Premature ventricular contraction detection and estimation of daily burden by an insertable cardiac monitor

Abstract

Background

Premature ventricular contraction (PVC) burden is a clinically important metric in the context of PVC-induced cardiomyopathy and is commonly obtained via ambulatory electrocardiogram (ECG) monitoring.

Objective

The purpose of this analysis is to characterize the performance of a novel PVC detection algorithm capable of identifying single PVCs and PVC sequences (couplets and triplets) for estimation of 24-hour PVC burden in an insertable cardiac monitor (ICM).

Methods

Performance of the ICM algorithm for detecting PVCs was validated by replaying 748 patient-triggered ICM-recorded ECG episodes from 184 patients through the ICM device. To assess performance over longer ambulatory periods, a validated software model equivalent of the implemented ICM algorithm was evaluated against a 24-hour Holter dataset of 89 patients. The model also was used to evaluate performance on an established reference library from the Massachusetts Institute of Technology and Beth Israel Hospital (MIT-BIH Arrhythmia Database) as a basis of comparison with other published algorithms.

Results

Beat-level validation on the ICM-stored episode dataset yielded a gross PVC sensitivity of 80.1% with a specificity of 99.7%. The correlation between 24-hour Holter burden and ICM algorithm PVC burden was r = 0.95. The sensitivity for identifying patients with PVC burdens ≥10% was 84%, with a patient-level positive predictive value (PPV) of 100%. Beat-level sensitivity of the PVC algorithm evaluated against the MIT-BIH dataset was 87.9% with a PPV of 96.4%.

Conclusion

The ICM algorithm reliably detects PVCs with high sensitivity and specificity. Twenty-four-hour PVC burden measurements demonstrated a strong correlation with a gold standard 12-lead Holter and may provide utility for identifying patients at risk for worsening left ventricular function.