Diagnostic value of implantable loop recorder in patients undergoing cryoballoon ablation of atrial fibrillation

Abstract Background Due to limited data, implantable loop recorders (ILR) are not currently recommended by the guidelines to routinely monitor patients after atrial fibrillation (AF) ablation. Aims To validate the diagnostic value of ILR after AF ablation, modern generation ILRs (LINQ) were implanted in patients scheduled for cryoballoon ablation of AF (CBA). Methods We included 29 patients with frequent and symptomatic episodes of paroxysmal AF. ILR was implanted 3 months prior to CBA, and data were collected before and for 6 months after the procedure. The device was programmed to maximize sensitivity of AF/ atrial tachycardia (AT) detection. All EGM recordings were “manually” assessed and annotated as true AF, pseudo AF, unrecognized AF, and episodes with no EGM available. Duration and episode‐based standard performance metrics were evaluated. Results A total number of 5,842 episodes were recorded. A total of 4,403 episodes were true AF, 453 episodes were pseudo AF, and 986 episodes had no EGM available. The device did not recognize 144 episodes of AF. Duration‐based sensitivity was 95.2%, duration‐based specificity 99.9%, duration‐based PPV 99.2%, duration‐based NPV 99.9%, episode‐based sensitivity 98.0%, and episode‐based PPV 91.0%. Misdiagnosis happened in 1 in 10 episodes. Total data review time was 166 hr. Conclusions Implantable loop recorders is a valuable tool in evaluation of AF episodes in patients undergoing CBA. However, for high precision all recorded episodes need to be evaluated “manually.” The memory storage space is too low for frequent AF episodes, resulting in overwriting of stored EGMs and data loss.


| ME THODS
The study enrolled 29 patients scheduled for CBA. ILR Reveal LINQ™ (Medtronic Inc.) was implanted 3 months prior to scheduled CBA. The device was inserted in an electrophysiology laboratory. It was placed in the left parasternal area at the level of the 4th−5th intercostal space 45 degrees to the sternal border, as instructed by the manufacturer. The incision and insertion tools supplied by the manufacturer were used. Patients received local anesthesia; no procedure-related complications were observed (Wong et al., 2016). Patients were followed for the total time of 9 months after ILR insertion, consisting of 3 months prior and 6 months after CBA. The ILR was programmed to maximize sensitivity of AF/ atrial tachycardia (AT) detection to precisely asses the real number of AF episodes and calculate the burden of arrhythmia. ILRs were programmed as follows: AT/AF episodes = ALL, sensitivity = 35 µV and blanking = 150 ms, "ectopy rejection" algorithm = disabled. The device makes a rhythm classification every 2 min; therefore, it is able to detect AF lasting at least 2 min. Up to 14 episodes of AF can be stored with ECG tracing; afterward, the earliest episode is overwritten by most recent episodes. The longest AF episode is always preserved. The transmission of episodes from device memory was conducted daily via remote monitoring system. The device was able to record pauses with time duration longer than 3 s. In case of symptoms, patient could manually activate the ILR with a remote controller and store up to 4 episodes with time duration of 7.5 min (1 min prior and 6.5 min after device activation). All patients were supplied with remote monitoring system (MyCareLink™, Medtronic Inc). Data stored by the device were daily sent to the Carelink ® web system. All data were evaluated on day-to-day basis by a clinician. ECG recordings of all AF episodes that were registered automatically and also ECGs of symptom-related events were assessed and annotated as true AF or pseudo AF. All recordings were independently reviewed by two cardiologists, and in case of disagreement, third expert electrophysiologist was consulted.

| Statistical analysis
All episodes lasting at least 2 min were analyzed and divided into true AF, false AF, unrecognized AF, and episodes with no EGM available.

What's new?
• For the first time, such a large number of arrhythmia episodes have been evaluated automatically by the device and verified by an observer.
• We showed that the data gathered by the device cannot be fully trusted as misclassifications were not infrequent.
• Some unique ILR recordings are presented. Episode-based sensitivity ( Episode PPV was calculated, but no NPV could be established, as a true-negative episode cannot be defined .

TA B L E 1 Duration and episode detection performance
Duration-based metrics were also calculated for 3 months prior to and 6 months after CBA (Table 1.

| RE SULTS
Twenty-nine patients with paroxysmal atrial fibrillation resistant to antiarrhythmic drug therapy were included into the study (20 men, mean age 55.6 years old, range 34-72). All patients had uneventful ILR insertion and successful CBA (all 4 veins were isolated in every patient). CBA resulted in significant reduction of AF burden from 9.4% to 0.2%; the baseline characteristics of the enrolled patients and CBA outcomes are listed in Table 2.
Therefore, the final number of AF true-positive episodes that were possible to evaluate was 4,403 (8,222.6 hr). Additionally in 144 episodes (420.0 hr) initially diagnosed by the ILR as AT, the final diagnosis was changed to AF after evaluation.
The reason for incorrect diagnosis of AF was in most cases supraventricular extrasystoles or external electric interference (noise) -Figures 1, 2, and 3. There were also some episodes of T-wave oversensing ( Figure 4). The ILRs had mainly difficulties to differentiate arrhythmias in short episodes that lasted up to 4 min.
In some cases, the device could not recognize AF when it was present ( Figures 5, 6, and 7). In total, the ILRs did not diagnose accurately 137 episodes of AF, diagnosing atrial tachycardia (AT) instead ( Figure 8).
Patients were advised to send the data on daily basis, yet often the number of episodes recorded by the ILR was so high that the device was not able to store all EGMs. Therefore, it was not possible to "manually" assess all episodes.
It took approximately two minutes to adjudicate each available EGM. With total number of 5,000 EGMs, 10,000 min was required (about 166 hr) to evaluate them all.
The performance-based metrics are summarized in Table 1.

| D ISCUSS I ON
In the current study, we assessed the clinical value of ILR in arrhythmia monitoring and categorization in patients undergoing CBA and showed that approximately 10% of recorded arrhythmia episodes were misclassified by the automatic, device-based method. For precise and reliable arrhythmia assessment, all episodes needed to be verified "manually." In the current study, the most recent generation of ILR-Reveal LINQ was tested. Despite small size, it is packed with upgraded software when compared to older generations of ILR. The new generation device was supposed to overcome many issues reported for its predecessor (Gunda et al., 2015;Hindricks et al., 2010). It had very good duration-based sensitivity and specificity ( Table 1) that stands in line with earlier observations . The longer the device was active and the longer were the AF episodes, the better were sensitivity and specificity. In our study, we observed patients with high number of AF episodes that transferred to large number of EGMs. This might be the reason for less false-positive episodes than indicated in earlier studies (Damiano et al., 2016). To make the device more sensitive to AF, ectopy rejection algorithm was programmed off. In our study, we attempted to record each AF episode,  2 min. This observation is also in concordance with earlier studies .
Current ESC guidelines define episode of AF as lasting at least 30 s (Lip et al., 2019). The 2-min loop that is used to check heart rhythm by ILR will obviously skip some short episodes of arrhythmia. Those short runs of AF are not uncommon. They were often recorded when patient pressed symptoms button, and the device stored longer EGM. It is also not possible to evaluate how many clinically silent AF episodes lasting less than 2 min happened.
Regarding issues with limited storage and inaccuracy when assessing short episodes, it is still frustrating that some AF episodes remain undetected despite continuous monitoring. The clinical significance of short AF episodes remains uncertain (Steinberg, O'Connell, Li, & Ziegler, 2018). The total number of AF episodes, including these lasting less than 2 min, is definitely greater than indicated by the ILRs.

| Clinical implications
The However, ILR data can explain the nature of patient symptoms by providing a firm correlation; occasionally, AF-like symptoms are caused by a different arrhythmia (mainly atrial extrasystoles in our experience), and AF ablation/re-ablation would be a mistake. We believe that with 10% arrhythmia misclassification rate by the ILR, "manual" verification of automatic arrhythmia categorization should be considered when important decisions are made on the basis of ILR data. With reliable information regarding AF recurrences, it could be possible to decide in some patients with borderline CHA 2 DS 2 -VASc scale score whether the anticoagulation therapy should be continued or might be stopped. It is still debatable whether a male with 1 point or a female with 2 points in that scale should be anticoagulated. In such cases, reliable ILR data could provide additional guidance. The same holds true for antiarrhythmic drugs use; accurate information about AF episodes, including asymptomatic ones, could guide optimal pharmacological treatment.

| Limitations
The main limitation of the study is relatively small study group.
However, in study population there was a large number of observed episodes.

F I G U R E 7
Episode of AF incorrectly diagnosed as AT