© 2004 by European Society of Cardiology
Arrhythmia detection by dual-chamber implantable cardioverter defibrillators
A review of current algorithms
aNancy University Hospital Nancy, France; bELA Medical, Le Plessis-Robinson France
Manuscript submitted 7 November 2003. Accepted after revision 19 February 2004.
*Corresponding author. Service Cardiologie, Hôpital Brabois, Rue du Morvan, 54500 Vandoeuvre-les-Nancy, France. Tel.: +33-383-153244; fax: +33-383-153856. E-mail address: E-mail address: e.aliot{at}chu-nancy.fr (E. Aliot).
 |
Abstract |
|---|
 Top Abstract IntroductionDetection criteriaImplemented algorithmsBiotronik SMARTELA PARAD+Guidant Atrial View and...St Jude morphology...Medtronic logical tableRecent clinical resultsDiscussionConclusionsAcknowledgementsReferences |
|---|
This article reviews the arrhythmia detection criteria currently available in dual-chamber implantable cardioverter defibrillators (ICDs), and describes the implementation and performance of various detection algorithms. Nearly all criteria implemented in single-chamber devices appear to have been included in dual-chamber ICDs. However, two different strategies can be distinguished: the first adds dual-chamber inhibition criteria to a single-chamber detection configuration; the second is a new approach entirely based on a dual-chamber detection scheme. Despite widely available clinical data, an analysis of the implemented detection indexes, arrhythmia characteristics (induced vs spontaneous, minimum duration), device programming (detection rate, programme maintenance and tuning during follow-up), and different storage capabilities among various ICD models, leaves the results of these studies ultimately ambiguous. New algorithms are under study, but only protocols using a single set of arrhythmias and the same programming for all devices may allow relevant comparisons of the performances of detection algorithms. Furthermore, a criterion is required to distinguish reliably between haemodynamically stable and unstable tachycardias, not simply based on rate, but including the underlying cardiac function.
Key Words: implantable cardioverter defibrillator, dual-chamber, arrhythmia detection algorithm
|
Introduction |
|---|
|
TopAbstractIntroductionDetection criteriaImplemented algorithmsBiotronik SMARTELA PARAD+Guidant Atrial View and...St Jude morphology...Medtronic logical tableRecent clinical resultsDiscussionConclusionsAcknowledgementsReferences |
|---|
Fifteen years after the commercial release of the first generation of implantable cardioverter defibrillators (ICDs), dual-chamber devices were implanted in humans [1]
. This upgrade was necessary to (1) offer dual-chamber pacing, and (2) increase the accuracy of arrhythmia detection algorithms. Since then, various approaches have been developed to refine the detection and treatment of tachyarrhythmias. We review the clinical data recently published on all current detection criteria, and summarize the characteristics and performance of detection algorithms included in marketed dual-chamber devices. The accuracy of different analyses of detection and research perspectives are also discussed. |
Detection criteria |
|---|
|
TopAbstractIntroductionDetection criteriaImplemented algorithmsBiotronik SMARTELA PARAD+Guidant Atrial View and...St Jude morphology...Medtronic logical tableRecent clinical resultsDiscussionConclusionsAcknowledgementsReferences |
|---|
The design of dual-chamber devices was based on single-chamber experience. While nearly all criteria implemented in single-chamber ICDs have been included in dual-chamber devices, several features have been added to incorporate the atrial information:
- A comparison between the number of atrial and ventricular events during tachycardia.
- An analysis of atrioventricular (AV) association.
- The number and relative position(s) of atrial sensed events contained between ventricular events, and/or
- R-wave morphology, effect of premature stimulation, detection of long ventricular cycles.
The addition of these new criteria to improve the diagnostic specificity of the algorithm increased the risk of ventricular tachycardia (VT) under-detection, and explains the diversity of approaches adopted with respect to diagnostic sensitivity, according to the level of risk perceived by each manufacturer.
Definitions
The definitions of true positive (TP) vs false positive (FP), and of true negative (TN) vs false negative (FN) tachycardia detection are presented in Table 1. Accordingly, the following definitions are generally applied when examining the performance of detection algorithms:
|
- Sensitivity = TP/(TP + FN) = VT+/clinical VT.
- Specificity = TN/(TN + FP) = SVT/ST–/clinical SVT/ST.
- Positive predictive value (PPV) = TP/(TP + FP) = VT+/treated episodes.
- Negative predictive value (NPV) = TN/(TN + FN) = SVT/ST+/untreated episodes.
- VT prevalence=number of clinical VT/total number of documented episodes.
Detection based on single-chamber information
Ventricular rate
All ICDs use the mean ventricular rate as a first criterion for the detection of tachycardia or fibrillation. This criterion may be reached in various ways (number of consecutive short intervals, number of ventricular intervals shorter than a programmed tachycardia or fibrillation interval within a sliding window). When the number exceeds a programmed value, tachycardia or fibrillation is detected. This criterion intends mainly to (a) mitigate the risk of ventricular undersensing during ventricular fibrillation (VF) or, conversely, (b) limit the risk of inappropriate therapy delivery for frequent premature ventricular complexes occurring during sinus rhythm.
Sudden onset of ventricular activity
This criterion was developed to distinguish tachycardia with a gradual rate increase, as in the case of sinus tachycardia (ST), from true arrhythmic events, and may include two separate analyses: (a) detection of a sudden shortening of the ventricular interval at the onset of tachycardia, and (b) the atrial vs ventricular origin of this acceleration. The first analysis is implemented in nearly all devices, and is generally called "sudden onset". It may be based on monitoring of atrial intervals, as in the case of automatic fallback (mode switch) for the management of atrial fibrillation by pacemakers, or it may monitor ventricular intervals on a beat-by-beat comparison of the last interval with an average of several, or with a single preceding interval(s). Current ICDs generally monitor ventricular onset only. Sudden onset is highly specific in the rejection of ST [2,3]. However, its sensitivity is low when the transition from sinus rhythm to VT is subtle, for instance during exercise [4]. It may also be obscured by premature ventricular complexes occurring before the onset of tachycardia.
Ventricular stability
Ventricular interval stability is a criterion, which aims mainly at detecting episodes of atrial fibrillation (AF), during which the cardiac cycles are generally more unstable than during all other tachyarrhythmias monitored by ICDs. Different methods can be used to examine ventricular stability in a sliding window:
- Analysis of the shortest/longest intervals, standard deviation (SD), the number of cycles exceeding the mean ± SD value, the peak(s) within the histogram of ventricular intervals, etc.
- On a second level, which appears more specific, identification of intervals longer than a reference, updated beat-by-beat during tachycardia (PARAD+, ELA Medical, Montrouge, France), or longer than the VT detection interval (Medtronic Inc., Minneapolis, MN).
Stability is highly specific for rejection of AF with a slow ventricular response [2], though does not reliably discriminate AF from VT at faster rates, as the RR intervals during AF become more stable [2,5,6]. Furthermore, the setting of this parameter to highly specific values may lead to false negative detection during unstable VT. If monomorphic VT becomes markedly irregular or has been slowed by antiarrhythmic drugs, stability may even hinder the delivery of VT therapy [7–9].
Atrial count, rate and stability
Dual-chamber devices could analyze the mean atrial rate by the same method as the ventricular rate. However, the inclusion of atrial blanking periods after ventricular sensing in all devices, except those manufactured by Medtronic, may interfere with the implementation of this criterion, since atrial events may be blanked. On the other hand, atrial far-field sensing may lead to an overestimation of the atrial rate. To enhance atrial sensing, recent devices include a very short atrial blanking period combined with a decrease in sensitivity. Detection algorithms must manage these risks. In practice, only devices capable of delivering therapy to the atria, such as the Medtronic Jewel AF [10], use the stability criterion to discriminate AF from organized atrial tachycardias in order to adapt the therapy to the tachyarrhythmia detected. Preliminary data have also recently been presented on a new algorithm developed by Guidant [11].
EGM width and morphology
Ventricular width and morphology criteria are based on the observation of a narrow QRS on the surface electrocardiogram, when impulses are normally conducted through the AV node and His-Purkinje system or, more generally speaking, of a variable QRS morphology as a function of variability in ambient rhythm. This surface electrocardiographic observation has been extended to endocardial electrograms (EGM), and implemented in implantable devices using the assumptions that a normally propagated EGM is narrower, and/or has a more rapid slew rate than an ectopic event, due to its faster conduction through the ventricles. These two hypotheses may be tested in the time domain, comparing duration or shape, or in the frequency domain, comparing the spectrum or energy content of EGM recorded during tachycardia vs a reference EGM recorded during sinus rhythm.
The first morphology-based criterion, used in the original AID-B device (Intec Systems Inc., Pittsburgh, PA) and later Cardiac Pacemakers Inc. (St Paul, MN) devices, was the probability density function. This criterion was based on the assumption that signals generated during VF spend more time away from the isoelectric baseline than EGM generated during tachycardia. It was generally programmed "off", and is now abandoned since it was associated with the frequent delivery of inappropriate therapies [12].
Today, three main important and different approaches have been implemented: (1) measurement of the endocardial R-wave width, (2) temporal analysis of EGM morphology, and (3) frequency domain analysis.
QRS width is a simple and intuitive criterion. However, it is highly sensitive to posture, noise interference such as myopotentials, lead maturation [13], changes in intraventricular conduction [14], and exercise [15,16]. Consequently, its predictive value is low [17].
Monitoring of morphology requires more complex calculations. In all algorithms currently in use, the last tachycardia EGMs are compared with a reference EGM memorized during sinus rhythm. The more similar the complexes, the greater the probability that the arrhythmia is of supraventricular origin. If the EGM does not match the sinus template, VT is diagnosed. Two approaches have been used: the first (Ventritex/St Jude Medical) compares the number and areas of the positive and negative deflections of EGMs recorded during sinus rhythm with those during tachycardia [18]. The second (Guidant: vector and timing correlation, VTC) calculates a coefficient of correlation between the reference and the tachycardia EGMs [19]. Correlation waveform analysis (CWA) is a classic temporal analysis which has been abundantly bench-tested, though never implemented in a marketed device [20]. These morphology criteria show some of the limitations of EGM width, and also carry the potential complication of signal truncation due saturation of the sensing amplifiers (clipping). If they are the only single chamber discriminators to distinguish a sudden-onset tachycardia, VT may be inappropriately detected when rate-dependent bundle-branch block develops during SVT [18,21].
Frequency domain analysis of the QRS is more complex, requiring considerably more computing power than time domain analysis, though is now possible with the most recent implantable technology.
An innovative criterion, comparing morphologies of EGMs during sinus rhythm with those in tachycardia in respect of differences between the coefficients of their wavelet transforms [22], has been implemented in single-chamber ICDs. It has yielded promising results in a prospective clinical study [23].
Time and frequency domain analyses can both be implemented with a neural network, which offers the advantage of being self-adaptive to time-dependent changes in the reference signal characteristics. A retrospective study of short-term data retrieved in 16 ICD recipients showed promising results and a distinct advantage of time-domain analysis over EGM width or CWA [24].
Another morphology criterion analyses atrial signals during 1/1 tachycardia to detect temporal or frequency changes in the EGM due to retrograde activation, compared with a normal sinus reference [25]. This criterion has, however, not been implemented, since the analysis is even more complex than in the ventricles, due to the low amplitude of the atrial signals and the higher risk of electrical interference.
Sustained rate duration
This feature has been added for safety in the case of inappropriate inhibition of therapy for long-duration VT. It systematically triggers therapy delivery when a rhythm satisfies the VT rate criterion for a programmed period of time. However, inappropriate therapy will be delivered for SVT if its rate remains above the detection rate for the programmed period [3,26]. The only remedy, in such a case, is to increase the duration of the sustained rate [26].
Detection based on the analysis of the atrioventricular relationship
Chamber of origin
This criterion is used for tachycardias with 1:1 atrioventricular relationships (ELA). The analysis of the chamber of origin consists of labelling the ventricular event(s) defining the first tachycardia interval either as conducted from the atrium, or as a premature ventricular complex. Therapy is inhibited if an onset of supraventricular origin is confirmed [27]. This criterion, classifying the tachycardia on the basis of a single or two ventricular events, may be adversely affected by flaws in atrial sensing [27–29].
A:V events or rate counting
This criterion is theoretically very powerful. The ventricular rate exceeds the atrial rate in approximately 95% of VTs [30–33]. Therefore, additional criteria may be necessary in only 5% of VTs. However, it cannot, alone, classify VT with a 1:1 AV relationship, and is associated with false negative detections in the case of dual tachycardias, for example AF + VT. As has been well documented in standard pacing, low amplitude atrial signals may be undersensed during AF, or simply blanked out by the refractory periods of the device, resulting in incorrect diagnosis of SVT as VT. Conversely, atrial oversensing during VT, usually due to VA far-field sensing, may interfere with this criterion and prevent the delivery of therapy [28,29].
Active stimulation
It has long been known that the response to sequences of late premature atrial stimuli during 1:1 tachycardia, can effectively discriminate VT from atrial or sinus tachycardia [34]. It advances the activation of the other chamber only if the stimulus is delivered in the chamber of origin of the tachycardia. However, the pacing stimulus may be inefficient, failing to modify the sequence of activation during tachycardia, and be the source of an erroneous diagnosis. Biotronik (Berlin, Germany) is the only manufacturer to have implemented this option (Enhanced SMART algorithm) in its dual-chamber ICDs [35,36].
Atrioventricular association
This criterion is used as operators do in the electrophysiology laboratory, exploring the origin of a tachycardia by comparing various endocardial activation sites. Analyzing AV timings, the device aims to classify the tachycardia as originating from the atria, or independently originating from the ventricles. The PR Logic algorithm (Medtronic) uses a similar though unique approach, which will be detailed below.
The AV association criterion monitors the stability of the PR or RP interval(s) during tachycardia, by measuring absolute variations in a sliding window (maximum–minimum value, SD, etc.), or by analyzing PR interval histograms, searching for peak(s) of stable conduction times [27,37]. This PR stability index is then correlated with a comparable RR or PP intervals stability index, to classify the tachycardia as atrial or ventricular in origin. AV association is a reliable diagnostic criterion for all tachycardias where a 1:1 PR or RP relationship is absent. However, PR or RP intervals and degree of AV or VA block may vary during the tachycardia. Far-field R-wave sensing during VT may also lead to an inaccurate diagnosis of 1:1 AV association, withholding therapy for dissociated VT [28,29]. A summary of the advantages and disadvantages of each criterion is presented in Table 2.
|
|
Implemented algorithms |
|---|
|
TopAbstractIntroductionDetection criteriaImplemented algorithmsBiotronik SMARTELA PARAD+Guidant Atrial View and...St Jude morphology...Medtronic logical tableRecent clinical resultsDiscussionConclusionsAcknowledgementsReferences |
|---|
We will now briefly describe detection algorithms included in currently available dual-chamber ICDs, and summarize the results of selected prospective clinical studies. The detection criteria available in various devices are listed in Table 3. Three of these algorithms implement unique features. Two of them, the Enhanced SMART active stimulation algorithm (Biotronik) and the PARAD+ analysis of chamber of origin (ELA), address the complex issue of 1:1 tachycardia classification. An original PR Logic algorithm, examining specific P:R patterns during tachycardia was developed by Medtronic.
|
|
Biotronik SMART |
|---|
|
TopAbstractIntroductionDetection criteriaImplemented algorithmsBiotronik SMARTELA PARAD+Guidant Atrial View and...St Jude morphology...Medtronic logical tableRecent clinical resultsDiscussionConclusionsAcknowledgementsReferences |
|---|
The SMART algorithm first compares the mean atrial with the mean ventricular cycle length (Fig. 1) [35,
36]. If the ventricular cycle length is shortest, VT is diagnosed. If the ventricular cycle length is longer, ventricular stability and the A:V ratio analysis allow the distinction of AF (unstable), from atrial flutter (stable, N:1 associated) from VT (stable, AV dissociated). If both intervals are identical, SMART analyzes the stability of the ventricular rhythm. If the RR interval is stable but PP is unstable, VT is confirmed. If the PP interval is stable, the function analyzes the stability of AV association. In the case of stable 1:1 association, the algorithm delivers active ventricular pacing sequences to determine which chamber pilots the tachycardia [35]. If the tachycardia is atrial in origin, the PP interval remains fixed and therapy is inhibited. Otherwise, VT therapy is activated if a sudden onset has been confirmed.
|
|
ELA PARAD+ |
|---|
|
TopAbstractIntroductionDetection criteriaImplemented algorithmsBiotronik SMARTELA PARAD+Guidant Atrial View and...St Jude morphology...Medtronic logical tableRecent clinical resultsDiscussionConclusionsAcknowledgementsReferences |
|---|
If a majority of RR intervals is detected within the VT zone, ventricular stability is analyzed in a first step, in search of a peak in the RR intervals histogram (Fig. 2) [1,
27,37]. If the rhythm is unstable, AF is diagnosed and therapy is withheld. If the rhythm is stable and is AV dissociated by comparing peak amplitudes of RR and PR intervals histograms, VT is diagnosed, unless the Long Cycle Search is activated, which inhibits therapy if a long ventricular cycle is detected [38,39]. If a N:1 PR association is found, the rhythm is classified as SVT. In the presence of 1:1 PR association, the function examines the mode of onset in the ventricles. If the rate acceleration is gradual, sinus tachycardia is diagnosed. If acceleration is classified as sudden, PARAD identifies the chamber of origin of onset and withholds therapy if in the atrium.
|
|
Guidant Atrial View and VTC |
|---|
|
TopAbstractIntroductionDetection criteriaImplemented algorithmsBiotronik SMARTELA PARAD+Guidant Atrial View and...St Jude morphology...Medtronic logical tableRecent clinical resultsDiscussionConclusionsAcknowledgementsReferences |
|---|
In the Atrial View algorithm, the first version of dual-chamber detection functions implemented in Ventak AV devices (Guidant Inc., St Paul, MN), single-chamber stability and onset criteria have been supplemented by atrial information [40]
. Once a fast rhythm is detected, ventricular and atrial rates are analyzed. If the ventricular rate is significantly faster than the atrial rate, VT is diagnosed. If it is not, enhancement criteria are applied. If the atrial rate is above the AF threshold, and if RR intervals are unstable, the rhythm is classified as SVT. If the atrial rate is below the AF threshold, therapy may still be withheld if no sudden ventricular onset was detected.
Guidant dual-chamber ICDs were recently updated with "vector timing and correlation" (VTC) a new EGM morphology criterion [19], which has been implemented in the Rhythm ID algorithm, available in the latest Vitality devices (Fig. 3). VTC is used if the atrial rate exceeds the ventricular rate. Therapy is inhibited if the QRS morphology, analyzed over 8 points per EGM, matches a QRS stored during sinus rhythm. Otherwise, therapy is delivered if the atrial rate is 
200 bpm, or if the ventricular interval varies by 20 ms. A refinement is now under study for 1:1 tachycardias, which analyzes the relative stability of PR and RP [41], an approach close to active stimulation.
|
|
St Jude morphology discrimination |
|---|
|
TopAbstractIntroductionDetection criteriaImplemented algorithmsBiotronik SMARTELA PARAD+Guidant Atrial View and...St Jude morphology...Medtronic logical tableRecent clinical resultsDiscussionConclusionsAcknowledgementsReferences |
|---|
The morphology discrimination (MD) criterion (Ventritex), implemented in single-chamber devices, has been supplemented with atrial information in the St Jude dual-chamber ICDs [42–
45]. The A/V rate branch + MD algorithm examines the relationship between atrial and ventricular rates, and classifies it as V<A, V=A, and V > A (Fig. 4). It then applies MD and interval stability if V<A, or MD and sudden onset if V=A. Therapy is systematically delivered if V > A.
|
This function has recently been updated with an automatic template feature (ATU) for real-time calibration of the sinus template, in order to overcome the limitations described earlier [46]
. |
Medtronic logical table |
|---|
|
TopAbstractIntroductionDetection criteriaImplemented algorithmsBiotronik SMARTELA PARAD+Guidant Atrial View and...St Jude morphology...Medtronic logical tableRecent clinical resultsDiscussionConclusionsAcknowledgementsReferences |
|---|
All functions we have reviewed thus far are dynamic, calculating beat-by-beat RR and PR intervals, ventricular stability, A:V ratio, AV conduction, etc. A different approach aims to analyze AV conduction assuming that nearly all tachycardias fit predetermined patterns.
The Medtronic PR Logic algorithm analyses the two previous RR intervals for each ventricular event [10,49–53]. Zero, a single, or multiple atrial events may be detected within the first or second half of each RR interval. The position of these events within the ventricular interval defines a pattern, which corresponds to a code (letter) attributed to each ventricular event, such that a sustained tachycardia is defined by a unique sequence (Fig. 5). This sequence is compared with the sequences of letters predefined for typical tachyarrhythmias to classify each episode as VT or SVT/ST to determine if therapy should be withheld or delivered. Clinical studies have shown this function to have a low specificity for 1:1 tachycardias, the maximum anterograde conduction duration being very sensitive to first degree AV block during exercise [33]. The recently released Enhanced PR Logic algorithm aims to correct this deficiency by enabling the setting of the anterograde boundary from 0.5 × RR to a longer value [53].
|
|
Recent clinical results |
|---|
|
TopAbstractIntroductionDetection criteriaImplemented algorithmsBiotronik SMARTELA PARAD+Guidant Atrial View and...St Jude morphology...Medtronic logical tableRecent clinical resultsDiscussionConclusionsAcknowledgementsReferences |
|---|
The clinical results from recent dual-chamber ICDs, summarized in Table 4A, consist of full-length peer-reviewed articles indexed in Medline up to March 2003. Abstracts of studies presented at recent international scientific meetings were also reviewed and are listed in Table 4B. Only studies reporting spontaneous events representing the whole spectrum of tachycardias, including sinus rhythm, were selected. The latter results generally present recent intermediate data pertaining to the Guidant Rhythm ID, Medtronic Enhanced PR Logic and St Jude MD + A/V with ATU functions. Since few papers present data on a per-patient basis, we summarized the results on a per-event basis only.
|
All algorithms were associated with a VT detection sensitivity >98%, and a specificity between 66% and 94%. However, as discussed later, these results must be interpreted with caution. The Biotronik Enhanced SMART (Biotronik) and PARAD+ (ELA) had the highest specificity (93–94%) [54–
56]. Though the performance of the Medtronic Enhanced PR Logic is expected to be higher than its previous version, its specificity remains to be evaluated in-depth. The highest PPV was reported in studies of the Smart algorithm (Biotronik) [36,54], but we will see below that this variable is highly dependent on VT prevalence. Though the patient populations studied are limited, the performance of the new St Jude ATU appears to have significantly improved compared with the earlier version of the A/V + MD algorithm [46]. While the bench results of the new Guidant Rhythm ID, reported by Gold et al., are very promising [19], these observations need to be confirmed in larger prospective clinical studies.
Additional observations have examined specific tachycardias. In a study of the new Guidant Atrial Rhythm Classification (ARC), a specific algorithm developed for devices delivering atrial therapies, Morris et al. reported a 94% accuracy in the classification of 50 AF and atrial flutter episodes [11]. Likewise, Swerdlow et al. stratified the PR Logic data, and found this algorithm to be powerful (98% appropriate classification) in the detection of 132 episodes of AF, though less accurate for atrial tachycardias (88% in 190 episodes) [10]. Similar results were reported by Gillberg et al. [53], warranting the development of the Enhanced PR Logic to improve the classification of 1:1 tachycardia.
The detection of AF, a weakness of the PARAD algorithm [27,28], prompted the inclusion of long cycle analysis in PARAD+, which analyzes all intervals in the tachycardia [39]. Mletzko et al. compared the two algorithms in a set of clinical AF and VT episodes, and showed a significant improvement in the detection of AF by PARAD+ [59]. The specificity of the algorithm increased from 78% to 94% for AF with a slow ventricular response.
Finally, the bench study by Hintringer is the only comparison of the performance of four devices in the same set of 71 SVT/ST and 15 VT episodes recorded during electrophysiological studies [60]. The specificities of the Guidant Ventak AV (Atrial View), Biotronik Phylax AV (SMART), Medtronic Gem DR (PR Logic), and ELA Defender IV (PARAD+) were 11%, 12%, 20% and 28%, respectively. These low values were due to the inclusion of multiple AV junctional tachycardias, rarely observed in ICD populations, and must be considered carefully as events were simulated.
|
Discussion |
|---|
|
TopAbstractIntroductionDetection criteriaImplemented algorithmsBiotronik SMARTELA PARAD+Guidant Atrial View and...St Jude morphology...Medtronic logical tableRecent clinical resultsDiscussionConclusionsAcknowledgementsReferences |
|---|
Comparisons among published studies
While the performances of various ICD arrhythmia detection algorithms vary significantly among studies, meaningful comparisons are difficult to make for several reasons. (1) Preliminary studies report results of bench testing on induced arrhythmias, whereas clinical studies examine the response to spontaneous events. Bench testing does not precisely account for ICD-system differences in electrodes or sense amplifiers, and libraries of induced arrhythmias do not allow the evaluation of arrhythmia onsets and duration. In addition, induced arrhythmias may be markedly different from daily-life events in their rates and AV conduction characteristics. (2) All ICDs have limited memories, and the specific types of data storage determine which arrhythmias are reported. (3) When slowing below the detection rate, single sustained episodes of SVTs, particularly AF, are often counted as multiple episodes in ICD data sets. (4) Detection algorithms are not systematically applied in the same rate zones among the various ICD models. Rate-only detection is associated with a 100% sensitivity except in the case of ventricular undersensing, and a 0% specificity. Comparing a mix of all detection zones may not reflect the actual capabilities of devices that are properly programmed. (5) Detection parameters include settings of minimum rates and duration of VT and VF detection rates, which may vary widely. For example, the rate of nearly 100% of ST occurring in ICD recipients is <140 bpm [34,
61], and the majority of events documented below this rate is ST (Fig. 6). Whether these events are included vs excluded in the analysis profoundly influence the results of a study, according to the performance of the device with respect to ST. On the other hand, setting the detection rate at values >150 bpm limits the detections to mostly VT precluding an analysis of the system specificity with regard to ST or atrial tachycardias. (6) Some studies address narrowly defined types of rhythms, for example slow tachycardias [56], or abnormal supraventricular rhythms such as AF [10,11,59]. Their results are distinctly dependent on the type and rate of documented tachycardias. (7) Whether the VT detection specificity or PPV should be reported remains uncertain. Specificity analyzes the performance of the classification algorithm with respect to SVT/ST and measures the proportion of SVT/ST inappropriately treated by the device, whereas PPV measures the appropriateness of all delivered therapies for VT or SVT [62]. While PPV may be more clinically appropriate to rank "appropriate ICD therapy", it is highly dependent on the detected/treated SVT/ST vs VT ratio, thus, on the programmed detection rate. This explains the lesser dispersion of published PPV than of VT specificity, since the incidence of treated SVT/ST is regularly much smaller than the incidence of VT therapies (Tables 4A and B). On the other hand, specificity is highly dependent on the population of documented events, and on the ST/SVT ratio.
|
These theoretical considerations can be illustrated by an imaginary algorithm associated with a 99% sensitivity and 90% specificity. Two different outcomes according to the detection rate settings are shown in Table 5. A mere decrease in the detection rate, while keeping the same sensitivity and specificity, results in a decrease in PPV from 91% to 66.6%.
|
Analysis of detection accuracy vs therapy appropriateness
The choice between analysis of detection and appropriateness of therapy is another important distinction. In the case of ATP therapy, which immediately follows VT detection, both are generally comparable. However, shock therapy needs charging of the capacitors after VT detection. This prolongation of the classification duration by a few seconds may markedly modify the results, for example by allowing inhibition of therapy for AF, which had been initially detected as VT. In this case, analysis of specificity and negative predictive values were accurate with regard to therapy delivery, though inaccurate with regard to initial detection. The programming of tachycardia persistence (minimum duration to be stored and/or treated by the device) will have comparable effects on the results.
Other analytical considerations
Should data be analyzed per-episode, or per-patient? While, from a theoretical point of view, a per-episode analysis may be more accurate and allow stratifications by types or rates of events, from a clinical point of view a per-patient analysis appears preferable. A single shock for AF may not be "statistically" significant, though is excessive for the patient who has received it.
Some studies apply nominal detection settings [27], while others allow the reprogramming of the device for each patient during follow-up. More uniformity in the study design is desirable.
In most studies, a few patients contribute a large number of analyzed episodes. Statistical methods, such as the Generalized Estimating Equation should be used to correct raw performance measures and eliminate the bias introduced by these patients [63]. The positive predictive accuracy (probability that a treated episode is VT) of the Medtronic PR Logic fell from 88% to 78% when corrected [33]. A comparable analysis has not yet been reported for other algorithms, mainly because it needs to be applied to very large study populations.
|
Conclusions |
|---|
|
TopAbstractIntroductionDetection criteriaImplemented algorithmsBiotronik SMARTELA PARAD+Guidant Atrial View and...St Jude morphology...Medtronic logical tableRecent clinical resultsDiscussionConclusionsAcknowledgementsReferences |
|---|
Detection functions benefit from 20 years of clinical experience with single-chamber ICD technology. The development of dual-chamber detection algorithms in the mid-1990s was a major advance. When customizing the detection for each patient, most algorithms can obviate the delivery of nearly all inappropriate shocks, while preserving a high VT sensitivity. Few algorithms have proven their efficacy at "as-shipped" settings in prospective clinical studies. Since physicians are becoming increasingly busy and have limited time to programme specific parameters, algorithms should be adaptable to all patient profiles at nominal settings.
The definition of "appropriate" therapy may undergo a profound transformation in upcoming years. Some episodes of very rapid VT may be well tolerated and effectively treated by ATP alone. Conversely, slow VT may be haemodynamically unstable, in which case, shock therapy should be promptly delivered. Rate detection seems to have reached its limits in these haemodynamically driven therapeutic indications. Specific sensors need to be developed and incorporated in the detection decision trees [64–67], opening the way to highly promising new applications.
|
Acknowledgements |
|---|
|
TopAbstractIntroductionDetection criteriaImplemented algorithmsBiotronik SMARTELA PARAD+Guidant Atrial View and...St Jude morphology...Medtronic logical tableRecent clinical resultsDiscussionConclusionsAcknowledgementsReferences |
|---|
The authors want to thank Olivier Bazillais (Guidant France), Luc Cheminot (St Jude France) and Philippe Ortiz (Biotronik France) for their advice, and Rodolphe Ruffy for reviewing the manuscript.
|
References |
|---|
|
TopAbstractIntroductionDetection criteriaImplemented algorithmsBiotronik SMARTELA PARAD+Guidant Atrial View and...St Jude morphology...Medtronic logical tableRecent clinical resultsDiscussionConclusionsAcknowledgementsReferences |
|---|
[1] Lavergne T, Daubert JC, Chauvin M, et al. Preliminary clinical experience with the first dual chamber pacemaker defibrillator. Pacing Clin Electrophysiol 1997; 20: 182–188.[CrossRef][Medline]
[2] Swerdlow CD, Chen PS, Kass RM, et al. Discrimination of ventricular tachycardia from sinus tachycardia and atrial fibrillation in a tiered-therapy cardioverter-defibrillator. J Am Coll Cardiol 1994; 23: 1342–1355.[Abstract]
[3] Brugada J, Mont L, Figueiredo M, et al. Enhanced detection criteria in implantable defibrillators. J Cardiovasc Electrophysiol 1998; 9: 261–268.[ISI][Medline]
[4] Himmrich E, Horn T, Przibille O, et al. Ventricular tachycardia with low onset incidence, reasons and relevance. Europace Suppl 2002; 3: A97 [Abstract].
[5] Kuehlkamp V, Mewis C, Suchalla R, et al. Rate dependence of R–R stability during atrial fibrillation and ventricular tachyarrhythmias. Circulation 1998; 98: I-173 [Abstract].
[6] Kettering K, Dörnberger V, Lang R, et al. Enhanced detection criteria in implantable cardioverter defibrillators: sensitivity and specificity of the stability algorithm at different heart rates. Pacing Clin Electrophysiol 2001; 24: 1325–1333.[CrossRef][Medline]
[7] Le Franc P, Kus T, Vinet A, et al. Underdetection of ventricular tachycardia using a 40 ms stability criterion: effect of antiarrythmic therapy. Pacing Clin Electrophysiol 1997; 20: 2882–2892.[CrossRef][Medline]
[8] Garcia-Alberola A, Yli-Mayry S, Block M, et al. RR interval variability in irregular monomorphic ventricular tachycardia and atrial fibrillation. Circulation 1996; 93: 295–300.
[9] Swerdlow CD, Ahern T, Chen PS, et al. Underdetection of ventricular tachycardia by algorithms to enhance specificity in a tiered-therapy cardioverter-defibrillator. J Am Coll Cardiol 1994; 24: 416–424.[Abstract]
[10] Swerdlow CD, Schoels W, Dijkman B, et al. Detection of atrial fibrillation and flutter by a dual-chamber implantable cardioverter-defibrillator. Circulation 2000; 101: 878–885 For the Worldwide Jewel AF Investigators.
[11] Morris MM, Palreddy S, Lovett EG, Lang DJ. Evaluation of atrial rhythm classification algorithm. Europace Suppl 2002; 3: A253 [Abstract].
[12] Langer A, Heilman MS, Mower MM, Mirowski M. Considerations in the development of the automatic implantable defibrillator. Med Instrum 1976; 10: 163–167.[Medline]
[13] Favale S, Nacci F, Galati A, et al. Electrogram width parameter analysis in implantable cardioverter defibrillators: influence of body position and electrode configuration. Pacing Clin Electrophysiol 2001; 24: 1732–1738.[CrossRef][Medline]
[14] Unterberg C, Stevens J, Vollmann D, et al. Long-term clinical experience with the EGM width detection criterion for differentiation of supraventricular and ventricular tachycardia in patients with implantable cardioverter defibrillators. Pacing Clin Electrophysiol 2000; 23: 1611–1617.[CrossRef][Medline]
[15] Duru F, Schönbeck M, Lüscher TF, Candinas R. The potential for inappropriate ventricular tachycardia confirmation using the intracardiac electrogram (EGM) width criterion. Pacing Clin Electrophysiol 1999; 22: 1039–1046.[CrossRef][Medline]
[16] Klingenheben T, Sticherling C, Skupin M, Hohnloser SH. Intracardiac QRS electrogram width—an arrhythmia detection feature for implantable cardioverter defibrillators: exercise induced variation as a base for device programming. Pacing Clin Electrophysiol 1998; 21: 1609–1617.[Medline]
[17] Zellenhoff C, Nebeling D, Przibille O, Himmrich E. A new ICD morphology criterion for tachycardia classification. Europace Suppl 2002; 3: A41 [Abstract].
[18] Duru F, Bauersfeld U, Rahn-Schonbeck M, Candinas R. Morphology discriminator feature for enhanced ventricular tachycardia discrimination in implantable cardioverter defibrillators. Pacing Clin Electrophysiol 2000; 23: 1365–1374.[CrossRef][Medline]
[19] Gold MR, Shorofsky SR, Thompson JA, et al. Advanced rhythm discrimination for implantable cardioverter defibrillators using electrogram vector timing and correlation. J Cardiovasc Electrophysiol 2002; 13: 1092–1097.[CrossRef][ISI][Medline]
[20] Lin D, DiCarlo LA, Jenkins JM. Identification of ventricular tachycardia using intracavity electrograms: analysis of time and frequency domain patterns. Pacing Clin Electrophysiol 1988; 11: 1592–1606.[Medline]
[21] Wieneke H, Dagres N, Caspari G, et al. Intracardiac electrographic morphology stability assessed by a new template matching algorithm. Europace Suppl 2002; 3: A40 [Abstract].
[22] Swerdlow CD, Brown ML, Lurie K, et al. Discrimination of ventricular tachycardia from supraventricular tachycardia by a downloaded wavelet-transform morphology algorithm: a paradigm for development of implantable cardioverter defibrillator detection algorithms. J Cardiovasc Electrophysiol 2002; 13: 432–441.[CrossRef][ISI][Medline]
[23] Merrill JJ, Dorian P, Compton S, et al. Multi-center experience with wavelet dynamic VT/SVT discrimination. Pacing Clin Electrophysiol 2003; 26:4 513 [Abstract].
[24] Rojo-Alvarez JL, Garcia-Alberola A, Arenal-Maiz A, et al. Automatic discrimination between supraventricular and ventricular tachycardia using a multilayer perceptron in implantable cardioverter defibrillators. Pacing Clin Electrophysiol 2002; 25: 1599–1604.[Medline]
[25] Jenkins JM and Caswell SA. Detection algorithms in implantable cardioverter defibrilators. Proc IEEE 1996; 84: 428–445.[CrossRef]
[26] Wieckhorst A, Buchwald AB, Unterberg C. Optimized programming of sustained rate duration in patients with implantable cardioverter-defibrillators and diagnosed atrial fibrillation. Z Kardiol 1999; 88: 426–433.[Medline]
[27] Sadoul N, Jung W, Jordaens L, et al. Diagnostic performance of a dual-chamber cardioverter defibrillator programmed with nominal settings: a European prospective study. J Cardiovasc Electrophysiol 2002; 13: 25–32.[CrossRef][ISI][Medline]
[28] Israel CW, Grönefeld G, Iscolo N, et al. Discrimination between ventricular and supraventricular tachycardia by dual chamber cardioverter defibrillators: importance of the atrial sensing function. Pacing Clin Electrophysiol 2001; 24: 183–190.[CrossRef][Medline]
[29] Nitzsche R, Henry C, Deniaud C. Can we believe in atrial signals to classify arrhythmias? Europace Suppl 2000; 1:Suppl. D D65 [Abstract].
[30] Morris MM, Marcovecchio A, KenKnight BH, et al. Retrospective evaluation of detection enhancements in a dual-chamber implantable cardioverter defibrillator: implications for device programming. Pacing Clin Electrophysiol 1999; 22: 849 [Abstract].[Medline]
[31] Stein KM, Hess M, Hannon C, et al. Simultaneous atrial and ventricular tachyarrhythmias in defibrillator recipients: does AF beget VF? J Am Coll Cardiol 1999; 22: 115A [Abstract].
[32] Swerdlow CD, Gunderson BD, Gillberg JM, et al. Discrimination of concurrent atrial and ventricular tachyarrhythmias from rapidly-conducted atrial arrhythmias by a dual-chamber ICD. Pacing Clin Electrophysiol 1999; 22: 775 [Abstract].
[33] Wilkoff BL, Kühlkamp V, Volosin K, et al. Critical analysis of dual-chamber implantable cardioverter-defibrillator arrhythmia detection: results and technical considerations. Circulation 2001; 103: 381–386.
[34] Jenkins J, Noh KH, Bump T, et al. A single atrial stimulus can distinguish sinus tachycardia from 1:1 paroxysmal tachycardia. Pacing Clin Electrophysiol 1986; 9: 1063–1068.[Medline]
[35] Revishvili A. Dual-chamber implantable cardioverter-defibrillator with active discrimination of supraventricular tachycardia. Prog Biomed Res 1999; 2: 166–171.
[36] Theuns D, Klootwijk AP, Kimman GP, et al. Initial clinical experience with a new arrhythmia detection algorithm in dual chamber implantable cardioverter defibrillators. Europace 2001; 3: 181–186.
[37] Nair M, Saoudi N, Kroiss D, Letac B. Automatic arrhythmia identification using analysis of the atrioventricular association. Application to a new generation of implantable defibrillators. Participating Centers of the Automatic Recognition of Arrhythmia Study Group. Circulation 1997; 95: 967–973.
[38] Singer I, Barold S, Camm J. Non-pharmacological therapy of arrhythmias for the 21st century. The state of the art 2000; Armonk, NY Futura Publishing Co. Inc.
[39] Aliot E, Nitzsche R, De Chillou C, et al. How to further increase the specificity of a dual-chamber tachycardias classification algorithm? Arch Malad C
ur et Vaiss 1998; 91: 175 [Abstract].
[40] Sticherling C, Schaumann A, Klingenheben T, Hohnloser SH. First worldwide clinical experience with a new dual chamber implantable cardioverter defibrillator. Advantages and complications. The Ventak AV II DR investigators. Europace 1999; 1: 96–102.
[41] Jenkins J, Thompson J, Lovett E. Separating ventricular tachycardia with 1:1 retrograde activation from supraventricular tachycardia with 1:1 conduction in dual-chamber implantable cardioverter defibrillators. Europace Suppl 2002; 3: A252 [Abstract].
[42] Luceri RM. Initial clinical experience with a dual chamber rate responsive implantable