Few data are available about beat-to-beat stability in monomorphic VTs. Absolute differences between successive intervals as high as 50 ms have been reported.13 Irregularities are greater at VT onset, and differences above 40 ms are common in the first 2 to 3 beats.12 Our results suggest that about 10% of all induced VTs maintain successive RR differences >20 ms after 4 seconds from onset, but most of them exhibit absolute differences <50 ms between 4 successive beats. Therefore, the hazard of underdetection by stability algorithms based on such short-term comparisons should be small, provided that spontaneous and induced VT episodes have similar variabilities.
Underdetection due to irregular RR intervals can be produced by two mechanisms: long RR intervals that exceed the VT detection interval and oscillations of the CL longer than the programmed STAB. Changes of TDI have important effects on the ability to detect irregular VTs when the stability criterion is not programmed, but only small differences with different TDIs are observed if usual STABs are set. This result suggests that when the stability function is operative, most underdetection instances are probably related to the second mechanism.
Even though most irregular VTs can be detected during the period of analysis (50 beats), the standard deviation of the first 10 to 20 RR intervals is shown to be higher in VTs that remain irregular,14 and this may result in detection delay. As suggested by our results, this delay is expected to be greater for slow VTs, because the VT variability increases with the mean CL.14 The clinical relevance of underdetection or detection delay during slow VT episodes by ICD devices is not completely understood.
Inappropriate detection of AF as VT remains a common problem of ICD devices. Stability algorithms that reject irregular arrhythmias should improve the ability to discriminate AF from VT, avoiding inappropriate therapies. A marked reduction in the detection of spontaneous AF episodes has been reported with programmed stability of 40 to 60 ms and NID of 12 to 16 beats.20 21 Nevertheless, the number of intervals required to detect VT had to be increased up to 20 to 24 in some patients in order to avoid inadequate AF detection.9 According to our results, this procedure might be hazardous in patients with high-variability VTs. Moreover, the use of lower TDI values in these patients would not reduce the number of detected AF episodes.
The tested algorithm exhibits a good rejection power (85% to 90%) for STABs of 50 ms and NID ≥12 beats. At a comparable mean heart rate (150 beats per minute), Swerdlow et al21 found 95% to 100% specificity with similar programmed values in patients with a Medtronic PCD defibrillator. The difference is probably nonsignificant, given the small number of patients with AF in both series. Since the beat-to-beat variability during AF decreases with the mean RR interval,24 more inappropriate detections are expected at high ventricular rates. A retrospective analysis of ICD-stored RR intervals showed that a programmed stability of 50 ms with NID of 12 beats would have prevented only 3 of 12 spontaneous documented episodes of AF that triggered VT therapy.15 The poor specificity obtained might be explained partially by a higher mean heart rate in this series (approximately 170 beats per minute) and stresses the need to slow down the ventricular response to improve the rejection power of stability algorithms.
Effect of Antiarrhythmic Drugs
The mechanisms of CL variability during VT are not clearly understood. Since most monomorphic VTs probably are due to reentry,25 rate-related changes in conduction velocity,12 in the configuration of the reentry pathway, or in myocardial refractoriness26 may be contributing factors to variations in CL. Antiarrhythmic agents are known to modify the conduction properties and the refractoriness of the reentry pathway and thereby could influence the VT stability. Preliminary data suggest that antiarrhythmic drugs may increase the variations of RR intervals during monomorphic VT.19 Our results support this hypothesis: 59% of patients with irregular VTs were taking antiarrhythmic drugs, versus only 13% with regular VTs. This may be a specific effect or may simply reflect a higher incidence of slow VTs in patients treated with antiarrhythmic drugs, since these agents usually decrease the mean tachycardia rate27 28 and slower VTs seem to be more irregular.14 Interactions between stability and antiarrhythmic drugs deserve further investigation but probably should be considered when programming the stability function in patients with ICDs.
In slow VTs with ventriculoatrial dissociation, sinus beats may capture the ventricles, resetting the tachycardia and leading to changes in RR intervals, as shown in Fig 1⇑. This phenomenon is also a potential cause for underdetection and may be more frequent in patients with antiarrhythmic treatment as a result of the lower mean VT rates.
The most important limitation of this study is the unknown comparability of the magnitude of beat-to-beat variations in induced and spontaneous episodes of VT and AF. Similar morphology and mean CL in induced and spontaneous VTs have been described,29 30 but the question of variability is not yet elucidated. Low reproducibility of the first RR interval differences in consecutively induced VT episodes has been reported,12 but neither a quantitative analysis nor a comparison with spontaneous episodes is available.
Many records analyzed in this study were done at a paper speed of 50 mm/s. The use of higher paper speeds or computer-based systems could have improved the accuracy of the measurements. Nevertheless, the analysis was made on the intracardiac signals, which usually have a very steep stroke, thus allowing correct measurements within 0.5 mm, equivalent to 10 ms. This is in the order of accuracy used by current devices.
Clinically, the stability algorithm usually has to reject tachycardia detection of AF during long periods of AF instead of the 50 RR intervals tested in this study. Hence, the percentage of rejected tachyarrhythmias caused by AF might be overestimated.
Only the stability criterion is simulated by the above tested algorithm. Interactions with other ICD programmable options (such as the “onset” function) as well as competition effects between VT and VF detection windows are not considered in this study but might have clinical relevance.
The TDI has been calculated as a percentage of the mean tachycardia rate, following the technical manual recommendations for “polymorphic” VTs of a commonly used ICD device.31 Other reported approaches21 might lead to different results.
We have simulated only one of many possible stability algorithms. The selection was made considering that a similar principle is used by some available ICDs and thereby the results can be compared with reported clinical data. Whether modifications of the selected approach or use of a different one might improve the efficacy of VT detection remains to be investigated.
Selected Abbreviations and Acronyms
|NID||=||number of intervals to detect VT|
|TDI||=||tachycardia detection interval|
Dr Garcı́a-Alberola was supported in part by a grant from DGICYT-(Ministerio de Educación y Ciencia), Madrid, Spain.
Innere Medizin C, Kardiologie/Angiologie, Albert-Schweitzerstr 33, D-48129 Münster, FRG.
- Received March 15, 1995.
- Revision received July 10, 1995.
- Accepted August 29, 1995.
- Copyright © 1996 by American Heart Association
Josephson ME. Clinical Cardiac Electrophysiology: Techniques and Interpretations. Philadelphia, Pa: Lea & Febiger; 1993:708-715.
Olson WH, Bardy GH, Mehra R, Keimel JG, Almquist C, Biallas RM. Onset and stability for ventricular tachyarrhythmia detection in an implantable pacemaker, cardioverter and defibrillator. Comp Cardiol.. 1986;13:167-170.
Wietholt D, Block M, Isbruch F, Borggrefe M, Shenasa M, Breithardt G. Clinical experience with antitachycardia pacing and improved detection algorithms in a new implantable cardioverter-defibrillator. J Am Coll Cardiol.. 1993;21:88-94.
Bardy GH, Troutman C, Poole JE, Kudenchuk PJ, Dolack GL, Johnson G, Hofer B. Clinical experience with a tiered-therapy, multiprogrammable antiarrhythmia device. Circulation. 1992;85:1689-1698.
Fromer M, Cardinal R, Pagé P, Nadeau R, Shenasa M. Variation in cycle length of induced ventricular tachycardia episodes in humans: incidence and electrophysiologic mechanisms. In: Shenasa M, Borggrefe M, Breithardt G, eds. Cardiac Mapping. Mount Kisco, NY: Futura; 1993:507-514.
Borggrefe M, Breithardt G. Predictive value of electrophysiologic testing in the treatment of drug-refractory ventricular arrhythmias with amiodarone. Eur Heart J.. 1986;7:735-742.
Medtronic Inc. Technical manual: The PCD Tachyarrhythmia Control Device Model 7217B. Minneapolis, Minn: Medtronic Inc; 1991:7-8.
Bardy GH, Hofer B, Johnson G, Kudenchuck PJ, Poole JE, Dolack GL, Gleva M, Mitchell R, Kelso D. Implantable transvenous cardioverter-defibrillators. Circulation. 1993;87:1152-1168.
Braunstein JR, Franke EK. Autocorrelation of ventricular response in atrial fibrillation. Circ Res.. 1961;9:300-304.
Josephson ME, Almendral JM, Buxton AE, Marchlinski FE. Mechanisms of ventricular tachycardia. Circulation. 1987;75:41-49.
Marchlinski FE. Characterization of oscillation in ventricular refractoriness in man after an abrupt increment in heart rate. Circulation. 1987;75:550-556.
Josephson ME. Clinical Cardiac Electrophysiology: Techniques and Interpretations. Philadelphia, Pa: Lea & Febiger; 1993:436-445.
Berger MD, Waxman HL, Buxton AE, Marchlinski FE, Josephson ME. Spontaneous compared with induced onset of sustained ventricular tachycardia. Circulation. 1988;78:885-892.
Cardiac Pacemakers Inc. Ventak P AICD Models 1600 and 1610: Physician’s Manual. St Paul, Minn: Cardiac Pacemakers Inc; 1989:32.