Detection of atrial fibrillation with implantable loop recorders in horses

Abstract Background Cardiac arrhythmias in horses are diagnosed by auscultation or electrocardiogram (ECG), which results in a low sensitivity for detecting arrhythmias that occur sporadically. Implantable loop recorders (ILRs) are small ECG devices placed subcutaneously, to automatically detect arrhythmias in human patients. Objectives To test ILRs ability to detect atrial fibrillation (AF) in horses. Furthermore, we hypothesised that anatomical location of the implant site might influence signal quality. Signal quality was evaluated both during exercise and over time. Study design Experimental study. Methods In five Standardbred mares, eleven ILRs were implanted subcutaneously in up to three different positions (Front: pectoral region, Left‐6: sixth left intercostal space and Ventral: xiphoid region) and AF induced. The R‐ and T‐wave amplitudes were measured in all positions over time during AF. AF burden automatically registered by the ILRs over a 2‐month period was compared with selected Holter ECG recordings. Results All three positions had stable R‐ and T‐wave amplitudes during the study period and were of sufficient quality to allow AF detection at rest. The position Left‐6 showed significantly higher R‐ and T‐wave amplitudes compared with the other positions. During submaximal exercise only the Left‐6 position was able to record ECG signals of diagnostic quality. No position yielded diagnostic signals at maximum exercise due to artefacts. Main limitations Few horses and ILRs included and no spontaneous AF episodes were studied. Conclusions This preliminary study indicates that ILRs can be used for AF detection in horses, but the anatomical location is important for optimal ECG quality. Despite insufficient quality during exercise, ILRs were suitable for AF detection at rest. Therefore, the ILR may be a valuable diagnostic tool for detecting paroxysmal AF in horses.


| INTRODUC TI ON
Cardiac arrhythmias occur relatively often in horses both during rest and exercise. Although most arrhythmias are without clinical significance, 1,2 atrial fibrillation (AF) is a common pathological arrhythmia affecting performance in horses. [2][3][4][5][6] Diagnosis of AF is based on cardiac auscultation, confirmed by an electrocardiogram (ECG) showing irregular RR intervals with normal QRS complexes, absence of P waves and the presence of f waves. A Holter ECG may be obtained in cases where paroxysmal AF (PAF) is suspected. Also heart rate monitors are able to distinguish AF from sinus rhythm (SR). 7 However, for practical reasons, the duration of heart rate monitor recordings or Holter ECGs is limited to a few days and electrodes can be cumbersome to keep in place. As a result, short-lasting AF episodes may not be diagnosed. This is particularly problematic as long-lasting AF becomes more resistant to treatment, and early recognition of the disease is therefore essential. 8 Interestingly, a recent study has shown changed treatment strategy in people with PAF diagnosed with ILRs to prevent chronic AF and co-morbidities. 9 Implantable loop recorders (ILRs) can overcome time limitations of Holter ECGs. First applied almost 20 years ago in human patients experiencing unexplained syncope, 10 now their application has been extended to detect AF. 11,12 The ILRs are subcutaneously implanted devices that continuously analyse an ECG of the patient.
Based on changes in RR interval, ILRs are able to automatically detect predefined arrhythmic events such as bradycardia, tachycardia, AF and asystole. Unlike Holter recordings, no continuous ECG is stored, rather clinicians are provided with a report on detected arrhythmias including short ECG examples. In humans, ILRs are used to detect PAF or AF recurrence in asymptomatic patients, with greater diagnostic yield than conventional Holter ECG. [11][12][13][14][15] In veterinary medicine, ILRs have been used in dogs with unexplained syncope in order to identify cardiac causes, [16][17][18] and in three horses to exclude cardiac arrhythmias as the triggering cause of collapse. 19 The aim of the current study was to test ILRs ability to detect AF in horses. We hypothesised that anatomical location of the implant site might influence signal quality. Furthermore, signal quality was evaluated both during exercise and over time.

| Experimental protocol
Five Standardbred mares with a mean age of 8.6 years (range 5-13 years) and a mean bodyweight of 471 kg (range 320-572 kg) were included. This study used a subset of horses that were used to evaluate the longitudinal effect of AF in horses (timeline, Figure 1) and the horses were subjected to euthanasia and post-mortem examination at the end of the study. Atrial fibrillation was induced by tachypacing the right atrium with pacemakers implanted in local anaesthesia and leads were placed in the right atrium through the right cephalic vein as previously described. 20 The Medtronic Reveal Linq ILRs (Medtronic Inc.) (size 44.8 × 7.2 × 4.0 mm, weight 2.5 g) were implanted at different anatomical positions along with pacemaker implantation (Figure 1).
ILRs were inserted through a 1 cm incision into a subcutaneous pocket, anchoring the device to the muscular plane. They were kept in place by the surrounding tissue without further fixation.
Skin was stapled using two skin staples (Appose VLC Auto suture, Coviden) and covered with sterile adhesive dressing (Sorbact, ABIGO Medical). The operation (from incision to closure) lasted less than 5 minutes. The horses were treated with 1 mg/kg bodyweight (bwt) flunixin intravenously (Finadyne ® , MSD Animal Health) for 3 days. Four horses underwent an incremental treadmill testing to fatigue when in persistent AF as previously described. 6 During exercise, ILRs were manually activated (a 10-minute ECG trace can be recorded after manual activation) to cover the latest part of the warm-up period, the exercise test to fatigue and immediate post-exercise period. The horses were equipped with a head collar and an elastic girth to attach the surface ECG.

| Implantable loop recorder
Implantable loop recorders continuously obtain ECG signals to detect RR intervals. Detection of arrhythmia (bradycardia, tachycardia, AF and asystole) is based on these RR intervals. The AF detection algorithm uses a 2-minute ECG interval to detect variation in the RR intervals. The variation is measured by plotting the current RR interval vs the RR interval of the preceding heart beat in a Lorenz Plot ( Figure 2). To refine AF detection, additionally the ILR searches for a P wave between two R waves. A list containing all types of arrhythmia and its respective number of occurrence is stored automatically for the total lifetime of the device. Additionally, the reveal LINQ can store a list containing up to 30 episodes in greater detail, including date and time of onset, duration as well as mean and maximal heart rate. This list is complemented with up to 14 ECG traces (examples are shown in Figure 2). The ECGs carry annotations under each detected R wave and can be used to retrace the ILRs diagnose. If more episodes occur, the ECG of the oldest episodes will be overwritten, but the list containing the total count will not be affected by that.
Total duration of AF episodes is stored as AF burden (percentage in AF per day). Recorded episodes can be downloaded using a programming device or transferred automatically to Medtronic's care link server every 24 hours using a home monitoring device. Battery life is typically 3 years depending on the model. For further details F I G U R E 1 Schematic illustration of timeline of the study (above), anatomical position of the three implantable loop recorders (middle) and implantation procedure (below). ILR, implantable loop recorder; PM, pacemaker implantation; CV, cardioversion. *Interrogation of the ILR, "Front": Superficial pectoral region, "Left-6": The sixth left intercostal space at the level of the shoulder joint, "Ventral": The level of the xiphoid process in a cranio-caudal direction. Picture of the implantable loop recorder used (Medtronic reveal LINQ). Implantation sequence (A-D) and interrogation with the programmer (E, F) are shown I L R a n d P M i m p l a n t a t i o n on the functionality of the ILR, the manufacturer's manual can be consulted.
In this study, each ILR was interrogated immediately after inser- to be turned on. For this study we only focused on automatic detection of AF (threshold ≥6 minutes). If manually activated, the ILRs were set to record 10 minutes of ECG (9 minutes before and 1 minute after the manual activation).

| Data collection
Interrogations with data transfer from the ILR were performed

| Data analysis
Exported PDF report files were used for offline analysis using PDF-X Change (Version 2.5, Tracker Software Products Ltd.).
The AF burden (AF hours/d) was extracted from the AF burden histogram and two physicians reviewed all episodes that were classified as AF using ECG examples from the ILRs to ensure correct detection. ECG recordings were available at 25 mm/s sweep speed with a calibration pulse at the start of each recording.
PDF-X change was used to perform calibration while measuring the ECG recordings.
As the ILR algorithm relies on R-wave detection, the amplitude of the R and T waves was measured from 10 consecutive heart beats from registered AF episodes obtained approximately

| RE SULTS
All 11 ILRs were successfully implanted without complication.
However, during the interrogation after 1 month, one ILR at ventral position was defective and no data were recorded. Therefore, only data from two ILRs at the Ventral position were analysed. In total, 10 ILRs could be used for analysis and mean duration of ILR time per horse from insertion until the last interrogation of the ILR was 70 days (range 52-85 days).

| Anatomical location and signal quality at rest
The R-and T-wave amplitudes for the three ILR positions are shown in Table 1 Table 1). Amplitudes remained stable over time since no significant differences between R-and T-wave amplitude were observed between the measurements obtained approximately 1 and 2 months after implantation (Table 1). ILRs were removed postmortem. The connective tissue in contact with ILRs showed mild calcification. Otherwise, no signs of inflammation could be observed.

| Signal quality during exercise
In the four horses tested on the treadmill, the ILRs sensed R waves

| Atrial fibrillation burden
AF burden showed similar durations for the different ILR positions in the individual horses. An example of AF burden is given in Figure 2.
Overall, all ILRs could detect AF episodes on days where the horses were in AF (diagnosed by auscultation or Holter ECG). The AF burden registered at different location differed slightly.
All horses had been in sinus rhythm for 3 weeks with ILRs implanted (day −21 to 0). None of the ILRs detected a false-positive AF episode during this period. However, horses were kept in box rest the first 2 weeks and allowed paddock turn out in week 3.

| Artefacts and heartbeat misclassification
Undersensing and/or oversensing of R waves ( Figure 2) in ECG traces recorded with detected AF were rare. Undersensing was typically observed when intermittent beats with lower R-wave amplitude were present. Further classification of these beats was not possible from the recorded ECG traces. Oversensing occurred when the R/T-wave amplitude ratio was low. This was typically the case for T waves with high amplitude that were misclassified as R waves, regardless whether the T wave was positive or negative.
In individual cases, episodes of 10-20 seconds lasting artefacts due to movement or muscular activity led to intermittent over-and undersensing of R waves (Figure 2). Since this phenomenon was short lasting, it did not lead to the detection of false-positive AF episodes. Overall, the Front and Ventral ILR were most prone to muscle artefacts.

| D ISCUSS I ON
This study showed that ILR implantation in horses is feasible and safe and that the obtained ECG amplitudes are stable over time. It is possible to detect experimentally induced AF, but the anatomical TA B L E 1 Mean ± SD of R-and T-wave amplitudes from 10 consecutive beats and the percentage of QRS complexes with preceding f waves during AF for the three implantable loop recorders approximately 1 and 2 mo after implantation position is crucial for a high-quality ECG recording and correct detection. Movement artefacts increased during exercise and thereby reduced the diagnostic quality of the recordings.

R (mV) T (mV) f waves R (mV) T (mV) f waves R (mV) T (mV) f waves
The ILR's arrhythmia detection algorithms are R-wave dependent, therefore, high R-wave amplitudes (to ensure a good signal-to-noise ratio) and a high R/T amplitude ratio are essential to correctly identify arrhythmias. The left lateral position (Left-6) was superior to the other positions used in the present study. At this location the ILR could potentially interfere with the saddle girth, although in most horses the girth would be located caudal to the position. The only previous reported study in horses using ILR implanted the ILR at the front pectoral region to detect cardiac arrhythmias as an explanation for collapse.
The ILR used in the previous study did not detect arrhythmias during the two recorded episodes of collapse. 19 Our study suggests that this position is suboptimal as the Front position was associated with low R-wave amplitude and a high prevalence of artefacts.
The quality of the ECG recordings from the ILRs remained stable over time, which support the anticipation that these recorders are capable of recording sufficient quality ECGs over longer time in horses as it is seen in humans. 21 However, artefacts were seen and could potentially lead to false-positive registration of AF. ILR can therefore not replace the manual analysis of a Holter ECG, which is the Golden Standard for AF detection, but the ILR could be useful in suspected cases of intermittent AF as they provide continuous arrhythmia detection over long time.

ACK N OWLED G EM ENTS
This study was conducted at the University Hospital for Large Animals, Department of Veterinary Clinical Sciences, Faculty of Health and Medical Sciences, University of Copenhagen. We gratefully acknowledge all staff members for their support and the caretaking of the animals.

CO N FLI C T O F I NTE R E S T S
No competing interests have been declared.

AUTH O R CO NTR I B UTI O N S
R. Buhl designed the study, analysed the data and drafted the manuscript. E. Hesselkilde designed the study, collected the data and analysed the data. H. Carstensen designed the study and collected the data. M. Fenner collected the data and analysed the data. T.
Jespersen was a major contributor writing the manuscript. J. Tfelt-Hansen assisted in designing the study and performing the surgery. S. Sattler designed the study, analysed the data and drafted the manuscript. All authors read and approved the final version of the manuscript.

E TH I C A L A N I M A L R E S E A RCH
The

OWN E R I N FO R M E D CO N S E NT
Written owner-informed consent was obtained prior to the study.

DATA ACCE SS I B I LIT Y S TATE M E NT
The data are available from the corresponding author on reasonable request.