Session S62.3

Atrial Fibrillation Detection by a Subcutaneous Monitoring Device

G Hindricks, M Taborsky, P Wohlgemuth, G Rieger,
F Beckers, B Albers*

Medtronic SQDM
Arnhem, Netherlands

The Reveal XT subcutaneous insertable cardiac monitor detects Atrial Fibrillation (AF) from the irregularity of the ventricular rhythm. The rhythm is classified every 2 minutes, based on a Lorenz plot of consecutive RR-intervals. A noise rejection algorithm is incorporated to avoid inappropriate detections due to noise and myopotentials. Initial validation of the detection algorithm was performed using standard surface ECG arrhythmia databases and resulted in an episode sensitivity of 99% and a positive predictive value (PPV) of 76%. These standard databases mainly contain well defined sinus rhythm and AF and almost no noise. The present study is a first assessment of the AF detection performance using subcutaneous signals from implanted Reveal XT devices. Patients implanted with a Reveal XT device were monitored by a dedicated Holter recorder, providing a maximum of 46 hours of surface ECG and uplinked data from the implanted device (broadband ECG, sense markers and an irregularity classification based on the Lorenz plot). From the uplinked data, AF detections were reconstructed. True positive (TP), false positive (FP) and false negative (FN) detections were identified, using the annotated surface ECG as a reference. In this study 82 Holter recordings from 60 patients were collected, including a total of 2982 hours of valid recording. In 17 patients (28%) true AF was observed, yielding 83 AF episodes of = 2 minutes and 198.7 hours of AF. 46 patients (77%) had no FPs. The mean detection sensitivity for AF episodes =2 minutes was 90.6% (CI: 77.6% - 100%), mean detection sensitivity for AF episodes =4 minutes was 98.1%. Of the 14 FNs, 13 were from a single patient with a highly regular arrhythmia pattern during AF. PPV of the episode detection was 55.1% (CI: 36.8% - 73.3%). Among all FPs, the majority (88%) were observed in 4 patients and were due to irregular rhythms (sinus arrhythmia, frequent extrasystoles). Only 3% of the FPs were attributed to noise and/or myopotentials. First subcutaneous data from implanted devices show a limited effect of noise and myopotentials on AF detection. In this population, inappropriate AF detections were clustered within a small minority of patients with irregular rhythms that were not annotated as AF but considered to be clinically relevant.

(Abstract Control Number: 104)