Abstract 11561: Real-World Performance of an Internal Loop Recorder to Detect Atrial Fibrillation

• Published in: Circulation (New York, N.Y.) Vol. 138; no. Suppl_1 Suppl 1; p. A11561
• Main Authors: Gianni, Carola, Chen, Qiong, Gedikli, Ömer, MacDonald, Bryan C, Della Rocca, Domenico G, Trivedi, Chintan, Mohanty, Sanghamitra, Al-Ahmad, Amin, Burkhardt, J D, Gallinghouse, G J, Hranitzky, Patrick M, Horton, Rodney P, Di Biase, Luigi, Sanchez, Javier E, Natale, Andrea
• Format: Journal Article
• Language: English
• Published: by the American College of Cardiology Foundation and the American Heart Association, Inc 06.11.2018
• ISSN: 0009-7322; 1524-4539

BackgroundInternal loop recorders (ILRs) are increasingly used in AF management, therefore it is important to know the yield of a specific monitoring device/algorithm to detect and monitor AF. The purpose of this study was to investigate the accuracy of the Confirm RX (Abbot Medical) AF detection algorithm in patients with cryptogenic stroke, suspected or known AF, and syncope.MethodsConsecutive patients with an implanted Confirm RX ILR programmed to detect AF episodes ≥2 min were included in the study. Each AF episode was automatically classified by the device using proprietary discrimination criteria. A single reviewer annotated all AF episodes in which electrograms were stored; the duration of each episode was 150 s (30 s before and 120 s after the onset of the episode). The AF burden was defined as the proportion of the time that the patient was in AF over the total monitored time.ResultsA total of 1296 AF episodes were detected in 21 patients (48% female, 67±13 years old) over 65±13 days. Stored electrograms were available for review in 195/1296 detected episodes. The median AF episode duration was 10.8 (7.3-19.3) min; 110 (52%) episodes were ≥10 min, 24 (12%) ≥30 min, and 13 (7%) ≥1 hr. Overall, the algorithm correctly classified 54/195 AF episodes. The main reason for inappropriate AF detection was the presence of frequent PACs (80%), non-sustained AT (6%), sinus arrhythmia (11%), frequent PVCs (2%) and AFL with variable conduction (1%). Using a per episode analysis in the overall population, the positive predictive value (PPV) was 26% for all AF episodes (≥ 2 min), increasing to 32%, 63%, and 77% for detected AF episodes ≥ 10 min, ≥ 30 min, and ≥ 1 hr respectively. When considering the AF burden in the known AF population, the algorithm adjudicated 33.0 hr of true positive vs 29.4 hr of true negative AF time, accounting for 1.3% true vs 1.1% false AF burden.ConclusionIn a mixed population, the Confirm ILR is not accurate in diagnosing AF episodes of ≥ 2 min, with a high rate of inappropriately detected AF episodes. Therefore, when using the nominal programmed settings, careful review of automatically detected episodes is necessary to accurately diagnose AF and correctly quantify the AF burden. Of note, increasing the duration of the detected AF episode, increases the Confirm ILR diagnostic yield, and reprogramming might be useful to reduce the workload of AF episode adjudication.