Wearable smartwatches for atrial fibrillation detection and burden estimation after ablation: comparison with continuous monitoring

Figures

• 

  Figure 1. Kaplan–Meier analysis of arrhythmia-free survival. One-year arrhythmia-free survival after atrial fibrillation (AF) ablation as detected by implantable cardiac monitor (ICM; black) and by simulated wearable algorithms: Apple Watch AF Burden (blue), Apple Watch Irregular Rhythm Notification (IRN; red), and Fitbit Irregular Heart Rhythm Detection (IHRD; green). Wearables detect significantly more recurrences than standard non-invasive monitors, narrowing the gap with continuous ICM surveillance. The purple lines indicate survival estimates from simulated non-invasive ambulatory ECG monitoring strategies (24-hour, 48-hour, 7-day, and 14-day AECG) at 3, 6, and 12 months. P values represent log-rank comparisons with ICM.

• 

  Figure 2. Comparison of AF burden estimates across monitoring strategies. (A) AF burden for the entire study population calculated using the implantable cardiac monitor (ICM; Black), the Apple Watch AF burden algorithm (Blue), the Apple Watch IRN algorithm (Red) and the Fitbit IHRD algorithm (Green). (B) AF burden for those patients with AT/AF recurrence (non-zero AF burden) calculated with the ICM, and for the Apple Watch AF burden algorithm, the Apple Watch IRN algorithm and the Fitbit IHRD algorithm. (C) ICM-derived AF burden in those patients with missed recurrence detection with the Apple Watch AF burden algorithm, the Apple Watch IRN algorithm and the Fitbit IHRD algorithm.

• 

  Figure 3. Correlation between wearable-estimated and ICM-derived AF burden. Scatter plots showing correlation between AF burden estimated by wearable algorithms and the reference ICM-derived burden. Pearson correlation coefficients (r) are displayed for each algorithm. All wearable algorithms demonstrate excellent correlation (r > 0.97) with the gold-standard ICM-derived AF burden. Dotted black lines represent lines of identity.

• 

  Figure 4. Effect of daily watch usage on detection and AF burden accuracy. (A) One-year arrhythmia-free survival as a function of daily watch usage (1–15 h/day) for each wearable algorithm. (B) Sensitivity for arrhythmia recurrence detection compared with ICM. (C) Correlation (Pearson r) between wearable- and ICM-derived AF burden across usage levels. (D) Median ICM-derived AF burden among patients with missed recurrence detection by daily watch usage. Longer daily wearable use substantially improves detection sensitivity without compromising accuracy of AF burden. Dotted lines indicate values for simulated non-invasive monitoring strategies.

• 

  Figure 5. Impact of patient activity level on detection and burden estimation. (A) One-year arrhythmia-free survival by simulated patient activity level (0%, 25%, 50%, and 75% probability of missing checks). (B) Sensitivity for recurrence detection compared with ICM. (C) Correlation (Pearson r) between wearable- and ICM-derived AF burden across activity levels. (D) Median ICM-derived AF burden among patients with missed recurrence detection at each activity level. Detection performance remains robust despite high patient activity, with minimal clinically relevant missed episodes. Dotted lines indicate simulated non-invasive monitoring results.