Key points are not available for this paper at this time.
This editorial refers to ‘Tachycardia detection performance of implantable loop recorders: results from a large ‘real-life’ patient cohort and patients with induced ventricular arrhythmias’ by K. Volosin et al., on page 1215. More than a decade ago an implantable loop recorder (ILR) was launched to detect bradyarrhythmias as a cause of syncope, dizziness, and feeling unwell. It is now recognized as an indispensable tool when serial Holter recordings, long-term external event loop recording, tilt test, or an electrophysiological study fail to unmask the cause of syncope.1 The same holds true for unexplained palpitations. In about a quarter of patients with syncope, which remains unexplained despite various diagnostic methods, ILR disclosed a bradyarrhythmic event during syncope,2 whereas in about 10% tachyarrhythmias could be related to syncope.3 Growing experience brought out that ILR showed a superior diagnostic power to tilt table testing and programmed electrical stimulation in unexplained syncope and dizziness, and therefore conquered a prominent diagnostic position in this condition.4 Initially, ILR was not designed for the independent automated detection of various types of tachyarrhythmias due to the limited diagnostic programmability and the small storage capacity. A recent upgrade of diagnostic algorithms for atrial fibrillation improved the accuracy of ILR.5 However, oversensing of myopotentials and P-waves, frequent atrial extrasystoles, and R-wave undersensing undermine its diagnostic performance in a small percentage (15%) of patients with atrial fibrillation. Remote monitoring and follow-up of the performance of implanted cardiac devices through the Internet emerged at the same time as ILR deployment. Apart from timely identification of technical failure of leads and device pacing and sensing algorithms, the detection of asymptomatic arrhythmias and true arrhythmic events related to syncope and other symptoms constituted a major diagnostic therapeutic leap forwards, leading to appropriate treatment measures. Combining remote monitoring with ILR to transmit the automated or manually stored electrocardiogram (ECG) rhythm strips to the server and thereafter to the clinic was the following logical step. Currently in the Netherlands ∼11 000 patients, predominantly implantable cardioverter-defibribllator (ICD) patients, use the Medtronic CareLink server, and ∼400 of them transmit ILR data. In this issue of the Journal, Volosin et al.6 report on the diagnostic performance of tachycardia by application of new ILR diagnostic algorithms (Medtronic Reveal). The performance was tested in a large, unselected patient cohort, and secondly in an external ILR model to assess the diagnostic performance of true ventricular tachycardia (VT) and ventricular fibrillation (VF) induced during ICD insertion. Comparable with other cardiac devices, detection and diagnosis of tachycardia is solely based on interval detection and comparison, the nominal (but programmable at discretion) number of successive QRS complexes, their stability, and the onset and offset pattern of tachycardias. Fulfilling of a set of programmed parameters includes the presence of tachycardia and storage in the ILR memory with a programmable duration of the episode. Because the primary study target was correct VT and VF detection, the conventional VT and fast VT zones of the ICD programmes were applied. Conceivably, because VT cannot be discriminated from various types of supraventricular tachycardia (SVT), the diagnosed VT or fast VT could also be SVT with similar rates. In a nearly 4-month observation period, the ILR identified 1909 tachycardia episodes of maximum 27 min ECG storage in 15% of 2190 patients with enabled programming for tachycardia detection. Tachycardias with RR intervals with maximum range of 520–250 ms (VT) were correctly detected in 78%, with false classification due to noise and P- or T-wave oversensing leading to double counting in the remaining episodes.6 Tachycardias with shorter RR intervals ranging from 400 to 150 ms (fast VT) were correctly detected in ∼8% of episodes, while incorrect detection could be attributed to the same reasons as in VT episodes. In total, a correct detection of all tachycardia episodes was found in about 68%. Notably, the percentages of correct detection did not differ between ILR programmed according the nominal (manufacturer) and individually programmed parameters of tachycardia detection. Operator inspection of all 1909 tachycardia episodes showed only 12 true VT, including two fast VT in 10 (0.5%) patients. In principle, the study design excluded information on missed VT episodes (false-negative performance) due to absence of other recording methods as long-term Holter recording. In the set-up of detection performance of induced VT or VF, three ECG leads were positioned in such manner that the recording mimicked various potential vectors of the conventional ILR chest implantation site.6 The correct classification of the 1475 available, induced VF and VT episodes was ∼95% whereas misclassification arose because VT escaped the programmed detection rate or due to sensing problems or conflicting noise detection. In the opinion of investigators, the 100% correct detection was not achievable because in some cases too short recordings of induced VT or VF were analyzable to permit correct detection. After optimization of parameters to also detect slow VT (130 b.p.m.), correct classification increased to 99%. The authors concluded that the new algorithms of the Reveal ILR accurately detected VT and VF but inspection of the detected tachycardia episodes by the cardiologist or allied professional is required to verify the type of tachycardia. This effort appears, however, minimal in view of the very low incidence of true tachycardia including VT, in daily cardiology practice. If this indeed is the case, verification of these one-channel ECG rhythm strips with tachycardia is a precarious action. An automated ILR false-positive diagnosis of tachycardia due to noise, double counting, and QRS mal-sensing is easily visible because printed annotations of detection and automated diagnosis are supportive. However, the discrimination between SVT and VT in a one-lead ECG challenges the operator because diagnostic rules applicable to tachycardia recorded in the rest or ambulatory 12-lead ECG are missing.7 New algorithms are incorporated that automatically adjust QRS wave sensing threshold, improve noise rejection,8 and better detect asystole, bradyarrhythmia, and tachyarrhythmia with a 85% reduction of incorrectly detected arrhythmia episodes compared with previous versions at the expense of undetected true arrhythmia episodes of ∼2%.9 Despite these diagnostic improvements and remote monitoring compensating for the limited ILR storage capacity, the low correct detection performance (∼68%) of all tachycardias reported by Volosin et al.6 is not satisfactory. This can be partly attributed to the unfavourable ILR chest position in many patients to record optimal ECG signals. It is worth mentioning that the authors did not report any diagnostic problem with the discrimination between SVT and VT in this one-channel ECG. This outcome raises the question whether the correct detection of time intervals between successive QRS complexes and counting of beats with shorter intervals, as currently used in pacemakers and ICDs, match sufficiently the high requirements for automated tachycardia discrimination. Automated QRS wave shape analysis with digital algorithms is already available10 and can be implemented in ILR devices to improve the tachycardia detection rate in conjunction with remote monitoring counterbalancing the small ILR storage capacity. Of importance is the issue of nominal vs. elected programming of parameters to detect VT and fast VT reported in this study. A clear difference in the number of correct detections of tachycardia between both parameter settings could not be observed, although nominal settings led to a clearly smaller number of detected episodes per patient-year than with parameters chosen by the physician. The difference can be explained by individually selected lower cut-off heart rates and shorter duration of tachycardias resulting in more detected episodes. Because the pre-set (‘out of the box’) parameters in this study appear to have valid diagnostic power, this put forward the question whether deviation of these settings (at the discretion of a physician) is really needed. Probably, individual programming should be restricted to subgroups of patients that will most benefit from individual tailoring. For example, a recent study of cardiac resynchronization therapy (CRT)11 demonstrated that only 23–45% of CRT recipients obtained greater acute haemodynamic benefit by individual adaptation of the AV and VV delays while in the remaining patients the nominal values did not. When automated tachycardia analysis with ILR emerges as a very effective method, programming of diagnostic parameters is a major issue and manufacturers are invited to adapt nominal values and parameters to upcoming scientific evidence. Though the diagnostic performance of the ILR regarding unexplained syncope and dizziness and palpitations is well recognized, tachycardia detection needs further technical development and clinical studies to classify the ILR as a smart one. Moreover, cost-effectiveness studies and assessment of patient and physician satisfaction with ILR with remote monitoring are needed to convince health-care providers that this approach can be reimbursed and spread out for daily practice. Willem Einthoven, the founder of the ECG and for this reason a Nobel Prize winner, would be very pleased with this ECG application but not fully satisfied because of still present shortcomings. Conflict of interest: none declared.
Norbert M. van Hemel (Thu,) studied this question.