Atrial fibrillation early detection by remote monitoring in patients with a subcutaneous implantable defibrillator: a single centre prospective study
S Viani, A Canu, M Parollo, S Sbragi, M Giannotti Santoro, G Grifoni, R De Lucia, A Di Cori, L Segreti, L Paperini, T Cellamaro, S Ribechini, G ZucchelliAbstract
Background
The Subcutaneous Implantable Defibrillator (S-ICD) is a guideline established therapy for the treatment of arrhythmic sudden cardiac death in patients without pacing indications. The last A219 S-ICD device generation is completely monitorable from remote (RM) and is provided with a dedicated algorithm to detect atrial fibrillation (AF) episodes.
Purpose
to prospectively investigate the accuracy of early detection of AF episodes by RM in S-ICD patients and its clinical impact in a large volume centre for treatment of cardiac arrhythmias.
Methods
from November 2017 to August 2025, 106 patients (78 M; mean age 54.7±13 years), without a previous AF diagnosis, were implanted with a S-ICD at our institution and regularly followed by RM.
Results
at a mean FU of 43.4±25 months, 508 AF alerts were generated for 34 patients. 117 episodes of "true" AF, lasting more than 30 s, were detected in 13 patients (12%). All the patients, apart 2, were asymptomatic. A DOAC therapy was instituted in all the cases and a rhythm control therapy as deemed indicated. 391 "false" AF episodes were detected in 21 patients. Frequent VEB was the leading cause of misdetection (62%) followed by SVEB (33%) and noise (one patient). No clinical AF episode remained undetected and no clinical event of cerebrovascular o peripheral embolization occurred at FU. In our S-ICD population PPV, NPV, Specificity and Sensitivity for "true" AF episodes detection were 38%, 100%, 77% and 100% respectively.
Conclusions
The incidence of new onset AF episodes is not negligible in a typical S-ICD population. S-ICD RM allows the early detection of "true" AF episodes, even when asymptomatic, with a very high sensitivity and NPV, leading to significant therapeutic adjustments and possibly avoiding AF complications. However, due to the lower PPV and specificity, the algorithm should be improved to reduce the burden of RM notification workload.