Real-world performance of an enhanced atrial fibrillation detection algorithm in an insertable cardiac monitor

Abstract

Background: Insertable cardiac monitors (ICMs) are used for long-term ECG monitoring. The Reveal LINQ ICM has an improved atrial fibrillation (AF) detection algorithm.

Objective: The purpose of this study was to investigate the algorithm's real-world performance in patients with syncope, cryptogenic stroke, and known AF.

Methods: Consecutive patients with implanted ICM and AF detection parameters automatically set and maintained depending on the indication for monitoring were included. A single reviewer annotated all stored episodes after ICM implant. A second reviewer annotated a random sample of 10% of all detected AF episodes. The episode detection positive predictive value as well as true and false detection rates were determined for AF episodes of different durations.

Results: The study enrolled 3759 patients (1604 [43%] with syncope, 1049 [28%] with known AF, 1106 [29%] with cryptogenic stroke). Overall, 20,659 AF episodes were detected in 1020 patients. The gross episode detection positive predictive value was 84%, 73%, and 26% for all episodes (≥2 minutes) and improved to 97%, 95%, and 91% for detected AF episodes ≥1 hour in the syncope, known-AF, and cryptogenic stroke patient cohorts, respectively. The true (and false) detection rate was 0.23 (0.05), 3.8 (1.4), and 0.23 (0.65) per patient-month of monitoring for the syncope, known-AF, and cryptogenic stroke patient cohorts, respectively. Limiting ECG storage to the longest detected AF episode significantly reduced the burden of episode adjudication without significantly compromising the identification of patients with true AF.

Conclusion: The performance of LINQ ICM is dependent on the AF incidence rate in the population being monitored, the programmed sensitivity of AF algorithm, and the duration of detected AF episodes.

Keywords: Atrial fibrillation; Diagnosis and monitoring; Insertable cardiac monitor.